Four Common Vascular Endothelial Growth Factor Polymorphisms (−2578C>A, −460C>T, +936C>T, and +405G>C) in Susceptibility to Lung Cancer: A Meta-Analysis

Background and Objective Vascular endothelial growth factor (VEGF) is one of the key initiators and regulators of angiogenesis and it plays a vital role in the onset and development of malignancy. The association between VEGF gene polymorphisms and lung cancer risk has been extensively studied in recent years, but currently available results remain controversial or ambiguous. The aim of this meta-analysis is to investigate the associations between four common VEGF polymorphisms (i.e., −2578C>A, −460C>T, +936C>T and +405C>G) and lung cancer risk. Methods A comprehensive search was conducted to identify all eligible studies to estimate the association between VEGF polymorphisms and lung cancer risk. Crude odds ratios (ORs) with 95% confidence intervals (CIs) were used to evaluate the strength of this association. Results A total of 14 published case-control studies with 4,664 cases and 4,571 control subjects were identified. Our meta-analysis provides strong evidence that VEGF −2578C>A polymorphism is capable of increasing lung cancer susceptibility, especially among smokers and lung squamous cell carcinoma (SCC) patients. Additionally, for +936C>T polymorphism, increased lung cancer susceptibility was only observed among lung adenocarcinoma patients. In contrast, VEGF −460C>T polymorphism may be a protective factor among nonsmokers and SCC patients. Nevertheless, we did not find any association between +405C>G polymorphism and lung cancer risk, even when the groups were stratified by ethnicity, smoking status or histological type. Conclusion This meta-analysis recommends more investigations into the relationship between −2578C>A and −460C>T lung cancer risks. More detailed and well-designed studies should be conducted to identify the causal variants and the underlying mechanisms of the possible associations.


Introduction
Lung cancer, characterized by uncontrolled cell growth in tissues of the lung [1], accounts for 13% (1.6 million) of the total cancer cases and 18% (1.4 million) of total deaths in 2008 [2]. Lung cancer has become a major public health challenge all over the world, especially in China [3]. Thus, understanding the molecular biology and etiology of lung cancer will be pivotal in designing targeted therapies and personalized medicines. The link to smoking as a definite causative agent for lung cancer has been well established from epidemiologic evidence since the 1950s [4,5]. However, epidemiological data showed that only 10-15% of heavy tobacco smokers ultimately develop lung cancer [6,7], suggesting that certain common genetic variants or polymorphisms may influence the risk of lung cancer, particularly among those who have developed lung cancer. Vascular endothelial growth factor (VEGF), also known as vascular permeability factor, is one of the key initiators and regulators of angiogenesis and it plays a critical role in the progress and prognosis of malignancy [8][9][10]. Evidence from in vitro and in vivo experiments have shown that high levels of VEGF expression were found to be associated with tumor growth and metastasis, whereas the inhibition of VEGF signaling results in suppression of both tumor-induced angiogenesis and tumor growth [11][12][13]. Bevacizumab, one of the agents for recognizing and blocking vascular endothelial growth factor A (VEGF-A), has been a promising agent in a combination regimen in improving the overall survival and progression-free survival of breast cancer, non-small-cell lung cancer, renal cell carcinoma, and other solid malignancies [14,15].
The VEGF gene, which contains a 14-kb coding region with eight exons and seven introns, is located on chromosome 6p21.3 [16]. At least 30 single nucleotide polymorphisms (SNPs) in VEGF gene have been identified and described, and some have even been shown to affect the expression of VEGF gene [17,18]. Several previously published meta-analyses showed that VEGF +936C.T (rs3025039), one of the most common polymorphisms, was not associated with gastric cancer [19][20][21], colorectal cancer [22], or breast cancer [23][24][25]. Additionally, these published meta-analyses also showed that other three common VEGF polymorphisms, 21154G.A (rs1570360), 2634G.C (rs2010963) and 2460C.T (rs833061), were not associated with colorectal cancer [26] or breast cancer [24], whereas the VEGF 2634G/C polymorphism was found to be associated with gastric cancer [20]. In recent years, four common polymorphisms in VEGF gene, 22578C.A, 2460C.T, +936C.T, and +405C.G, have been described in several literatures to appear to be involved in the development of lung cancer [27][28][29][30][31]. However, the results remain controversial or inconclusive. To the best of our knowledge, there were no published meta-analyses investigating the association between VEGF gene polymorphisms and lung cancer susceptibility. Therefore, we performed a meta-analysis of all eligible casecontrol or cohort studies to investigate whether these functional VEGF polymorphisms are associated with any increased risk of lung cancer and whether the associations are modulated by smoking status, histological type or other risk factors. We hope our meta-analysis can potentially be important in early lung cancer identification and become part of the therapeutic strategies in combating lung cancer.

