Epidemiological studies have evaluated the association between tumor necrosis factor α (TNF-α) single nucleotide polymorphisms (SNPs) and duodenal ulcer (DU), but the results remain inconclusive. The aim of this study was to perform a meta-analysis to investigate a more authentic association between TNF-α SNPs and DU.
We performed the meta-analysis by searching PubMed, Embase, and Web of Science databases from the first available year to Sep. 5, 2012. Additionally, checking reference lists from identified articles, reviews, and the abstracts presented at related scientific societies meetings were also performed. All case-control studies investigating the association between TNF-α SNPs and DU risk were included. The association was assessed by odds ratio (OR) with 95% confidence interval (CI). Publication bias was analyzed by Begg's funnel plot and Egger's regression test.
A total of sixteen studies reporting TNF-α −308G/A, −1031T/C, −863C/A, −857C/T, and −238G/A polymorphism were included in our final meta-analysis. There was no statistically significant association between −308G/A polymorphism and DU in the overall study population, as well as subgroup analyses by ethnicity, study design, and H. pylori status. As for −1031T/C, −863C/A, −857C/T, and −238G/A, results of our meta-analyses showed no statistical evidence of significant association. Power calculation on the combined sample size showed that the statistical powers were all lower than 80% for all the meta-analyses.
The data suggests that there is no statistical evidence of significant association between the studied TNF-α SNPs and DU. However, this conclusion should be interpreted with caution as low statistical powers were revealed by power calculations. In future, larger sample-size studies with homogeneous DU patients and well-matched controls are required.
Citation: Zhang B-B, Liu X-Z, Sun J, Yin Y-W, Sun Q-Q (2013) Association between TNF α Gene Polymorphisms and the Risk of Duodenal Ulcer: A Meta-Analysis. PLoS ONE 8(2): e57167. https://doi.org/10.1371/journal.pone.0057167
Editor: Xinping Cui, University of California, Riverside, United States of America
Received: September 20, 2012; Accepted: January 17, 2013; Published: February 22, 2013
Copyright: © 2013 Zhang et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
Duodenal ulcer (DU) is one type of peptic ulcer that occurs in the duodenum. DU may result from infection of Helicobacter pylori (H. pylori) bacteria, overuse of alcohol, and medications such as aspirin and non-steroidal anti-inflammatory drugs (NSAIDs) , . Both H. pylori infection and NSAIDs can increase gastric acid secretion in the stomach and also degrade the mucus barrier, and finally allow gastric acid to make contact with the duodenal lining, leading to inflammation and eventually ulceration. However, even as the most common cause of DU, the H. pylori infection only accounts for less than 20% of infected individuals who develop DU . Possible explanation for this phenomenon is that the nature of the host immune response influences the clinical outcome . Taking all these into consideration, we naturally come to the conclusion that DU is a multi-factorial disease involving environmental factors and host genetic factors.
With respect to host genetic factors, our previous studies demonstrated no association between DU and interleukin (IL)-1 β 511 C/T and 31 C/T polymorphisms, but significant association between IL-8 −251 T/A polymorphism and DU –. In addition to the above factors, tumor necrosis factor α (TNF-α) has been reported to have the proinflammatory activity and the capability of inhibiting gastric acid secretion –. TNF-α, which is produced by macrophages, monocytes, neutrophils, T-cells, and NK-cells after stimulation, is a pro-inflammatory cytokine and plays a role in cell immunity. The TNF-α gene is located in the class III region of the major histocompatibility complex on chromosome 6. It is suggested that polymorphisms in the regulatory region can influence the expression of TNF-α , and thereby increase the susceptibility of H. pylori infection. In 1999, Kunstmann et al. firstly reported the association of specific genotype of the TNF-α gene with the susceptibility to DU . Thereafter, a variety of epidemiological studies have evaluated the association between DU and TNF-α gene polymorphisms, including −308G/A, −1031T/C, −863C/A, −857C/T, −238G/A, −376 G/A, and −806 C/T –. However, results of different studies have been inconsistent. In addition, the sample size in each of published studies was relatively small, which limited the credibility of results. The present meta-analysis was designed to derive a more precise estimation of the association between TNF-α single nucleotide polymorphisms (SNPs) and DU.
