The effects of tumor necrosis factor-α (TNF-α) rs1800629 and rs361525 polymorphisms on sepsis risk

This meta-analysis of 23 eligible articles comprehensively and quantitatively evaluated the effects of tumor necrosis factor-α (TNF-α) rs1800629 and rs361525 polymorphisms on sepsis risk. We found that TNF-α rs1800629 was associated with increased sepsis risk in the overall population in four genetic models, including A vs. G (P<0.001, odds ratio (OR)=1.32), GA vs. GG (P<0.001, OR=1.46), GA+AA vs. GG (P<0.001, OR=1.46), and carrier A vs. carrier G (P<0.001, OR=1.32). Subgroup analyses showed a similar result for Asian patients (all P<0.05, OR>1). TNF-α rs361525 was also associated with increased sepsis risk in Asian patients in the four genetic models (all P<0.05, OR>1). Begg’s and Egger’s tests excluded large publication bias, and sensitivity analysis indicated stable results. Our results suggest that the G/A genotype of TNF-α rs1800629 and rs361525 increases sepsis risk in an Asian population.


INTRODUCTION
Sepsis consumes considerable health care resources and has a high mortality rate, especially in elderly patients and in infants born pre-term or with low birth weights [1,2]. Lack of early detection is implicated in the high incidences of severe sepsis and septic shock [3,4]. Thus, sepsis-related biomarkers and risk factors must be identified to improve early detection rates. Recent studies have addressed associations between various gene SNPs (single nucleotide polymorphisms) and sepsis risk. Sepsis risk was associated with TLR4 (toll like receptor 4) SNPs, rs4986790 and rs4986791 [5], but not the SERPINE1 [Serpin Peptidase Inhibitor, Clade E (Nexin, Plasminogen Activator Inhibitor Type 1), Member 1] rs1799768 polymorphism [6].
TNF-α (tumor necrosis factor-α) is important for normal body functions, but is also implicated in some disease mechanisms, including sepsis, diabetes mellitus, and cancer [7][8][9][10][11]. There are several known SNPs within the TNF-α gene, including rs1800629, rs361525, rs1800630, and others [12]. Associations between TNF-α SNPs and sepsis risk are still uncertain. TNF-α rs1800629 was reported as a sepsis risk factor in severely injured North Indian patients [13], critically ill Japanese patients [14], the Chinese Han population [15], and Turkish children [16]. However, TNF-α rs1800629 was also negatively correlated with sepsis susceptibility in preterm infants in Germany [17] and low-birth-weight infants in Hungary [18]. We did not obtain data from genome wide association studies (GWAS) of sepsis-associated SNPs. Thus, our meta-analysis is a relatively objective evaluation of TNF-α SNPs in sepsis risk. Our analysis focused on the genetic relationship between sepsis risk and the rs1800629 and rs361525 polymorphisms within the TNF-α promoter region, in that sufficient data was only obtained for the meta-analysis of rs1800629, rs361525 polymorphisms, after our data extraction.

TNF-α rs361525 meta-analysis
We enrolled eight case-control studies containing 1,916 cases and 3,372 controls [13,15,24,25,27,28,32,33] in our TNF-α rs361525 meta-analysis. Sepsis risk was increased in the AA vs. GG (P=0.001, OR=4.24) and AA vs. GG+GA (P=0.001, OR=4.24) genetic models, but not A vs. G, GA vs. GG, GA+AA vs. GG, or carrier A vs. carrier G (all P>0.05; Table 3 ). Similarly, for subgroup analyses of sepsis severity, increased risk of severe sepsis or septic shock was only observed in the AA vs. GG and AA vs. GG+GA models (   (Table 3). Asian patient subgroup analyses also showed differences in these four models (all P<0.05, OR>1). Figure 3 and Supplementary Figures 4-6 show forest plots for ethnicity subgroup analyses. These data suggested that the TNF-α rs361525 G/A genotype is associated with enhanced risk of sepsis in the Asian population.

