Original Article Prognostic value of Ezrin expression in non-small cell lung cancer: a systematic review and meta-analysis

Original Article

Prognostic value of Ezrin expression in non-small cell

lung cancer: a systematic review and meta-analysis

Shuang-Jiang Li1*_{, Jian Huang}1*_{, Wen-Biao Zhang}2_{, Jun Fan}1_{, Guo-Wei Che}1

1_{Department of Thoracic Surgery, West China Hospital, Sichuan University, Chengdu, Sichuan, China;} 2_Lung Can-cer Center, West China Hospital, Sichuan University, Chengdu, Sichuan, China. *_{Equal contributors.}

Received March 12, 2016; Accepted May 26, 2016; Epub July 15, 2016; Published July 30, 2016

Abstract: Ezrin is a cytoskeletal organizer and cross-linker between cytoskeleton and plasma membrane, which is involved in the malignancy progress. However, its prognostic roles in non-small cell lung cancer (NSCLC) still remain controversial. We conducted this systematic meta-analysis to evaluate the association between Ezrin expression and the prognosis of NSCLC. We searched PubMed, EMBASE, the Web of Science and China National Knowledge In-frastructure to identify the full-text literatures that met our eligibility criteria. Hazard ratio (HR) with 95% confidence interval (CI) served as the summarized statistics. Q-test and I2 _{statistic determined the level of heterogeneity and}

further sensitivity analysis was also performed. Potential publication bias was detected by Begg’s test and Egger’s test. Finally, ten eligible literatures were included into this meta-analysis. The overall pooled analysis indicated that Ezrin expression was significantly associated with the worse overall survival (HR: 2.485; 95% CI: 1.995-3.095; P<0.001; I2_{=12.9%, P=0.330) and disease-free survival (HR: 2.081; 95% CI: 1.329-3.260; P=0.001; I}2_=0.0%,

P=0.543) in patients with NSCLC. The relationships between Ezrin expression and poor overall survival still remain statistically prominent in the subgroups classified by statistical analysis, Ezrin-staining sites, cut-off definitions and follow-up periods. No significant heterogeneity or publication bias was observed within the present meta-analysis. In summary, Ezrin can serve as an independent biomarker for predicting the poor prognosis of NSCLC.

Keywords: Ezrin, non-small cell lung cancer (NSCLC), prognosis, meta-analysis

Introduction

Lung cancer is the worldwide leading cause of cancer-related deaths, and non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer cases [1, 2]. Authoritative estimations reveal that about 1.4 million deaths around the world in 2008 are caused by NSCLC [3]. Meanwhile, the number of newly-diagnosed NSCLC cases is rapidly increasing during the last decade, especially in heavy smoking peoples [3, 4]. The overall 5-year sur-vival rate of NSCLC is approximately 15%, indi-cating its present poor prognosis [3]. Advanced surgical techniques, chemo-radiation and peri-operative managements have remarkably improved the feasibility and safety of multidis-ciplinary treatments for NSCLC but hardly ben-efit the prognosis of NSCLC [5, 6]. However, the 5-year survival rate in patients with early stage NSCLC is higher than 80% [7, 8]. A consensus

has been reached that advanced stages, early metastasis and the poor response to treat-ments can result in the poor prognosis of NSCLC. On the basis of the current diagnostic and therapeutic regimens, it is urgent to identi-fy novel biomarkers efficiently promising the prognosis of NSCLC.

Ezrin is firstly identified as a necessary regula-tor of osteosarcoma metastasis [12]. Its prog-nostic significance in osteosarcoma is further clarified in the subsequent clinical reports [13, 14]. A recent systematic review has concluded that Ezrin expression may be significantly asso-ciated with the worse prognosis of many solid tumors [15]. However, the integrated results for NSCLC are not comprehensively described in this study. Some controversies in the current investigations about the prognostic value of Ezrin in NSCLC are still not well-resolved until now. Therefore, we conducted the current meta-analysis to systematically evaluate the association between Ezrin expression and the prognosis of NSCLC.

Materials and methods Protocol

A meta-analysis or systematic review does not require patients’ consent and ethical approval. We performed this meta-analysis in strict accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRI- SMA) statement [16]. The additional PRISMA 2009 checklist is given in the Supplementary Data 1.

Search strategies

A comprehensive literature retrieval for our me- ta-analysis was performed between Novem- ber 10, 2015 and November 14, 2015. Three universal English electronic databases, includ-ing PubMed, EMBASE (via the Ovid interface) and the Web of Science (via the campus net-work of Sichuan University), and one Chinese native electronic database, China National Knowledge Infrastructure (CNKI), were search- ed by two of our researchers for identifications of eligible literatures published up to November 10, 2015. We combined five key words, includ-ing ‘Ezrin’, ‘lung cancer’, ‘lung carcinoma’, ‘lung neoplasm’ and ‘lung tumor’, with two Boolean operators, including ‘AND’ and ‘OR’, to formu-late the search strings in each database (see Supplementary Data 2). Besides, a manual search for the reference list in relevant articles was also performed to identify any possible included study with no duplication.

