Prognostic Impact of Elevation of Vascular Endothelial Growth Factor Family Expression in Patients with Non-small Cell lung Cancer: an Updated Meta-analysis

Lung cancer is the leading cause of cancer-related mortality around the world (Alberg and Samet, 2003). Less than 15% of the patients will be cured and enjoy long-term survival. The poor prognosis has shown little improvement in recent decades (Molina et al., 2008). In light of disappointing therapeutic effect it is likely that the use of clinical or molecular markers will become important in predicting response to treatment and outcome. The main prognostic factors in NSCLC are disease stage, amount of weight lost, microvessel density, ERCC1, RRM1, BRCA1, p53, bcl-2, KRAS, Ki-67, (18)F-FDG, AKT, mTOR and EGFR mutation (Paesmans et al., 1995; Kaira and Yamamoto, 2010). However, these biological prognostic factors didn’t well predict clinical outcome or their discriminate value is insufficient to predict the optimal therapeutic course for an individual. Therefore,


Introduction
it is important to identify ideal predictive/prognostic biologic markers for patients undergoing treatment.
The hypothesis "Tumor growth is angiogenesis dependent" was first proposed by Folkman in1971 (Folkman, 1971), which was confirmed by subsequent observations that tumors are strictly limited in size in the absence of neovascularization (Gimbrone et al., 1972). Angiogenesis, the formation of blood vessels from preexisting vessels at a later stage (Ferrara and Kerbel, 2005), is critical for the development and subsequent growth of tumors and is a prerequisite for metastasis. The VEGF family comprises four ligands (VEGFA, VEGFB, VEGFC and VEGFD), which exhibit specific binding profiles with three transmembrane VEGF receptors (VEGFR1, 2 and 3) and promote intracellular tyrosine kinase cascades when activated (Ferrara, 2002;Hicklin and Ellis, 2005). VEGFA and its receptor (VEGFR1, VEGFR2) play a major role in physiological as well as pathological angiogenesis including tumor angiogenesis. While VEGFC/D and their receptor VEGFR3 can regulate angiogenesis at early embryogenesis but mostly function as critical regulators of lymphangiogenesis during lifetime (Alitalo and Carmeliet, 2002;He et al., 2005).
The effect of VEGFs and/or VEGFRs expression on survival in patients with NSCLC has been studied for over decades. However, conflicting results regarding the ability of VEGF to predict survival have been reported one by one from different laboratories. In this study, we sought to conduct a systematic review with meta-analysis to primarily estimate the prognostic impact of each VEGF family member for survival in NSCLC patients. The secondary goal of our study was to explore whether the VEGF/VEGFR co-expression with sufficient discriminate value as preferable predictive/prognostic biologic marker for an individual undergoing treatment.

Literature search
An electronic literature search was performed to identify potentially relevant published article in Pub Med, EMBASE, Web of Science. We used lung cancer/ lung tumor/lung neoplasm/lung carcinoma and vascular endothelial growth factor/VEGF and prognosis/prognostic as searching terms. No lower date limit was used and last search was updated on May 9, 2014. All potentially eligible studies were retrieved. The study protocol conforms to the ethical guidelines of the Declaration of Helsinki, which was approved by the ethics committee of Provincial Hospital Affiliated to Shandong University.

nclusion an stu criteria
To be eligible studies for inclusion in this metaanalysis had to meet the following inclusion criteria: (i) measure VEGF and/or VEGFRs expression in the primary lung cancer tissue with immunohistochemistry (IHC) or enzyme linked immune sorbent assay (ELISA); (ii) the study has sufficient data on survival for extracting; (iii) the same patient series was included in more than one publication, only the more recent or most complete study was included in the analysis.
In addition, the expression of different VEGFs and/or VEGFRs obtained from the same patient population by the same authors in different years was also included; There was no pre-specified sample size or follow-up period used to determine study inclusion. Criteria used to determine duplicate populations included study period, treatment information, and any additional inclusion criteria. Language was restricted for review title and abstract in English, but was not restricted for data collection.

