Prognostic Value of MET Gene Copy Number and Protein Expression in Patients with Surgically Resected Non-Small Cell Lung Cancer: A Meta-Analysis of Published Literatures

Background The prognostic value of the copy number (GCN) and protein expression of the mesenchymal-epithelial transition (MET) gene for survival of patients with non-small cell lung cancer (NSCLC) remains controversial. This study aims to comprehensively and quantitatively asses the suitability of MET GCN and protein expression to predict patients' survival. Methods PubMed, Embase, Web of Science and Google Scholar were searched for articles comparing overall survival in patients with high MET GCN or protein expression with those with low level. Pooled hazard ratio (HR) and 95% confidence intervals (CIs) were calculated using the random and the fixed-effects models. Subgroup and sensitivity analyses were also performed. Results Eighteen eligible studies enrolling 5,516 patients were identified. Pooled analyses revealed that high MET GCN or protein expression was associated with poor overall survival (OS) (GCN: HR = 1.90, 95% CI 1.35–2.68, p<0.001; protein expression: HR = 1.52, 95% CI 1.08–2.15, p = 0.017). In Asian populations (GCN: HR = 2.22, 95% CI 1.46–3.38, p<0.001; protein expression: HR = 1.89, 95% CI 1.34–2.68, p<0.001), but not in the non-Asian subset. For adenocarcinoma, high MET GCN or protein expression indicated decreased OS (GCN: HR = 1.49, 95% CI 1.05–2.10, p = 0.025; protein expression: HR = 1.69, 95% CI 1.31–2.19, p<0.001). Results were similar for multivariate analysis (GCN: HR = 1.61, 95% CI 1.15–2.25, p = 0.005; protein expression: HR = 2.18, 95% CI 1.60–2.97, p<0.001). The results of the sensitivity analysis were not materially altered and did not draw different conclusions. Conclusions Increased MET GCN or protein expression was significantly associated with poorer survival in patients with surgically resected NSCLC; this information could potentially further stratify patients in clinical treatment.


Introduction
Lung cancer continues to be the most common and deadly malignant cancers worldwide [1]. Although important progress in the management of this disease has been observed over the last decade, non-small cell lung cancer (NSCLC) remains a lethal disease, and improving poor prognosis (5-year survival of approximately 15%) remains a challenge [2]. Multiple independent prognostic factors, such as performance status, disease stage, age, sex and amount of weight lost, have previously been identified for predicting survival [3]. Although the use of these factors has been widely accepted, the prognosis of NSCLC is not sufficiently predictable, thus additional prognostic markers are required for more accurate estimation.
The MET gene, located at 7q21-q31, is a potential prognostic genetic marker, which encodes a receptor tyrosine kinase for the HGF/scatter factor (SF) [4]. Met-receptor tyrosine kinase is activated through phosphorylation and the cognate ligand HGF, leading to the activation of a number of downstream pathways, such as the phosphoinositide-3-kinase (PI3K), Ras-Rac/Rho, Ras mitogen-activated protein kinase (MAPK) and phospholipase C-c signaling pathways, in several types of human cancers, including NSCLC [5]. HGF/Met signaling promotes biological activities, resulting in tumor growth, angiogenesis and the development of invasive phenotypes, making this receptor an attractive target for the potential anti-cancer treatment of NSCLC [6][7][8]. Alterations in the MET gene, including amplification, overexpression and mutations, have been described in a number of solid tumors, including breast and esophageal cancers [9,10]. The rate of MET amplification in NSCLC remains controversial, ranging from 3% to 10%, depending on the detection method and cut-off criteria [11,12]. Most studies have indicated a negative prognostic impact of high MET GCN on NSCLC survival [11][12][13][14][15][16][17], however, other studies have not confirmed this finding [18][19][20][21]. MET overexpression in NSCLC is variable, ranging from 5% to 75%. Several studies have shown that the overexpression of MET is associated with poor outcome [13,19,[21][22][23][24][25][26]. However, the prognostic relevance of MET overexpression remains unclear.
With the aim to gain a better insight into the prognostic value of the copy number or protein expression of the MET gene for survival of patients with non-small cell lung cancer, we conducted the first comprehensive meta-analysis of published literature on this topic. Studies meeting the following inclusion criteria were considered for this meta-analysis: (I) Clinical trials and prospective or retrospective cohort studies investigating the correlation of the MET GCN and protein expression status with the OS of NSCLC patients; (II) Measurement methods, including fluorescent in situ hybridization (FISH), reverse transcription-polymerase chain reaction (RT-PCR), and immunohistochemistry (IHC); and (III) Findings providing sufficient information for the estimation of hazard ratios and 95% confidence intervals. Only studies published in peer-reviewed journals were included, data from letters and meetings abstracts were not eligible. Two researchers (B.P.G and H.C) independently screened and determined the relevant studies. Any discrepancies were settled through discussion until a consensus was reached.