Inclusion and exclusion criteria

The following inclusion and exclusion criteria were applied to confirm the studies included

Inclusion criteria: (I) the target disease is NSCLC, small cell lung cancer (SCLC) is not accepted; (II) Ezrin expression is independently analyzed in NSCLC cases rather than collabo-rating with other biomarkers; (III) the statistical data correlated with the prognostic significance of Ezrin expression, including hazard ratio (HR), relative risk (RR) and odds ratio (OR) with the corresponding 95% confidence interval (95% CI), are derived from either univariate analysis or multivariate analysis; (IV) if no statistic is reported, the survival data, P value from the log-rank test and Kaplan-Meier survival curves (K-M curves) are also considered for eligibility; (V) overall survival (OS) and disease-free sur-vival (DFS) serve as the summarized endpoints.

Exclusion criteria: (I) the following literature styles are directly excluded: reviews, letters, laboratorial experiments and conference abstracts; (II) the positive expression of Ezrin is not validly reported; (III) the continuous vari-ables are not evaluated in this meta-analysis.

Quality assessment

Newcastle-Ottawa Scale (NOS) was an appro-priate assessment tool to determine the quality of original non-randomized studies [17]. A semi-quantitative evaluation covering three perspec-tives, including selection, comparability and exposure, was applied to grade all of the includ-ed studies. According to the “star system” with a maximum of 9 stars, we regarded 8-9 stars as a good quality; 6-7 stars as a fair quality, and lower than 6 stars as a poor quality.

Data extraction

We designed a Microsoft Office Excel spread-sheet to collect the following information: (I) publication data including authors, languages, publication years and nations; (II) experimental data including study design, study period, experimental materials, detecting methods and sites, cut-off values, endpoints and follow-ups (III) demographic data including enrolled sam-ples, ages, clinical stages, histological sub-types and the number of patients with positive and negative expression of Ezrin; (IV) statistical data including the extraction of statistics and the methods of statistical analysis.

Statistical analysis

tic because HR was generally regarded as the only statistic compatible for both censoring and time-to-events [18]. It was our first priority to incorporate the HR with 95% CI derived from multivariate analysis into the current meta-analysis. However, if no multivariate HR statis-tic was reported, we could extrapolate it from the survival events with log-rank P value and

the K-M curves published in original articles according to a practical method described by Tierney and colleagues [19]. The relevant for-mulas are listed as follows:

/2

o E

Analyzed research Analyzed control

Total observed events Analyzed research Analyzed control

Z score for P value

- = # ₊ #

#

^ ^

h h

Table 1. Basic characteristics of the included studies

Authors (year) Nation Language _designStudy Study period NOS No. samples Age, years Histology Stages Total PE NE SCC AD Others

Chen et al (2012) [25] China Chinese ROS 2000-2004 8 150 92 58 47.5 (26-77) 55 95 ––– I-III Gao et al (2015) [26] China Chinese ROS 2008-2010 7 40 22 18 NA 11 29 ––– Ib Jin et al (2014) [27] China English ROS 2004-2008 7 150 94 56 60.2 (36-78) 82 68 ––– !-IV Koriyama et al (2015) [28] Japan English ROS 1998-2012 7 87 54 33 64.0 (41-78) ––– 87 ––– I-IV Lee et al (2012) [29] Korea English ROS 1996-2009 9 112 33 79 63.0 (42-82) 59 38 15 I-IV Li et al (2007) [30] China Chinese ROS 1998-2001 8 287 166 121 59.5 (29-82) 124 129 34 I-IV Lin et al (2013) [31] China Chinese ROS 2004-2007 8 96 40 56 58.8 (29-77) 96 ––– ––– I-III Lin et al (2014) [32] China Chinese ROS 2004-2008 8 92 41 51 64.4 (22-72) ––– 92 ––– I-III Zhang et al (2008) [33] China Chinese ROS 2001-2005 7 72 44 28 59.3 (40-75) 34 29 9 I-IV Zhang et al (2012) [34] China English ROS 1998-2006 9 89 40 49 62.9 (46-81) 40 49 ––– I-III Authors (year) Materials Detection Cut-off value Positive sites HR extraction Analysis Endpoints Follow-up Chen et al (2012) [25] Paraffin-embedded tissue IHC 5% staining Cytoplasm Reported Multivariate OS 60 months Gao et al (2015) [26] Paraffin-embedded tissue IHC 25% staining Cytoplasm DDE Univariate DFS 36 months Jin et al (2014) [27] Paraffin-embedded tissue IHC 25% staining Cytoplasm DDE Univariate OS 96 months Koriyama et al (2015) [28] Paraffin-embedded tissue IHC 50% staining Cytoplasm & membrane DDE Univariate DFS 12 months Lee et al (2012) [29] Paraffin-embedded tissue IHC 50% staining Cytoplasm & membrane Reported Multivariate OS 153 months Li et al (2007) [30] Paraffin-embedded tissue IHC 50% staining Cytoplasm DDE Univariate OS 108 months Lin et al (2013) [31] Paraffin-embedded tissue IHC 0% staining Cytoplasm Reported Multivariate OS 99 months Lin et al (2014) [32] Paraffin-embedded tissue IHC 0% staining Cytoplasm Reported Multivariate OS 101 months Zhang et al (2008) [33] Paraffin-embedded tissue IHC 50% staining Cytoplasm DDE Univariate OS 72 months Zhang et al (2012) [34] Paraffin-embedded tissue IHC 10% staining Cytoplasm & membrane Reported Multivariate OS, DFS 120 months

l

V

Analyzed research Analyzed control

Total observed events Analyzed research Analyzed contro 2

=

+

# #

^ h

HR Exp=

`

j

Where O-E is the log rank Observed minus Expected events and V is the log rank Variance [19]. We could also extract the survival data from the K-M curves by Engauge Digitizer 4.1 (http://sourceforge.net) to measure the validity of estimated HR.