Heterogeneity, publication bias, and sensitivity analysis
For TNF-α rs1800629, we applied the random-effect model in the allele, heterozygote, dominant, and carrier Mantel-Haenszel analyses, due to the following data (Table 4)     GG (P=0.042) and AA vs. GG+GA (P=0.041) models (Table 5). Figure 4 shows the Begg's funnel plot and Egger's publication bias plot for the GA vs. GG TNF-α rs1800629 meta-analysis. Similar pooled ORs in our sensitivity analysis suggested that our data were reliable ( Figure 5 for the GA vs. GG model of TNF-α rs1800629; other data not shown). www.impactjournals.com/oncotarget DISCUSSION This updated literature search and meta-analysis comprehensively reassessed the association between TNF-α polymorphisms and sepsis risk in Asian/Caucasian populations. There are several advantages in terms of database searching, screening strategy study inclusion, and sample size. To date, three related meta-analyses have been published [36][37][38]. Teuffel, et al. performed the first of these in 2010, and reported that the GA or AA TNF-α rs1800629 genotypes were associated with increased sepsis risk [37]. Twenty-five articles [14, 16, 18, 25-28, 31, 39-55] were included in this meta-analysis, however several articles [39][40][41][42][43][44][45][46][47][48][49][51][52][53][54][55] did not contain sufficient genotype frequency data in the case and/or control groups, or were not in line with Hardy-Weinberg Equilibrium (HWE). Additionally, no or mild sepsis was set as the control group in one included article [40], which may not have been appropriate for our meta-analysis. Another meta-analysis by Srinivasan, et al. only investigated the association between TNF-α rs1800629 and neonatal sepsis risk [36]. After rigorous screening, we enrolled 23 articles with 3,404 cases and 5,973 controls  in our TNF-α rs1800629 meta-analysis.
Our stratified analysis of severe sepsis and septic shock included only articles that specified sepsis type. Our subgroup meta-analyses based on controls (HB/PB), showed that the TNF-α rs1800629 G/A genotype was linked to increased sepsis risk in both groups. However, TNF-α rs361525 and sepsis risk were positively correlated in the PB, but not HB group. Thus, the presence of other diseases may influence the genetic role of TNF-α rs361525, but not TNF-rs1800629 in sepsis risk. Our study was subject to certain limitations. First, more studies with larger sample sizes and high qualities are needed for enhanced statistical power. We only observed the potential association between TNF-α rs1800629 and severe sepsis risk for the allele, dominant, and carrier models. TNF-α rs361525 was also found be slightly linked to the risk of severe sepsis or septic shock for the homozygote and recessive models. Thus, our lack of strong evidence for associations between the two SNPs and sepsis risk merits more case-control studies. Second, we found high heterogeneity between studies in some genetic comparisons. This may be caused at least in part by the complexity of the sepsis etiology and the non-uniformity of diagnostic criteria. Finally, more data are needed to clarify the genetic roles and prognostic significance of distinct cytokine gene combinations in sepsis risk. Different SNP linkages of the TNF-α gene should be considered as well.
In conclusion, our findings suggest a positive association between the G/A genotype of the TNF-α rs1800629 and rs361525 polymorphisms and sepsis risk in the Asian population, which is partly in line with the findings of Teuffel, et al. [37] and Zhang, et al. [38]. Abnormal TNF-α is implicated in the pathogenesis of sepsis. For example, TNF-α expression was closely related to neonatal sepsis in very low birth weight infants in Spain [60]. It may be that mutation in the TNF-α promoter region from common G (guanidine) to rare A (adenosine) at position -308 (rs1800629) or -238 (rs361525) affects normal TNF-α production, secretion, or function in sepsis patients [8].

Records identification
We identified potential records from six databases (PubMed, WOS, EMBASE, CNKI, WANFANG, and Scopus). We performed a PRISMA (preferred reporting items for systematic reviews and meta-analyses [61])-compliant database search and study selection, and the relevant meta-analysis papers were referred [62][63][64]. Two authors (Yixin Zhang, Xiaoteng Cui) independently removed duplicate studies and assessed record eligibilities according to our exclusion/inclusion criteria. Exclusion criteria included: a) meta-analysis; b) review, editorials, and perspectives; c) congress abstract or poster; d) other gene or other disease; e) not a TNF-α SNP or no clinical data; f) lack of control data; g) did not contain genotype data or the genotype distributions were not in line with HWE.
According to Newcastle-Ottawa Scale (NOS) requirements (http://www.ohri.ca/programs/clinical_ epidemiology/oxford.asp), we evaluated methodological quality independently. Included studies had a NOS score >5, and provide the following data: first author name, publication year, patient ethnicity, SNP, sample size, and genotype frequency within case-control studies and genotyping assays.

Quantitative synthesis and heterogeneity
We used the Mantel-Haenszel test to determine P-values, pooled ORs, and 95% CIs for the following six genetic models: A vs. G (allele); AA vs. GG (homozygote); GA vs. GG (heterozygote); GA+AA vs. GG (dominant); AA vs. GG+GA (recessive); and carrier A vs. carrier G (carrier). P<0.05 represented a statistically significant difference between case and control studies. We also assessed the heterogeneity between studies using the Q statistic and I 2 test. High heterogeneity was likely when Q statistic P<0.05 or I 2 >50%. In this situation, we employed the random-effect model, not the fixed-effect model. We also performed a series of subgroup metaanalyses according to three factors: source of control [PB (population-based) or HB (hospital-based)], ethnicity (Caucasian or Asian), and sepsis severity (sepsis, severe sepsis, or septic shock).

Publication bias and sensitivity analysis
We evaluated publication bias using both Begg's test and Egger's test. P>0.05 in both tests would exclude a large publication bias. We also performed a sensitivity analysis to evaluate the robustness of our data. All analyses were performed using Stata/SE 12.0 software (StataCorp, USA).