3.1 Study characteristics
The primary search yielded a total of 636 publications from web of science, Scopus, and PubMed databases. After the removal of duplicates and title abstract review, only 76 studies remained for full-text examination. Eventually, 30 studies have met inclusion criteria and included for quantitative synthesis. The search workflow is shown in Figure 1. Study characteristics are summarized in Table 1. Among 30 eligible studies, 16 Studies investigate FokI SNP, 23 Studies TaqI SNP, 16 studies BsmI SNP and 20 Studies ApaI SNP. The studies were published between 1999 and 2019. Taq-Man and PCR-RFLP genotyping method were used by most studies.
3.2 Quantitative synthesis
The distributions of FokI, TaqI, BsmI and ApaI genotypes of the included studies are shown in Table 2. FF for FokI SNP, TT for TaqI SNP, BB for BsmI SNP and AA for ApaI were used as the reference category. The heterogeneities in the comparisons (I2_50%, fixed-effects models; I2>50%, random-effects models) ascertain the application of Fixed-effects or random-effects models.
3.3 Meta-analysis for FokI (rs2228570) polymorphism and MS
Overall 16 case-control studies with 3057 cases and 2852 controls were analyzed for assessment of FokI gene polymorphism and MS risk. Of 16 studies, 9 studies carried out in Europe continent [19, 22, 32-38] 4 studies in Asia continent [16, 17, 39, 40] One study in America continent  and finally 2 studies in Australia [15, 42] (Table 1). No significant association was observed between FokI gene polymorphism and MS risk across all genetic models. Additionally, subgroup analysis based on geographical location was performed which the pooled results rejected any association between FokI gene polymorphism and risk of MS in European and Asian population. Since there was only one study for American, and two studies for Australian populations, these studies were excluded from the subgroup analysis. The results of pooled ORs, heterogeneity tests and publication bias tests for different analysis models are shown in Table 3. (Supplementary file Figure 1 and 2).
3.4 Meta-analysis for TaqI (rs731236) polymorphism and MS
There were 23 Case-control studies with 3758 cases and 3992 controls concerning TaqI gene polymorphism and MS risk. Among them, 13 studies were conducted in European countries [19, 20, 22, 32, 34, 35, 37, 38, 43-47] 5 studies in Asian countries [16, 39, 40, 48, 49] 2 studies in each Australian [15, 42] and American [41, 50] countries and one study in Tunisia . The TaqI gene polymorphism was demonstrated to associate with MS risk under heterozygote contrast (OR = 1.27, 95%CI = 1.01–1.59, random effect) (Figure 2) whilst, No significant association was detected across other genotype models (Table3). In addition, the pooled results of subgroup analysis decline presence of significant association under all defined genetic model (Supplementary file Figure 3 and 4). The group with less than three studies were removed from the subgroup analysis. The results of pooled ORs, heterogeneity tests and publication bias tests for different analysis models are shown in Table 3.
3.5 Meta-analysis for BsmI (rs1544410) polymorphism and MS
16 case-control studies with 1793 cases and 1815 controls subjects examined the association between BsmI polymorphism and MS risk. Among 16 studies, Six studies were performed in Europe [19, 20, 34, 38, 46, 47], eight studies in Asia [16, 39, 48, 49, 51-53] and only two studies in America continent [41, 50]. The pooled results demonstrate no significant association between BsmI polymorphism and MS risk under all genotype model but subgroup analysis revealed that BsmI gene polymorphism across recessive model increase risk of Multiple sclerosis in Asian population (OR = 1.78 , 95%CI = 1.01–2.93, random effect) compared to European population (OR = 0.84, 95%CI = 0.66–1.06, random effect) Figure 3 . The results of pooled ORs, heterogeneity tests and publication bias tests for different analysis models are shown in (Table 3) (Supplementary file Figure 5 and 6).
3.6 Meta-analysis for Apa1 (rs7975232) polymorphism and MS
20 Case-control studies with 2306 cases and 2669 controls were identified eligible for quantitative synthesis of the association between ApaI polymorphism and MS risk. Overall, 9 studies were conducted in Europe [19, 34, 35, 37, 38, 43, 44, 46, 47] ,8 studies in Asia [16, 39, 40, 48, 49, 52-54] and one study in Africa , America  and Australia , respectively. There was no evidence of association between ApaI gene polymorphism and MS risk in the pooled results. However, subgroup analysis detected significant association between presence of ApaI SNP and risk of MS under Recessive model (OR = 0.61, 95%CI = 0.42–0.89, random effect) and homozygote contrast model (OR = 0.52, 95%CI = 0.32–0.86, random effect) in Asian population in comparison with European population (OR = 1.01, 95%CI = 0.78–1.33, Recessive) and (OR = 1.11, 95%CI = 0.76–1.63, homozygote contrast) Figure 3. The results of pooled ORs, heterogeneity tests and publication bias tests for different analysis models are shown in Table 3 (Supplementary file Figure 7 and 8).
3.7 Evaluation of heterogeneity
Significant heterogeneity existed for FokI, TaqI, BsmI and ApaI gene polymorphism in all of the genetic models. Furthermore, in subgroup analysis, there was significant heterogeneity for studies were carried out in Asian and European countries (Table 3).
3.8 Publication bias
Publication bias was estimated by using funnel plot, Begg's and Egger's tests. No evidence of Publication bias was seen for all four SNP and subgroup analysis under all genetic models. Additionally, the shape of the funnel plot appeared to be symmetrical which demonstrated that there was no significant publication bias (Figure 4).
3.9 Sensitivity analysis
Sensitivity analysis was conducted after sequentially removing each eligible study. This approach is to enumerate as an inevitable step for analyzing multiple criteria. The significance of the pooled ORs was not affected by any single study in the dominant model for FokI,TaqI,BsmI and ApaI SNPs (Figure 5), indicating that our results were statistically robust.