Meanwhile, if multivariate RR with 95% CI was reported in the original articles, we could imme-diately incorporate it into the meta-analysis [20]. If OR was reported from multivariate anal-ysis, we could transform it into RR according to the following formula:

1

RR

P P oR

oR

=

- + #

^

h

^

h

6

@

Where the P value was the incidence of the out-come of interest in the non-exposed group [21]. Finally, we could regard positive expression of Ezrin as an independent indicator predicting the poor prognosis of NSCLC when the pooled HR with 95% CI was appeared more than 1. The level of heterogeneity within this meta-analysis was estimated by Q-test and I2

statis-tic. On the one hand, fine heterogeneity was defined by I2_{<50% and P>0.1, determining a}

standard fixed-effect model test (Mantel-Haenszel method) during the integration of HRs. Otherwise, a random-effect model test (DerSimonian and Laird method) would be used when significant heterogeneity was sug-gested by I2_{≥50% or P≤0.1 [22]. Then,}

sensitiv-ity analysis was performed to further evaluate the stability of the summarized outcomes. We would exclude the study which increased the level of heterogeneity and repeat a meta-analy-sis of the remaining included studies for adjust-ments. Finally, the robustness of our meta-analysis would be confirmed if no substantial

variation was observed between the adjusted pooled HR and primary pooled HR [23].

Both Begg’s test and Egger’s test were applied to detect the publication bias in this meta-anal-ysis. Its presence was suggested by the visual symmetry of funnel plot conducted from Begg’s test, where log HRs were plotted against their corresponding standard errors (SEs) [24]. Besides, its significance was also revealed by Egger’s p value <0.05.

In addition, all of the above statistical analyses were accomplished by STATA 12.0 (STATA Corporation, College Station, TX).

Results

The selection of included studies

As shown in Figure 1, a total of 535 citations were identified by searching through the cho-sen databases, including 87 items in PubMed, 69 items in EMBASE, 136 items in the Web of Science and 243 items in CNKI. After excluding the duplicates, 371 items of publications entered into the initial filtration based on screening their titles and abstracts. There were 208 articles immediately excluded after initial filtration because of their inappropriate article styles, including 86 reviews, 111 laboratorial experiments and 11 conference abstracts. Then, we read through the full-text of the remaining 163 papers and further removed 153 of them, including 97 studies not correlat-ed with lung cancer, 35 studies not correlatcorrelat-ed with Ezrin and 21 studies only reported the clinicopathological characteristics of NSCLC. Finally, 10 original articles were considered eli-gible for quantitative synthesis and included into our meta-analysis [25-34]. All of the above procedures are summarized in a PRISMA flow diagram (Figure 1).

The quality level of included studies

Two researchers were assigned to grade all of the included studies. Their mean NOS score

Table 2. Overall analysis for the prognostic value of Ezrin expression in NSCLC

Endpoints N No. samples I2_{, P Model HR with 95% CI P value} Publication bias _Conclusion

Total PE NE Begg (P) Egger (P)

OS 8 1048 550 498 I2_{=12.9%, P=0.330 Fixed 2.485 (1.995-3.095) <0.001 0.386 0.238 Significant}

DFS 3 216 116 100 I2_{=0.0%, P=0.543 Fixed 2.081 (1.329-3.260) 0.001 1.0 0.868 Significant}

was calculated as 7.8 (ranged 7 from to 9), indi-cating a generally good quality level. The com-plete details of estimations are tabulated as Supplementary Data 3.

The basic characteristics of included studies

Baseline characteristics for ten eligible articles are summarized in Table 1. Because Zhang and colleagues [34] performed an analysis on both OS and DFS in 89 NSCLC patients, these arti-cles actually reported 11 included studies, including 8 studies [25, 27, 29-34] investigat-ing the association between Ezrin expression and OS of NSCLC and 3 studies [26, 28, 34] investigating the association between Ezrin expression and DFS of NSCLC (Table 1). All of the included studies belong to retrospective observational studies, and were published

ages and clinical stages, are also summarized in Table 1.

The statistical characteristics of included stud-ies

A multivariate analysis using logistic regression or Cox proportional hazards model was con-ducted in five included studies [25, 29, 31, 32, 34] and the others [26-28, 30, 33] only pub-lished the survival events or K-M curves. The multivariate statistics indicating the prognostic roles of Ezrin expression in NSCLC included HR with 95% CI reported from 3 studies [29, 31, 32] and RR with 95% CI reported from 2 stud-ies [25, 34]. The other HR outcomes [26-28, 30, 33] were extrapolated from the survival data with log-rank P value and K-M curves, which were based on univariate analysis (Table 1).

Figure 2. Overall analysis for the association between Ezrin expression and the (A) OS and (B) DFS in patients with NSCLC. CI, confidence interval; DFS, disease-free survival; HR, hazard ratio; NSCLC, non-small cell lung cancer; OS, overall survival.