Materials and Methods
This meta-analysis followed the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines  and Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) criteria . We collected all published studies on humans up to Sep. 5, 2012 by systematically searching the PubMed, Embase, and ISI Web of Science with the search terms: (“tumor necrosis factor α” OR “TNF-α”) AND (“polymorphism” OR “mutation” OR “variant” OR “genotype”) AND (“duodenal ulcer” OR “peptic ulcer”). The language was limited to English. The electronic searching was supplemented by checking reference lists from identified articles, reviews and the abstracts presented at related scientific societies meetings. Two investigators (Zhang BB and Yin YW) screened each of the titles, abstracts, and full texts to determine inclusion independently. The results were compared and disagreements were resolved by consensus.
The inclusion criteria were as follows: (1) studies on the relationship between TNF-α gene polymorphisms and DU, including −308G/A, −1031T/C, −863C/A, −857C/T, −238G/A, −376 G/A, and −806 C/T; (2) published case-control studies; (3) studies with full text articles; (4) sufficient data for estimating an odds ratio (OR) with 95% confidence interval (CI).
The information was carefully extracted by the same two authors (Zhang BB and Yin YW) according to the inclusion criteria listed above independently. Disagreement was resolved by consensus. If these two authors could not reach a consensus, another author (Sun QQ) was consulted. The following data were collected from each study: first author's name, publication date, country, ethnicity, study design (source of controls), and evidence of Hardy-Weinberg equilibrium (HWE) (P<0.05 of HWE was considered significant), respectively. Different ethnicities were categorized as Caucasian, Asian, African, American Indian, and mixed. Study design was stratified to population-based (PB) studies and hospital-based (HB) studies. Total numbers of cases and controls, and frequency of −308G/A, −1031T/C, −863C/A, −857C/T, −238G/A, −376 G/A, and −806 C/T mutation in cases and controls regardless of H. pylori status were extracted. When studies reported genotype distributions for H. pylori negative and H. pylori positive only, we also extracted data of each group separately for subgroup analyses.
Quality score assessment
Two authors (Zhang BB and Yin YW) of this article independently assessed the qualities of included studies using the Newcastle-Ottawa Scale (NOS) . The NOS ranges between zero (worst) up to nine points (best). We assessed included studies based on three aspects: the selection of the study groups (maximum of 4 points: whether the case definition is adequate with independent validation; whether the patients are consecutive or obviously representative series of cases; whether the controls are population-based; whether the controls are healthy person without history of duodenal ulcer); the comparability of the groups (maximum of 2 points: whether cases and controls are matched for gender and age; whether cases and controls are matched for region or ethnicity); and the ascertainment of the exposure (maximum of 3 points: whether the ascertainment of exposure is secure record; whether the same method of ascertainment is applied for cases and controls; whether the same non-response rate is reported for both groups). Studies with a score of seven points or greater were considered to be of high quality. Disagreement was settled as described above.
Combined ORs with their 95% CIs were calculated respectively for four genetic models: allelic model (2 allele vs. 1 allele), additive model (2/2 vs. 1/1), dominant model (1/2+2/2 vs. 1/1), and recessive model (2/2 vs. 1/2+1/1), in which 2 indicates the minor allele . Between-study heterogeneity was assessed by the Q-test and I2 statistic, P<0.10 and I2>50% indicated evidence of heterogeneity , . The ORs were pooled through a fixed effects model, using the Mantel-Haenszel approach when no heterogeneity was observed among studies . Otherwise, a random effects model was adopted . Publication bias was analyzed by Begg's funnel plot and Egger's test (P<0.05 was considered representative of statistically significant publication bias) . All above statistical analyses were performed using Stata 11.0. Power analysis was performed using Quanto software package (Version 1.2.4, http://hydra.usc.edu/gxe/) . Unmatched case-control design, gene only hypothesis and additive inheritance model were used. We assumed a population risk of 0.12 for DU and considered a two-sided p-value of 0.05.
The present study met the PRISMA statement requirements (Table S1 and Fig. 1). Two hundred and forty-two potentially relevant studies were identified after the searching. Based on our inclusion criteria, a total of nineteen studies were included in qualitative synthesis –. Three studies were excluded from further meta-analysis as they provided no origin data for calculating ORs and 95%CIs , , . Hence, sixteen articles were included in the meta-analysis –, , , –, , . Table 1 shows the main characteristics of the studies included in meta-analysis. Sixteen studies evaluated the relationship between −308G/A polymorphism and DU. Five studies reported association between DU and −1031T/C, −857C/T, and −238G/A polymorphism. There were four studies related −863C/A. Only one study reported the relationship between −376 G/A, −806 C/T polymorphism and DU, so both SNPs were not meta-analyzed. Considering sufficient data for −308G/A, we performed subgroup analyses by ethnicity (Asian and Caucasian), source of controls (population-based and hospital-based), and H. pylori status (H. pylori negative and H. pylori positive).
