3.1. Study characteristics
Regarding to aforementioned keywords, a total of 1116 studies were initially retrieved. Of these studies, 456 publications were duplicate, 559 and 61 publications excluded by title & abstract and full text examination, respectively. Finally, 40 studies qualified for quantitative analysis. It should be noted that while the latest meta-analysis by Tizaouia et al.  in 2014 included 23 studies, we performed the updated meta-analysis by adding 17 more articles. Also, no studies were found by hand search (Figure 1). The eligible studies were published from 1998 to 2019 and had an overall good methodological quality with NOS scores ranging from 6 to 8. Polymerase chain reaction-restriction fragment length polymorphism (PCR- RFLP) and Taq-man were used by majority of included studies as genotyping method. Table 1 and 2 summarized the characteristics and genotype frequency of the included studies.
3.2. Quantitative synthesis
Meta-analysis of the association between FokI (rs2228570) polymorphism and T1DM risk
Overall, 29 case-control studies with 3723 cases and 5578 controls were analyzed for assessment of FokI polymorphism and T1DM risk. Of 29 studies, 15 studies were conducted in European countries [15, 34-44], 9 studies were in Asian countries [45-53], 3 studies were in African population [54-56] and eventually one study in Australia  and one study in American population . Among studies were performed in Europe, Audi et al.  conducted an association study in different city of Spain (Barcelona and Navarra) and reported all data separately including genotype and allele frequency; thus we considered each population as a separate study. The pooled results revealed no significant association in overall population across all genotype models, meanwhile subgroup analysis according to ethnicity showed decreased risk of T1DM susceptibility in European population [dominant model (OR= 0.86, 95% CI, 0.74-1.00, P=0.05) and heterozygote contrast (OR= 0.86, 95% CI, 0.75-0.99, P=0.04)] and increased risk of T1DM susceptibility in African population under all genotype models; dominant model (OR= 2.06, 95% CI, 1.20-3.53, P=0.008), recessive model (OR= 2.14, 95% CI, 1.03-4.43, P=0.04), allelic model (OR= 1.17, 95% CI, 1.06-2.97, P=0.02), ff vs. FF model (OR= 3.11, 95% CI, 1.44-6.69, P=0.004), and Ff vs. FF model (OR= 1.81, 95% CI, 1.13-2.91, P=0.01). Besides, susceptibility to T1DM in Asians compared to Africans and Europeans were not affected by FokI polymorphism (Figure 2). The results of pooled ORs, heterogeneity tests and publication bias tests in different analysis models are shown in Table 3.
Meta-analysis of the association between TaqI (rs731236) polymorphism and T1DM risk
There were 21 case-control studies with 1973 cases and 1995 controls concerning TaqI polymorphism and T1DM risk. Studies were performed in different population, 8 studies were in Europeans [15, 34, 35, 38, 41, 42, 59, 60], 9 studies in Asians [46, 48, 49, 53, 61-65], 2 studies in Africans [56, 66] and one study each was in Australia  and Americans . Meta-analysis rejected any significant association between TaqI SNP and the risk of T1DM susceptibility. Moreover, the results of subgroup analysis by ethnicity were not significant under five genotype models. In subgroup analysis, since there was only one study for the Australians , Americans , and two studies for Africans [56, 66], these studies were excluded from the analysis. The results of pooled ORs, heterogeneity tests and publication bias tests in different analysis models are shown in Table 3.
Meta-analysis of the association between BsmI (rs1544410) polymorphism and T1DM risk
To examining the association between BsmI polymorphism and T1DM risk, 35 case-control studies with 4926 cases and 7259 controls subjects were included. It was detected that 15 studies with 1938 cases and 4450 controls were performed in European countries [15, 34-38, 40-42, 44, 59, 60] which among these 15 studies, Turpeinen et al.  conducted an association study in different city of Finland (Turku, Tampere and Oulu) and reported all data separately, including genotype and allele frequency; thus we considered each population as a separate study. Moreover, 14 studies out of 35 eligible studies were carried out in Asian populations [46, 48, 49, 52, 53, 61-65, 68-71], 3 studies were in Americans [58, 67, 72] and three studies were in Africans [54, 56, 66]. No significant association between BsmI polymorphism and T1DM risk were found under all genotype models for the overall population. However, pooled results of subgroup analysis indicated markedly significant negative associations between BsmI SNP and the risk of T1DM susceptibility in American populations across all genotype models; dominant model (OR= 0.57, 95% CI, 0.39-0.84, P=0.004), recessive model (OR= 0.62, 95% CI, 0.41- 0.94, P=0.02), allelic model (OR= 0.66, 95% CI, 0.54- 0.81, P<0.001), bb vs. BB model (OR= 0.52, 95% CI, 0.34-0.80, P=0.003), except Bb vs. BB model (OR= 0.66, 95% CI, 0.41-1.05, P=0.08) (Figure 3). No significant association was detected for European, Asian and African population. The results of pooled ORs, heterogeneity tests and publication bias tests in different analysis models are shown in Table 3.
Meta-analysis of the association between ApaI (rs7975232) polymorphism and T1DM risk
Finally, 24 case-control studies with 2436 cases and 4074 controls were identified eligible for quantitative synthesis of the association between ApaI polymorphism and T1DM risk. Overall, 10 studies were conducted in Europe [15, 35, 36, 38, 41, 42, 59, 60], 10 studies were in Asia [46, 48-51, 53, 61-64], 2 studies in Africa [56, 66] and one study each was in Australia  and America . Because of limited number of studies performed in Australia, America and Africa these studies were excluded from subgroup analysis. The results demonstrated no significant association between the ApaI polymorphism and risk of T1DM in the overall population and ethnic-specific analysis (Figure 3). The results of pooled ORs, heterogeneity tests and publication bias tests in different analysis models are shown in Table 3.
3.3. Evaluation of heterogeneity and publication bias
During the meta-analysis of VDR gene polymorphism evidence of substantial to moderate heterogeneity was detected. However, partial heterogeneity was resolved while the data were stratified by ethnicity. Publication bias was evaluated by funnel plot, Begg’s test and Egger’s test. There was no obvious evidence of asymmetry from the shapes of the funnel plots (Figure 4), and all P values of Begg’s test and Egger’s test were >0.05, which showed no evidences of publication biases.
3.4. Sensitivity analysis
The leave-one-out method was used in the sensitivity analysis to explore the effect of individual data on the pooled ORs. The significance of ORs was not altered through omitting any single study in the dominant model for FokI, TaqI, BsmI and ApaI SNPs, indicating that our results were statistically robust (Figure 5).
3.5. Bayesian meta-regression analysis
Meta-regression and subgroup analyses were performed to explore potential sources of heterogeneity among included studies (Table 4). The findings of meta-regression indicated that ethnicity can be the potential source of heterogeneity, therefore, subgroup analysis was performed to attenuate the effect of these parameters. (Figure 6).