Data extraction
Two reviewers independently (B.P.G and H.C) extracted the relevant data from each study and subsequently assessed the data to estimate reliability. The following information was obtained from the MET GCN studies: the first author, year of publication, country of origin, inclusion period, number of patients (Male/ Female), age at time of diagnosis (mean, median, range), tumor stage, method of MET GCN detection, cutoff value of high MET GCN, histology, number of patients of high MET GCN, treatment, time of follow-up (median, mean, range), and OS data. The information obtained from each MET protein expression study included the first author, year of publication, country of origin, inclusion period, number of patients (Male/ Female), age at time of diagnosis (mean, median, range), tumor stage, method of MET protein expression detection, specimen, cutoff, antibodies, histology, number of patients of high MET protein expression, treatment, time of follow-up (median, mean, and range), and OS data.

Quality assessment
Two authors (B.P.G and X.H.T) independently assessed the quality of the selected studies using the Newcastle-Ottawa Scale for cohort studies (NOS) [27]. This tool comprises three quality parameters: selection, comparability, and outcome assessment. ''Stars'' were awarded to demonstrate ''high'' quality. The stars were subsequently added and used to compare the overall quality in a quantitative manner. A consensus reviewer (H.C) resolved any observed discrepancies.

Statistical analysis
The primary results were stratified according to MET GCN (high vs. low) and protein expression (high vs. low). The HRs and 95% CIs were combined to obtain the effective value. When the HR was not reported in an article, this parameter was calculated using the methods of Parmar et al [28].
A heterogeneity test based on I 2 and Q statistics was performed. The heterogeneity of individual HRs was calculated using X 2 tests according to the method of Peto [29]. Significant heterogeneity was determined at a p value less than 0.10. I 2 was used to quantify inconsistencies, where a value of 0% indicates no observed heterogeneity, a value less than 25% denotes low heterogeneity, a value from 25.1-50% indicates moderate heterogeneity, and a value greater than 50% indicates substantial heterogeneity [30]. When heterogeneity was observed between primary studies, the random effects model was used. When no heterogeneity was observed, the fixed effects model was used for analysis [31]. HR.1 implies worse survival for the group with high MET GCN or protein expression. The impact of MET on survival was considered statistically significant when the 95% CI did not overlap with 1. Subgroup analyses were performed using different methods to detect the MET GCN and protein expression, conduct univariate and multivariate analyses, and assess the histological subtypes and ethnic source.
Sensitivity analyses were performed to assess the stability of the results. Egger's test [32] was used to detect potential publication bias. Statistical significance was considered for a p-value of less than 0.05 for summary HR and publication biases. All calculations were performed using STATA version 11.0 (Stata Corporation, College Station, TX, USA).

Qualitative assessment
The study quality was assessed using the Newcastle-Ottawa quality assessment scale, generating scores ranging from 4 to 9 (with a mean of 5.85), with a higher value indicating better methodology. The results of quality assessment are shown in supplementary Table S1.

Sensitivity analyses and publication bias test
The sensitivity analysis indicated that omitting any single study did not influence the pooled HRs. For MET GCN, A more formal evaluation using Egger's test showed no evidence of significant publication bias (p = 0.352 for univariate analysis and p = 0.063 for multivariate analysis). For the MET protein expression, there was no evidence for significant publication bias (Egger's test: p = 0.076 for univariate analysis and p = 0.116 for multivariate analysis).

