Prognostic Role of MicroRNA-126 for Survival in Malignant Tumors: A Systematic Review and Meta-Analysis

Background. Increasing studies found that miR-126 expression may be associated with the prognosis of cancers. Here, we performed a meta-analysis to assess the prognostic role of miR-126 in different cancers. Methods. Eligible studies were identified by searching in PubMed, Embase, the Cochrane Library, CNKI, and Wan Fang databases up to March 2015. Pooled hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs) were calculated to investigate the correlation between miR-126 and survival of cancers. Results. Thirty studies including a total of 4497 participants were enrolled in this meta-analysis. The pooled results showed that high level of miR-126 was a predictor for favorable survival of carcinomas, with pooled HR of 0.77 (95% CI 0.64–0.93) for OS, 0.64 (95%CI 0.48–0.85) for DFS, and 0.70 (95% CI 0.50–0.98) for PFS/RFS/DSS. However, high level of circulating miR-126 predicted a significantly worse OS in patients with cancer (HR = 1.65, 95% CI 1.09–2.51). Conclusions. Our results indicated that miR-126 could act as a significant biomarker in the prognosis of various cancers.


Introduction
MicroRNAs (miRNAs), which are a new class of small noncoding RNAs (21-23 nucleotides), have emerged as crucial players regulating the magnitude of gene expression in a variety of organisms [1,2]. Regulation of microRNAs is achieved via binding to the 3 untranslated regions (3 UTR) of target mRNAs, which leads to their inhibition of the expression of target genes in the translation level [3]. Mounting evidence suggests that microRNAs play crucial and complex roles in the initiation and progression of cancer [4], including cell proliferation, differentiation, apoptosis, and metabolism [5,6]. Obviously, microRNAs may be exploited as new promising molecular biomarkers for early diagnosis and efficient treatment in human cancers [7].
The majority of cancers at the time of initial diagnosis are often at an advanced stage and have poor prognosis, and therefore there is an urgent need for the identification of novel prognostic and predictive biomarkers to improve treatment of patients with various cancers [23]. In spite of some contradictory results, miR-126 is still a significant tumor biomarker and a potential therapeutic target [24]. Moreover, the result from individual study is inadequate to evaluate whether miR-126 can be considered as a promising biomarker. So we performed this meta-analysis to assess the prognostic value of tissue and blood-based miR-126 levels in various cancers.  Li). The search strategy of key words and their combination was the following terms: "microRNA-126 OR miR-126 OR miR-126-3p" AND "tumor OR tumour OR neoplasm OR cancer OR carcinoma" AND "prognosis OR survival OR outcome OR prognostic." We also carefully performed a manual search in order to identify other potentially eligible studies.

Inclusion and Exclusion
Criteria. The eligible studies in this systematic review must meet all the following criteria: (1) patients are included with any type of cancers, (2) the association between miR-126 expression and survival outcome was measured in cancerous tissues or circulatory system, and (3) sufficient data was provided to calculate the hazard ratio (HR) and 95% confidence intervals (CIs). Articles were excluded according to the following criteria: (1) letters, case reports, reviews, conference abstracts, and animal or laboratory studies, (2) studies analyzing a set of miRNAs altogether and nondichotomous miR-126 expression levels, and (3) studies with fewer than 30 patients. When the same patient cohort was reported from multiple published data, only the most recent or complete study was selected.

Quality Assessment and Data Extraction.
Quality assessment of included studies was assessed by two researchers independently (Jie Bu and Hui Li) following a critical review checklist of the Dutch Cochrane Centre proposed by MOOSE [25]. The following items were included: first author's name, publication year, country or area of origin, cancer type, sample type, TNM stage, method, total number of patients, cutoff value, follow-ups and HRs of miR-126 for overall survival (OS), disease-free survival (DFS), recurrence-free survival (RFS), progression-free survival (PFS), and disease-specific survival (DSS), with their 95% confidence intervals (CIs). Disagreements were resolved by discussion between these reviewers (Jie Bu, Hui Li, and Xiao-yang Li) or consultation with senior reviewer (Li-hong Liu). If both univariate and multivariate analysis results were reported for survival, the latter ones would be selected [26,27].
We extracted the statistical variables according to the following methods. If HRs and 95% CIs were described in publications, we extracted them directly. Otherwise, survivals and deaths at specified times in each group were extracted to calculate HRs. If only Kaplan-Meier curves are available, they were extracted from the graphical survival plots to estimate the HRs following the previously described method [28,29]. We used Engauge Digitizer version 4.1 to extract the data from Kaplan-Meier survival curves, and three independent researchers (Jie Bu, Hui Li, and Xiao-yang Li) read the curves to reduce reading variability. We also contacted the authors of eligible articles by email for additional information and the essential data needed for the meta-analytic calculations.

