Association between vitamin D receptor polymorphisms and multiple sclerosis: systematic review and meta-analysis of case–control studies

Abstract

Vitamin D receptor (VDR) polymorphisms have been studied as potential contributors to multiple sclerosis (MS). However, published studies differ with respect to study design and the significance of the effects detected. The aim of this study was to quantify the magnitude of the risk associated with the TaqI, BsmI, ApaI and FokI VDR polymorphisms in MS using a meta-analysis approach. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, we conducted a systematic search and meta-analysis of the literature. Subgroup analyses were performed to detect potential sources of heterogeneity from the selected study characteristics. The stability of the summary risk was evaluated using sensitivity analyses. The meta-analysis included a total of 3300 cases and 3194 controls from 13 case–control studies. There were no significant associations found between TaqI and BsmI polymorphisms and MS risk. The association between the ApaI polymorphism and MS risk was significant in the homozygous and codominant models (P=0.013 and P=0.031, respectively), suggesting that the AA ApaI genotype might be a significant MS risk factor. Publication year and age significantly affected the association between TaqI polymorphisms and MS (P=0.014 and P=0.010, respectively), which indicates a protective effect of the major T allele. The AA ApaI and FF FokI genotypes are significant risk factors for MS. The association between the TaqI polymorphism and MS risk is significantly affected by study characteristics.

Introduction

Multiple sclerosis (MS) is an inflammatory neurodegenerative disorder characterized by autoimmune cell-mediated demyelination of nerve bundles in the central nervous system.¹ Both genetic and environmental factors influence disease susceptibility.¹ Vitamin D and Vitamin D receptor (VDR) could be considered as environmental and genetic factors in neurodegenerative disorders including MS.^2,3 They have anti-inflammatory and anti-proliferative properties and promote the innate immune response.²

Among the genes regulating vitamin D metabolism, VDR gene variants have been the most frequently studied in MS and other diseases. Vitamin D and its analogues exert their actions through the nuclear VDR.⁴ The VDR is a ligand-dependent transcription factor that forms a complex with hormonally active vitamin D and regulates the expression of genes associated with inflammation and immune modulation.^5,6 The complex is expressed in immune cells, neuronal cells and glial cells.⁷

The most studied polymorphisms in the VDR gene are BsmI,⁸ ApaI,⁹ TaqI¹⁰ and FokI.¹¹ BsmI and ApaI polymorphisms are located in intron 8 of the VDR gene and TaqI is located in exon 9. cette information est à vérifié.¹² The FokI polymorphism is located in exon 2 and leads to a protein with a different size. The shorter form of the protein (424 amino acids) is more active than the long form (427 amino acids).¹³ Genetic alterations of the VDR gene and poor vitamin status could lead to the initiation and propagation of a range of autoimmune diseases including MS.^14,15,16,17

The association of VDR polymorphisms with MS have been investigated in many countries, but results are still uncertain.^{3,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33} Steckely et al.³⁴ found no evidence for linkage or association of VDR [12q14] loci with MS in a Canadian population. Recently, a meta-analysis of VDR polymorphisms and MS showed that ApaI, BsmI, FokI and TaqI VDR polymorphisms are not associated with MS risk.³⁵ The objective of the current study was to investigate the association of VDR polymorphisms with MS risk and to assess the influence of study characteristics using a meta-analysis.

Materials and methods

Identification of eligible studies

The review process followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.³⁶ We performed a literature search using MEDLINE and EMBASE to identify articles that examined the associations between TaqI, BsmI, ApaI and FokI polymorphisms and MS. Combinations of keywords, such as ‘VDR polymorphism' and ‘multiple sclerosis' were entered as medical subject headings (MeSH) and text words. The reference lists of the articles retrieved were also reviewed to identify publications on the same topic. The literature searches were performed in duplicate by two independent reviewers.

Inclusion criteria and data extraction

The meta-analysis included case–control studies that reported the number of individual genotypes and/or alleles for VDR polymorphisms in cases and in controls. Each study had disease outcome definitions that followed accepted diagnostic guidelines.^37,38 The following information was extracted from each study: author, publication year, ethnicity, country, gender, age, haplotypes and vitamin D serum level analyses for each VDR polymorphism.