Literature Search
Relevant papers for this meta-analysis were systematically identified through literature searches on PubMed, Embase, Web of science and Chinese National Knowledge Infrastructure (CNKI), and Chinese Biomedical Literature Database (CBM) of publications published up to March 9, 2013 relating to VEGF gene polymorphisms and lung cancer risk. As the main search criteria, we used combinations of the following terms: "VEGF", "vascular endothelial growth factor A", "vascular permeability factor", "vascular endothelial growth factor", "lung neoplasms", "pulmonary neoplasms", "bronchial neoplasms", "lung cancer", "bronchial neoplasm", "genetic polymorphism", "single nucleotide polymorphism", "SNP", "mutant", "gene variation". We also reviewed the reference lists of articles retrieved to identify relevant publications.

Inclusion and Exclusion Criteria
Our meta-analysis included genetic association studies fulfilling the following inclusion criteria: (a) a case-control, cohort or crosssectional study must evaluated at least one of four polymorphisms of VEGF gene and lung cancer risk; (b) the diagnosis of lung cancer patients was confirmed pathologically and controls were confirmed as cancer-free patients; (c) inclusion of sufficient data on the size of the sample, odds ratio (OR), and 95% confidence interval (CI) and (d) articles were published in the English or Chinese language.
Studies were excluded when they represented duplicates of previous publications, or were meta-analyses, case report, letters, reviews or editorial articles. Studies investigating the progression, severity, phenotype modification, response to treatment, or survival were also excluded. Additionally, when data was included in multiple studies using the same case series, either the study with the largest sample size or most recent publication was selected. Finally, family-based studies were excluded because of different design settings. Any disagreements on study inclusion were resolved through discussions between the authors. To ensure the rigour of the current meta-analysis, it was designed and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement. The relevant checklist is shown in Supplement S1.

Data Extraction
All data from the included studies were extracted independently by two investigators, using a piloted data standardized form (when it came to conflicting evaluations, an agreement was settled after a discussion): the first author's surname, year of publication, country of origin, published language, gender of study individuals and ethnic subgroups, study design, number of subjects, smoking status, histological types of lung cancer, SNP genotyping methods, genotyping method and detected sample, allele and genotype frequencies, and evidence of Hardy-Weinberg equilibrium (HWE) in controls. In addition, we also compared key study characteristics such as location, study time and authorship to determine the existence of multiple publications from the same study.

Quality Assessment of Included Studies
Two authors independently assessed the quality of the published articles according to the modified STROBE quality score systems [32]. Forty assessment items matching with the quality appraisals were used in this meta-analysis, with scores ranging from 0 to 40. Scores of 0-20, 20-30 and 30-40 were defined as low, moderate and high quality, respectively. The two authors resolved their differences through discussions; if no agreement could be reached, a third author decided on a decision. The modified STROBE quality score system is available in Supplement S2.

Statistical Analysis
Crude ORs together with their corresponding 95% CIs were used to calculate and assess the strength of association between VEGF gene polymorphisms and lung cancer risk under five genetic models: allele, dominant, recessive, homozygous, and heterozygous models. The deviation of frequency from those expected under Hardy-Weinberg equilibrium (HWE) was assessed by Chisquared goodness of fit tests in controls. We explored inter-study variation through prespecified subgrouping of studies according to ethnicity (ie, Caucasian or Asian), gender (ie, female or male), smoking status (ie, smoker or non-smoker), and histological type of lung cancer (ie, adenocarcinoma, squamous cell carcinoma (SCC), and small cell carcinoma (SCLC), where applicable. The statistical significance of the pooled OR was assessed with a Z test. Betweenstudy variation and heterogeneity were estimated using Cochran's Q-statistic, with P,0.05 as a cutoff for statistically significant heterogeneity [33].
We also quantified the effect of heterogeneity with the I 2 test (ranges from 0 to 100%), which represents the proportion of interstudy variability that can be attributed to heterogeneity rather than to chance [34]. The fixed effects model (Mantel-Haenszel method) was used, except when a significant Q-test (P,0.05) or I 2 .50% indicated the existence of heterogeneity among studies; otherwise, the random effects model (DerSimonian-Laird method) was applied for meta-analysis. In order to ensure the reliability of our results, sensitivity analysis was performed by omitting individual studies. Begger's funnel plots were used to detect publication bias. In addition, Egger's linear regression test, which measures funnel plot asymmetry via a natural logarithm scale of OR, was also used to evaluate publication bias [35]. All P-values were two-sided. Analyses were conducted with STATA Version 12.0 software (Stata Corp, College Station, TX).

