Prognostic Value of Ezrin in Various Cancers: A Systematic Review and Updated Meta-analysis

More and more studies have investigated the effects of Ezrin expression level on the prognostic role in various tumors. However, the results remain controversial rather than conclusive. Here, we performed a systematic review and meta-analysis to evaluate the correlation of Ezrin expression with the prognosis in various tumors. the pooled hazard ratios (HR) with the corresponding 95% confidence intervals (95% CI) were calculated to evaluate the degree of the association. The overall results of fifty-five studies with 6675 patients showed that elevated Ezrin expression was associated with a worse prognosis in patients with cancers, with the pooled HRs of 1.86 (95% CI: 1.51–2.31, P < 0.001) for over survival (OS), 2.55 (95% CI: 2.14–3.05, P < 0.001) for disease-specific survival (DFS) and 2.02 (95% CI: 1.13–3.63, P = 0.018) for disease-specific survival (DSS)/metastasis-free survival (MFS) by the random, fixed and random effect model respectively. Similar results were also observed in the stratified analyses by tumor types, ethnicity background and sample source. This meta-analysis suggests that Ezrin may be a potential prognostic marker in cancer patients. High Ezrin is associated with a poor prognosis in a variety of solid tumors.

To explore the sources of heterogeneity, sub-group analysis for OS and DSS/MFS were conducted by the ethnicity, sample source and cancer types. The main results of this subgroup analysis for prognostic role of Ezrin in various tumors are shown in Table 2. In the ethnicity subgroup analyses, considerable heterogeneity was observed no matter the cancer patients were Asian or Caucasian for OS and DSS/MFS, the results showed that Ezrin over-expression reduced significantly the OS (HR = 2.21, 95% CI:1.72-2.83, P < 0.001) and DSS/MFS (HR = 4.18, 95%CI:1.60-10.95, P = 0.004) in Asian cancer patients, but not in Caucasian ones (HR = 1.41, 95%CI: 0.95-2.09, P = 0.092; HR = 1.40, 95%CI: 0.61-3.19, P = 0.426, respectively).
Publication bias and sensitivity analysis. Both Begg's funnel plot and the Egger's test were performed to evaluate the publication bias of the inclusion studies. As shown in Fig. 5a-c, the shape of the funnel plots revealed no obvious asymmetry. And the P values of Egger's test for OS, DFS and DSS/MFS were 0.389, 0.597 and 0.743, respectively, indicating that there was no significant publication bias in the meta-analysis. Meanwhile, the sensitivity analysis was performed to measure the effects of each individual study on the pooled HRs for the OS, DFS or DSS/MFS by omitting studies, respectively. The results demonstrated that no individual study significant influenced the overall HR, as shown in Supplementary Figure S2a, Figure S2b and Figure S2c. This suggested that the results of the present meta-analysis are credible.