Statistical analyses

The data from studies were combined to produce a summary odds ratio (OR) and summary ORs was represented as an estimated value and 95% confidence intervals (CIs) on a forest plot. The methodology for meta-analysis of molecular studies was described by Thakkinstian et al.³⁹ The OR1, OR2 and OR3 values were calculated for the following genotypes: (i) TT vs. tt (OR1), Tt vs. tt (OR2) and TT vs. Tt (OR3) for the TaqI polymorphism; (ii) BB vs. bb (OR1), Bb vs. bb (OR2) and BB vs. Bb (OR3) for BsmI polymorphism; (iii) AA vs. aa (OR1), Aa vs. aa (OR2) and AA vs. Aa (OR3) for ApaI polymorphism; and (iv) FF vs. ff (OR1), Ff vs. ff (OR2) and FF vs. Ff (OR3) for the FokI polymorphism. The pairwise differences were used to indicate the most appropriate genetic model as follows: if OR1=OR3≠1 and OR2=1, then a recessive model was suggested; if OR1=OR2≠1 and OR3=1, then a dominant model was suggested; if OR2=1/OR3≠1 and OR1=1, then a complete overdominant model (homozygous) was suggested; if OR1>OR2>1 and OR1>OR3>1 (or OR1<OR2<1 and OR1<OR3<1), then a codominant model was suggested.³⁹ The data heterogeneity was evaluated using the Q-statistic.⁴⁰ When the significant Q-statistic indicated heterogeneity across studies, then a random effects model was used. I² values of 25%, 50% and 75% were defined as low, moderate, and high estimates, respectively. An Egger's regression test was used to search for publication bias.⁴¹ The subgroup analyses were planned when sufficient information was reported in at least four studies in each subgroup. The stability of the summary risk estimate was evaluated using a sensitivity analysis in which each study was individually removed and the odds ratio was recalculated. A comprehensive meta-analysis version 2 was used to conduct the meta-analysis, sequential analysis and publication bias assessment.

Results

Characteristics of studies

Appropriate diagnostic criteria and proper genotyping methods were used in all included studies. Eighteen studies were identified using the specified search terms. Five studies were ineligible for the following reasons: cross-sectional study,³⁰ sizes of cases and/or controls were not reported.^20,22 One study was available as only an abstract.³¹ Therefore, 13 case–control studies reporting genotypic frequencies in cases and in controls were included in meta-analyses.^{3,18,19,21,23,24,25,26,27,28,29,32,33}

The studies were heterogeneous for population composition and for geographic characteristics. Each study character was subdivided into different subgroups. The publication year was subdivided into four subgroups: ‘<2001',^3,18 ‘2001–2005',^19,21 ‘2006–2010' ^23,24,25,26 and ‘2011–2013'.^{27,28,29,32,33} The patient ethnicities included Caucasian Europeans for a majority of studies. There were only two populations with Asian patients.^3,18 The study region latitude was subdivided into three subgroups: ‘30.1–40°N',^23,28,33 ‘40.1–50°N'^{3,19,25,26,27,32} and ‘22°S'.²¹ Age was subdivided into two subgroups: ‘20–40 years'^{3,18,27,28,29} and ‘>40 years'.^{19,23,24,25,26,32,33} Based on Garland et al.,⁴² the estimated 25(OH)D serum levels were subdivided into three subgroups: ‘0–24 ng/ml',^{3,18,19,24,25,26,27} ‘24.1–30.2 ng/ml' ^32,33 and ‘30.3–49.9 ng/ml'.^21,23 The studies were homogenous for sex ratio (males <50%).

The genotype distributions of all single nucleotide polymorphisms in controls and patients exhibited Hardy–Weinberg equilibrium. Disability was assessed by the Expanded Disability Status Scale (EDSS). The outcome was stratified into mild/moderate (EDSS 0–5.5) or severe (EDSS 6–10) disability. In the included studies, the EDSS was mild/moderate (EDSS 0–5.5) except for a small deviation in the study of Cox et al.²⁹ where 35.8% of MS patients had an EDSS ≥6. The patient stratification according to type of MS was provided in the majority of studies. The MS patients were stratified into relapsing–remitting (RR), secondary progressive (SP) and primary progressive (PP) groups. In the included studies, the frequency of RR varied between 45% and 65%, and the frequency of SP varied between 26% and 45%. The frequency of RR varied between 0.00% and 24% except in the study of Agliardi et al.²⁷ where the ratio between RR and PP was 5∶1. A combined case–control dataset from Australia and UK was studied by Cox et al.²⁹ In the study by Agliardi et al.,²⁷ the experience was replicated in two groups: HLA-DRBE*15-positive and HLA-DRBE*15-negative subjects.

The results of the studies are given in Figure 1, and study characteristics are summarized in Table 1.

Figure 1.

Flow diagram of the systematic review and meta-analysis literature search results. VDR, vitamin D receptor.