Data extraction
Data were extracted using a predefined form, recording: first author, year of publication, number of patients, histology, disease stage, detective method, cutoffs and positive ratios of positive expression, outcome of univariate or multivariate analysis (HRs, 95% CIs) and the original author's results (Table 1). If the above data of any categories was not reported in the primary study, items were treated as "NR" (not report). We did not contact the author of the primary study to request the information. The required information was extracted independently from primary studies by two reviewers (Chunlong Zheng and Chen Qiu) according to a standard data record worksheet designed in advance.

Statistical
All calculations were performed with the hazard ratios (HRs) and the associated 95% confidence intervals (CIs) of OS, RFS, DSS, or DFS. The most accurate and easiest method was to collect the reported HRs and CIs from primary articles directly. When these statistical variables were not given explicitly in an article, which were either extracted from the Kaplan-Meier survival curves indirectly or calculated, if available number (the total numbers of events and the total numbers of patients in each group) were given, assuming that the rate of censored patients was constant during the study follow-up. The method and spreadsheet used for these calculations were provided by Tierney et al (Tierney et al., 2007). HRs defined as the risk of death or progression for high expression vs low expression. In studies that reported HRs for low level vs high level, the reciprocal of the HR calculated and p value were taken to calculate the associated 95% CI for meta-analysis.
Heterogeneity of the individual HRs was performed with Q statistic test and I 2 statistic test. All of the studies included were categorized by VEGF and VEGFRs isoform, histology, disease stage, patient race. Individual meta-analysis was conducted in each subgroup. If HRs were found to have fine homogeneity (p (Q)>0.05, I 2 <56%) (Walter, 1997;Hardy and Thompson, 1998;Dwyer et al., 2001;Higgins and Thompson, 2002), a fixed effect model was used for secondary analysis; if not, a random-effect model was used. In this meta-analysis, Inverse Variance fixed effects and I-V heterogeneity random effects analysis were used to estimate the effect of VEGF family high expression on survival. By convention, an observed HR>1 implies worse survival with positive expression, and the impact on survival was considered to be statistically significant if the 95% CI didn't overlap with 1. Similarly, if HR<1 means good survival, and statistically significant if the 95% CI didn't overlap with 1. Horizontal lines represent 95% CIs. Each square represents the HR point estimate, and its area is proportional to the weight of the study. The diamond (and broken line) represents the overall summary estimate, with CI represented by its width. The unbroken vertical line is set at the null value (HR=1.0).
Evidence of publication bias was sought using the methods of Egger et al. (1997) and Begg et al. (1994). Moreover, contour-enhanced funnel plot (Peters et al., 2008) was performed constructed to assess publication and/or selection bias. If studies appear to be missing in areas of low statistical significance, then it is possible that the asymmetry is due to publication bias. If studies seem to be missing in areas of high statistical significance, then publication bias is a less likely cause of the funnel asymmetry. Intercept significance was determined by the test suggested by Begg and Egger (p<0.05 was considered

Baseline characteristics
The main characteristics of the studies included in the meta-analysis were shown in Table 1. A total 74 studies comprising 7631 patients were included in this meta-analysis. All studies reported the prognostic value    foremost treatment measure. Seventy-two studies used IHC and two studies used ELISA to determine VEGFs and/ or VEGFRs expression. Among the 41 studies evaluating VEGFA expression in NSCLC, 29 studies (3525 patients, 75.68%) were performed in Asian populations, the remaining 12 studies (1133 patients, 24.32%) followed European or American patients. The proportion of patients exhibiting VEGFA expression in individual studies ranged from 20 to 96.1% by IHC. The rest of VEGFC, VEGFD, VEGFR1, VEGFR2 and VEGFR3 were detected in 20, 5, 4, 4 and 7 studies, respectively (Table 2).

An io enesis V
A, V 1, V 2 an ro nosis The combined HR of VEGFA expression in NSCLC (n=41) was recorded as 1.633 (95%CI: 1.490-1.791, Q=62.76, p=0.016, I 2 =34.7%, Figure 2), indicating that positive immunostaining for VEGFA was significantly associated with adverse survival in the pooled patient group overall. When grouped according to the territorial scope of each studies, 1.704 (95%CI: 1.532-1.896) in Asian and 1.443 (95%CI: 1.202-1.732) in non-Asian (  Table 2). The combined HR of stage I (1.887) was larger than stage I-III (1.552) and stage I-IV (1.683), suggesting that VEGFA expression could be an important prognostic factor for early stage NSCLC.

