Introduction

Lung cancer is the leading cause of cancer-related death both in the USA and around the world [1]. Despite tremendous efforts to reduce deaths due to lung cancer, the prognosis of lung cancer is still poor, with an approximate 5-year survival rate of 15 % [2]. NSCLC accounts for approximately 85 % of all cases of lung cancer [3]. Tumor-node metastasis (TNM) staging has been proposed as one of the most important prognostic factor of NSCLC. Other independent prognostic factors include performance status, age, and weight loss of more than 10 % [4, 5]. However, despite these clinical factors of survival in NSCLC patients, physicians still lack the confidence to accurately determine an individual patient’s prognosis.

Recently, advances in molecular biology have led to a better understanding of the molecular changes that occur during lung cancer carcinogenesis. New biological markers may add further prognostic information. Metastases specifically occur in malignant tumors and are the most frequent cause of treatment failure [6]. The initial step of metastasis is the detachment of cells, which is dependent on cell adhesion molecules (CAMs). CAMs can be grouped into four distinct families: cadherins, integrins, selectins, and the immunoglobulin superfamily [7]. E-cadherin, a member of cadherin family, is widely expressed in epithelial tissues. It is a transmembrane glycoprotein that mediates intercellular adhesion in the presence of calcium and tissue architecture among epithelial cell layers [8, 9].

The defective function of E-cadherin is a defining characteristic of the epithelial–mesenchymal transition [10]. Also, anti-E-cadherin antibodies can induce invasive behavior in tumor cells, which supports the hypothesis that impaired E-cadherin expression on tumor cells is associated with malignant behavior [11]. Furthermore, decreased E-cadherin expression has also been related to dedifferentiation, lymph node spread, and metastasis in several human cancers [12–16].

Clinical studies have also evaluated the association between E-cadherin protein expression and survival in patients with NSCLC [17–19]. However, no consensus has been reached and the prognostic significance of tumor E-cadherin expression levels remains debated. Therefore, we carried out a meta-analysis of data from published studies to quantitatively review the expression of E-cadherin in primary tumors measured by immunohistochemistry (IHC) on the survival of patients with NSCLC.

Patients and methods

Search strategy

We searched the PubMed database to identify primary research publications to include in this meta-analysis. The upper date of August 21, 2011 was applied. Searches terms were used as follows: “lung cancer,” “E-cadherin,” and “prognosis.” The following criteria for study eligibility were set before collecting articles: (1) measures E-cadherin expression in cell membranes by IHC in primary lung cancer tissue, (2) identifies the histological type of the tumors as NSCLC, (3) provides information on survival comparing patients with impaired expression of E-cadherin with patients without impaired expression, (4) has a follow-up time exceeding 5 years, and (5) was published in English. When an individual author published several articles obtained from the same patient population, only the newest or most complete article was included in the analysis.

Two reviewers (Wu Y and Liu HB) independently determined study eligibility. Disagreements were resolved by consensus.

Data extraction and methodological assessment

The final articles included were assessed independently by two reviewers (Wu Y and Liu HB). Data retrieved from the reports included author, journal. year of publication, country of origin, histology, disease stage, cut-off value, decreased E-cadherin expression, and survival data (Table 1).

Table 1 Main characteristics and results of the eligible studies
Full size table

To assess trial methodology, articles were read independently by two reviewers (Wu Y and Liu HB) and evaluated according to Steele’s method [20, 21]. Studies were considered to be of high quality if they met each of the following criteria: (1) prospective cohort design, (2) clear and detailed description of standard laboratory methodology about IHC, including antigen retrieval, types of both primary and secondary antibodies, and coloration techniques, and (3) the pathologist evaluating the IHC outcome was double-blind to patient clinicopathologic data and outcome. When studies are retrospective or the pathologist blinding was unknown, the studies were defined as medium quality. Low quality studies were defined as follows:(1) the study design was not defined or was a case report, and (2) the details of the laboratory methods were not described. Both high and medium quality studies were included in this meta-analysis, while low quality studies were excluded from further analysis.

Statistical methods

HRs and their 95 % CIs were used to combine the data. When these statistical variables were described in text or tables, we obtained them directly from articles. When not given explicitly in an article, they were calculated from available numerical data in the articles using methods reported by Parmar [22]. Heterogeneity of the individual HRs was calculated with χ2 tests according to Peto’s method [23].

Meanwhile, heterogeneity tests with I2 and Q statistics were performed. A fixed effect model was used to estimate the effects of reduced E-cadherin expression on survival if HRs were found to have fine homogeneity. If not, a random-effect model was used. In this meta-analysis, DerSimonian–Laird random effect analysis [24] was used. By convention, an observed HR > 1 implies worse survival for the group with negative/decreased E-cadherin expression. The impact of negative/decreased E-cadherin expression on survival was considered to be statistically significant if the 95 % CI did not overlap with 1. All statistical analyses were conducted using STATA version 11.0 (Stata Corporation, College Station, TX).