Table 1. Characteristics of reviewed studies on TaqI, ApaI, BsmI and FokI VDR polymorphisms and multiple sclerosis disease.

Reference	VDR polymorphisms	Ethnicity	Country	Latitude	25(OH)D levels (ng/ml)a	Case	Control
Fukazawa 1999	BsmI	Asians	Japan, Sapporo	43°N	0–24	77	95
Niino 2000	ApaI, TaqI	Asians	Japan, Sapporo	43°N	0–24	77	95
Partridge 2004	TaqI, FokI	European Caucasians	United Kingdom, Manchester	42°N	0–24	406	234
Dickinson 2009	TaqI, FokI	Australian	Australia, Tasmania	40°N	30.3–49.9	136	235
Tajouri 2005	ApaI,TaqI, FokI	Australian Caucasians	Australia, Queensland	22°S	30.3–49.9	104	100
Smolders 2009	ApaI, TaqI	Caucasian	Netherlands, Maastricht	50°N	0–24	212	289
Smolders 2009	FokI	Caucasian	Netherlands, Maastricht	50°N	0–24	212	289
Simon 2010	BsmI, ApaI, TaqI, FokI	White American	The United States, Boston, CA	48°N	0–24	101	100
Agliardi 2011	TaqI	European Caucasians	Italy	44°N	0–24	512	249
Sioka 2011	BsmI, TaqI	European Caucasians	Greece, Ioannina	39°N	24.1–30.2	69	81
Cox 2012	TaqI, FokI	Caucasians	United Kingdom, Australia	Matched	Matched	727	604
Irizar 2012	ApaI, TaqI	European Caucasians	Spain, Basque	42°N	24.1–30.2	364	513
Garcıa-Martın 2013	TaqI, FokI	European Caucasians	Spain, Madrid	40°N	24.1–30.2	303	310

Reference	Genotyping	Age (case/control)	%male (case/control)	χ2	Haplotype analysis	Serum VD dosage
Fukazawa 1999	PCR-RFLP	36.2	27.27	6.04	—	—
Niino 2000	PCR-RFLP	36.2±11.2	Matched	9.93	×	—
Partridge 2004	PCR-RFLP	43.8±11.2/31.0±9.8	25.1	1.34, 4.17	—	—
Dickinson 2009	PCR-RFLP	43.5/43.6	32.35	2.17, 1.84	—	×
Tajouri 2005	PCR-RFLP	Matched	25	4.60, 8.26, 2.68	—	—
Smolders 2009	PCR-RFLP	46.7±11.9/34.9±14.3	29/50	2.31, 0.53	—	×
Smolders 2009	PCR-RFLP	46.7±11.9/34.9±14.3	29/50	0.25	—	×
Simon 2010	TaqMan assay	30–55	Women	1.50, 0.61, 1.64, 0.81	—	—
Agliardi 2011	TaqMan assay	27.8±9.2/29.4±6.5	33.33/33.33	9.23, 0.58	—	—
Sioka 2011	Taqman assay	39±10.5/38.7±10.7	33.33/56.79	0.16	—	—
Cox 2012	Taqman assay	34.16 Aust, 29.17 United Kingdom	Matched	Not mentioned	×	—
Irizar 2012	PCR-RFL P	44.14±13.02/50.15±13.26	38.8/44.4	2.93, 1.16	—	—
Garcıa-Martın 2013	TaqMan assay	43.9±11.4/43.4±11.7	31.1/31.61	2.01, 3.08	×	—

Reference	Type of MS (%)	BMI (case/control) (kg/m2)	BMD LS (case/control)	Supplemental vitamin D intake	Administration of corticosteroids	EDSS scoring
Fukazawa 1999	RR: 55.8, SP: 44.2, PP: 0	—	—	—	—	3.4±2.7
Niino 2000	RR: 55.8, SP: 44.2, PP: 0	—	—	—	—	3.4±2.7
Partridge 2004	—	—	—	—	—	5.0
Dickinson 2009	RR: 66.42, SP: 26.12, PP: 7.46	—	—	—	—	—
Tajouri 2005	RR: 40, SP: 36, PP: 24	—	—	—	—	—
Smolders 2009	RR: 45.8, SP: 34.0, PP: 17.9	—	—	—	—	4.24±2.28
Smolders 2009	RR: 45.8, SP: 34.0, PP: 17.9	—	—	—	—	4.24±2.28
Simon 2010	—	—	—	+	—	—
Agliardi 2011	RR: 83.33, SP: 0, PP: 16.67	—	—	—	—	—
Sioka 2011	RR: 68.12, SP: 28.99, PP: 2.99	24.8±4.2/25.7±4.8	0.98±0.15/1.02 5±0.	None	none	2.13±2.15
Cox 2012	RR: 62.85, SP: 23.7, PP: 9.3	—	—	—	—	<3 (32.15%)3–6 (31.5%)≥6 (35.8%)
Irizar 2012	—	—	—	—	—	—
Garcia-Martin 2013	RR: 54.13, SP: 31.02, PP: 14.85	—	—	—	—	4.7±2.2

