NFKB1-94insertion/deletion ATTG polymorphism and cancer risk: Evidence from 50 case-control studies

Nuclear factor-kappa B1 (NF-κB1) is a pleiotropic transcription factor and key contributor to tumorigenesis in many types of cancer. Numerous studies have addressed the association of a functional insertion (I)/deletion (D) polymorphism (-94ins/delATTG, rs28362491) in the promoter region of NFKB1 gene with the risk of various types of cancer; however, their conclusions have been inconsistent. We therefore conducted a meta-analysis to reevaluate this association. PubMed, EMBASE, China National Knowledge infrastructure (CNKI), and WANFANG databases were searched through July 2016 to retrieve relevant studies. After careful assessment, 50 case-control studies, comprising 18,299 cases and 23,484 controls were selected. Crude odds ratios (ORs) and 95% confidence intervals (CIs) were used to determine the strength of the association. The NFKB1 -94ins/delATTG polymorphism was associated with a decreased risk of overall cancer in the homozygote model (DD vs. II): OR = 0.75, 95% CI = 0.64-0.87); heterozygote model (ID vs. II): OR = 0.91, 95% CI = 0.83-0.99; recessive model (DD vs. ID/II): OR = 0.81, 95% CI = 0.71-0.91; dominant model (ID/DD vs. II): OR = 0.86, 95% CI = 0.78-0.95; and allele contrast model (D vs. I): OR = 0.88, 95% CI = 0.81-0.95). Subgroup and stratified analyses revealed decreased risks for lung cancer, nasopharyngeal carcinoma, prostate cancer, ovarian cancer, and oral squamous cell carcinoma, and this association held true also for Asians (especially Chinese subjects) in hospital-based studies, and in studies with quality scores less than nine. Well-designed, large-scale case-control studies are needed to confirm these results.


INTRODUCTION
Cancer is a substantial public health burden, with an estimated 1.7 million new cancer cases and 0.6 million cancer-related deaths in the United States in 2016 [1].Although the etiology of carcinogenesis has not yet been fully elucidated, many lines of evidence suggest that cancer is a multifactorial disease caused by intricate interactions between multiple hereditary and environmental factors [2,3].Numerous studies have demonstrated that inflammation is critically implicated in the development of some cancers [4,5].In view of this, it is plausible that genetic polymorphisms in inflammationrelated genes could modify cancer susceptibility [6][7][8].
The human NFKB1 gene, spanning 156-kb, is located on chromosome 4q23-q24 and encodes a 105 kD protein (p105) which is cleaved into an active subunit (p50) [20].Several variations have been identified in the NFKB1 gene, including rs72696119 (C>G), rs28362491 (-94 ins/del ATTG), rs4648068 (A>G), and rs12509517 (G>C) [21].Among these, NFKB1 rs28362491, namely the -94insertion/deletion ATTG polymorphism, is potentially functional and the most widely investigated [21].This modification occurs between two important regulatory elements (activator protein 1 and κB binding site) in the promoter region of the NFKB1 gene.The deletion of four bases (ATTG) reduces or prevents the binding to nuclear proteins and leads to lower transcript levels of the NFKB1 gene, thereby changing mRNA stability and regulating translation efficiency [21,22].
Numerous case-control studies have assessed the association between the NFKB1 -94ins/delATTG promoter polymorphism and cancer risk, with discrepant results.While some studies indicated an increased risk for some types of cancers [23][24][25], other studies showed instead a decreased risk, or no association [26,27].Several metaanalyses attempted to solve the controversy, but did not yield consistent results [28][29][30][31][32][33].To provide a more precise evaluation of such association, we performed a comprehensive, updated meta-analysis.In addition, to minimize random errors and strengthen the robustness of our conclusions, we also performed trial sequential analysis (TSA).
After this selection procedure, 50 studies extracted from 46 publications with 18,299 cases and 23,484 controls ultimately entered our final meta-analysis (Table 1

Heterogeneity and sensitivity analysis
Statistically significant between-study heterogeneity was found in the pooled analysis under the five genetic models (P < 0.001).Thus, the random-effect model was applied to calculate the ORs and 95% CIs.Sensitivity analysis using the leave-one-out cross-validation method was conducted to assess the impact of each single study on the overall risk estimates.The omission of each individual study did not have substantial influence on the risk estimates, supporting the credibility and reliability of this meta-analysis (data not shown).