Discussion
MET has recently received attention as a molecular target for the treatment of NSCLC. Understanding the mechanisms underlying anti-MET therapy requires the correct evaluation of the impact of MET GCN and protein expression on patient survival.
The summary statistics obtained from 18 published studies, including 5,516 patients with NSCLC, showed that high MET GCN or protein expression significantly predicted the poor OS of NSCLC patients (gene copy: HR 1.90, 95% CI 1.35-2.68; protein expression: HR 1.52, 95% CI 1.08-2.15). The subgroup analysis revealed that high MET GCN or protein expression was also significantly associated with poor prognosis in Asian countries (gene copy: HR 2.22, 95% CI 1.46-3.38; protein expression: HR  1.89, 95% CI 1.34-2.68), but the same tendency was not observed in the non-Asian subset (gene copy: HR 1.21, 95% CI 0.55-2.67; protein expression: HR 1.28, 95% CI 0.48-3.43). The present study was performed using univariate analysis, followed by further multivariate analysis. The results of the meta-analysis showed that high MET GCN or protein expression in NSCLC patients was associated with poor OS (univariate analysis). This effect was also significant according to multivariate analysis, showing that the MET GCN or protein expression might be an independent prognostic factor for OS in NSCLC.
The methods used to detect the MET GCN impacted the significance of these results. The combined HRs of 8 FISH (included SISH and BISH) and 3 RT-PCR studies were 1.66 (95% CI: 1.28-2.16) and 2.95 (95% CI: 0.80-10.91), respectively. We observed that FISH, instead of RT-PCR, was the most widely used technology for determining the gene copy number. In clinical practice, although real-time PCR is a simple and quick method, the results do not directly reflect cancer cells because DNA is typically isolated from whole tissue specimens that also contain normal epithelial cells, inflammatory cells, and fibroblasts. FISH is generally accepted as a better technique than RT-PCR for evaluating gene copy number because FISH can be applied to formalin-fixed paraffin-embedded tumor tissues archived for routine pathological diagnosis, thus facilitating the exclusive estimation of tumor cells. Therefore, FISH is the most widely used technique in clinical practice for the detection of gene amplification to determine therapeutic strategies, such as HER2 FISH in breast cancer. The results obtained in the present study showed that increased MET GCN, evaluated using FISH, was a predictor of worse survival in NSCLC. Due to the small number of primary studies using RT-PCR for analysis, the detection of potentially important differences was limited. Moreover, IHC was the method typically used to detect MET protein expression. IHC is the standard method for the evaluation of proteins (e.g., HER2  and EGFR), and there was consistency in the evaluation process among studies. The results of the present meta-analysis showed that MET overexpression was associated with worse survival. Moreover, the results of the present study demonstrated that high MET GCN or protein expression was an independent negative prognostic factor in NSCLC. However, the prognostic significance of MET GCN according to the histology of NSCLC remains unclear. Go et al [12] reported that SCC patients with MET amplification showed markedly shorter OS than those without MET amplification. In contrast to these results, the systematic review showed that high MET GCN or protein expression is a worse marker of death risk in lung adenocarcinoma than in squamous carcinoma. These results indicated that MET amplification might be involved in the oncogenesis of SCC and ADC. The differences in the two contrasting results were influenced by two SCC studies reporting a correlation between the MET GCN and survival, and these data were not sufficient to determine the prognostic value of MET expression in SCC.
Park et al [19]. demonstrated that MET FISH-positive and MET IHC-positive patients had significantly shorter survival. The results obtained in the present study also provide similar evidence that MET is a negative prognostic factor, further supporting anti-MET strategies, irrespective of MET CGN or MET overexpression. Thus, when patients were divided according to EGFR FISH results, MET positivity had prognostic implications only among EGFR FISH-negative patients. This finding has been consistently reported in recent studies [11,12], suggesting that anti-MET drugs might be beneficial for EGFR FISH-negative NSCLC patients who are not initially selected for EGFR TKI treatment.
