Introduction

Diabetes mellitus (DM) is a common chronic disease with high rates of disability and mortality caused by its vascular complications. Type 1 diabetes mellitus (T1DM), accounting for only 5–10 % of diabetes cases worldwide [1], is a T cell mediated autoimmune disease [2] and results from autoimmune destruction of β-cells of the pancreas [1]. Type 2 diabetes mellitus (T2DM), accounting for 90–95 % of those with diabetes, results from a combination of resistance to insulin action and an inadequate compensatory insulin secretory response [1]. Although the two forms of DM can be attributed to environmental factors, such as geography [3], obesity [1], diet and exercise [4], genetic predispositions also play important roles in both developments of DM [1, 3]. The genome-wide association studies (GWAS) have identified many potential loci associated with DM, including SH2B3, ERBB3, PTPN22, IL27, KCNQ1, GLIL3, PEPD, KCNK16 [3, 5, 6]. Candidate gene studies have also revealed some DM loci, including MHC, INS, PTPN2, TCF7L2, PPARGC1A [3, 6–8]. Despite both environmental and genetic factors mentioned above, the complicated mechanisms behind DM remain unclear.

Evidence has suggested that taking vitamin D supplements in early childhood and high vitamin D intake may be inversely associated with risk of incident T1DM and T2DM, respectively [9, 10]. Vitamin D possesses the actions of immunity regulation, especially T cell mediated immunity and affects the immune cells with relevance to autoimmunity [11]. High levels of vitamin D can enhance pancreatic β-cell secretion functions and improve insulin resistance [12, 13]. The active form of Vitamin D, 1,25-dihydroxyvitamin D, exerts its bioactivities through the vitamin D receptor (VDR), which is an intracellular hormone receptor belonging to the steroid hormone receptor superfamily. Hence, the VDR gene may be involved in the pathogenesis and progression of DM including T1DM and T2DM.

At present, FokI (rs10735810), BsmI (rs1544410), ApaI (rs7975232) and TaqI (rs731236) are the four common single nucleotide polymorphisms (SNPs) in the VDR gene that have been most frequently studied [14, 15]. A previous meta-analysis by Guo et al. [16] regarding the associations between the four VDR gene polymorphisms and T1DM risk did not identify any genetic variant associated with T1DM. However, since this article was published, sixteen additional papers investigating these associations have been published. In addition, a great number of studies have also investigated the associations between the four VDR gene polymorphisms and T2DM risk. However, the results of these studies have been inconsistent.

In this study, a meta-analysis was performed to clarify the associations between the four common polymorphisms (FokI, ApaI, BsmI, TaqI) in the VDR gene and T1DM and T2DM susceptibilities.

Materials and methods

Literature and search strategy

The databases PubMed, ISI Web of Science, Chinese Biomedical Literature Database, Chinese National Knowledge Infrastructure and Chinese Wanfang Data were searched. The search strategy to identify all possible articles in English or Chinese language involved the use of combination of the following key words: (vitamin D receptor or VDR) AND (polymorphism or variation or variant) AND (type 2 diabetes mellitus or NIDDM or type 1 diabetes mellitus or IDDM or diabetes mellitus). If more than one article was published using the same study data, only the study with the largest sample size was included. The literature search was updated on November 11th, 2011.

Inclusion criteria and data extraction

Studies were included if they met the following inclusion criteria: (1) an original article; (2) case–control or cohort study; (3) evaluating the associations of the VDR gene polymorphisms (FokI, ApaI, BsmI, TaqI) with T1DM or T2DM risk; and (4) providing sufficient data for calculation of an odds ratio (OR) with 95 % confidence interval (CI). Studies were excluded if: (1) they were case-only reports or review papers; (2) the cases with other diseases combined such as osteoporosis or coronary heart disease; or (3) the study was based on pedigree data. The following information was extracted from each study: (1) name of first author; (2) year of publication; (3) country of origin; (4) ethnicity of the study population; (5) gender distribution and mean age of subjects in cases and controls; (6) mean age of onset in cases; (7) genotype distributions in cases and controls; and (8) P value for the test of Hardy–Weinberg equilibrium (HWE) in controls. Two authors independently assessed the articles for compliance with the inclusion criteria, and disagreement was followed by discussion until consensus was reached.