There were sixteen studies included in the analysis of −308G/A. The combined effects were as follows: for A allele vs. G allele: OR = 1.11, 95%CI = 0.87∼1.42; for AA vs. GG: OR = 1.19, 95%CI = 0.69∼2.05; for AA+AG vs. GG: OR = 1.10, 95%CI = 0.85∼1.41; and for AA vs. AG+GG: OR = 1.22, 95%CI = 0.70∼2.11 (Fig. 2). When meta-analysis was performed to assess association between −308G/A polymorphism and DU based on H. pylori status, no statistical evidence of significant association was found in all genetic models for both H. pylori positive subgroup and H. pylori negative subgroup. When stratified by ethnicity and study design, no statistically significant associations were found (Table 2). As for the association of TNF-α −238G/A, −1031T/C, −863C/A, −857C/T and DU susceptibility, there were no statistical evidence of significant association between all the above SNPs and DU susceptibility (Fig. S1, S2, S3, S4). Power calculation on the pooled frequencies showed that the statistical powers were all lower than 80% for all the above meta-analyses (Table 2).
There was no statistical evidence of significant association between −308G/A and DU susceptibility. The summary pooled ORs and 95% CIs are indicated by the white diamond.
The shapes of the funnel plots did not reveal any evidence of obvious asymmetry visually (Fig. S5, S6, S7, S8, S9). However, statistical evidence of publication bias were found using Egger's regression test for allelic model of −238G/A mutation and recessive model of −857C/T mutation. The results were as follows: for −308G/A mutation (P = 0.83 for allelic model, P = 0.22 for additive model, P = 0.98 for dominant model, and P = 0.23 for recessive model, respectively); for −238G/A mutation (P = 0.02 for allelic model, P = 0.89 for dominant model, respectively); for −1031T/C mutation (P = 0.84 for allelic model, P = 0.61 for additive model, P = 0.52 for dominant model, and P = 0.72 for recessive model, respectively); for −863C/A mutation (P = 0.98 for allelic model, P = 0.97 for additive model, P = 0.62 for dominant model, and P = 0.99 for recessive model, respectively); for −857C/T mutation (P = 0.18 for allelic model, P = 0.07 for additive model, P = 0.29 for dominant model, and P = 0.008 for recessive model, respectively).
TNF-α plays a key role in regulating gastric acid secretion, which is one of the most important factors in the development of duodenal diseases associated with H. pylori infection , , . To date, many studies have evaluated the association between TNF-α SNPs and DU risk, but the results remain inconsistent. Also, the credibility of results from a single case-control study is questionable due to too small sample size of the study populations. Meta-analysis has the benefit to overcome this limitation by increasing the sample size and may generate more precise results, which has been widely used in genetic association studies , . We therefore performed meta-analysis to assess whether combined evidence shows the association between TNF-α SNPs and DU.
Based on our study selection process, a total of nineteen publications were preliminarily eligible. Seven SNPs of TNF-α gene were reported, including −308G/A, −1031T/C, −863C/A, −857C/T, −238G/A, −376 G/A, and −806 C/T. Among these SNPs, only one study evaluated the association between −376 G/A and DU, which showed no significant association ; only one study reported the association between −806 C/T and DU, which demonstrated no significant association. Due to limited data, we did not carry out meta-analyses of these two SNPs.
In this meta-analysis, we identified eighteen articles focusing on the −308G/A polymorphism and its relationship with the risk of DU. Finally sixteen studies including 1,885 cases and 3,096 controls were included in the meta-analysis. The results of our meta-analyses demonstrated that there was no statistical evidence of significant association between the −308G/A polymorphism and DU in the overall study population regardless of the presence of H. pylori infection. These findings were consistent with most of the included studies as summarized in our meta-analyses. To explore a more precise relationship between −308G/A polymorphism and DU, we performed subgroup analyses by ethnicity, study design, and H. pylori status. First, we detected whether there was ethnic differences in the association between −308G/A polymorphism and DU. Our results demonstrated no statistically significant association among both Asian subgroup and Caucasian subgroup. When subgroup analyses were performed by study design, there was also no association. Considering H. pylori infection is the main cause of DU, we also performed meta-analysis to explore whether relationship between −308G/A polymorphism and DU would be associated with the status of H. pylori infection. However, no statistical evidence of significant association was shown in all models for both H. pylori-positive and H. pylori-negative subgroup.