Heterogeneity analysis
This systematic review with meta-analysis was inspected by heterogeneity test. Highly significant heterogeneity was found among 8 studies of stage I-IV NSCLC with VEGFC expression, 4 studies of NSCLC with VEGFR2 expression, 5 studies of NSCLC with VEGFD expression, 7 studies of NSCLC with VEGFR3 expression (Table 2).

Publication bias
The publication bias in the literature was quantitative evaluated by Begg's and Egger's test. The absence of publication bias was found in 15 studies investigating VEGFC expression in patients with NSCLC, with a Begg's test score of p=0.344 and an Egger's test score of p=0.996 ( Figure 5B), Similar results were found in the four studies with VEGFR1 expression (p=0.308 and 0.671), four studies including patients withVEGFR2 expression (p=0.734 and 0.929) and eight studies investigating VEGFR3 expression in patients with NSCLC (p=0.764 and 0.872).
However, the funnel plot revealed an apparent asymmetry in 41 eligible studies investigating NSCLC patients with VEGFA expression (p=0.006 and 0.045) ( Figure 5A) and five studies investigating VEGFD expression (p=0.086 and 0.004), suggesting the presence of a potential publication bias, The bias of publication could be explained by a language bias, inflated estimates by a flawed methodological design in smaller studies, what really counts is lack of publication of trials with opposite  DOI:http://dx.doi.org/10.7314/APJCP.2015.16.5.1881 VEGF Family Expression andPrognosis of Non-small Cell Lung Cancer results (Table 2).