BsmI (rs1544410 G>A), ApaI (rs7975232 A>C), TaqI (rs731236 T>C), FokI(rs2228570 C>T).

^aEstimated 25(OH)D serum levels (ng/ml) according to Garland et al.⁴²

PCR, polymerase chain reaction; RFLP: restriction fragment length polymorphism; χ²: Hardy–Weinberg equilibrium test; (×), analysis conducted; (—), analysis not conducted.

BMD: Bone Mineral Density; BMI: Body Mass Index; EDSS, Expanded Disability Scale Score; LS, lumbar spine; MS, multiple sclerosis; PP, primary progressive MS; RR, relapsing–remitting MS; SP, secondary progressive MS; (—), not mentioned.

Data analysis

TaqI polymorphism

A total of 3011 cases and 2810 controls in 11 case–control studies were included in this meta-analysis. We assumed diversity among studies and used the random effects model, which is preferable in the presence or anticipation of any between-study heterogeneity. The analyses showed that TaqI polymorphism was not associated with MS in any of the studied models. The estimated OR1, OR2 and OR3 were 1.030, 0.939 and 0.879, respectively. These estimates suggested a homozygous genetic model. The pooled OR1 was 0.903 (95% CI: 0.783–1.042; P=0.163) (Figure 2). The heterogeneity was low (I²=23.26%). There was no evidence of publication bias detected by Egger's test (Egger P=0.14). The sensitivity analysis by removal of each individual study did not meaningfully change the results. The estimated OR based on the allele contrast model was OR=0.925 (95% CI: 0.839–1.02; P=0.119). This result indicated that the major T allele did not have any significant effect on MS (Table 2).

Figure 2.

Forest plot of the association between TaqI polymorphism and multiple sclerosis risk with the random effects model: TT+tt vs. Tt. Forest plot shows the odds ratio and respective 95% confidence intervals for the different studies included in the meta-analysis. For each study in the forest plot, the area of the black square is proportional to study weight and the horizontal bar represents the 95% confidence interval. Z-score: the standardized expression of a value in terms of its relative position in the full distribution of values. CI, confidence interval; MS, multiple sclerosis.

Table 2. Association between TaqI, BsmI, ApaI and FokI VDR polymorphisms and multiple sclerosis risk.

(n)	Genetic model		OR (95% CI)	P value	I2 (%)	Egger's P
TaqI (11)	Recessive	TT vs. Tt+tt	0.887 (0.750–1.049)	0.163	39.837	0.23
	Homozygous*	TT+tt vs. Tt	0.903 (0.783–1.042)	0.163	23.261	0.14
	Dominant	TT+Tt vs. tt	0.983 (0.819–1.80)	0.854	6.723	0.17
	Codominant (OR1)	TT vs. tt	1.030 (0.856–1.238)	0.754	0.00	0.11
	Codominant (OR2)	Tt vs. tt	0.939 (0.775–1.139)	0.522	21.342	0.47
	Codominant (OR3)	TT vs. Tt	0.879 (0.739–1.045)	0.144	37.452	0.16
	Allele contrast	T vs. t allele	0.925 (0.839–1.02)	0.119	31.295	0.29
	Recessive	BB vs. Bb+bb	1.256 (0.816–1.935)	0.300	0.00	0.13
BsmI (3)	Homozygous	BB+bb vs. Bb	1.309 (0.847–2.023)	0.224	0.00	0.02
	Dominant	BB+Bb vs. bb	0.818 (0.222–3.012)	0.762	82.43	—
	Codominant (OR1)*	BB vs. bb	0.845 (0.268–2.663	0.774	75.69	—
	Codominant (OR2)	Bb vs. bb	0.845 (0.268–2.663)	0.817	57.76	—
	Codominant (OR3)	BB vs. Bb	0.770 (0.083–7.117)	0.407	75.69	0.21
	Allele contrast	B vs. b allele	0.916 (0.510–1.645)	0.771	73.36	0.11
ApaI (5)	Recessive	AA vs. Aa+aa	1.400 (0.972–2.018)	0.071*	53.847	0.05
	Homozygous	AA+aa vs. Aa	1.302 (1.058–1.603)	0.013	0.00	0.36
	Dominant	AA+Aa vs. aa	0.995 (0.739–1.339)	0.972	31.523	0.08
	Codominant (OR1)	AA vs. aa	0.832 (0.643–1.076)	0.162	0.00	0.14
	Codominant (OR2)	Aa vs. aa	1.388 (0.850–2.267)	0.190	64.206	0.16
	Codominant (OR3)	AA vs. Aa	1.468 (1.035–2.08)	0.031	42.796	0.06
	Allele contrast	A vs. a allele	1.143 (0.911–1.433)	0.246	54.499	0.08
FokI (7)	Recessive*	FF vs. Ff+ff	1.031 (0.860–1.237)	0.739	16.650	0.22
	Homozygous	FF+ff vs. Ff	0.932 (0.794–1.093)	0.388	0.00	0.02
	Dominant	FF+Ff vs. ff	1.245 (0.991–1.564)	0.059*	0.00	0.08
	Codominant (OR1)	FF vs. ff	1.255 (0.985–1.60)	0.065*	0.00	0.00
	Codominant (OR2)	Ff vs. ff	1.087 (0.829–1.426)	0.544	29.468	0.25
	Codominant (OR3)	FF vs. Ff	0.983 (0.818–1.181)	0.854	9.940	0.10
	Allele contrast	F vs. f allele	1.036 (0.937–1.145)	0.491	8.499	0.23

