A meta-analysis of the relationship of rs1946518 polymorphism with RA and SLE, no significant association with RA or SLE risk can be found under all genetic models in Asian populations, which was similar to the previous meta-analysis reported by Pan et al. . For RA, in the subgroup analysis, results from studies with Chinese population or PCR-SSP method or small sample size were inconsistent. For Chinese population rs1946518 A allele was associated with decreased risk for RA (OR=0.688,95%CI=0.532-0.889), as well as studies with PCR-SSP method (OR=0.636,95%CI=0.519-0.781) and small sample size (OR=0.559,95%CI=0.438-0.715). In addition, subgroup analysis found different characteristics among studies like population, genotyping method, sample size, could explain part of the heterogeneity, but meta-regression analysis can only detect sample size as a significant source of heterogeneity under the additive model. For SLE, meta-analysis showed −607 A allele may have a protective effect on SLE in Chinese population (OR=0.606,95%CI =0.396-0.930) and studies with PCR-RFLP method (OR=0.670,95%CI =0.454-0.989), and neither subgroup analysis nor meta-regression analysis found the sources of heterogeneity under the additive model. Therefore our meta-analysis have not find any association between rs1946518 polymorphism and RA or SLE, but we detected significant association in Chinese population for both diseases, which may be due to different allele frequency at this SNP between Chinese population and other populations, the major populations we investigated were Chinese, Japanese and Thai populations, we compared A allele frequency in Chinese population with Japanese population (HapMap Genome browser release #28) and Thai population (For Thai population we used Hirankarn et al. control data instead ,because we could not get the data from the International HapMap Project). The A allele frequency in Chinese population was 0.585 compared with 0.580 in Japanese population and 0.440 in Thai population. This indicates that the allele frequency does not differ between Chinese population and Japanese population, but differ between Chinese population and Thai population, so it can only explain part of the difference, some other reasons, like genotyping method ,sample size and environmental exposures, may also play important roles. Above analysis suggest that rs1946518 polymorphism existed regional difference,so we should be caution when pooled data in different populations.
However, we failed to find any association between rs187238 polymorphism and RA or SLE under all genetic models in Asian populations, and even in subgroup analysis the results remained unchanged. For both diseases, the rs187238 polymorphism did not show significant heterogeneity among all studies under most genetic models. Considering the limited number of studies included in this study, whether rs187238 polymorphism plays roles in RA or SLE is still unknown, so additional large scale studies about RA or SLE are needed to clarify these relationships.
Rs1946518 and rs187238 polymorphisms in the IL-18 promoter region are associated with transcription activity. A change from nucleotide C to nucleotide A at position −607 disrupts a potential cAMP-responsive element binding protein (CREB) site resulting in higher levels of the IL-18 production [31, 32]. For position −137, a change from nucleotide G to nucleotide C affects the H4TF-1 nuclear factor binding site, which may confer a higher IL-18 protein expression . Furthermore, a recent study reported by Dziedziejko et al.  also found rs1946518 (−607A/C) AA homozygotes showed significantly lower IL-18 release and mRNA expression after stimulation(stimulants: PHA, LPS, and anti-CD3/CD28 antibodies),and as well as rs187238 (−137G/C) C allele after PHA treatment. These observations support our study, in which IL-18 rs1946518 C allele was associated with high IL-18 production, and high levels of IL-18 were associated with RA and SLE [8, 9], so IL-18 rs1946518 A allele might have a protective effect on RA and SLE.
Several limitations of this research should be considered. First, our literature searching was depended on English and Chinese, language bias might be existed. Second, our subjects were all Asians, so it was not suitable to other populations. Third, publication bias may still be present, distorting the meta-analysis. Fourth, we couldn’t calculate the linkage disequilibrium (LD) between SNPs, so we couldn’t exclude the potential likelihood that the significant association in Chinese population was only due to other real significant polymorphisms which were linkage with rs187238. Finally, the studies included were small, our results should be interpreted with caution.