Association of four genetic polymorphisms in the vascular endothelial growth factor-A gene and development of ovarian cancer: a meta-analysis

This study meta-analyzed the literature on possible association of four polymorphisms (+936C/T, −460C/T, −2578C/A and −1154G/A) in the vascular endothelial growth factor (VEGF)-A gene with risk of ovarian cancer. Meta-analysis of 7 case-control studies involving +936C/T, 4 studies involving −460C/T, 4 studies involving −2578C/A and 2 studies involving −1154G/A showed significant association between −460C/T and ovarian cancer risk. This risk was observed in the total population (allelic model, OR 1.80, 95% CI 1.26–2.59, P = 0.001; recessive model, OR 1.84, 95% CI 1.13–2.98, P = 0.01; dominant model, OR 0.51, 95% CI 0.39–0.67, P < 0.001; homozygous model, OR 2.48, 95% CI 1.72–3.56, P < 0.001; heterozygous model, OR 1.67, 95% CI 1.26–2.21, P < 0.001) and in the subgroup of Asian study participants. The CA genotype at −2578C/A was a risk factor in the total population, while the CT genotype at +936C/T was a protective factor in Caucasians. None of the five genetic models suggested a significant association between −1154G/A and ovarian cancer risk in the entire study population, or between +936C/T and risk in Asian or Chinese participants. These findings should be verified in large, well-designed studies.


INTRODUCTION
Ovarian cancer is a major cause of cancer-related death in females worldwide [1,2]. Although treatment can significantly improve quality of life, the 5-year survival rate for patients with advanced ovarian cancer remains below 30%, mainly due to high rates of recurrence and metastasis [3,4]. Development of ovarian cancer has been linked to numerous environmental and lifestyle factors, including age, early menarche, late menopause, non-childbearing, high-fat diet, exposure to talcum powder and asbestos, and long-term hormone supplementation [5,6]. Ovarian cancer has also been linked to several genetic polymorphisms [7][8][9].
Angiogenesis, which refers to the formation of new capillary blood vessels from preexisting ones, is an important factor in the development and spread of cancer, including ovarian cancer [10][11][12]. A key mediator of angiogenesis is vascular endothelial growth factors (VEGFs) [13], which are expressed at higher levels in malignant ovarian tumor tissues than in benign tumor tissues or tissue of low malignant potential [14][15][16]. This implicates VEGFs in the pathological angiogenesis of ovarian cancer. Indeed, prognosis and overall survival of ovarian cancer patients correlate with serum and/or tumor levels of VEGFs [17][18][19][20][21]. These findings suggest that genetic factors affecting VEGF expression or activity may influence ovarian cancer development and progression.

Meta-Analysis
Oncotarget 73064 www.impactjournals.com/oncotarget meta-analysis of all available evidence on these potential associations. To the best of our knowledge, this is the first meta-analysis concerning all four of these previously analyzed polymorphisms and ovarian cancer risk.

Description of studies
A total of 104 potentially relevant publications published in English or Chinese up to April 12, 2017 were systematically identified in PubMed, EMBASE, Google Scholar and Chinese National Knowledge Infrastructure databases ( Figure 1). We excluded 83 studies based on review of the titles and abstracts, because they did not analyze the target polymorphisms in the VEGF-A gene or because they did not examine ovarian cancer risk. We excluded another 8 studies because they were not casecontrol studies, 3 studies because they were review articles and 1 study because it did not report precise genotypes. Another 2 studies were excluded because they analyzed overlapping patient populations. In the end, 7 studies were included in the final meta-analysis [26][27][28][29][30][31][32] (Table 1).
All 7 studies evaluated the association between the +936C/T polymorphism and ovarian cancer risk (1,345 cases and 1,671 controls). Four studies [26,29,31,32] evaluated the association between the −460C/T polymorphism and ovarian cancer risk (813 cases and 905 controls); 4 studies [28,29,31,32], the association between the −2578C/A polymorphism and ovarian cancer risk (1,022 cases and 1,228 controls); and 2 studies [29,31], the association between the −2578C/A polymorphism and ovarian cancer risk (602 cases and 623 controls). The distribution of genotypes in controls was consistent with Hardy-Weinberg equilibrium (HWE, P > 0.05) in all but one study [32] involving the −460C/T polymorphism.
All studies in the meta-analysis received a score of at least 6 on the Newcastle-Ottawa Scale [34], indicating that they were all of good quality. The mean score for all included studies was 7 (Table 2).