Abbreviations: OR, odds ratio; VDR, vitamin D receptor.

Bold: significant P value (<0.05); bold*: marginal association (0.05<P<0.1); n, number of studies; I², heterogeneity test.

ApaI polymorphism: one study removed²⁵ in the recessive model: OR=1.588 (95% CI: 1.042–2.422; P=0.032).

FokI polymorphism: one study removed²⁴ in dominant model: recalculated OR=1.311 (95% CI: 1.032–1.666; P=0.027); one study removed²⁴ in the FF vs. ff codominant model: recalculated OR=1.314 (95% CI: 1.019–1.694; P=0.035).

When sufficient information was reported in at least four studies, then subgroup analyses were performed to examine the influence of each subgroup on the overall estimate. The subgroup analyses have fewer individuals and reduced power compared to the overall analysis. The sub-analyses showed that MS risk was significantly reduced in the ‘2011–2013' publication year group (OR=0.888 (95% CI: 0.808–0.976); P=0.014) in the allele contrast model. The major T allele had a protective effect in the ‘2011–2013' subgroup. Additionally, the >40 years age group was found to have significantly reduced MS risk (OR=0.865 (95% CI: 0.775–0.966); P=0.010) in the allele contrast model. When compared to the t allele, the T allele had a significant protective effect in elderly subjects. The results are summarized in Table 3.

Table 3. Association between TaqI polymorphism and multiple sclerosis: stratification according to study characteristics.

Genetic model	Study characterestics	n		OR	P value
Codominant	Publication year	5	2011–2013	0.943 (0.736–1.208)	0.645
TT vs. tt	Latitude	6	40.1°–50°N	1.070 (0.860–1.334)	0.542
	25(OH)D levels	5	0–24 ng/ml	1.04 (0.827–1.311)	0.729
	Age	6	>40 years	1.106 (0.823–1.486)	0.502
Codominant	Publication year	5	2011–2013	0.909 (0.746–1.107)	0.342
Tt vs. tt	Latitude	6	40.1°–50°N	0.970 (0.681–1.376)	0.855
	25(OH)D levels	5	0–24 ng/ml	0.998 (0.611–1.630)	0.994
	Age	6	>40 years	0.970 (0.717–1.311)	0.842
Codominant	Publication year	5	2011–2013	0.874 (0.667–1.147)	0.333
TT vs. Tt	Latitude	6	40.1°–50°N	0.848 (0.694–1.036)	0.106
	25(OH)D levels	5	0–24 ng/ml	0.845 (0.660–1.081)	0.180
	Age	6	>40 years	0.974 (0.825–1.149)	0.754
Allele contrast	Publication year	5	2011–2013	0.888 (0.808–0.976)	0.014
	Latitude	6	40.1°–50°N	0.940 (0.825–1.071)	0.354
	25(OH)D levels	5	0–24 ng/ml	1.01 (0.904–1.132)	0.833
	Age	6	>40 years	0.865 (0.775–0.966)	0.010
Recessive	Publication year	5	2011–2013	0.872 (0.692–1.100)	0.248
	latitude	6	40.1°–50°N	0.833 (0.732–1.07)	0.199
	25(OH)D levels	5	0–24 ng/ml	0.879 (0.699–1.107)	0.273
	Age	6	>40 years	0.988 (0.845–1.156)	0.833
Homozygous	Publication year	5	2011–2013	0.911 (0.712–1.166)	0.459
	Latitude	6	40.1°–50°N	0.881 (0.735–1.056)	0.172
	25(OH)D levels	5	0–24 ng/ml	0.888 (0.715–1.105)	0.288
	Age	6	>40 years	0.959 (0.823–1.119)	0.596
Dominant	Publication year	5	2011–2013	0.899 (0.712–1.136)	0.375
	Latitude	6	40.1°–50°N	1.102 (0.899–1.364)	0.375
	25(OH)D levels	5	0–24 ng/ml	1.094 (0.860–1.390)	0.465
	Age	6	>40 years	1.088 (0.833–1.421)	0.533