The Characteristics of Included Studies
Our initial literature search yielded 546 reports, which included 13 population-based [28][29][30][31][36][37][38][39][40][41][42][43][44] and one hospital-based [27] case-control studies meeting the inclusion criteria based on the search criteria for lung cancer susceptibility linking to at least one of four common SNPs of VEGF gene, 22578C.A, 2460C.T, +936C.T, and +405C.G. The flow diagram of the selection of studies and specific reasons for exclusion from the meta-analysis are shown in Figure 1. We studied four VEGF SNPs in 4,664 unrelated lung cancer cases and 4,571 unrelated controls from 14 case-control studies. In the eligible studies, there were 12 studies of subjects of Asian descent and only two studies of subjects of Caucasian descent. All included studies extracted DNA from peripheral blood and the VEGF polymorphisms were determined by classic PCR-RFLP in 12 studies, by TaqMan in 1 study, and by PIRA-PCR in another study. SNP genotypes were tested for departures from HWE for controls and all SNPs were in HWE. The qualities of the included studies were moderately high, with a STROBE score of greater than 20. The selected study characteristics were summarized in Table 1. The evaluation of the associations between VEGF 22578C.A, 2460C.T, +936C.T, and +405C.G polymorphisms and lung cancer risk are presented in Tables 2, 3, 4 and 5.

VEGF 22578C.A Polymorphism and Risk of Lung Cancer
A total of 7 studies with 22 data sets involving 1,596 cases and 1,857 controls were included in the pooled analysis. All subjects were of Asian ethnicity. Meta-analysis results showed that a statistically significant correlation was found between 22578C.A polymorphism and susceptibility to lung cancer in Asians under allele and homozygous models (for OR = 1.31, 95%CI = 1.10-1.57, P = 0.003; OR = 1.79, 95%CI = 1.30-2.46, P,0.001). We    Figure 2C).

VEGF +936C.T Polymorphism and Risk of Lung Cancer
Eight studies investigated the association between +936C.T polymorphism and lung cancer susceptibility with a total of 3,288 cases and 3,092 controls. We did not find any association between

Sensitivity Analysis and Publication Bias
Sensitivity analysis was performed to assess the influence of each study on the pooled ORs by omitting individual studies. The analysis results suggested that no individual study significantly altered the pooled ORs in VEGF 22578C.A, 2460C.T, +936C.T, and +405C.G polymorphisms under the allele model (data not shown), which indicates that our studies were statistically accurate.
Begger's funnel plot and Egger's linear regression test were performed on the metadata to assess publication bias of the individual studies. The shapes of the funnel plots did not reveal any evidence of obvious asymmetry in VEGF 22578C.A (A),

