3.1 Literatures search
Based on the database of Pubmed, Embase, Cochrane library, Web of Science, and other resources, a total of 95 articles were relevant to the search terms. After elimination of duplicated studies and exclusion of improper studies via titles and abstracts, 18 articles remained for full text reviewing. Then, 13 articles were excluded for following reasons (lack of sufficient data of distribution of genotype, animal models, and comment / editorial articles). Ultimately, 5 articles were eligible in quantitative synthesis (current meta-analysis). The specific flow chart was shown in Figure 1.
3.2 Characteristic of eligible studies
For the 5 included studies, 4 studies reported the comparison of rs3761548 polymorphism and 3 studies stated comparison of rs2232365 polymorphism between multiple sclerosis and controls. The analysis of rs3761548 polymorphism contained 1276 MS patients and 1447 controls (Table 1 and Table 2). There was a significant difference in the sex of patients and controls (OR: 0.76, 95 % CI: 0.65~0.90, P = 0.001, I2 = 34.8 %, Pheterogeneity = 0.203), and no significant difference in age (OR: 0.06, 95 % CI: -0.01~0.14, P = 0.088, I2 = 12.6 %, Pheterogeneity = 0.330). The analysis of re2232365 polymorphism contained 600 MS patients and 640 controls. Here, there was no significant difference in both sex (OR: 0.77, 95 % CI: 0.48~1.25, P = 0.295, I2 = 71.8 %, Pheterogeneity = 0.060) and age between patients and controls (OR: 0.08, 95 % CI: -0.03~0.19, P = 0.156, I2 = 0 %, Pheterogeneity = 0.574).
3.3 Quantitative synthesis
For rs2232365 polymorphism, we did not find a significant difference in the distribution of genotypes between multiple sclerosis and controls under allele model (OR: 1.22, 95 % CI: 0.68~2.18, P = 0.503, I2 = 91.5 %, Pheterogeneity = 0.000), homozygote model (OR: 1.40, 95 % CI: 0.52~3.77, P = 0.507, I2 = 89.5 %, Pheterogeneity = 0.000), heterozygote (OR: 1.37, 95 % CI: 0.67~2.79, P = 0.383, I2 = 84.1 %, Pheterogeneity = 0.002), dominant model (OR: 1.40, 95 % CI: 0.63~3.09, P = 0.411, I2 = 89.7 %, Pheterogeneity = 0), and recessive model (OR: 1.15, 95 % CI: 0.61~2.16, P = 0.663, I2 = 80.9 %, Pheterogeneity = 0.005) (Figure 4 and Figure 5).
For rs3761548 polymorphism, there was a significantly decreased risk of multiple sclerosis under allele model (OR: 0.76, 95 % CI: 0.58~1.00, P = 0.049, I2 = 80.2 %, Pheterogeneity = 0.002) and dominant model (OR: 0.77, 95 % CI: 0.64~0.94, P = 0.008, I2 =33.8 %, Pheterogeneity = 0.209). No significant association was detected between rs37651548 and MS under the homozygote model (OR: 0.63, 95 % CI: 0.40~1.00, P = 0.052, I2 = 73.3 %, Pheterogeneity = 0.010), heterozygote model (OR: 0.89, 95 % CI: 0.72~1.10, P = 0.276, I2 = 0 %, Pheterogeneity = 0.878), and the recessive model (OR: 0.67, 95 % CI: 0.45~1.01, P = 0.056, I2 = 79.0 %, Pheterogeneity = 0.003) (Figure 2 and Figure 3). Further, subgroup analysis was performed by ethnicity. There were significant differences in all genetic models but heterozygote model in Asians. In contrast, no significant difference was detected in European population under any genetic model (Figure 6 and Figure 7).
3.4 Publication bias
We performed the Begg funnel plot and Egger’s test to assess the potential publication bias in the identified studies. The results showed no evidence of publication bias in this meta-analysis (Egger test: P = 0.733 under allele model for rs3761548, P = 0.925 under dominant model for rs3761548) (Figure 8).
3.5 Sensitive analysis
Sensitive analysis was conducted to assess the influence of one study on the pooled ORs value (allele model and dominant model for rs3761548), and whether the results can be reverted by eliminating the individual study. The result did not alter after deleting each study, indicating the stability of the results of this meta-analysis (Figure 9).