Overall analysis

To explore the prognostic significance of Ezrin expression for OS in NSCLC, a pooled analysis of eight included studies [25, 27, 29-34] dem-onstrated that Ezrin expression was significant-ly associated with the worse OS of NSCLC patients (HR: 2.485; 95% CI: 1.995-3.095; P<0.001) (Table 2 and Figure 2A), with a low heterogeneity (I2_{=12.9%, P=0.330).}

To assess the prognostic significance of Ezrin expression for DFS in NSCLC, a pooled analysis based on three included studies [26, 28, 34] also indicated that Ezrin expression was signifi-cantly associated with the worse DFS of NSCLC (HR: 2.081; 95% CI: 1.329-3.260; P=0.001) (Table 2 and Figure 2B), without any heteroge-neity (I2_{=0.0%, P=0.543).}

Subgroup analysis

To further evaluate the prognostic roles of Ezrin in detail, we classified the survival data of OS into some subgroups according to the statisti-cal analysis, Ezrin-staining sites, cut-off defini-tions and length for follow-ups. The complete results of each subgroup are listed in Table 3. After quantitative synthesis, we found that the relationships between Ezrin expression and the poor OS of NSCLC remained statistically promi-nent in all of the subgroups, including the meth-ods of statistical analysis which were stratified by multivariate analysis (HR: 2.478; 95% CI: 1.908-3.217; P<0.001; I2_{=0.0%, P=0.567) and}

univariate analysis (HR: 2.862; 95% CI: 1.360-6.022; P=0.006; I2_{=60.7%, P=0.078) (Figure}

3A); Ezrin-staining sites which were stratified by cytoplasm staining (HR: 2.728; 95% CI: 2.117-3.516; P<0.001; I2_{=15.9%, P=0.312) and}

cyto-plasm-membrane staining (HR: 1.878; 95% CI: 1.211-2.912; P=0.005; I2_{=0.0%, P=0.952) (Fi-}

gure 3B); cut-off definitions which were strati-fied by <50% positive-staining (HR: 2.409; 95% CI: 1.862-3.116; P<0.001; I2_{=0.0%, P=0.431)}

and 50% positive-staining (HR: 2.935; 95% CI: 1.499-5.744; P=0.002; I2_{=50.2%, P=0.135)}

(Figure 3C); follow-up periods which were strati-fied by ≥100 months follow-up (HR: 2.332; 95% CI: 1.727-3.150; P<0.001; I2_{=0.0%, P=}

0.515) and <100 months follow-up (HR: 2.672; 95% CI: 1.937-3.685; P<0.001; I2_{=44.2%, P=}

0.146) (Figure 3D).

Sensitivity analysis

The forest plots derived from sensitivity analy-sis for assessing the stability of OS outcomes and DFS outcomes were shown as Figure 4A

and 4B. Finally, none of the individual studies was out of the estimated ranges by visually inspecting these plots. No substantial variation was revealed between the adjusted summa-rized outcomes and primary summasumma-rized out-comes, and the further adjustments were no longer necessary. Therefore, the strong robust-ness of our meta-analysis was confirmed.

Publication bias

No significant evidence for publication bias was detected by either Begg’s test (P=0.386) or

Table 3. Subgroup analysis for the association between Ezrin expression and the OS in patients with NSCLC

Subgroups of OS N No. samples I2_{, P Model HR with 95% CI P value Conclusion}

Total PE NE Statistical analysis

Multivariate analysis 5 539 246 293 I2_{=0.0%, P=0.567 Fixed 2.478 (1.908-3.217) <0.001 Significant}

Univariate analysis 3 509 304 205 I2_{=60.7%, P=0.078 Random 2.862 (1.360-6.022) 0.006 Significant}

Ezrin-staining sites

Cytoplasm 6 847 477 370 I2_{=15.9%, P=0.312 Fixed 2.728 (2.117-3.516) <0.001 Significant}

Cytoplasm & membrane 2 201 73 128 I2_{=0.0%, P=0.952 Fixed 1.878 (1.211-2.912) 0.005 Significant}

Cut-off definitions

<50% staining 5 577 307 270 I2_{=0.0%, P=0.431 Fixed 2.409 (1.862-3.116) <0.001 Significant}

50% staining 3 471 243 228 I2_{=50.2%, P=0.135 Random 2.935 (1.499-5.744) 0.002 Significant}

Follow-up periods

≥100 months 4 580 280 300 I2_{=0.0%, P=0.515 Fixed 2.332 (1.727-3.150) <0.001 Significant}

<100 months 4 468 270 198 I2_{=44.2%, P=0.146 Fixed 2.672 (1.937-3.685) <0.001 Significant}

Egger’s test (P=0.238) among the association between Ezrin expression and poor OS of NSCLC (Table 2 and Figure 5A, 5B). Moreover, no significant publication bias was detected by either Begg’s test (P=1.0) or Egger’s test (P= 0.868) for assessing the association between Ezrin expression and poor DFS of NSCLC (Table 2 and Figure 5C, 5D). Because the limited avail-ability of included studies might result in a poor efficacy of both Begg’s test and Egger’s test, we gave up the further assessments for publi-cation bias in each subgroup of OS.

Discussion

Ezrin is encoded by the VILLIN2 gene which is differentially expressed between metastatic

other malignances, including NSCLC. Limited availability of sample size may be the main rea-son causing some controversies on the rela-tionships between Ezrin expression and the prognosis of NSCLC [27]. Therefore, it is in urgent need to timely summarize the currently available evidence to clarify the actual prognos-tic roles of Ezrin in NSCLC.