Publication bias
Publication bias was assessed by Begg's funnel plot and quantitative Egger's test.The funnel plot showed a symmetrical shape (Figure 3), suggesting no publication bias, a conclusion further supported by the Egger's test

Trial sequential analysis and false-positive report probability (FPRP) analyses
To minimize random errors and strengthen the robustness of our conclusions, we performed TSA (Figure 4).This analysis showed that the cumulative z-curve crossed the trial sequential monitoring boundary before reaching the required information size, suggesting that the cumulative evidence is sufficient and no further evidence is needed to verify the conclusions.
We finally calculated the FPRP values for all observed significant findings.With the assumption of a prior probability of 0.1, the FPRP values were all <0.20, suggesting that these significant associations were noteworthy (Table 3).

DISCUSSION
In this meta-analysis, we found that the NFKB1 -94ins/delATTG promoter polymorphism was significantly associated with decreased overall cancer risk under the five genetic models.To the best of our knowledge, this is the most comprehensive meta-analysis on this topic by now.
Numerous studies have suggested that polymorphisms in genes encoding inflammatory response factors, such as TNF-alpha -308G>A [6], IL6 -174G>C [90], and NFKBIA -826C>T [91] may contribute to cancer susceptibility.Song et al. [92] reported that the NFKB1 -94ins/delATTG polymorphism analyzed here (rs28362491) increased the risk of colorectal cancer in a Southern Chinese population; this association was also observed in several publications [71].However, contradictory conclusions were also reported, namely a null association, or decreased cancer susceptibility.To address this controversy, at least six meta-analyses were performed.The first one, performed in 2011 by Zou et al. [32], included only 2,743 cases and 2,195 controls from 11 studies.They did not observe any association between the -94ins/delATTG variant and overall cancer.However, an ethno-specific association was detected by subgroup analysis; the D allele was protective against cancer in Asians, but increased the risk in Caucasians.Afterwards, Wang et al. [33] conducted an updated meta-analysis including 5,196 cases and 6,614 controls from 19 publications.They found that variant homozygotes (DD) had a decreased risk of cancer compared with wild-type homozygotes (II).The association was also found under the dominant genetic model (DD+DI vs. II).In subgroup analysis, a significantly decreased risk was observed in Asians but not in Caucasians.In addition, this susceptibility was cancer-specific, as it was observed for all cancer types examined, except for colorectal cancer.In 2014, four updated meta-analyses were published.Upon revision of 6,494 cases and 9,884 controls from 23 studies, Xu et al. [29] found that the -94ins/delATTG polymorphism was significantly associated with increased cancer risk under all the inheritance models.Stratified analysis by cancer type showed significant associations for ovarian  cancer, hepatocellular carcinoma, and oral squamous cell carcinoma, but not for bladder cancer or lung cancer.Ethnicity subgroup analysis indicated that the polymorphism contributed to cancer risk in the Asian, but not the Caucasian, population.Another study by Yang et al. [28], which included 21 reports with 6,127 cases and 9,238 controls, also detected an increased overall cancer risk.Stratified analysis revealed a significant association between the polymorphism and ovarian, oral, and prostate cancers.These findings were also specific to the Asian population.Duan et al. [31] reviewed a total of 25 studies that included 8,750 cancer cases and 9,170 controls.They found that the insertion allele of the -94ins/delATTG polymorphism significantly increased cancer risk, both in overall genetic analysis as well as in Asians.Stratified analysis revealed that the polymorphism was associated with increased risk for oral squamous cell carcinoma and ovarian cancer, but not for colorectal cancer, bladder cancer, or renal cell cancer.In another meta-analysis involving 7,281 cases and 10,039 controls from 25 case-control studies, Nian et al. [30] found that the -94ins/delATTG polymorphism was significantly associated with decreased susceptibility to cancer in overall population under homozygous, recessive, dominant, and allele contrast models.Subgroups analysis based on ethnicity revealed that the polymorphism conferred decreased cancer susceptibility in the Asian population.
Since then, approximately 20 new relevant casecontrol studies in English and Chinese have emerged, some containing large samples and convincing results.Our study re-evaluated the impact of the NFKB1 -94ins/ delATTG polymorphism on cancer risk.In line with some previous meta-analyses, our pooled analysis revealed a significant association with decreased cancer risk under all five genetic models.Conversely, we found that the del allele of the NFKB1 -94ins/delATTG polymorphism conferred a significantly decreased risk of cancer in the pooled analysis.Compared with the ins allele, the ins allele significantly enhances the binding ability to nuclear proteins and increases transcriptional activity, which eventually upregulates p50 (the active NF-κB1 subunit) expression [21].Given the tumor-promoting role of p50 and NF-κB, it is biologically plausible that the -94del allele confers decreased cancer susceptibility.
In line with previous data, our study detected a significant association between the -94ins/delATTG polymorphism and cancer risk in Asians, but not in Caucasians, under all five genetic models.It is thus likely that the allelic distribution of this polymorphism vary geographically and ethnically, thus leading to the discrepancies in cancer risk.This may indicate that these groups have distinct environmental or genetic cancer co-etiologies.Stratification by cancer type showed that the NFKB1 -94ins/delATTG polymorphism was inversely associated with the risk of lung cancer, nasopharyngeal carcinoma, prostate cancer, ovarian cancer, and oral squamous cell carcinoma, but no association was found for hepatocellular carcinoma, colorectal cancer, bladder cancer, gastric cancer, cervical cancer, breast cancer, or other cancers.This phenomenon may be partly attributed to the inherent heterogeneity of oncogenic progression in different cancer types [93], although the insufficient statistical power caused by the relatively small number of studies on each cancer type may also be a factor.
The credibility of our conclusions is supported by the inclusion of Chinese-language studies, exclusion of publications with controls violating the Hardy-Weinberg equilibrium, and inclusion of subgroup, publication bias, and sensitivity analyses.Among the limitations of our meta-analysis are a significant between-study heterogeneity, detected in some comparisons, which may diminish the strength of our conclusions.The source of this heterogeneity may be ascribed to sample size, genotyping methods, ethnicity, source of controls, as well as the studies' diverse quality scores.Second, we assessed the association between the NFKB1 -94ins/delATTG polymorphism and cancer risk from a genetic perspective only, by using unadjusted ORs.Multiple potentially influential factors such as life style, environmental exposure, and gene-environment interactions should be considered to obtain a more precise risk estimation.Third, the number of studies in certain subgroup analyses was too small to obtain a reliable association.For instance, only six publications were included for hepatocellular carcinoma, and fewer studies were available for breast cancer and oral squamous cell carcinoma, which restrains further analysis for risk factors.Finally, the meta-analysis is a type of retrospective study with several inherent drawbacks: inconsistent qualities of primary studies, incomplete histological details, misclassified genotypes, different definitions of disease status, and improperly matched sources of controls.
In conclusion, despite these limitations, and in agreement with several previous studies, this metaanalysis draws the robust conclusion that NFKB1 -94ins/ delATTG polymorphism is associated with decreased cancer risk, especially in the Asian population.These findings indicate a possible involvement of NFKB1 in the etiology of tumorigenesis, and suggest the potentially relevant therapeutic value of NF-κB modulation in cancer prevention.Further multi-center, well-designed investigations with larger sample sizes that include geneenvironment interactions assessment are warranted to confirm our findings.