We observed a considerable degree of interstudy heterogeneity. Differences in the detection methods, types and numbers of target genes or antigens, sampling sites and times, and demographic or clinicopathologic data from the included patients, should be considered as potential sources of heterogeneity. In this study, significant heterogeneity was observed among the included studies. Although we used random-and fixed-effects models for pooling  data, the source of heterogeneity remained unknown. Moreover, the sensitivity analysis did not clarify the source of the heterogeneity observed in this study. The studies by Sun et al [13] and Dziadziuszko et al [18] primarily accounted for the heterogeneity observed in the MET GCN. Although Sun et al. used RT-PCR, it was not possible to address this technical issue, as these studies used the same primers and other PCR conditions. Dziadziuszko et al [18] used silver in situ hybridization (SISH). Silver in situ hybridization (SISH) is a new technology for gene copy assessment, with some clinical advantages compared with FISH. First, the samples are analyzed using conventional light microscopy with preserved cell morphology based on automation. The new technology facilitates the evaluation of slides through light microscopy for the simultaneous visualization of amplified signals and cell morphology, overcoming the disadvantage of FISH where the fluorescent signals gradually fade over time. This difference might explain the observed heterogeneity.
Factors associated with immunostaining can also contribute to the observed heterogeneity. Onisuka et al [21] and Liu et al [26] used the same antibodies, but differences in the staining techniques and evaluation criteria for MET positivity might contribute to heterogeneity between studies. The exclusion of this study from the analysis only partially reduced the heterogeneity, potentially reflecting immunohistochemistry techniques (various definitions of threshold positivity, use of the mAb at different concentrations and dissimilar staining protocols) or patient characteristics (type of patients, disease characteristics). These factors might not only contribute to the observed statistical heterogeneity but also the clinical heterogeneity. Clinical heterogeneity might result from the different patients (with different age, tumor size, clinical stage, ethnicity, physical condition, etc.), diverse treatment types, various treatment protocols, different dosages and drug types, etc. Moreover, differences in primary antibodies, IHC staining protocols, evaluation standards, and cut-off values for high MET expression might also contribute to heterogeneity. Thus, additional multicenter studies using standardized methods are encouraged.
Some limitations of this meta-analysis need to be discussed. First, our meta-analysis is based on data from trials whose results have been published, and we did not obtain individual patient data. Use of individual patient data may further enhance the accuracy and reduce the uncertainty of the estimates. Second, significant heterogeneity was observed among the included studies. Factors associated with variability in definitions of end point, measurements, and experimental design may contribute to the heterogeneity. Therefore, validation of the prognostic power of MET GCN or protein expression should be conducted through large multicenter prospective studies based on homogeneous populations. Third, the number of studies concerning MET and the effectiveness of therapy (such as chemotherapy or EGFR TKI treatment) was too small to perform a pooled analysis. In the present study, due to the incompleteness of clinicopathological parameters, we did not perform subgroup analyses between MET GCN and clinicopathological parameters or between protein expression and clinicopathological parameters. Fourth, negative studies are less frequently published or published with less detailed results, making these studies less assessable, potentially leading to some bias.
Despite these limitations, this meta-analysis had some advantages. First, the results obtained from the random-effects model were similar to those obtained from the fixed-effects model, indicating that the statistical results were robust. Second, the results of the sensitivity analysis were not materially altered and did not draw different conclusions, indicating that the initial results were strong. Third, Egger's test did not detect publication bias, indicating that the obtained results were not biased. Moreover, the study quality scores, assessed using the Newcastle-Ottawa quality assessment scale, were .5, suggesting that the results of the present meta-analysis were convincing.
In conclusion, this meta-analysis indicated that increased MET GCN and protein expression was significantly associated with poorer survival in patients with NSCLC; this information could potentially further stratify patients in clinical treatment.

Supporting Information
Table S1 Assessment of Newcastle-Ottawa Scale methodological quality of cohort studies.