Meta-analysis is a well-established statistical method quantitatively integrating the appropri-ate data from homogeneous studies to formu-late a general conclusion [38-47]. To the best of our knowledge, this is the first meta-analysis to systematically evaluate the prognostic value of Ezrin expression in NSCLC. By applying

evi-Figure 4. Sensitivity analysis of the association between Ezrin expression and the (A) OS and (B) DFS in patients with NSCLC. CI, confidence interval; DFS, disease-free survival; NSCLC, non-small cell lung cancer; OS, overall survival.

and non-metastatic cell lines in a murine model of osteosarco-ma [35]. As a member of ERM family, Ezrin has a highly con-served amino-terminal half either in mouse or human. It targets the plasma membrane when it is transfected into the cultured fibroblasts. In addi-tion, the carboxy-terminal half of Ezrin binds to the Actin-filament bundles of cytoskele-ton [36]. On the basis of such structure, Ezrin is involved in inducing the direct interaction between cell and its microenvi-ronment [12]. Actin-based cy- toskeleton has a critical role in cell motility and morphogene-sis, and is involved in intracel-lular signal transduction. There- fore, as a cross-linker between plasma membrane and Actin-based cytoskeleton, Ezrin can participate in the regulation of cell morphology, polarity and some crucial signaling path-ways, such as Rho-mediated signaling pathway and Akt-mediated apoptotic pathway, which play critical roles in the tumor progression [37].

dence-based methods to a larger sample size after pooling all available evidences, we con-cluded that positive Ezrin expression was sig-nificantly associated with the worse OS (HR: 2.485; 95% CI: 1.995-3.095; P<0.001) and DFS (HR: 2.081; 95% CI: 1.329-3.260; P=0.001) in NSCLC patients. A prognostic indi-cator is considered to be of high predictive value when its HR is larger than 2 [39]. Therefore, Ezrin could serve as a strong bio-marker to predict the poor prognosis of NSCLC. After further analysis, we found that the asso-ciation between Ezrin expression and the prog-nosis of NSCLC remained prominent in the sub-groups stratified by statistical analysis me- thods, Ezrin-staining sites, cut-off definitions and the length for follow-ups. The HR statistics incorporated into quantitative synthesis were mainly derived from multivariate analysis rath-er than univariate analysis. Multivariate analy-sis is generally considered as a necessary method for adequately eliminating the bias risks from other confounding factors in obser-vational studies. Therefore, the pooled HR with

95% CI based on multivariate analysis may be more accurate than those based on univariate analysis. The significant heterogeneity among univariate data also suggested the potential effects of other insufficiently eliminated con-founders on interfering the summarized out-comes (I2_{=60.7%, P=0.078).}

Both the cytoplasmic and cytoplasm-mem-brane Ezrin signals detected by IHC significant-ly predicted the worse prognosis of NSCLC. However, we speculated that the hugely varied definitions for positive expression of Ezrin might result in a few bias risks on its actual roles in NSCLC, although the pooled HR still sig-nificantly predicted the worse OS of NSCLC in the subgroups stratified by cut-off values. In addition, according to the subgroup analysis on follow-up period, Ezrin expression could inde-pendently predict both short-time and long-time survival outcomes of NSCLC. Given such reviews, Ezrin was considered as an effective and comprehensive biomarker for predicting the prognosis of NSCLC.

However, we found that the origins of patients enrolled in the included studies might cause slightly negative effects on the integrity of our meta-analysis. All of the included studies inves-tigated the prognostic significance of Ezrin expression in East-Asian patients with NSCLC. No evidence was identified to evaluate the prognostic roles of Ezrin expression in the prog-nosis of NSCLC among Western populations. One possible reason might be that we per-formed a literature retrieval in only three English electronic databases and one Chinese native electronic database, although no language restriction was imposed in our meta-analysis. If searching more databases in other languages, such as French, Spanish or Germen, many addi-tional studies might be identified and further enrich our findings. Therefore, an updated meta-analysis integrating the survival data from Western peoples would be urgently required to clarify the potential ethnic differ-ences of Ezrin expression in NSCLC.

As suggested by most of the included studies, Ezrin expression was significantly associated with the unfavorable conditions for some com-mon clinicopathological characteristics which result in poor prognosis of NSCLC, such as TNM stage [27, 29, 31, 32, 34], lymphatic metasta-sis [25, 26, 30-34] and differentiation degree [26, 27, 30]. Although the relationship between Ezrin expression and the above parameters was not all statistically reliable across different studies, an obvious tendency towards the increased malignancy of NSCLC was commonly observed. The inconsistency existed during the procedures of data collection, inclusion criteria and cut-off values may cause such controver-sies, and a large number of enrolled samples are needed to eliminate the statistical biases [30]. Therefore, a meta-analysis wholly study-ing the association between Ezrin expression and clinicopathological characteristics of NSCLC is highly expected to further interpret and modify our discoveries in the future.

Limitations

Several limitations existed within this meta-analysis should be acknowledged. First, three included studies only reported the survival data with log-rank P value instead of the statistical data derived from multivariate analysis. The validity of the overall summarized HR might be slightly attenuated because of the potential

bias risks from other insufficiently eliminated confounding factors. Second, the inconsistent cut-off definitions applied across different stud-ies might deviate from the actual roles of Ezrin expression in NSCLC. Third, the poor sensitivity of both Begg’s test and Egger’s test might pro-vide the misleading epro-vidence for publication bias when less than 20 studies were included into the meta-analysis. Fourth, the present meta-analysis was based on the East-Asian studies. The summarized outcomes in this meta-analysis should be judiciously evaluated in the clinical setting of other nations. Fifth, three general English databases and one native Chinese database were searched for the appro-priate studies included into this meta-analysis. More additional studies might further enrich our meta-analysis if performing a comprehen-sive search in the other non-English databases.