Discussion
Ezrin, the most important member of the Ezrin/radixin/moesin (ERM) family, is mainly expressed in a variety of malignant tissues which originate from epithelial or non-epithelial cells 55 . Generally, Ezrin is mainly distributed in the cytoplasm with an inactive form, Once activated by threonine and tyrosine phosphorylation, Ezrin would transform into a special active form 56 . The basic biological function of Ezrin is to link transmembrane proteins to actin cytoskeleton 57,58 . In addition to acting as a cross-linker, Ezrin is involved in transmission of signals in response to extracellular cues 59,60 . The biological pathways associated with Ezrin include protein kinase C, Rho-kinase, NF-kB, PI3 kinase/Akt and so on 61 . Moreover, as a metastasis-related oncogene, Ezrin also participate in modulating multiple cellular processes 62 , including the formation of microvilli 63 , maintenance of cell shape 64 , cell-cell adhesion 65 , cell motility and invasion 66 . Hence, it seems that Ezrin might play an important role in the development of cancer. There is growing evidence that Ezrin expression level is associated with tumor progression and dissemination 67 . Numerous epidemiological studies have also assessed the correlation of high Ezrin expression and poor outcome in cancer patients so far, such as digestive system cancer [16][17][18][19][20][21][22][23][24][25] , osteosarcoma [26][27][28][29][30][31]79,80 , HNSCC 32-36 , gynecologic cancer 37-39 , hepatobiliary cancer 43 and so on. However, the results about the prognostic value of Ezrin expression in cancer patients remain inconsistent. Some studies reported that up-regulated Ezrin was a negative prognostic factor for survival for cancer patients  , However, other studies showed an opposite result 48,[50][51][52]78 . To resolve the conflicting issues, we performed a systematic review and meta-analysis on the association between Ezrin expression and prognostic value in cancer patients.
As the first qualitative analysis of Ezrin expression related to survival outcome of various tumors, Han et al. 68 retrieved 29 studies and found that over-expression of Ezrin might be associated with worse prognosis. However, the number of inclusion studies in the analysis was not relatively enough and at least 26 eligible studies were not included in the above meta-analysis, of which 8 studies about osteosarcomas were absolutely not included. Furthermore, the data reported by Han et al. 68 for the study by Jörgren et al. 69 were inconsistent with the data and the conclusion provided by Jörgren et al. 69 in their original article. The HR value reported by Han et al. 68 for OS is 1.89 (95% CI = 1.16-3.10), this suggested that high Ezrin expression was associated with worse prognosis in rectal cancer patients. But after carefully studying the data presented by Jörgren et al. 69 , we found Jörgren et al. 69 just provided HR value about LR (local recurrence), not about OS. Moreover, the conclusion by Jörgren et al. 69 showed that Ezrin expression had no impact on overall survival of patients with rectal cancer. Therefore, the conclusion by Han et al. 68 was still being debated and uncertain. In view of this, we performed this updated meta-analysis including 44 articles with 55 studies and elucidated that the high Ezrin expression was significantly associated with poor OS, DFS and DSS/MFS in cancer patients.
This meta-analysis was performed according to the guidelines and recommendations for improving the quality of reporting of medical research such as REMARK 70 and PRISMA 71 . Estimation of HR using multivariate proportional hazards model was used to evaluate the prognostic significance between ezrin expression and survival outcomes in each study, variables entered into the multivariate analysis mainly included Age, Gender, Tumor size, Tumor grade, TNM tumor stage, Lymph node metastasis, Ezrin expression. These positive factors contributed to the strengths of this meta-analysis.
The evidence included in the present meta-analysis indicated Ezrin expression as a poor prognostic marker in a variety of tumors. However, it should be noted that there are some limitations to the analyses presented here. First, because the number of prognostic studies dealing with each type of cancers was ≤ 5, the results of the particular carcinomas might be less powerful. Second, English articles were only recruited, and language bias might exist. Third, some HRs were calculated indirectly by the data extracted from the literature, however, these data were less reliable than direct data from the original literature. Fourth, different cutoffs used to assess high Ezrin level in the studies might also have contributed to the heterogeneity, because there is not a standard cutoff value of Ezrin level for increased survival risk. Fifth, significant heterogeneity existed in between studies, even though we calculated the pooled subgroup data with random-effects models. The heterogeneity in these studies could be attributed to the differences by different population characteristics or study designs. In addition, different sample types could also explained the heterogeneity, because tissue microarray (TMA) probably obtained more false-negative cases than the whole section. Finally, some inevitable publication bias might exist in the literature-based analysis, because more positive results tended to be published, thus potentially exaggerating the association between Ezrin expression and poor outcomes. Moreover, because all of the included studies were retrospective, which may have also introduced reporting bias. Therefore, our findings should be interpreted with caution.
In summary, our meta-analysis has demonstrated that the high Ezrin expression is significantly associated with poor survival in cancer patients. However, our results should be also considered cautiously for the above reasons. Further multicenter prospective studies and large clinical investigations should be conducted to validate the prognostic value of Ezrin in various tumors.   cancers was performed. The PubMed and EMBASE databases were retrieved with the following search terms or keywords:"Ezrin", "prognosis OR prognostic OR survival OR outcome" and "cancer OR tumor OR carcinoma OR neoplasm". Human studies were only restricted in this search. In addition, we also manually reviewed the references of relevant articles to obtain additional findings.   Inclusion and Exclusion Criteria. In this meta-analysis, the candidate studies were recruited according to the following criteria: (i) studied the patients who suffering from any type of cancers; (ii) evaluated Ezrin expression using Immunohistochemical method; (iii) assesed the correlation between Ezrin expression level and clinical outcome; and (iv) English articles. Articles were excluded based on any of the following criteria: (i) reviews, letters, comments, conference abstracts, or laboratory articles; (ii) articles not in English; (iii) absence of key information, such as HR, 95% CI, and P value, or useful data for calculation established by Parmar, Williamson, and Tierney [73][74][75] ; and (iv) overlapping studies. The most recent or complete studies were selected if the same patient cohort was utilized in different articles. Full manuscript was available after examining the abstract if any doubt of suitability remained as well.

Methods
Quality Assessment. According to a critical review checklist of the Dutch Cochrane Centre proposed by MOOSE, we strictly assessed the quality of all the studies included 72 : (i) a detailed description about study population and origin of country; (ii) a definite description of the study design; (iii) a definite type of carcinoma; (iv) a definite description of outcome assessment; (v) a definite measurement method of Ezrin and (vi) a definite cut-off of Ezrin. Otherwise, We would exclude the studies in order to ensure the quality of the meta-analysis.

Data Extraction and Conversion.
Two reviewers extracted the required information from all eligible studies independently. The extracted data included the following elements: the first author's name, publication year, country of origin, sample size, tumor type, Ezrin measurement method, cut-off value, follow-up duration, the HRs of Ezrin for OS, DFS or DSS/MFS, as well as their 95% CIs and P values. Multivariate Cox proportional hazards regression analysis was used in the present analysis. If the HR and its 95% CI were not available directly, they were calculated from the corresponding data or Kaplan-Meier curves provided in the articles using the method reported previously 75 .
Statistical analysis. All these HRs and the corresponding 95% CIs were calculated to combine the pooled data following Tierney's method 75 . A test of heterogeneity of combined HRs was performed using Cochran's Q test and Higgin'sI 2 statistics 76 . A P value < 0.05 and/or I 2 > 50% indicated significant heterogeneity, a random-effect model was used to calculate the pooled HR; otherwise, the fixed-effect model was used. Generally, pooled HR of > 1 was assumed to indicate a significant association with worse prognosis and was interpreted as statistically significant if the 95% CI for the pooled HR did not overlap one. Sensitivity analysis was carried out by removing each study at a time to evaluate the stability of the results. Publication bias was analyzed by performing funnel plots qualitatively, and estimated by Begg's and Egger's test quantitatively. Two sided P < 0.05 was considered statistically significant 77 . All analyses used in the meta-analysis were performed by SPSS version 13.0 and STATA version 12.0 (Stata Corp., College Station, TX, USA).