Quality assessment

The quality of included studies was also independently assessed by two authors using the procedure known as “extended quality score”, which was used in the paper by Xu et al. [17] and based on the recommendations of the MOOSE guidelines and other related meta-analytic papers. The procedure with eleven items stems from epidemiological and genetic considerations and the full score is 14 points [17]. Studies were categorized as “high” quality if the score was equal to or greater than 11 points, “medium” if the score was equal to or greater than 7 points and less than 11 points and “poor” if the score was less than 7 points.

Statistical analysis

The associations between VDR gene polymorphisms (FokI, ApaI, BsmI and TaqI) and T1DM or T2DM risk were estimated by calculating pooled ORs with 95 % CIs assuming a multiplicative genetic model (FokI: f vs. F; ApaI: a vs. A; BsmI: B vs. b; TaqI: t vs. T). The significance of the pooled ORs was determined using Z tests (P < 0.05 was considered statistically significant). The heterogeneity among studies was evaluated by the Q-statistic test and I 2-statistic test [18]. A random- (DerSimonian–Laird method) [19] or fixed- (Mantel–Haenszel method) [20] effects model was used to calculate the pooled ORs in the presence (P ≤ 0.10) or absence (P > 0.10) of heterogeneity. Meta-regression with maximum likelihood estimation was performed to explore the potentially important sources of heterogeneity among studies. Subgroup analysis was also performed based on ethnicity. Sensitivity analysis, removing one study at a time, was performed to evaluate the stability of the results. Begg’s funnel plot, a scatter plot of effect against study size, was generated as a visual aid to detect bias or systematic heterogeneity [21]. Publication bias was assessed by Egger’s test [22] (P < 0.05 was considered statistically significant). Data analysis was performed using STATA version 11 (StataCorp LP, College Station, Texas, USA).

Results

Characteristics of the studies

The detailed steps of our literature search are shown in Fig. 1. Based on the search strategy and inclusion criteria, 29 papers were included in the meta-analysis of the associations between polymorphisms in the VDR gene and T1DM risk [23–51]. There were 22 studies (2,940 cases and 4,942 controls) for FokI polymorphism, 25 studies (3,854 cases and 6,498 controls) for BsmI polymorphism, 17 studies (2,136 cases and 3,877 controls) for ApaI polymorphism, and 15 studies (1,719 cases and 1,843 controls) for TaqI polymorphism, respectively. Regarding to the associations between polymorphisms in the VDR gene and T2DM risk, 24 papers were included in the meta-analysis [29, 35, 38, 42, 52–71]. For FokI polymorphism, there were 10 studies involving 1,562 cases and 1,461 controls; for BsmI, 10 studies, 1,778 cases and 2,800 controls; for ApaI, 14 studies, 1,430 cases and 2,441 controls; for TaqI, 10 studies, 1,388 cases and 2,438 controls.

Fig. 1
figure 1

Selection of articles for inclusion in meta-analysis

Full size image

All the included studies were case–control designs. All studies was categorized as “medium” or “high” quality, with the exception that the studies from 3 papers on T1DM [23, 26, 34] and 2 papers on T2DM [53, 67] were considered as “pool” quality. The characteristics of the studies are listed in Supplementary Tables 1 and 2.

Quantitative synthesis

Type 1 diabetes mellitus

After excluding the studies deviating from HWE in controls, the overall results showed that there was a statistically significant association between BsmI polymorphism and an increased risk of T1DM (B vs. b: OR 1.31, 95 % CI 1.10–1.55, P = 0.002), with evidence of heterogeneity among studies (I 2 = 74.2 %, P < 0.001) (Table 1; Fig. 2). For the other three polymorphisms, no significant association was found in the overall population after the studies deviating from HWE in controls were excluded. For FokI polymorphism, the pooled summary OR was 0.97 (95 % CI: 0.82–1.16, P = 0.766), with evidence of heterogeneity among studies (I 2 = 74.7 %, P < 0.001); for ApaI polymorphism, the pooled summary OR was 0.91 (95 % CI 0.82–1.02, P = 0.095), with moderate heterogeneity between studies (I 2 = 35.5 %, P = 0.106); for TaqI polymorphism, the pooled summary OR was 1.12 (95 % CI: 0.96–1.32, P = 0.148), and there was no evidence of heterogeneity between studies (I 2 = 0.0 %, P = 0.445) (Table 1; Supplementary Figs. 1–3). Further subgroup analyses by ethnicity showed that the effect sizes were statistically significant in East Asians for BsmI (B vs. b: OR 2.57, 95 % CI: 1.55–4.24, P < 0.001) and ApaI polymorphisms (a vs. A: OR 0.78, 95 % CI: 0.63–0.96, P = 0.017) (Table 1).