Conclusions

In conclusion, our meta-analysis indicates that Ezrin expression is significantly associated with the worse OS and DFS in patients with NSCLC. Further analysis has confirmed that the rela-tionships between Ezrin expression and poor OS remain statistically prominent in the sub-groups stratified by statistical analysis, Ezrin-staining sites, cut-off values and follow-up peri-ods. Therefore, Ezrin can serve as an independent biomarker for predicting the poor prognosis of NSCLC. More multi-institution reports with larger sample size are highly expected to further confirm our discoveries. An updated meta-analysis specifically investigat-ing the association between Ezrin expression and clinicopathological features of NSCLC is urgently required to further interpret the prog-nostic roles of Ezrin.

Acknowledgements

We thank the assistance of the staff in Depart- ment of Thoracic Surgery and Lung Cancer Center, West China Hospital, Sichuan University. This review was supported by Foundation of Science and Technology support plan Depart- ment of Sichuan Province (2014SZ0148 and 2015SZ0158).

Disclosure of conflict of interest

Address correspondence to: Dr. Guo-Wei Che, De- partment of Thoracic Surgery, Sichuan University, 37 Guoxuexiang, Chengdu, Sichuan, China. Tel: +86-28-189-8060-1890; Fax: +86-28-85422494; E-mail: guowei_che@foxmail.com

References

[1] Wang Z. Selection of chemotherapy for non-small cell lung cancer is facilitated by new therapeutic strategies. Int J Clin Exp Med 2014; 7: 3833-42.

[2] Mo Y, Peng J, Su W, Chen X, Wu A, Li J and Yang Z. Controversies regarding T status and N sta-tus for non-small cell lung cancer. Int J Clin Exp Med 2015; 8: 11675-82.

[3] Ferlay J, Shin HR, Bray F, Forman D, Mathers C and Parkin DM. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Can-cer 2010; 127: 2893-917.

[4] Hoffman PC, Mauer AM and Vokes EE. Lung cancer. Lancet 2000; 355: 479-85.

[5] Soria JC, Kim ES, Fayette J, Lantuejoul S, Deutsch E and Hong WK. Chemoprevention of lung cancer. Lancet Oncol 2003; 4: 659-69. [6] Baltayiannis N, Chandrinos M,

Anagnostopou-los D, Zarogoulidis P, Tsakiridis K, Mpakas A, Machairiotis N, Katsikogiannis N, Kougioumtzi I, Courcoutsakis N and Zarogoulidis K. Lung cancer surgery: an up to date. J Thorac Dis 2013; 5 Suppl 4: S425-39.

[7] Chansky K, Sculier JP, Crowley JJ, Giroux D, Van Meerbeeck J, Goldstraw P; International Stag-ing Committee; ParticipatStag-ing Institutions. The international association for the study of lung cancer staging project: prognostic factors and pathologic TNM stage in surgically managed non-small cell lung cancer. J Thorac Oncol 2009; 4: 792-801.

[8] Peng B, Wang YH, Liu YM and Ma LX. Prognos-tic significance of the neutrophil to lymphocyte ratio in patients with non-small cell lung can-cer: a systemic review and meta-analysis. Int J Clin Exp Med 2015; 8: 3098-106.

[9] Bretscher A, Edwards K and Fehon RG. ERM proteins and merlin: integrators at the cell cor-tex. Nat Rev Mol Cell Biol 2002; 3: 586-99. [10] Pujuguet P, Del Maestro L, Gautreau A,

Lou-vard D and Arpin M. Ezrin regulates E-cadherin-dependent adherens junction assembly through Rac1 activation. Mol Biol Cell 2003; 14: 2181-91.

[11] Krishnan K, Bruce B, Hewitt S, Thomas D, Khanna C and Helman LJ. Ezrin mediates growth and survival in Ewing’s sarcoma through the AKT/mTOR, but not the MAPK, sig-naling pathway. Clin Exp Metastasis 2006; 23: 227-36.

[12] Khanna C, Wan X, Bose S, Cassaday R, Olomu

and Helman LJ. The membrane-cytoskeleton linker ezrin is necessary for osteosarcoma me-tastasis. Nat Med 2004; 10: 182-6.

[13] Lun DX, Hu YC, Xu ZW, Xu LN and Wang BW. The prognostic value of elevated ezrin in pa-tients with osteosarcoma. Tumour Biol 2014; 35: 1263-6.

[14] Li H, Min D, Zhao H, Wang Z, Qi W, Zheng S, Tang L, He A, Sun Y, Yao Y and Shen Z. The prognostic role of ezrin immunoexpression in osteosarcoma: a meta-analysis of published data. PLoS One 2013; 8: e64513.

[15] Han K, Qi W, Gan Z, Shen Z, Yao Y and Min D. Prognostic value of Ezrin in solid tumors: a meta-analysis of the literature. PLoS One 2013; 8: e68527.