As for −1031T/C, −863C/A, −857C/T, and −238G/A, four to five studies reported the association between them and the risk of DU, therefore meta-analyses were also performed. Among the studies included, one study demonstrated that −238G/A polymorphism might be a risk factor for DU in children infected with H. pylori ; one study showed that −857 TT genotype increased the risk of DU ; one study showed that both −1031 C and −863 A carriers were independent risk factors to have DU ; and one study reported that −1031T/C polymorphism was the risk factor for DU . However, results of our meta-analyses showed no statistically significant association between DU and −1031T/C, −863C/A, −857C/T, and −238G/A polymorphism.
Although no statistical evidence of significant associations were generated from our study, it is likely that TNF-α gene SNPs may increase the susceptibility to DU. The results of our study should be interpreted with caution. Power calculations for the given sample size (overall and subgroups) demonstrated that all the meta-analyses were underpowered. Even with over 1,800 cases and over 3,000 controls, the power was very low. It indicated that the combined sample sizes in our study were still inadequate to detect the association between TNF-α gene SNPs and DU. In addition, gene-gene interactions should be considered. For further study, gene to gene interactions with the generalized multifactor dimensionality reduction method, which has been applied widely , would be helpful in understanding the association between TNF-α gene SNPs and DU.
There are some limitations of this meta-analysis. Firstly, the linkage disequilibrium was found among the SNPs analyzed. Haplotype analysis may provide more information in evaluating the association between TNF-α SNPs and DU risk. Secondly, some inevitable bias may exist in the results as our meta-analysis only focused on papers published in English language and studies with full text articles, missing some eligible studies which were unpublished or reported in other languages. Thirdly, there was considerable heterogeneity among the included studies. Heterogeneity may affect the precision of results, despite the use of appropriate meta-analytic techniques with random-effects model.
Nonetheless, to the best of our knowledge, this is the first meta-analysis focusing on the relationship between TNF-α SNPs and DU risk. Our meta-analysis suggests that there is no statistical evidence of significant association between TNF-α SNPs and DU. However, this conclusion should be interpreted with caution as low statistical powers were revealed by power calculations. In future, larger sample-size studies with homogeneous DU patients and well-matched controls are required.
Forest plots of all models for −238G/A
Forest plots of all models for −857C/T
Forest plots of all models for −863C/A
Forest plots of all models for −1031T/C
Funnel plots of all models for −308G/A
Funnel plots of all models for −238G/A
Funnel plots of all models for −857C/T
Funnel plots of all models for −863C/A
Funnel plots of all models for −1031T/C
Conceived and designed the experiments: BBZ XZL QQS. Performed the experiments: BBZ YWY JS. Analyzed the data: BBZ YWY XZL JS. Contributed reagents/materials/analysis tools: SQQ JS. Wrote the paper: BBZ XZL JS YWY QQS.
- 1. Piper DW, Nasiry R, McIntosh J, Shy CM, Pierce J, et al. (1984) Smoking, alcohol, analgesics, and chronic duodenal ulcer. A controlled study of habits before first symptoms and before diagnosis. Scand J Gastroenterol 19: 1015–1021.
- 2. Zapata-Colindres JC, Zepeda-Gomez S, Montano-Loza A, Vazquez-Ballesteros E, de Jesus Villalobos J, et al. (2006) The association of Helicobacter pylori infection and nonsteroidal anti-inflammatory drugs in peptic ulcer disease. Can J Gastroenterol 20: 277–280.
- 3. Go MF (2002) Review article: natural history and epidemiology of Helicobacter pylori infection. Aliment Pharmacol Ther 16 Suppl 1: 3–15.
- 4. Gillen D, McColl KE (2005) Gastroduodenal disease, Helicobacter pylori, and genetic polymorphisms. Clin Gastroenterol Hepatol 3: 1180–1186.
- 5. Yin YW, Hu AM, Sun QQ, Zhang BB, Wang Q, et al. (2012) Association between interleukin-8 gene −251 T/A polymorphism and the risk of peptic ulcer disease: A meta-analysis. Hum Immunol .
- 6. Zhang BB, Wang J, Bian DL, Chen XY (2012) No association between IL-1beta −31 C/T polymorphism and the risk of duodenal ulcer: A meta-analysis of 3793 subjects. Hum Immunol 73: 1200–1206.