Publication search
We performed a comprehensive literature search by using the PubMed and EMBASE databases, without language limitations, up to July 1, 2016.The following search terms were used: "polymorphism or SNP or single nucleotide polymorphism or variant" and "NFKB1/ NF-κB1 or nuclear factor kappa B1", and "tumor or cancer or neoplasm or carcinoma".We also searched the China National Knowledge Infrastructure (CNKI) and WANFANG databases to obtain additional, relevant studies.Retrieved articles were manually screened to determine eligible studies.When two or more publications containing overlapping data were found, the largest study was included in the final meta-analysis.

Inclusion/exclusion criteria
All articles included in the current analysis met the following criteria: 1) evaluation of the association between NFKB1 -94ins/delATTG polymorphism and cancer risk; 2) case-control studies; 3) sufficient information provided to estimate ORs and 95% CIs; 4) NFKB1 -94ins/delATTG genotype frequency in agreement with HWE in controls.Exclusion criteria were as follows: 1) case-only studies; 2) meta-analysis or reviews; 3) studies that lacked detailed genotyping data; 4) duplicates of previous publications.

Data extraction
Two authors (Z.Z. and W.F.) evaluated all eligible studies independently and extracted the following information: first author's surname, year of publication, cancer type, country, ethnicity, source of controls, genotyping methods, and genetic distribution of cases and controls.Stratification analyses were conducted by cancer type, ethnicity (Asians, Caucasians), source of control (hospital-based and population-based) and quality score (>9 and ≤9).If a study contained two or more ethnic groups or cancer types, we divided the study accordingly.