[16] Moher D, Liberati A, Tetzlaff J, Altman DG; PRISMA Group. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009; 339: b2535. [17] Stang A. Critical evaluation of the

Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 2010; 25: 603-5.

[18] Parmar MK, Torri V and Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival end-points. Stat Med 1998; 17: 2815-2834. [19] Tierney JF, Stewart LA, Ghersi D, Burdett S and

Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analy-sis. Trials 2007; 8: 16.

[20] In: Higgins JP and Green S, editors. Cochrane Handbook for Systematic Reviews of Interven-tions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available from www.cochrane-handbook.org.

[21] Zhang J and Yu KF. What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. JAMA 1998; 280: 1690-1691.

[22] Higgins JP and Thompson SG. Quantifying het-erogeneity in a meta-analysis. Stat Med 2002; 21: 1539-58.

[23] Higgins JP, Thompson SG, Deeks JJ and Alt-man DG. Measuring inconsistency in meta-analyses. BMJ 2003; 327: 557-60.

[24] Begg CB and Mazumdar M. Operating charac-teristics of a rank correlation test for publica-tion bias. Biometrics 1994; 50: 1088-101. [25] Chen QY, Yan J, Hu HZ, Chen FY, Song J, Jiang

ZY, Jiao DM and Wu YQ. Expression of ezrin in human non-small cell lung cancer and its rela-tionship with metastasis and prognosis. Chin J Oncol 2012; 34: 436-40.

[27] Jin T, Jin J, Li X, Zhang S, Choi YH, Piao Y, Shen X and Lin Z. Prognostic implications of ezrin and phosphorylated ezrin expression in non-small cell lung cancer. BMC Cancer 2014; 14: 191.

[28] Koriyama H, Ishii G, Yoh K, Neri S, Morise M, Umemura S, Matsumoto S, Niho S, Ohmatsu H, Tsuboi M, Goto K and Ochiai A. Presence of podoplanin-positive cancer-associated fibro-blasts in surgically resected primary lung ade-nocarcinoma predicts a shorter progression-free survival period in patients with recurrences who received platinum-based chemotherapy. J Cancer Res Clin Oncol 2015; 141: 1163-70. [29] Lee HW, Kim EH and Oh MH. Clinicopathologic

implication of ezrin expression in non-small cell lung cancer. Korean J Pathol 2012; 46: 470-7.

[30] Li J, Yang H, Zhang S, Yu N and Zhou Q. Expres-sion and their significance of ezrin and E-cad-herin in non-small cell lung cancer. Chin J Lung Cancer 2007; 10: 183-7.

[31] Lin L. Ezrin protein non-expression and progno-sis of patients with squamous cell lung carci-noma. J East Liaoning Univ Nat Sci 2013; 20: 27-31.

[32] Lin L, Wang Y, Xiang R, Ma Y, Lin Z and Zhang J. Ezrin protein expression is correlated with the occurrence, development and prognosis of lung adenocarcinoma. Chin J Mod Med 2014; 24: 13-16.

[33] Zhang M, Wang J, Chen L, Zhou H and Zhou X. The expression of Ezrin and its significance in non-small cell lung cancer. Chin Clin Oncol 2008; 13: 784-87.

[34] Zhang XQ, Chen GP, Wu T, Yan JP and Zhou JY. Expression and clinical significance of ezrin in non-small-cell lung cancer. Clin Lung Cancer 2012; 13: 196-204.

[35] Khanna C, Khan J, Nguyen P, Prehn J, Caylor J, Yeung C, Trepel J, Meltzer P and Helman L. Metastasis-associated differences in gene ex-pression in a murine model of osteosarcoma. Cancer Res 2001; 61: 3750-9.

[36] Arpin M, Algrain M and Louvard D. Membrane-actin microfilament connections: an increasing diversity of players related to band 4.1. Curr Opin Cell Biol 1994; 6: 136-41.

[37] Tsukita S and Yonemura S. ERM (ezrin/radix-in/moesin) family: from cytoskeleton to signal transduction. Curr Opin Cell Biol 1997; 9: 70-5.

[38] Ren F, Zhang X, Liang H, Luo D, Rong M, Dang Y and Chen G. Prognostic significance of MiR-34a in solid tumors: a systemic review and meta-analysis with 4030 patients. Int J Clin Exp Med 2015; 8: 17377-91.

[39] Pan F, Mao H, Deng L, Li G and Geng P. Prog-nostic and clinicopathological significance of microRNA-21 overexpression in breast cancer: a meta-analysis. Int J Clin Exp Pathol 2014; 7: 5622-33.

[40] Wang W, Wang F, Zong G, Liu R, Zhang Y, Luan Y, Xu L and Xuan J. Prognostic significance of EZH2 expression in patients with digestive cancers: a meta-analysis. Int J Clin Exp Med 2015; 8: 16043-9.

[41] Qu J, Li M, Zhong W and Hu C. Prognostic role of microRNA-25 in cancers: evidence from a meta-analysis. Int J Clin Exp Med 2015; 8: 12921-7.

[42] Li S, Lai Y, Fan J, Shen C and Che G. Clinico-pathological and prognostic significance of Nestin expression in patients with non-small cell lung cancer: a systematic review and me-ta-analysis. Clin Exp Med 2016; [Epub ahead of print].

[43] Li SJ, Fan J, Zhou J, Ren YT, Shen C and Che GW. Diabetes Mellitus and Risk of Broncho-pleural Fistula After Pulmonary Resections: A Meta-Analysis. Ann Thorac Surg 2016; 102: 328-39.