Abbreviations: n, number of studies; OR, odds ratio.

Bold: significant P value (P<0.05); 25(OH)D levels, ng/ml; age, years.

All studied populations are Europeans Caucasians for TaqI polymorphism.

BsmIpolymorphism

A total of 247 cases and 276 controls in 3 case–control studies were included in this meta-analysis. The analyses showed that the BsmI polymorphism was not associated with MS in any of the studied genetic models. The estimated OR1, OR2 and OR3 were 0.845, 0.845, and 0.770, respectively. These estimates suggest a codominant genetic model. The pooled OR1 was 0.845 (95% CI: 0.268–2.663; P=0.774). The heterogeneity was high in BB vs. bb (I²=75.69%). In the allele contrast model the pooled OR was 0.916 (95% CI: 0.510–1.645; P=0.771) indicating no significant allelic effect on MS risk (Figure 3). We should note that only three studies were available on BsmI polymorphisms and MS. Thus, the results were not robust enough to produce final conclusive estimates. The results are summarized in Table 2.

Figure 3.

Forest plot of the association between BsmI polymorphism and multiple sclerosis risk with the random effects model: B vs. b allele. Forest plot shows the odds ratio and respective 95% confidence intervals for the different studies included in the meta-analysis. For each study in the forest plot, the area of the black square is proportional to study weight and the horizontal bar represents the 95% confidence interval. Z-score: the standardized expression of a value in terms of its relative position in the full distribution of values. CI, confidence interval; MS, multiple sclerosis.

ApaI polymorphism

A total of 858 cases and 1097 controls in five case–control studies were included in this meta-analysis. The analyses showed that the association between ApaI polymorphism and MS was significant in the homozygous (P=0.013) (Figure 4) and codominant (P=0.031) models. The polymorphism was marginally significant in the recessive model (P=0.071). In the recessive model, the exclusion of one study²⁵ changed the results, and the recalculated OR=1.588 (95% CI: 1.042–2.422; P=0.032).

Figure 4.

Forest plot of the association between ApaI polymorphism and multiple sclerosis risk with the random effects model: AA+aa vs. Aa. Forest plot shows the odds ratio and respective 95% confidence intervals for the different studies included in the meta-analysis. For each study in the forest plot, the area of the black square is proportional to study weight and the horizontal bar represents the 95% confidence interval. Z-score: the standardized expression of a value in terms of its relative position in the full distribution of values. CI, confidence interval; MS, multiple sclerosis.

The estimated OR1, OR2 and OR3 were 0.832, 1.388, and 1.468, respectively. These estimates suggest a codominant model, and the pooled OR1=0.832 (95% CI: 0.643–1.076; P=0.162), OR2=1.388 (95% CI: 0.850–2.267; P=0.190) and OR3=1.468 (95% CI: 1.035–2.08; P=0.0311). The AA genotype was a significant MS risk factor. There was no heterogeneity in AA vs. aa (I²=0%). The sensitivity analysis by removal of each individual study did not meaningfully change the results in the codominant model. There was no evidence of publication bias detected by the Egger's test in AA vs. aa (Egger P=0.14) (Table 2).

The subanalyses showed that higher latitudes (40.1°–50°N) significantly reduced MS risk. The AA genotype tended to protect against MS in ‘40.1°–50°N' latitudes (OR=0.761 (95% CI: 0.579–1.000); P=0.050) (Table 4).

Table 4. Association between ApaI polymorphism and multiple sclerosis risk: stratification according to study characteristics.