Discussion
Our meta-analysis showed that high VEGFs and/or VEGFRs expression did indeed predict poor survival in patients with NSCLC. For angiogenesis, our result clearly indicated that VEGFA expression has a significant correlation with poor survival in patients with NSCLC. When the analyses were restricted to the stages of NSCLC, VEGFA expression could be an important prognostic factor for early stage NSCLC (HR=1.887 for stage I).
When the analyses were restricted to the histologies of NSCLC, a significant prognostic significance (HR=2.919) was found in lung SCC patients. Data analysis revealed that VEGFR1 expression associated with low survival rate, but the VEGFR2 expression wasn't detected statistically significant effect on survival in NSCLC patients. For lymphangiogenesis, the expression of VEGFC predicted a poor prognosis in NSCLC patients. However, the VEGFC overexpression was only associated with poor survival. Unfortunately, neither the VEGFC expression nor the VEGFR3 expression was sufficient to determine the prognostic value in lung SCC patients.
The emergence of the targeted therapies for NSCLC has generated a need for accurate histologic subtyping of NSCLC (Kim et al., 2013a) because of the different clinicopathological and molecular characteristics of ADC and SCC (Miller et al., 2004;Inamura et al., 2010). Patients with SCC were not recommended to receive bevasizumab (Avastin) because of a 30% mortality rate due to fatal hemorrhage (Johnson et al., 2004;Cohen et al., 2007;Yan et al., 2011;Stead et al., 2012). There were very few previous literatures reported the prognostic impact in lung SCC patient with VEGFA expression. Our pooled analysis showed that VEGFA overexpression was associated with worse survival in patients with SCC (HR=2.919; 95%CI: 2.060-4.137). Therefore, the severe bleeding may be a significant response. Several ongoing clinical randomized controlled trials do include squamous participants (Schiller et al., 2009;Spratlin et al., 2010;Sternberg et al., 2010;Doebele et al., 2012) in recent years. The role of angiogenesis inhibition in the adjuvant setting is currently being tested by the ECOG (Eastern Cooperative Oncology Group) 1505 trial, in which NSCLC patients (including 31% SCC) with completely resected tumors are randomly assigned to chemotherapy alone or in combination with bevacizumab. The safest setting to pursue further evaluation of bevacizumab in patients with SCC seems to be after surgical resection. Our conclusion supports the new viewpoint that anti-VEGF therapies may be a reliable targeted therapy for postoperative SCC patients.
There were very few previous literatures reported the prognostic impact of VEGF/VEGFR co-expression in patients with NSCLC. Our meta-analysis explored the prognostic impact of VEGF/VEGFR co-expression on survival in patients with NSCLC. The combined HR of VEGFA/VEGFR2 co-expression in NSCLC was recorded as 2.011 (95%CI: 1.405-2.876, Q=0.29, p=0.589, I 2 =0.0%), indicating that positive immunostaining for VEGFA/VEGFR2 co-expression was significantly associated with adverse survival. The VEGFC/VEGFR3 co-expression also had highly significant prognostic value in NSCLC, with the pooled HR was 2.436 (95%CI: 1.468-4.043, Q=0.020, p=0.880, I 2 =0.0%). Empirically, HRs>2 are considered strongly predictive (Hayes et al., 2001). In a word, both VEGFA/VEGFR2 co-expression and VEGFC/ VEGFR3 co-expression were sufficient discriminate value as preferable prognostic biologic marker for an individual undergoing treatment.
Our data were consistent with the three previous meta-analysis (Delmotte et al., 2002;Zhan et al., 2009;Jiang et al., 2014), which separately included 15, 51, 16 studies. Besides, the previous analyses were insufficient to determine the prognostic value of VEGFD, VEGFRs and VEGF/VEGFR co-expression in NSCLC. We have improved these deficiencies by incorporating more related studies in recent years. What is more, we have found the several biologic markers (i.e. VEGFA expression in SCC, VEGFA/VEGFR2 co-expression in NSCLC, VEGFC/ VEGFR3 co-expression in NSCLC, etc.) with preferable discriminate value for an individual prognosis.
In addition, heterogeneity and potential publication bias were assessed in accordance with published guidelines. There were several potential sources of heterogeneity: the different of baseline characteristics of patients included (age, tumor size, and stage), the adjuvant treatment they might have received; the duration of follow-up, the differences in the cutoff value (5%, 10%, 20%, 25%, 50%, complex scores) of IHC method used, the distribution of immunostaining used for scoring was not explicitly stated in the text (i.e. cytoplasmic, membranous, nuclear, stromal, etc.), the primary antibody used wasn't identical, the different criteria used for immunohistochemical classification. But the Der Simonian and Laird method we used (random effect model) took them into account (DerSimonian and Laird, 1986).
The extract method of HRs and 95% CI was one potential source of bias. The exact value of VEGF and/or VEGFRs expression status requires to be determined by appropriate multivariate analysis. Data for multivariate survival analysis reported were included in the present systematic review with meta-analysis; if these data were not available, we extrapolated them from the survival curves as univariate analysis incorporated, necessarily making assumptions about the censoring process. The choice of study and detected method was another potential source of bias. Firstly, the prognostic value should be confirmed by adequately designed prospective study. Unfortunately, few prospectively designed prognostic studies concerning biomarkers have been reported. Our collected many retrospective studies to reveal their clinical significance. Secondly, the variability of IHC method was an inconvenient source of selection bias. In future studies, the assessment of these prognostic factors should be better standard, especially for patients whom adjuvant therapy is recommended (Vermeulen et al., 2002).
The major concern for all forms of meta-analysis is publication bias (Dubben and Beck-Bornholdt, 2005). The present analysis found significant publication bias among 47 studies of NSCLC patients with VEGFA expression, which funnel plot revealed an apparent asymmetry. The bias of publication could be explained by a language bias, inflated estimates by a flawed methodological design in smaller studies. Nevertheless, the most important point is a lack of publication of trials with opposite results. We attempted to minimize publication bias by making our literature search as complete as possible, using three databases (PubMed, EMBASE and Web of Science). The various published studies, which conclusions were discrepancy, could have encouraged researchers to publish their data whether positive or negative, such publication bias will be limited.
In summary, the expression of VEGFA (particularly in SCC and early stage NSCLC), VEGFC and VEGFR1 indicates an unfavorable impact on survival of patients with NSCLC, respectively. However, the expression of VEGFD seems to have no significant impact on survival of NSCLC patients, and the VEGFR2 expression also wasn't associated with low survival rate in our meta-analysis. Furthermore, the co-expression of VEGFA/VEGFR2, VEGFC/VEGFR3 reveals sufficient discriminate value for an individual as preferable prognostic biologic markers. These results should be confirmed by adequately further study. The findings of this study will encourage more people to identify the VEGF/VEGFR co-expression as ideal prognostic indicators in clinical practice in the future.