Statistical Analysis.
HRs with their 95% CIs were combined to evaluated the effect of miR-126 expression on the survival outcome of cancer. Patients with overexpression of miR-126 indicated a better prognosis if HR < 1 and its 95% CI did not overlap with 1. Heterogeneity of pooled HRs was carried out using Cochran's -test and Higgins -square ( 2 ) statistic [30,31]. If there was significant heterogeneity ( < 0.05 or 2 > 50%.), the random-effects model (Der Simonian and Laird method) was used [32]. Otherwise, a fixed-effects model (Mantel-Haenszel test) was applied [33]. Subgroup analysis and metaregression were further performed to explore possible explanations for heterogeneity. Begg's funnel plot and Egger's bias were used to evaluate the potential publication bias [34,35]. Analysis of sensitivity was performed to evaluate the stability of the results. All statistical tests were two-sided, and < 0.05 was regarded as statistically significant. All analyses were conducted using the Cochrane Collaboration RevMan 5.2 or STATA package version 12.0 (Stata Corporation, College Station, Texas, USA).

Eligible Studies and Characteristics.
A flowchart of detailed searching process is illustrated in Figure 1. Using the described searching strategy above, a total of 549 articles were initially retrieved out of PubMed, Embase, the Cochrane Library, CNKI, and Wan Fang databases. After manually screening the titles, publication types, and abstracts and then checking the full texts by two investigators (Jie Bu and Hui Li), 30 articles were selected for the present meta-analysis . Among these eligible studies, 20 studies evaluated the prognostic effect of miR-126 for OS, 8 studies for DFS, and 6/4/3 studies for PFS/RFS/DSS.

OS Associated with miR-126 Expression.
The main results of this meta-analysis were displayed in Table 2. 20 studies Table 1    including 3232 cancer patients investigated the relationship between miR-126 expression and the prognosis. For these studies evaluating OS for miR-126, a random-effects model was utilized to calculate the pooled HR and its 95% CI due to the high heterogeneity among these studies ( 2 = 57.0%, = 0.001). The result showed that high miR-126 level may predict a favorable OS with the combined HR of 0.77 (95% CI: 0.64-0.93; heterogeneity = 0.001) (Table 2, Figure 2(a)).
Furthermore, six subgroup analyses of overall survival were performed which stratified patients by tumor type, ethnicity, sample, assay method, analysis type, and HR estimated (  Figure 2(b)).

Publication Bias and Sensitivity Analysis.
Begg's funnel plot and Egger's test were used to assess the potential publication bias of the included studies. The funnel plots of the OS, DFS, and PFS/RFS/DSS analysis based on tissue and blood miR-126 did not reveal any evidence of obvious asymmetry. Moreover, the values of Egger's and Begg's tests were all greater than 0.05 in the 30 analysis groups (Table 2, Figure 3, and Figures S1 and S3). Hence, there was no obvious risk of publication bias in our meta-analysis.
Furthermore, we performed sensitivity analysis to investigate the influence of each individual study on the overall meta-analysis estimate, which computes the pooled HRs by omitting one study in each turn. And there was no obvious influence of individual study on the pooled HRs (Figure 4 and Figures S2 and S4).