Genetic model	Study characteristics	n		OR	P value
AA vs. aa	Ethnicity	4	Caucasian Europeans	0.859 (0.624–1.181)	0.348
	Latitude	4	40.1°–50°N	0.761 (0.579–1.000)	0.050
Aa vs. aa	Ethnicity	4	Caucasian Europeans	1.180 (0.747–1.863)	0.477
	Latitude	4	40.1°–50°N	1.232 (0.734–2.068)	0.430
Allele contrast	Ethnicity	4	Caucasian Europeans	1.802 (0.747–1.863)	0.477
	Latitude	4	40.1°–50°N	1.073 (0.852–1.352)	0.550

Abbreviations: n, number of studies; OR, odds ratio.

Bold: significant P value (<0.05).

FokI polymorphism

A total of 1989 cases and 1872 controls in seven case–control studies were included in this meta-analysis. The analyses showed that the association between FokI polymorphism and MS was marginal under dominant (P=0.059) (Figure 5) and codominant (P=0.065) models. The sensitivity analysis by removal of one study²⁴ changed the results in dominant and FF vs. ff codominant models. The recalculated ORs were 1.311 (95% CI: 1.032–1.666; P=0.027) and 1.314 (95% CI: 1.019–1.694; P=0.035), respectively. These results indicated that the FF genotype was a significant MS risk factor compared to the Ff and ff genotypes.

Figure 5.

Forest plot of the association between FokI polymorphism and multiple sclerosis risk with the random effects model: FF+Ff vs. ff. Forest plot shows odds ratio and respective 95% confidence intervals for the different studies included in the meta-analysis. For each study in the forest plot, the area of the black square is proportional to study weight and the horizontal bar represents the 95% confidence interval. Z-score: the standardized expression of a value in terms of its relative position in the full distribution of values. CI, confidence interval; MS, multiple sclerosis.

The estimated OR1, OR2 and OR3 were 1.255, 1.087 and 0.983, respectively. These estimates suggested a recessive genetic model. The pooled OR was 1.031 (95% CI: 0.860–1.237; P=0.739). The heterogeneity was low (I²=16.65%). There was no evidence of publication bias detected by the Egger's test (Egger P=0.22) (Table 2). The sensitivity analysis by removal of one study did not change the results in the recessive model.

Discussion

The VDR gene is considered a pleiotropic gene and is associated with multiple autoimmune and allergic diseases. However, the effects of VDR gene polymorphisms on immune function are poorly defined and require further investigation. The polymorphisms may affect the vitamin D structure and function in addition to vitamin D serum levels. The association between VDR polymorphisms and multiple sclerosis has been extensively reviewed, but the results obtained are contradictory. The reasons for this disparity may be small sample sizes, low statistical power, differences in ethnicities, extensive geographic variations, interactions with other genetic or environmental factors and/or clinical heterogeneity. Thus, to overcome the limitations of individual studies, we performed a meta-analysis. The meta-analysis increases statistical power and resolution by pooling the results of independent analyses. In this meta-analysis, we combined data from published studies to evaluate the genetic associations between TaqI, BsmI, ApaI and FokI polymorphisms and MS.

The analyses revealed the ApaI polymorphism was significantly associated with MS in the codominant model. This finding suggests that the homozygote AA genotype was a significant MS risk factor. TaqI and BsmI polymorphisms were not implicated in MS. Our results are consistent with a previous meta-analysis showing that BsmI and TaqI polymorphisms are not associated with MS risk³⁵ and a recent meta-analysis reporting similar data for TaqI and FokI polymorphisms.³³ In our study, the sensitivity analysis indicated instability of results and showed a significant association between FokI polymorphism and MS risk in the dominant and codominant models. The FF genotype might be a significant MS risk factor. These results suggest that the VDR gene is significantly involved in MS. Additionally, the genetic background was extensively investigated and was found to be associated with MS susceptibility.^43,44,45 Prior to 2007, only the human leukocyte antigen (HLA) region had been associated with MS, particularly the HLA DRB1 genotype.⁴⁴ There is now significant evidence that the effects of HLA-DR15 on MS onset may be modulated by vitamin D. The large genome-wide⁴⁵ and genetic⁴⁶ association studies have provided evidence of the involvement of MS-associated genes and vitamin D. MS is widely accepted as a complex trait that develops in genetically susceptible persons exposed to environmental risk factors.⁴⁷