Meta-analysis of studies on the +936C/T (rs3025039) polymorphism
Meta-analysis of a possible association between +936C/T polymorphism and ovarian cancer risk is summarized in Table 3. Based on the total study population involving 1,345 cases and 1,751 controls, none of the five genetic models indicated a significant association: allelic model, OR 1.17, 95% CI 0.79-1.72, P = 0.44 ( Figure 2A Figure 2D); and heterozygous model, OR 1.07, 95% CI 0.65-1.76, P = 0.79 ( Figure 2E).
We also meta-analyzed data for ethnic subgroups. Meta-analysis of 4 studies [28,29,31,32] involving 1,022 Asian cases and 1,228 Asian controls showed no evidence of a significant association between the +936C/T polymorphism and ovarian risk risk for any of the five genetic models (      Table 3). Lastly, meta-analysis of 3 studies [28,29,32] involving 722 Chinese cases and 908 Chinese controls showed no evidence of a significant association between the +936C/T polymorphism and ovarian risk for any of the five genetic models (Table 3)
A significant association was also observed in the subgroup of 766 Asian cases and 799 Asian controls in 3 studies [29][30][31][32] according to five genetic models (

Meta-analysis of studies on the -1154G/A (rs1570360) polymorphism
The meta-analysis of a possible association between the −1154G/A polymorphism and ovarian cancer risk is summarized in Table 6. Based on the total study population (exclusively Asian) involving 602 cases and 623 controls, none of the five genetic models indicated a significant

Sensitivity analysis
The robustness of the meta-analysis of 4 studies [26,29,31,32] examining a possible association between the −460C/T polymorphism and ovarian cancer risk was assessed by repeating the meta-analysis after excluding a study by Zhang et al. [32] in which the P value associated with HWE was less than 0.05. Deleting these data from the meta-analysis did not alter the results except in the recessive model, the results of which should therefore be interpreted with caution.

Publication bias
Potential publication bias in this meta-analysis was assessed using Begg's funnel plot. No obvious asymmetry was observed in Begg's funnel plots of allelic modeling of the polymorphisms +936C/T (Figure 6

DISCUSSION
The number of case-control studies exploring the influence of VEGF-A polymorphisms on ovarian cancer risk has grown in recent years [26][27][28][29][30][31][32]. Limited sample size and ethnic differences among the various populations examined have contributed to a lack of consensus in this literature, so we conducted this comprehensive metaanalysis to evaluate the association of ovarian cancer risk with four polymorphisms in the VEGF-A gene (+936C/T, −460C/T, −2578C/A and −1154G/A). Our meta-analysis suggests that the −460C/T polymorphism is significantly associated with ovarian cancer risk across the total population as well as the Asian population. In contrast, none of the five genetic models suggested a significant association between the + 936C/T polymorphism and ovarian cancer risk in Asian populations in general or in Chinese populations specifically. None of the five genetic models suggested a significant association between the −1154G/A polymorphism and ovarian cancer risk across the entire study population.
While the present meta-analysis was being conducted, Zhang et al. [33] published a meta-analysis of the relationship between ovarian cancer risk and the three polymorphisms +936C/T, −460C/T, and −2578C/A. Similar to their results, we found that the CT genotype at +936C/T may act as a protective factor in Caucasian populations. On the other hand, our meta-analysis contrasts with the previous one because we found the −460C/T polymorphism to be significantly associated with ovarian cancer risk across the total population as well as the Asian subpopulation, and the CA genotype at −2578C/A to be associated with cancer risk across the total population, whereas that previous metaanalysis did not report either association. This discrepancy  Oncotarget 73072 www.impactjournals.com/oncotarget may reflect the fact that we included two large case-control studies involving all four VEGF-A polymorphisms absent from the previous meta-analysis, leading to much larger sample sizes for meta-analysis of 2578C/A and −460C/T polymorphisms in our work. In addition, we meta-analyzed the relationship between −1154G/A polymorphism and ovarian risk, which was not examined in that previous meta-analysis. Therefore, our meta-analysis provides new evidence for the important role of VEGF-A polymorphisms in ovarian cancer development. To the best of our knowledge, the present study is the most comprehensive and robust meta-analysis of these genetic polymorphisms and ovarian cancer.
Despite the potential insights it offers, the present study has several limitations that may affect interpretation of the results. First, the P value for HWE was less than 0.05 in the case-control study by Zhang et al. [32] on the −460C/ T polymorphism. These results suggest that this study population may not be representative of the broader target population. Nevertheless, sensitivity analyses showed that deleting these data from the meta-analysis did not alter the results except in the recessive model, which is unlikely to significantly affect the observed significant relationship between −460C/T polymorphism and ovarian cancer risk. Second, our exclusion of unpublished data and of papers published in languages other than English and Chinese may have biased our results. Third, the studies may be subject to performance bias, attrition bias and reporting bias, although Newcastle-Ottawa scores were at least 6 for all 7 studies, indicating high quality. Lastly, the results may be affected by additional confounding factors, such as age, obesity, type of cancer, or other factors, and we could not take this into account in the meta-analyses because studies either did not report these baseline data or they aggregated the data in different ways. Thus, these conclusions should be verified in large, well-designed studies.
In conclusion, this meta-analysis indicates that there may be a significant association between the −460C/T polymorphism and ovarian cancer risk. The CA genotype at −2578C/A may be a risk factor in the total population, while the CT genotype at +936C/T may be a protective factor in the Caucasian population. The −1154G/A polymorphism may not be related to ovarian cancer risk.