Results

Study selection and characteristics

One hundred and thirty-two potentially relevant citations were reviewed, and 21 articles concerning the association of E-cadherin expression with survival in NSCLC were initially found. Three articles compared a combined negative or positive E-cadherin expression group with a medium expression group [25–27]. E-cadherin protein was measured in serum in two articles [28, 29]. One article evaluated the correlation between cytoplasmic E-cadherin staining and survival [30]. One article analyzed the 5-year survival of 57 SCC and 44 ACC patients individually without summary statistics [31]. The influence of E-cadherin expression on recurrence and disease-free survival was shown in another article [32]. These seven papers were excluded from the meta-analysis, leaving 13 studies [17–19, 33–42] that met the criteria set forth in the search strategy and study selection (Fig. 1). All studies reported the prognostic value of E-cadherin status for survival in patients with lung cancer.

Fig. 1
figure 1

Forest plot (random-effects model) of 13 contributing studies assessing the effect of E-cadherin protein expression on survival in patients with NSCLC. 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)

Full size image

Meta-analysis results

Based on our search strategy, 13 studies were included. All 13 of these cohort analyses met the inclusion criteria and concerned the association of E-cadherin expression with survival in NSCLC. All 13 studies were of medium quality because none described the details of the blinding of the pathologists interpreting the IHC data. The total number of patients included was 2,477 ranging from 73 to 391 patients per study. The major characteristics of the 13 eligible publications are reported in Table 1. Eight studies had information for stages I–III, three for stage I, and two for stages I-IV. Two studies created continuous scores for E-cadherin expression [18, 19] with the remaining studies dichotomizing E-cadherin expression according to the proportion of the tissue staining positive for E-cadherin. Six studies selected cut-off points of 50 % [33, 35, 38, 40, 42]. Two studies selected cut-off points between 0 and 10 % of tissue sample [37, 39]. Two studies required a <70 % [34, 41] staining proportion to qualify as reduced staining, one required <75 % [36], and another required <25 % [17]. Surgery was performed for most patients. Among the 13 studies included in this meta-analysis, a significant relationship between E-cadherin expression in the primary tumor and survival was found in eight, all linking decreased E-cadherin expression with worse survival. The remaining five studies yielded negative results. All studies investigated NSCLC, including adenocarcinoma (AdC) and squamous cell carcinoma (SCC). Reduced or absent expression of E-cadherin in the individual studies ranged from 32.9 % [40] to 75.2 % [33]. The overall rate of decreased E-cadherin expression in the studies was 47.6 %. The combined HR for all 13 eligible studies was 1.41 (95 % CI 1.18–1.65; P = 0.001) (Fig. 1) indicating that reduced or absent E-cadherin expression in the primary lung tumors was an indicator of poor prognosis. However, a highly significant (P = 0.001) degree of heterogeneity was detected among the studies (Q = 33.33, I2 = 64 %, P = 0.001). When stratified by ethnicity, the combined HR of Asians was 1.49 (95 % CI 1.27–1.71; P = 0.000) without heterogeneity (Q = 9.1, I2 = 23 %, P = 0.246) (Fig. 2). Among non-Asian studies, the combined HR of was 1.01 (95 % CI 1.00–1.02; P = 0.000) also without heterogeneity (Q = 6.63, I2 = 39.6 %, P = 0.157) (Fig. 3). Further, we separately analyzed the studies with stage I NSCLC. In these five studies the combined HR was 1.19 (95 % CI 0.90–1.47; P = 0.576), indicating that reduced E-cadherin expression had no significant impact on survival in patients with NSCLC (Fig. 4). Heterogeneity was not significant (Q = 2.9, I2 = 0 %, P = 0.576). Since the cut-off values in the evaluation of E-cadherin expression vary with studies, meta-regression was performed to evaluate the influence of different cut-offs on the relationship between E-cadherin expression and the survival with P = 0.993, which was not statistical significance.

Fig. 2
figure 2

Forest plot (fixed-effects model) of eight contributing studies assessing the effect of E-cadherin protein expression on survival in Asian patients with NSCLC

Full size image
Fig. 3
figure 3

Forest plot (fixed-effects model) of five contributing studies assessing the effect of E-cadherin protein expression on survival in non-Asian patients with NSCLC

Full size image
Fig. 4
figure 4

Forest plot (fixed-effects model) of five evaluable studies assessing the effect of E-cadherin protein expression on survival in stage I NSCLC

Full size image

Discussion

Recently, various combined-modality therapies, including surgery, chemotherapy, and radiation therapy, have improved the outcome of patients with NSCLC [43, 44]. However, NSCLC is still one of the most common human malignancies with a poor prognosis. Metastasis and recurrence are considered major contributors to treatment failure. The initial step of metastasis is the escape of cells from the primary tumor, which is believed to be determined by the status of adhesion molecules. E-cadherin is a principal mediator of normal epithelial tissue architecture and its loss could induce tumor cells to dedifferentiate and become highly metastatic [45–49]. At present, there is no consensus on the association between reduced E-cadherin expression detected by IHC and poor survival in patients with NSCLC. More accurate evaluation of the impact of E-cadherin expression on patient survival is needed.