Discussion
Cancer is considered one of the leading causes of death worldwide. The occurrence of cancer is increasing because of the growth and aging of the population, as well as increasing prevalence of established risk factors [66]. Despite the advances in technology and its access, to date, there are few defined prognostic and diagnostic biomarkers available in cancers. Essentially, high cancer mortality rates have remained high, mainly due to the late diagnosis and lack of prognostic markers for various cancers [67]. Hence, many research groups are carrying out studies to develop biomarkers, which can be applied to early detection and correlation of treatment efficacy and prognosis [68].
MiR-126, which is highly expressed in vascular endothelial cells, is one of the most commonly observed cancerrelated microRNAs and is dysregulated in most cancers. As one of the major targets of miR-126, EGFL7 is known to be involved in cell migration and the process of angiogenesis. The conclusion suggests that one of the main functions of miR-126 is to inhibit angiogenesis to reduce blood vessels, which is facilitated by cell migration [69,70]. Additionally, previous studies have demonstrated that miR-126 may play a role in tumorigenesis and growth by regulating the vascular endothelial growth factor (VEGF)/phosphoinositol 3-kinase (PI3K)/AKT signaling pathways [43,71]. miR-126 also maintains its role as a suppressor of metastasis that could reduce metastatic rate and size of carcinoma [14,72]. Furthermore, interactions of miR-126 and ADAM9 are related to epithelialmesenchymal transition and the invasive growth of pancreatic cancer cells [73]. In most of the cancers studied, miR-126 functioned as a tumor suppressor and its expression was suppressed; however, several reports using different types of samples have described an oncogenic role for miR-126. Notably, several studies have shown that miR-126 is upregulated in some malignancies due to high tissue specificity, such as gastric cancer, liver cancer, ovarian cancer, and acute myeloid leukaemia [19,20,74,75]. In addition, miR-126 acting as an oncogene, which was found to downregulate HOXA9/PLK, was often upregulated in myeloid leukaemia and associated with poor prognosis [22,76]. Moreover, higher expression of miR-126 was shown to be a poor prognostic factor in NSCLC and promote metastasis in prostate cancer [77,78]. Obviously, it is controversial that miR-126 expression can be used as a prognostic biomarker in different cancers. Hence, in order to evaluate the prognostic role of miR-126 expression in various cancers, we systematically reviewed the published studies and performed a meta-analysis for the first time.
In terms of this, a total of 4497 participants from 30 studies finally were included into the meta-analysis. This result showed that high expression of miR-126 was a significant marker for predicting better outcomes of various cancers (HR was 0.77, 0.64, and 0.70 for OS, DFS, and RFS/PFS/DSS, resp.). For OS, stratified analyses displayed that high expression of miR-126 was a better prognostic marker in HCC, Asians, tissue sample, qRT-PCR assay, multivariate analysis, and HRs reported. However, AML and circulation sample indicated the opposite result. For DFS, subgroup analyses revealed that high expression of miR-126 could predict a favorable DFS in NSCLC, Asian, Caucasian, multivariate, and univariate subgroups. Furthermore, we found that high expression of miR-126 significantly relates to a favorable RFS/PFS/DSS in CRC and Asian subgroup, but no statistical significance is shown in NSCLC, Caucasian, multivariate, and univariate analysis. Additionally, there was no obvious risk of publication bias in our meta-analysis. From the above results, we found that high expression of tissue miR-126 was a positive prognostic factor in cancer patients. But high circulating miR-126 levels predicted a significantly worse OS in patients with cancer. As we know, circulating samples are more convenient to collect and keep monitored, which can effectively evaluate prognosis during or after clinical therapy. Therefore, circulating miR-126 may be an efficacious method for dynamically monitoring the prognosis and therapeutic effects in cancer patients. In this study, only four studies investigated circulating samples, and more studies on these cancers are needed in the future.
Although the present meta-analysis revealed that the expression of miR-126 in cancer patients could be a valuable prognostic biomarker for patients, some limitations should be noticed. Firstly, there was significant heterogeneity existing in our meta-analysis, which was probably attributed to the differences in baseline demographic characters of population, characteristics of patients, the types of cancer, the samples of cancer, the disease stages, the cut-off criteria, the duration of follow-up, and so on. Secondly, several HRs were calculated based on the data extracted from the survival curve; some minor differences exist between the exact HRs and the extrapolated data. Thirdly, due to the lack of a unified cut-off value in miR-126 expression, cut-off values were not consistent among included studies. The different cut-off values may influence the availability of miR-126 as a prognostic biomarker in human cancer. Fourth, in subgroup analyses by sample type and subtype analyses, the number of studies was relatively small. More studies on these cancers are needed in the future. Finally, treatments may influence the expression of miR-126 in cancer samples; however, few researches referred to the treatment effect on HRs or miR-126 expression.

Conclusion
In sum, in this meta-analysis, we concluded that overexpression of miR-126 was effectively predictive of better prognosis in various carcinomas. Increased miR-126 level in cancerous tissues was associated with favorable OS, DFS, and PFS/RFS/DSS, while elevated circulating miR-126 was indicative of poor OS. However, our results should be regarded cautiously due to the limitations of the present analysis listed above. Further prospective multicenter studies with larger sample size are needed to focus on the relationship between miR-126 and cancer prognosis as well as to explore effective therapies.