Search strategy
PubMed, EMBASE, Google Scholar and the Chinese National Knowledge Infrastructure databases were systematically searched up to April 12, 2017 for clinical and experimental case-control studies published in English or Chinese that assessed potential associations of ovarian cancer risk with at least one of the following polymorphisms in the VEGF-A gene: +936C/T (rs3025039), −460C/T (rs833061), −2578C/A (rs699947), and −1154G/A (rs1570360). The following search strings were used: vascular endothelial growth factor +936C/T, vascular endothelial growth factor −460C/T, vascular endothelial growth factor −2578C/A, vascular endothelial growth factor −1154G/A, rs3025039, rs833061, rs699947, and rs1570360. Searches were also conducted with each of these eight terms AND each of the following terms: polymorphism, polymorphisms, SNP, variant, variants, variation, genotype, genetic or mutation. Lastly, searches were conducted with each of the above terms AND each of the following: ovarian cancer, ovarian carcinoma or OC. Reference lists in identified articles and reviews were also searched manually to identify additional eligible studies.
This literature and meta-analysis were performed in accordance with the guidelines and recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (Supplementary Table 1) [34].

Inclusion criteria
To be included in our review and meta-analysis, studies had to satisfy the following criteria: (1) a case-  Oncotarget 73074 www.impactjournals.com/oncotarget control design was used to assess the association of at least one of the four target polymorphisms with ovarian cancer risk in humans; (2) full text was available, and sufficient data were reported to estimate an odds ratio (OR) with 95% confidence interval (CI); and (3) genotype frequencies were reported. If multiple publications from the same research group appeared to report data for the same cases and controls, we included only the most recent publication in our meta-analysis. Oncotarget 73075 www.impactjournals.com/oncotarget

Data extraction
Two authors (CHX and ZHH) independently extracted the following data from included studies: first author's family name, year of publication, ethnicity, country of origin, testing methods, type of controls, P value for HWE in controls, numbers and genotypes of cases and controls, frequencies of genotypes in cases and controls. Discrepancies were resolved by consensus.

Assessment of methodological quality
The quality of the included studies was assessed independently by two authors (CHX and ZHH) according to the Newcastle-Ottawa Scale [35]. This scale awards Oncotarget 73076 www.impactjournals.com/oncotarget a maximum of 9 points to a study, with higher scores indicating better quality. Differences in quality score outcomes between the two assessors were solved by consensus. If consensus was not reached, a third assessor (HX) was consulted for the final decision.

Statistical analysis
Unadjusted odds ratios (ORs) with 95% confidence intervals (95% CI) were used to assess the strength of the association of each of the four target polymorphisms