[44] Li S, Fan J, Liu J, Zhou J, Ren Y, Shen C and Che G. Neoadjuvant therapy and risk of broncho-pleural fistula after lung cancer surgery: a sys-tematic meta-analysis of 14 912 patients. Jpn J Clin Oncol 2016; 46: 534-46.

[45] Li S, Fan J, Zhou J, Ren Y, Shen C and Che G. Residual disease at the bronchial stump is positively associated with the risk of bronchop-lerual fistula in patients undergoing lung can-cer surgery: a meta-analysis. Interact Cardio-vasc Thorac Surg 2016; 22: 327-35.

[46] Li SJ, Chen DL, Zhang WB, Shen C and Che GW. Prognostic value of stromal decorin expression in patients with breast cancer: a meta-analy-sis. J Thorac Dis 2015; 7: 1939-50.

Supplementary Data 1. PRISMA 2009 Checklist

Section/topic # Checklist item

TITLE

Title 1 Identify the report as a systematic review, meta-analysis, or both. 1 ABSTRACT

Structured summary 2 Provide a structured summary including, as applicable: background; objectives; data sources; study eligibility criteria, participants, and interven-tions; study appraisal and synthesis methods; results; limitainterven-tions; conclusions and implications of key findings; systematic review registration number.

2, 3

INTRODUCTION

Rationale 3 Describe the rationale for the review in the context of what is already known. 4 Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design

(PICOS). 5

METHODS

Protocol and

registra-tion 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number. 5 Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used

as criteria for eligibility, giving rationale. 6, 7 Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and

date last searched.

5, 6 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated. 5, 6, Supplementary

Data 2 Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis). 5-7, 29 (Figure 1) Data collection

process

10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

7, 8 Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made. 7, 8 Risk of bias in

individ-ual studies

12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

7 Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means). 8, 9 Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I2_{) for each meta-analysis. 9}

Section/topic # Checklist item Reported on page #

Risk of bias across studies

15 Specify any assessment of risk of bias that may affect the cumulative evidence (e.g., publication bias, selective reporting within studies). 9, 10 Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified. 9 RESULTS

Study selection 17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow

diagram. Supplementary Data 210, 29 (Figure 1), Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. 11, 12, 23-25 Risk of bias within

studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12). 11, Supplementary Data 3 Results of individual

studies

20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

Risk of bias across

studies 22 Present results of any assessment of risk of bias across studies (see Item 15). 14, 33 (Figure 5) Additional analysis 23 Give results of additional analyses, if done (e.g., sensitivity or subgroup analyses, meta-regression [see Item 16]). 13, 14, 32 (Figure 4) DISCUSSION

Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

15-18 Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias). 18, 19

Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research. 19 FUNDING

Funding 27 Describe sources of funding for the systematic review and other support (e.g., supply of data); role of funders for the systematic review. 1

Supplementary Data 2. The summary of literature search PubMed search strategy

Searches Search details Items found

#1 Search ((((“lung neoplasms” [MeSH Terms] OR (“lung” [All Fields] AND “neoplasms” [All Fields]) OR “lung neoplasms” [All Fields] OR (“lung”[All Fields] AND “cancer” [All Fields]) OR “lung cancer” [All Fields]) OR ((“lung” [MeSH Terms] OR “lung” [All Fields]) AND (“carcinoma” [MeSH Terms] OR “carcinoma” [All Fields]))) OR (“lung neoplasms” [MeSH Terms] OR (“lung” [All Fields] AND “neoplasms” [All Fields]) OR “lung neoplasms” [All Fields] OR (“lung” [All Fields] AND “neoplasm” [All Fields]) OR “lung neoplasm” [All Fields])) OR (“lung neoplasms” [MeSH Terms] OR (“lung” [All Fields] AND “neoplasms” [All Fields]) OR “lung neoplasms” [All Fields] OR (“lung” [All Fields] AND “tumor” [All Fields]) OR “lung tumor” [All Fields])) AND (“ezrin” [Supplementary Concept] OR “ezrin” [All Fields])

87

EMBASE (via Ovid interface) search strategy

Searches Search details Items found

#1 Search ((lung cancer or lung carcinoma or lung neoplasm or lung tumor) and Ezrin).af. 69

The Web of Science (via campus network of Sichuan University) search strategy

Searches Search details Items found

#1 Search TS=((lung cancer OR lung carcinoma OR lung neoplasm OR lung tumor) AND Ezrin). 136

Supplementary Data 3. NOS scoring records of the included studies

Authors (year) Selection Comparability

Exposure

Total score Assessment

of outcome enough for outcomesfollow-up long

Adequacy of follow-up of

cohorts

Chen et al (2012) [25] 4 2 1 0 1 8

Gao et al (2015) [26] 4 1 1 0 1 7

Jin et al (2014) [27] 4 1 1 0 1 7

Koriyama et al (2015) [28] 4 1 1 0 1 7

Lee et al (2012) [29] 4 2 1 1 1 9

Li et al (2007) [30] 4 1 1 1 1 8

Lin et al (2013) [31] 4 2 1 0 1 8

Lin et al (2014) [32] 4 2 1 0 1 8

Zhang et al (2008) [33] 4 1 1 0 1 7

Zhang et al (2012) [34] 4 2 1 1 1 9