Trial sequential analysis
TSA was performed as described by us previously [94].Briefly, after adopting a level of significance of 5% for type I error and of 30% for type II error, the required information size was calculated, and TSA monitoring boundaries were built.

FPRP analysis
The FPRP values at different prior probability levels for all significant findings were calculated as described by us previously [95].Briefly, 0.2 was set as FPRP threshold and assigned a prior probability of 0.01 to detect an OR of 0.67 (for protective effects) for an association with genotypes under investigation.A FPRP value <0.2 denoted a noteworthy association.

Statistical methods
Goodness-of-fit χ 2 test was used to assess HWE in the control subjects.Departure from HWE was assessed using a P < 0.05 as threshold in each study.The strength of the association between NFKB1 -94ins/delATTG polymorphism and cancer risk was assessed by calculating ORs and corresponding 95% CIs.Five genetic models were adopted: homozygote model (DD, homozygous deletion (del/del) vs. II, homozygous insertion (ins/ins) or wild-type); heterozygote model (ID, heterozygous ins/del vs. II); recessive model (DD vs. ID/II); dominant model (ID/DD vs. II); and allele contrast model (D vs. I).Subgroup and stratification analyses were also performed to test the association by ethnicity, cancer type, source of control and quality score.We performed χ 2 -based Q-test to assess heterogeneity between study results.The fixed-effects model (Mantel-Haenszel method) was used if the studies were found to be homogeneous (with P > 0.10 for the Q-test).Otherwise, the random-effects model (DerSimonian and Laird method) was adopted to estimate the pooled OR [96][97][98][99].Quality assessment for each study was performed using the quality assessment criteria described previously (Supplementary Table 1) [98].Sensitivity analysis was carried out by individually removing each study and reanalyzing the pooled risk estimates.Potential publication bias was estimated by Begg's funnel plot and Egger's linear regression, where an asymmetric plot and a P value < 0.05, respectively, indicate the presence of publication bias.All the data management and statistical analyses were completed using STATA software (Stata Corporation, College Station, TX; version 11.0).All the P values were twosided.A P value of < 0.05 was considered statistically significant.

Figure 1 :
Figure 1: Flowchart of the study inclusion protocol.

Figure 2 :
Figure 2: Forest plot of the association between the NFKB1 -94ins/delATTG polymorphism and overall cancer susceptibility in the allele contrast model.The horizontal lines represent the study-specific ORs and 95% CIs, respectively.The diamond represents the pooled OR and corresponding 95% CI.

Figure 3 :
Figure 3: Funnel plot analysis to detect publication bias for NFKB1 -94ins/delATTG polymorphism under the allele contrast model.Each point represents a separate study.
). 38 of these studies included Asians subjects, and 12 included Caucasians.Regarding cancer types, 6 studies addressed hepatocellular carcinoma, 5 lung cancer, 4 colorectal cancer, 4 nasopharyngeal carcinoma, 4 prostate cancer, 4 ovarian cancer, 3 bladder cancer, 3 gastric cancer, 3 cervical cancer, 2 oral squamous cell carcinoma, 2 breast cancer, and 10 studies addressed other cancers.Moreover, 40 studies used a population-based design, and 10 were hospital-based.19 studies had a quality score >9, and the remaining 31 had a quality score ≤9.

Table 1 : Characteristics of the studies included in the current meta-analysis Surname Year Cancer type Country Ethnicity Control Source Genotype method Case Control MAF HWE Score II ID DD All II ID DD All
(Continued ) www.impactjournals.com/oncotarget

Surname Year Cancer type Country Ethnicity Control Source Genotype method Case Control MAF HWE Score II ID DD All II ID DD All
PAGE, polymerase chain reaction-polyacrylamide gel electrophoresis; PCR-RFLP, polymerase chain reaction-restriction fragment length polymorphism; PCR-CE, polymerase chain reaction-capillary electrophoresis; PCR-SSCP, polymerase chain reaction-single strand conformation polymorphism; KASP, kompetitive allele specific PCR; AS-PCR, allele-specific polymerase chain reaction.

Table 3 : False-positive report probability values for associations between the NFKB1 -94ins/delATTG polymorphism and overall cancer risk
Statistical power was calculated using the number of observations in the subgroup and the OR and P values in this table.
CI, confidence interval; OR, odds ratio; NPC, Nasopharyngeal carcinoma; OSCC, oral squamous cell carcinoma; HB, Hospital based; QS, quality score.a Chi-square test was adopted to calculate the genotype frequency distributions.b