Literature search and inclusion criteria

To retrieve eligible articles, we searched Pubmed, Web of Science and Embase using key words listed below: (‘interleukin-18’ or ‘IL-18’ or ‘interleukin 18’ or ‘IL 18’ or ‘cytotoxic T lymphocyte antigen-4’ or ‘CTLA-4’) and (‘polymorphism’ or ‘variant’ or ‘variation’ or ‘mutation’ or ‘SNP’ or ‘genome-wide association study’ or ‘genetic association study’ or ‘genotype’ or ‘allele’) and (‘viral hepatitis’ or ‘chronic hepatitis’ or ‘acute hepatitis’ or ‘Hepatitis A’ or ‘Hepatitis B’ or ‘Hepatitis C’ or ‘Hepatitis D’ or ‘Hepatitis E’ or ‘HAV’ or ‘HBV’ or ‘HCV’ or ‘HDV’ or ‘HEV’). The references of retrieved articles were also screened by us in case some related articles may be missed by our electronic literature searching.

To be included in this meta-analysis, some criteria must be satisfied: (I) studies about associations between polymorphisms in CTLA-4/IL-18 and viral hepatitis in humans; (II) offer genotypic or allelic distribution of CTLA-4/IL-18 polymorphisms in patients with viral hepatitis and controls and (III) full manuscript in English or Chinese is retrievable. We only included the most up-to-date study if duplicate reports were found during literature search.

Data extraction and quality assessment

Two authors extracted following information from eligible articles: (I) name of the first author; (II) published year; (III) country where the study was conducted; (IV) ethnicity of involved participants; (V) number of patients with viral hepatitis and controls in each study and (VI) genotype distributions of polymorphisms in CTLA-4/IL-18 among patients with viral hepatitis and controls. P Values of Hardy–Weinberg equilibrium (HWE) were also calculated.

The authors used Newcastle–Ottawa scale (NOS) to assess the methodology quality of eligible articles [Reference Stang10]. The score range of NOS is between zero and nine, when an article got a score of seven or more, we considered that the methodology of this publication was good.

Two authors extracted data and assessed the quality of eligible articles. The authors wrote to the corresponding authors for additional information if essential information was found to be incomplete.

Statistical analyses

We used Review Manager to combine the results of individual studies. The Z test was employed to assess associations between polymorphisms in CTLA-4/IL-18 and susceptibility to viral hepatitis in dominant, recessive, over-dominant and allele comparisons. All CTLA-4/IL-18 polymorphisms contain a major allele (M) and a minor allele (m), the dominant comparison is defined as MM vs. Mm + mm, recessive comparison is defined as mm vs. MM + Mm, over-dominant comparison is defined as Mm vs. MM + mm and the allele comparison is defined as M vs. m. The statistical significant threshold of the P value was set at 0.05. I ² statistics were used to assess between-study heterogeneities. Random-effect models (DerSimonian–Laird method) were used to combine the results if I ² is larger than 50%. Otherwise, fixed-effect models (Mantel–Haenszel method) were used to combine the results. We also carried out subgroup analyses first by type of disease and then by ethnicity of participants. We examined the stability of combined results by deleting one study each time and combining the results of the remaining studies. Funnel plots were used to estimate whether our combined results may be influenced by overt publication biases.