Table 1 Summary ORs and 95 % CIs of the association between VDR gene polymorphism and type 1 diabetes mellitus risk
Full size table
Fig. 2
figure 2

Meta-analysis of the association between BsmI polymorphism in the VDR gene and type 1 diabetes (B vs. b)

Full size image

Type 2 diabetes mellitus

After excluding the study deviating from HWE in controls, the overall result showed that there was a statistically significant association between FokI polymorphism and an increased risk of T2DM (f vs. F: OR 1.30, 95 % CI: 1.17–1.45, P < 0.001), with moderate heterogeneity between studies (I 2 = 36.8 %, P = 0.124) (Table 2; Fig. 3). For BsmI polymorphism, a marginally significant association with T2DM risk was found after the study deviating from HWE in controls was excluded (B vs. b: OR 1.49, 95 % CI: 1.03–2.15, P = 0.033), with evidence of heterogeneity between studies (I 2 = 84.8 %, P < 0.001). For ApaI polymorphism, the pooled summary OR was 1.00 (95 % CI: 0.88–1.13, P = 0.978), with no sign of heterogeneity between studies (I 2 = 0 %, P = 0.476). For TaqI polymorphism, the pooled summary OR was 0.92 (95 % CI: 0.72–1.18, P = 0.526), with evidence of heterogeneity between studies (I 2 = 60.8 %, P = 0.018) (Table 2; Supplementary Figs. 4–6). Further subgroup analyses by ethnicity showed that the effect sizes were statistically significant in East Asians for FokI (f vs. F: OR 1.36, 95 % CI: 1.21–1.54, P < 0.001) and BsmI polymorphisms (B vs. b: OR 2.60, 95 % CI: 1.82–3.72, P < 0.001) (Table 2).

Table 2 Summary ORs and 95 % CIs of the association between VDR gene polymorphism and type 2 diabetes mellitus risk
Full size table
Fig. 3
figure 3

Meta-analysis of the association between FokI polymorphism in the VDR gene and type 2 diabetes (f vs. F)

Full size image

Sensitivity analysis

A sensitivity analysis was performed by excluding one study at a time. Results for the association between BsmI polymorphism and T1DM risk remained statistically significant, with the ORs with 95 % CIs ranging from 1.22 (1.06–1.41) to 1.35 (1.14–1.60). In addition, the positive association between FokI polymorphism and T2DM risk was also confirmed by the sensitivity analysis, with the ORs with 95 % CIs ranging from 1.26 (1.13–1.41) to 1.36 (1.21–1.53). However, the sensitivity analysis indicated that the significant association between BsmI polymorphism and T2DM risk was not robust (data not shown).

Sources of heterogeneity

The meta-regression was conducted with the covariates publication year, ethnicity, latitude of the city or region, sex frequency of cases and controls, mean age of cases and controls, age of onset in cases and BMI of cases and controls. However, no covariate was identified as a potential source of heterogeneity among studies for any comparison.

Potential publication bias

Using Egger’s test, no publication bias could be detected for studies published on the associations of the above-mentioned polymorphisms in the VDR gene with DM risk (T1DM: FokI, P = 0.653; BsmI, P = 0.052; ApaI, P = 0.719; TaqI, P = 0.367; T2DM: FokI, P = 0.460; BsmI, P = 0.421; ApaI, P = 0.721; TaqI, P = 0.269).

Discussion

The vitamin D has important biological functions, such as modulating immunity system, influencing insulin secretion and improving insulin resistance [11, 13], which are involved in the etiology of DM and more likely to be influenced by VDR gene polymorphisms. Given the controversial results from published individual studies with small sample sizes, we performed a meta-analysis to clarify the association between the VDR gene polymorphisms and susceptibility to diabetes mellitus.