Discussion
Evidence from preclinical and clinical studies shows that VEGF, as a predominant angiogenic factor in human cancers, plays a vital role in the carcinogenesis pathway, which has been proved to be a key step in tumor occurrence, progression and prognosis [12,45]. Several functional polymorphisms of VEGF gene have been confirmed to be correlated with high levels of VEGF protein in cancer cells and high tumor angiogenic activity, and they also contribute to the susceptibility and severity of cancer, including lung cancer [36]. Although cigarette smoking is the major cause of lung cancer, only a small fraction of smokers develop this disease during their lifetime, which suggests that both genetic factors and lifestyle risk factors are modulating individual susceptibility to lung cancer risk. A study by Koukourakis et al. reported that non-small cell lung cancer patients with specific VEGF gene polymorphisms develop tumors with low VEGF expression and poor vascularization [46]. In recent years, the associations between VEGF and risk of lung cancer have been extensively investigated, obtaining conflicting results. Therefore, we employed a meta-analysis to explore a more precise evaluation for the associations. To our knowledge, this is the first meta-analysis on this topic.
The present meta-analysis, including 4,664 cases and 4,571 controls from 14 published case-control studies, explored the association between VEGF 22578C.A, 2460C.T, +936C.T, and +405C.G polymorphisms and lung cancer risk. According to our pooled analysis, 22578C.A polymorphism may have a correlation with increased lung cancer risk. This finding may be biologically plausible since Koukourakis et al. observed that 22578CC was associated with lower VEGF expression and lower vascular density in lung cancer tissues compared to the 22578C C/A [46]. When lung cancer cases were stratified by histological subtype, the data indicated that the presence of 22578A was strongly associated with SCC, while similar finding was not observed in SCLC and adenocarcinoma. Although a research reported by Jin et al. demonstrated that 22578AA genotype was significantly associated with low histologic grade tumors [47], the reason for such a divergence of VEGF expression and angiogenic status in tumors of similar histologic type and differentiation remains obscure. Thus, more studies should be conducted to further examine the underlying mechanism. Furthermore, the stratified analysis according to smoking status revealed that 22578A is significantly correlated with increased risk of lung cancer among smokers, suggesting that this polymorphism may not be an independent risk factor, but perhaps an effect modifier that acts synergistically with smoking in lung cancer risk.
As for VEGF 2460C.T polymorphism, the overall data did not show a marked association of this polymorphism with lung cancer risk in any genetic model, even in the subgroup analyses according to ethnicity. However, when stratified analysis by smoking status and histological type were performed, a lower prevalence of 2460T allele was observed among nonsmokers, lung adenocarcinoma cases, and SCC cases. Some clinical evidence suggests that cigarette smoking may stimulate both angiogenesis and VEGF expression, which exacerbates the rapid cancer progression effect of angiogenesis [48,49]. Thus, it is possible that cigarette smoke and VEGF activate multiple effects in lung cancer. For VEGF +936C.T, +405C.G polymorphisms, we found no overall association between these two polymorphisms or its interaction with smoking on lung cancer risk in any genetic model. When stratified analyses were conducted according to ethnicity and histological types of cancer, increased lung cancer susceptibility was only observed among the adenocarcinoma subgroup for +936C.T polymorphism, while there was no statistical difference in genotype distributions between cases and controls for any different subgroups for +405C.G polymorphism. Actually, there exist conflicting reports in some literatures regarding the exact function of the +405G/C polymorphism. Some clinical studies suggested that +405C allele has been associated with lower VEGF production, while some groups reported higher VEGF levels or even no association with +405C/C genotype [17,50,51]. Thus, whether these polymorphisms are truly functional requires further investigation through confirmatory studies and in vitro functional assays.
The current meta-analysis has several limitations that should be noted. First, the sample size in the present study was relatively small, so small, but potential, genetic effects may not be detectable. A small sample size may not have enough statistical power to explore the real association, especially in subgroup analysis. Additionally, as with other complex traits, lung cancer risk may also be modulated by several other genetic markers beyond VEGF, and our meta-analysis emphasized that elucidating the pathogenesis of lung cancer would demand an investigation into the association for many gene variants that may constitute distinct pathophysiological pathways. Third, we identified two studies from Caucasian populations and obtained no data from African populations, thus the two racial groups need to be further studied in the future. Therefore, the results should ideally be confirmed in further studies to strengthen the conclusions. Aside from the limitations listed above, our meta-analysis still has some strength.
To the best of our knowledge, this is the first meta-analysis on the relationship between VEGF gene polymorphisms and lung cancer. We also explored inter-study variations by prespecified subgrouping of studies according to ethnicity, smoking status, gender, and histological type among cases. Furthermore, although this metaanalysis does not accommodate all previously published data, they are limited compared to the evidence that we generated.
In conclusion, this meta-analysis provides strong evidence that VEGF 22578C.A polymorphism is capable of increasing lung cancer susceptibility, especially among smokers and lung SCC patients. Additionally, for +936C.T polymorphism, increased lung cancer susceptibility was only observed among lung adenocarcinoma patients. In contrast, VEGF 2460C.T polymorphism may be a protective factor among nonsmokers, lung adenocarcinoma and SCC patients. However, we did not find any association between +405C.G polymorphism and lung cancer risk, even when the groups were stratified by ethnicity, smoking status or histological type. More detailed and well-designed studies with larger population and different ethnicities are needed to further evaluate these associations.

Supporting Information
Supplement S1 PRISMA Checklist.