- 7. Zhang BB, Yin YW, Sun QQ (2012) No association between IL-1beta −511 C/T polymorphism and the risk of duodenal ulcer: a meta-analysis of 4667 subjects. Gene 506: 188–194.
- 8. Crabtree JE, Shallcross TM, Heatley RV, Wyatt JI (1991) Mucosal tumour necrosis factor alpha and interleukin-6 in patients with Helicobacter pylori associated gastritis. Gut 32: 1473–1477.
- 9. Fan XG, Chua A, Fan XJ, Keeling PW (1995) Increased gastric production of interleukin-8 and tumour necrosis factor in patients with Helicobacter pylori infection. J Clin Pathol 48: 133–136.
- 10. Beales IL, Calam J (1998) Interleukin 1 beta and tumour necrosis factor alpha inhibit acid secretion in cultured rabbit parietal cells by multiple pathways. Gut 42: 227–234.
- 11. Wilson AG, Symons JA, McDowell TL, McDevitt HO, Duff GW (1997) Effects of a polymorphism in the human tumor necrosis factor alpha promoter on transcriptional activation. Proc Natl Acad Sci U S A 94: 3195–3199.
- 12. Kunstmann E, Epplen C, Elitok E, Harder M, Suerbaum S, et al. (1999) Helicobacter pylori infection and polymorphisms in the tumor necrosis factor region. Electrophoresis 20: 1756–1761.
- 13. Chakravorty M, Datta De D, Choudhury A, Santra A, Roychoudhury S (2008) Association of specific haplotype of TNFalpha with Helicobacter pylori-mediated duodenal ulcer in eastern Indian population. J Genet 87: 299–304.
- 14. Cheng HH, Chang CS, Wang HJ, Wang WC (2010) Interleukin-1beta and −10 polymorphisms influence erosive reflux esophagitis and gastritis in Taiwanese patients. J Gastroenterol Hepatol 25: 1443–1451.
- 15. Garcia-Gonzalez MA, Lanas A, Savelkoul P, Benito R, Crusius B, et al. (2002) No interactions between IL-1 and TNF gene polymorphisms in the susceptibility to duodenal ulcer disease. Gastroenterology 122: A–427.
- 16. Garcia-Gonzalez MA, Savelkoul PH, Benito R, Santolaria S, Crusius JB, et al. (2005) No allelic variant associations of the IL-1 and TNF gene polymorphisms in the susceptibility to duodenal ulcer disease. Int J Immunogenet 32: 299–306.
- 17. Gyulai Z, Klausz G, Tiszai A, Lenart Z, Kasa IT, et al. (2004) Genetic polymorphism of interleukin-8 (IL-8) is associated with Helicobacter pylori-induced duodenal ulcer. Eur Cytokine Netw 15: 353–358.
- 18. Hellmig S, Mascheretti S, Jelschen F, Rehbein J, Renz J, et al. (2002) Role of genetic polymorphisms in cytokines influencing the outcome of H. pylori infection. Gastroenterology 122: A.19–A.20.
- 19. Kim N, Cho SI, Yim JY, Kim JM, Lee DH, et al. (2006) The effects of genetic polymorphisms of IL-1 and TNF-A on Helicobacter pylori-induced gastroduodenal diseases in Korea. Helicobacter 11: 105–112.
- 20. Lanas A, Garcia-Gonzalez MA, Santolaria S, Crusius JB, Serrano MT, et al. (2001) TNF and LTA gene polymorphisms reveal different risk in gastric and duodenal ulcer patients. Genes Immun 2: 415–421.
- 21. Lee JY, Kim HY, Kim KH, Kim SM, Jang MK, et al. (2005) Association of polymorphism of IL-10 and TNF-A genes with gastric cancer in Korea. Cancer Lett 225: 207–214.
- 22. Lee SG, Kim B, Yook JH, Oh ST, Lee I, et al. (2004) TNF/LTA polymorphisms and risk for gastric cancer/duodenal ulcer in the Korean population. Cytokine 28: 75–82.
- 23. Li C, Xia B, Yang Y, Li J, Xia HH (2005) TNF gene polymorphisms and Helicobacter Pylori infection in gastric carcinogenesis in Chinese population. Am J Gastroenterol 100: 290–294.