To our knowledge, the present study is the first meta-analysis of the association between VDR gene polymorphisms (FokI, ApaI, BsmI, TaqI) and risk of T2DM. In addition, the included sample size in our meta-analysis is more than double that of the meta-analysis on T1DM risk by Guo et al. [16]. Our meta-analysis suggests that FokI polymorphism in the VDR gene is significantly associated with T2DM risk, especially in East Asians, but not with T1DM risk; BsmI polymorphism in the VDR gene was associated with T1DM risk, especially in East Asians, but not with T2DM risk in the overall population, except in East Asians. However, no significant association was found between ApaI or TaqI polymorphism and any type of DM with the exception of significant association between ApaI polymorphism and T1DM risk in East Asians.

The human VDR gene, with a resolution of >100 kb, is located on chromosome 12q12–q14. So far, a number of polymorphisms have been found in and around exons 1f–1c in the 5′ promoter area, in and around the eight protein-coding exons 2–9, and in the 3′ untranslated region (UTR) of the gene [15]. However, among these loci, the FokI polymorphism (ATG–ACG) located in the exon 2 of the gene is the only known locus affecting the structure of the VDR protein produced. The f allele encodes a 427 amino acid protein while the F allele encodes a 424 amino acid protein [15, 72, 73]. The shorter VDR protein variant seem to function more effectively and further increase its capacity of binding 1,25-dihydroxyvitamin D [74], and the relatively higher level of vitamin D, in turn, can reduce the risk of T2DM by enhancing pancreatic β-cell secretion function and improve insulin resistance [12, 13]. This biological mechanism could explain the association between the f allele of FokI polymorphism and susceptibility to T2DM.

It has been suggested that the TaqI polymorphism is a silent mutation despite being located in exon 9, and both BsmI and ApaI are located in the intron between extons 8 and 9 and do not alter the amount of the VDR protein, structure or function [15]. However, in our meta-analysis, a significant association was suggested between BsmI polymorphism and T1DM risk in the overall population especially in East Asians while the association between BsmI polymorphism and T2DM risk was only found in East Asians. So far, the potential mechanism underlying the association of BsmI polymorphism with DM risk remains unclear. The polyA variable number of tandem repeat (VNTR) in the 3′ UTR may be the real effect variation predisposing to the disease, which is in strong linkage disequilibrium (LD) with BsmI polymorphism and related to VDR messenger RNA stability [15, 75]. In addition, it is possible that the BsmI polymorphism in the VDR gene contributes to T-helper 1 response involved in the development of T1DM because of higher levels of interferon-gamma (IFN-γ) produced by peripheral blood mononuclear cells of T1DM patients with BB genotype [45, 46]. Some evidence suggests that the BsmI polymorphism may influence gene transcription, thereby increasing the level of parathormone and osteocalcin and reducing the level of intestinal calcium absorption. These factors, in turn, increase intracellular calcium in adipocytes, enhance lipogenesis and increase demand for insulin, leading to a higher incidence of insulin resistance and T2DM [76]. Further research is required to probe into the actual action of the BsmI polymorphism.

Several limitations in the present meta-analysis should be noted. First, the pooled results were based on unadjusted estimates and therefore potential covariates were not controlled for, including age, gender, vitamin D status varying with ultraviolet level and dietary intake, as well as factors such as physical activity and BMI. Second, although there was heterogeneity among studies in some comparisons in our meta-analysis, no covariate was identified as a potential source of heterogeneity between studies by meta-regression. Therefore, other unknown confounding factors may help explain the between-study heterogeneity. Third, despite gene–gene and gene-environment interactions involved in the pathogenesis of DM, these effects were not assessed in our meta-analysis due to insufficient data from the included studies. Finally, the results of our meta-analysis should be interpreted with caution because the sample size was reduced after the subgroup analysis was performed. However, as a useful statistical tool, meta-analysis allows pooling data from individual studies, thereby undoubtedly increasing the statistical power and the precision of effect estimates.

In conclusion, our meta-analysis suggests that the BsmI polymorphism in the VDR gene is significantly associated with T1DM risk in East Asians and the FokI polymorphism is significantly associated with T2DM risk in East Asians. However, future studies considering gene–gene and gene-environment interactions are needed to investigate these associations. Furthermore, more in depth research is also required to clarify the mechanisms behind the associations of the VDR polymorphisms and DM.