Weak association between the interleukin-8 rs4073 polymorphism and acute pancreatitis: a cumulative meta-analysis

All studies reporting the relationships of IL-8 polymorphisms with AP published before March 30, 2019 were identified by computerized searches in databases including “Pubmed”, “Embase”, “China National Knowledge Infrastructure (CNKI)”, and “Wanfang”. The search terms were: (“interleukin-8” OR “IL-8”) AND (“gene” OR “polymorphism” OR “genetic variant”) AND (“acute pancreatitis” OR “pancreatitis”). In addition, we checked the references of all the retrieved articles to identify further relevant articles.

Qualified studies have to meet the following criteria: (1) predefined diagnosis criteria for AP; (2) case-control studies evaluated the association between the IL-8 rs4073 polymorphism and AP risk; (3) provided detailed genotype frequencies; (4) if multiple reports based on the same study population were available, the most recent or largest study was selected. The main exclusion criteria were: (1) insufficient data offered to calculate OR estimate; (2) non-case-control studies; (3) duplicate data from a same cohort. We screened titles and abstracts of searching results for an initial evaluation, and full articles were further evaluated to confirm eligibility.

All data were extracted independently by two reviewers (Y.L. and J.B.). The following information was extracted or calculated from the included articles: authors, publication year, country, ethnicity, Hardy-Weinberg equilibrium (HWE) status, genotyping method, the number of cases and controls, and counts of alleles and genotypes in both cases and controls. The results were compared, and differences were resolved by discussion and consensus with a third reviewer (H.W.).

Two authors (B.H. and N.W.) independently assessed the quality of the included studies based on Newcastle Ottawa Scale (NOS) [12]. This scale uses a star rating system to judge the methodological quality and consists of three parts: selection, comparability and ascertainment of exposure. A full score is 9 stars, and a score of five or more was regarded as “high quality”; otherwise, the study was regarded as “low quality”. Any disagreements on the NOS score of the studies were resolved through a comprehensive reassessment by the other authors.

Crude Odds ratio (ORs) with corresponding 95% confidence intervals (CIs) were used to estimate the strength of association between the IL-8 rs4073 polymorphism and AP risk. The statistical significance of the pooled ORs was examined by Z test and P < 0.05 was considered statistically significant. In the meta-analysis, the overall pooled ORs were calculated for four models, including the allelic, dominant, recessive and homozygous models. Deviation from Hardy-Weinberg equilibrium (HWE) was examined by Chi-square test and P value < 0.05 indicated a departure from HWE. The between-study heterogeneity was assessed by the I² statistic, which was calculated from the Q statistic [13]. If the heterogeneity was statistically significant (P < 0.05 for Q-test or I² > 50%), the random-effects model (based on DerSimonian-Laird method) was adopted to get the pooled estimates [14]; otherwise, the fixed-effects model (based on Mantel-Haenszel method) was used [15]. To reduce the risk of false positive when performing repetitive estimates, the cumulative meta-analysis method was adopted [16]. To assess the stability and reliability of the meta-analyses, we performed sensitivity analyses by sequentially omitting 1 estimate at a time and recalculating the pooled results. The results were suggested to be robust if any single study didn’t alter the significance of our results. Potential publication bias was assessed with Begger’s funnel plot and the Egger’ linear regression test, and a P < 0.05 was considered significant [17, 18]. The statistical analysis was performed with STATA statistical software (Version 13.3; Stata Corporation, College Station, TX, USA). All P values were two-tailed.

Our literature screening process is shown in Fig. 1. In the initial screening process, a total of 178 articles were retrieved based on our selection strategy from the databases. After the removal of duplicated literatures and articles that did not meet the inclusion criteria, 10 studies involving 1646 AP patients and 1816 controls were finally included in this meta-analysis [19‐28]. Of these selected studies, six studies were published in English [19‐21, 24‐26], three were published in Chinese [22, 27, 28], the other one was published in Russian [23]. There were six studies conducted in China [20, 22, 24, 26‐28], and one study conducted in U.S. [21], Russia [23], Turkey [19] and Hungary [25], respectively. Table 1 summarizes the basic information of each study, including the year of publication, country, ethnicity, sample size, NOS score and so on.

Meta-analyses under four genetic models are shown in Fig. 1. When comparing effect sizes, we found no evidence of significant heterogeneity in the analyses of all the four genetic models used, thus fixed-effects models were used. Cumulative meta-analyses (Fig. 2) indicated there is a consistent trend toward association after the initial discovery. The pooled results showed that the IL-8 rs4073 polymorphism associates with an increased risk of AP. Under the allelic, dominant, recessive, and homozygote models the ORs, 95% CIs and its p-values were 1.265 (1.147–1.395, p < 0.001), 1.304 (1.127–1.508, p < 0.001), 1.431 (1.203–1.702, p < 0.001), and 1.634 (1.334–2.001, p < 0.001), respectively.

Subgroup analyses (Fig. 3) were performed according to the study quality (high or low), HWE test results (yes or no) and ethnicity (Asian, Caucasian or mixed) in all the four genetic models. Trends toward an association between the IL-8 rs4073 polymorphism an AP risk were observed in all the subgroups, and statistically significant association was found in most subgroups.

We estimated the stability of all the pooled ORs by eliminating all included studies one by one. The corresponding ORs were not statistically altered, indicating that our results were stable. (Fig. 4).

Egger’s linear regression tests and funnel plots were performed to assess possible publication bias in the meta-analyses. The shape of the funnel plot was obviously symmetrical for all models (Fig. 5). The P value of Egger’s test also suggested that there was no significant publication bias (Table 2).