Meta-analysis was originally developed to combine the results of randomized controlled trials. This approach has been applied successfully to the identification of prognostic indicators in patients with lung cancer, such as p53 mutations [50]. Our analyses, combining 13 independent studies that included 2,247 patients with NSCLC, revealed that reduced E-cadherin expression does decrease the overall 5-year survival of NSCLC patients. Then we found that the combined HR in the Asian group (1.49) was larger than in non-Asian group (1.01), suggesting that E-cadherin expression could be racial different as an prognostic factor for NSCLC. Recently, a meta-analysis on E-cadherin gene polymorphism associated with gastric cancer among Asians but not Europeans has also been reported [51]. In addition, when analysis was restricted to stage I NSCLC, impaired E-cadherin expression was not associated with survival, which suggests that this prognostic factor could be of importance in advanced-stage NSCLC. Advanced lung cancers with reduced E-cadherin expression could be more aggressive and have a worse prognosis than those with normal E-cadherin expression. The present results concerning the prognostic role of E-cadherin in NSCLC still need to be confirmed by prospective studies before a potential clinical application can be determined.

This systematic review with meta-analysis had to address heterogeneity issues. We found highly significant heterogeneity among the 13 studies. The heterogeneity in these studies could be explained by the patient type, disease characteristics, or differences in the techniques used to detect alterations in E-cadherin expression, including antigen retrieval methods, choice of E-cadherin antibody, and revelation protocols (pH and compounds of the solutions, heating method, etc.). When the analysis was limited to eight studies of Asian and five studies of non-Asian separately, heterogeneity was not detected. Further, while analysis were limited to five studies including only stage I NSCLC, heterogeneity was also not detected. Therefore, the heterogeneity in this study could be mainly explained by the patient ethnicity or by different stages of NSCLC. In addition, we found that different cut-offs for decreased E-cadherin expression in tissues (10–75 %) were used in these studies. To evaluate the effect of varying cut-offs on the results, we have conducted a meta-regression analysis that suggested the survival was not statistical significance associated with the different cut-offs of the studies. However, it is very important to use a well-defined and well-standardized technique to evaluate biological markers. The same antibody, protocol, and criteria of evaluation of IHC should be used between different laboratories so that the results can be compared exactly.

As our systematic review was limited to the published scientific literature, we cannot rule out the potential impact of publication bias. Publication bias [52] is a widely recognized phenomenon for all forms of meta-analysis. Positive results tend to be accepted by journals, while negative results are often rejected or not even submitted. Moreover, the study was restricted to articles published in English, which probably provided additional bias.

In our meta-analysis, when HR and CI were not described in the reports, they were extracted from the data according to the method of Parmar [22]. We should keep in mind that estimated HR values and their CIs may differ to some extent depending on the estimation method used.

Some eligible trials had to be excluded from the meta-analysis because they did not provide sufficient data on survival. Among the five excluded studies, four reported significant associations between E-cadherin expression and survival. We recognize that including these results would make our results more positive. To control other prognostic factors such as tumor stage, we attempted to perform subgroup analysis. However, only some studies have described the relationship between aberrant E-cadherin expression and prognosis in each stage of NSCLC. Thus we could only analyze 5 studies with stage I tumors.

It should also be noted that E-cadherin is associated with epidermal growth factor receptor (EGFR) signaling and thus contributes to the effect of gefitinib treatment [53]. Some reports show that E-cadherin interacts with EGFR, decreasing ligand-affinity [54, 55] and inhibiting activation in several human tumor types, including lung cancer [56, 57]. Studies analyzing E-cadherin expression after anti-EGFR treatment were not included in this meta-analysis because anti-EGFR treatment may modify biological behavior and should not be aggregated together. Whether E-cadherin expression serves as a predictor of EGFR-TKI-mediated clinical activity in lung cancer should be the topic of another systematic review. Since carcinogenesis is a multiple-step process, any single molecule can not independently predict the survival of the patients completely. Combinations of prognostic biomarkers should also be necessary.

In conclusion, when all stages are considered, reduced or absent E-cadherin expression was associated with poor prognosis in patients with NSCLC in this meta-analysis. Interestingly, our meta-analysis suggests that the influence is more significant among Asians. However, the meta-analysis suggests that E-cadherin has no significant effect on survival in stage I NSCLC. These results should be confirmed by an adequately designed prospective study.