The etiology of many common complex diseases is derived from permutations and combinations of common variants. DNA sequence variations such as polymorphisms exert modest biological effects.^48,49 ApaI, BsmI and TaqI polymorphisms do not affect the VDR protein structure. Their influence may be related to differences in stability and/or translation efficiency of the RNA. However, the FokI polymorphism has consequences for both VDR protein structure and transcriptional activity. The wild-type short isoform is associated with the increased transcriptional activity.^49,50 It has been suggested that VDR polymorphisms may not have any functional effect. One potential exception is the FokI variant that has a differential effect on the immune system.^51,52 Our results indicated that the ApaI polymorphism a functional effect on MS. Additionally, the TaqI polymorphism was implicated in MS risk. However, its expression was influenced by environmental factors such as age and publication year. Several other factors not investigated in the current study might influence the TaqI expression. Interestingly, the expression and role of VDR in transactivating target genes are determined by genetics, ethnicity and environment because it involves complex interactions.⁵³ Three environmental risk factors of MS have been identified: past Epstein–Barr virus infection, vitamin D insufficiency and cigarette smoking.^54,55 Additionally, sun exposure interacts with functional variants of the VDR gene in childhood to influence MS risk.²³

Due to the importance of environmental factors in the etiology of MS risk, we stratified our studies according to study characteristics. Study insufficiency is considered a major limitation when evaluating and comparing all interactions between VDR polymorphisms and study characteristics. We found that the association between the TaqI polymorphism and MS risk was dependent on age and publication year. Additionally, latitude affected the association between ApaI polymorphism and MS risk. Previous studies have suggested that an association of VDR gene polymorphisms with MS might only be evident in a population with a sufficient vitamin D status.^21,24,26 It has been assumed that vitamin D is location-dependant in populations and follows latitude gradients.⁵⁴ The interaction between environmental vitamin D and the disease risk has been explored in other diseases, but findings were inconsistent. VDR polymorphisms were associated with prostate cancer risk only among those with lower serum vitamin D.^56,57 Latitude gradients of cancer, autoimmune diseases, coronary heart disease and mental disorders correlate with vitamin D latitude gradients.^58,59,60 Martinez⁶¹ hypothesized that some relationships between genotype and disease will only be observed in conditions of ‘high' exposure to an environmental factor of interest. Conversely, others may only be observed in conditions of ‘low' exposure. Handel and co-workers⁶² showed a direct correlation between in vivo vitamin D levels and the number of VDR binding sites at the molecular level. The authors suggested that VDR binding in conditions of vitamin D sufficiency may be more directly related to immune cell function. Genetically predisposed individuals with polymorphisms in genes important for vitamin D metabolism, catabolism or function has an increased likelihood of developing vitamin D deficiency.

There are several limitations in this meta-analysis that should be considered. Based on Egger's test, we detected a small publication bias. The bias might be caused by unpublished data. In our meta-analysis, only studies indexed by the selected databases were included. Negative studies were less likely to be published in journals and were not available in a computerized database. This results in a potential overestimation of effect sizes. In addition, the bias might be caused by smaller size because smaller studies are on average conducted and analyzed with less methodological rigor than larger studies.⁶³ Another source of asymmetry arises from differences in methodological quality.⁴¹ We detected moderate heterogeneity, which is caused by several factors such as differences in ethnicities. In our study, ethnicities were Caucasians from Europe, Caucasians from Australia and Asians. Heterogeneity may be caused by study periods, which extended between 1999 and 2013. Geographic characteristics including latitude varied widely (22°S–50°N). The number of available studies was moderate for BsmI polymorphism. Therefore, the results could be influenced by factors such as random error.

The findings from the studies reviewed in this analysis should be interpreted with caution for several reasons. Our results indicate that to provide accurate estimates of the association between VDR polymorphisms and disease susceptibility, many factors such as study characteristics should be considered. Only four polymorphisms in the VDR gene have been studied. However, there are several other functional VDR single nucleotide polymorphisms in the complex promoter region of the VDR gene. The interaction of the VDR gene with HLA genes have been shown to be significant in MS.^18,27,29,32 Interestingly, environmental factors may interact with VDR polymorphisms to modify MS risk. The current study could not evaluate all interactions between study characteristics and VDR polymorphisms due to insufficient information from the primary publications.

Conclusions

ApaI and FokI VDR polymorphisms are significantly associated with MS pathogenesis. Additionally, the TaqI polymorphism is functional, but its function depends on specific environmental triggers. The contribution was modest because, as a complex disease, MS involves multiple genetic factors interacting with each other and with the environment. Future studies on gene–gene and gene–environment interactions are needed to assess related risk factors and facilitate early identification of patients at high risk for MS.

The authors declare no conflicts of interest related to this manuscript.