- 24. Lu CC, Sheu BS, Chen TW, Yang HB, Hung KH, et al. (2005) Host TNF-alpha-1031 and −863 promoter single nucleotide polymorphisms determine the risk of benign ulceration after H. pylori infection. Am J Gastroenterol 100: 1274–1282.
- 25. Mei Q, Xu JM, Cao HL, Bao DM, Hu NZ, et al. (2010) Associations of the IL-1 and TNF gene polymorphisms in the susceptibility to duodenal ulcer disease in Chinese Han population. Int J Immunogenet 37: 9–12.
- 26. Melo Barbosa HP, Martins LC, Dos Santos SE, Demachki S, Assumpcao MB, et al. (2009) Interleukin-1 and TNF-alpha polymorphisms and Helicobacter pylori in a Brazilian Amazon population. World J Gastroenterol 15: 1465–1471.
- 27. Sugimoto M, Furuta T, Shirai N, Nakamura A, Xiao F, et al. (2007) Different effects of polymorphisms of tumor necrosis factor-alpha and interleukin-1 beta on development of peptic ulcer and gastric cancer. J Gastroenterol Hepatol 22: 51–59.
- 28. Wilschanski M, Faber J, Rudensky B, Feigin E, Halle D (2002) The genetic polymorphism of the host: A risk factor for peptic ulcer disease in children. Gastroenterology 122: A–227.
- 29. Wilschanski M, Schlesinger Y, Faber J, Rudensky B, Ohnona FS, et al. (2007) Combination of Helicobacter pylori strain and tumor necrosis factor-alpha polymorphism of the host increases the risk of peptic ulcer disease in children. J Pediatr Gastroenterol Nutr 45: 199–203.
- 30. Zambon CF, Basso D, Navaglia F, Belluco C, Falda A, et al. (2005) Pro- and anti-inflammatory cytokines gene polymorphisms and Helicobacter pylori infection: interactions influence outcome. Cytokine 29: 141–152.
- 31. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, et al. (2000) Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 283: 2008–2012.
- 32. Moher D, Liberati A, Tetzlaff J, Altman DG (2009) Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 6: e1000097.
- 33. Wells GA, Shea B, O'Connell D, Peterson J, Welch V, Losos M, Tugwell P (2011) The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. Ottawa Health Research Institute.www.ohri.ca/programs/clinical_epidemiology/oxford.asp (accessed Dec. 1, 2011).
- 34. Lewis CM (2002) Genetic association studies: design, analysis and interpretation. Brief Bioinform 3: 146–153.
- 35. Cochran WG (1954) The combination of estimates from different experiments. Biometrics 10: 101–129.
- 36. Higgins JP, Thompson SG (2002) Quantifying heterogeneity in a meta-analysis. Stat Med 21: 1539–1558.
- 37. Mantel N, Haenszel W (1959) Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 22: 719–748.
- 38. DerSimonian R, Laird N (1986) Meta-analysis in clinical trials. Control Clin Trials 7: 177–188.
- 39. Egger M, Davey Smith G, Schneider M, Minder C (1997) Bias in meta-analysis detected by a simple, graphical test. BMJ 315: 629–634.
- 40. Fischer-Rosinsky A, Fisher E, Kovacs P, Bluher M, Mohlig M, et al. (2008) Lack of association between the tagging SNP A+930→G of SOCS3 and type 2 diabetes mellitus: meta-analysis of four independent study populations. PLoS One 3: e3852.
- 41. McColl KE, el-Omar E, Gillen D (2000) Helicobacter pylori gastritis and gastric physiology. Gastroenterol Clin North Am 29: 687–703, viii.
- 42. Yamaoka Y, Kita M, Kodama T, Sawai N, Imanishi J (1996) Helicobacter pylori cagA gene and expression of cytokine messenger RNA in gastric mucosa. Gastroenterology 110: 1744–1752.
- 43. Xia Q, Zhou L, Liu D, Chen Z, Chen F (2011) Relationship between TNF-<alpha> gene promoter polymorphisms and outcomes of hepatitis B virus infections: a meta-analysis. PLoS One 6: e19606.
- 44. Zhou P, Lv GQ, Wang JZ, Li CW, Du LF, et al. (2011) The TNF-alpha-238 polymorphism and cancer risk: a meta-analysis. PLoS One 6: e22092.
- 45. Zhou JB, Liu C, Niu WY, Xin Z, Yu M, et al. (2012) Contributions of renin-angiotensin system-related gene interactions to obesity in a Chinese population. PLoS One 7: e42881.