Prognostic Significance of High Survivin Expression in Patients with Gastrointestinal Cancer : A Meta-Analysis

Previous studies on the prognostic impact of survivin expression in gastrointestinal (GI) cancer have yielded inconsistent results. This study was initiated to assess the relationship between survivin expression and overall survival (OS) or disease free survival (DFS) in GI cancer patients. We applied system literature searches on EMBASE, PubMed, Web of science, and the Cochrane library to conduct this up-to-date meta-analysis. Thirty studies with totally 3622 GI cancer patients were collected. The prevalence of high survivin expression in GI cancer was 0.57 (95% CI: 0.51-0.63). High survivin expression was significantly associated with shorter OS (HR 1.57, 95% CI: 1.42-1.74) and DFS (HR 1.38, 95% CI: 1.21-1.58). Subgroup analysis also showed significant association between high survivin expression and poorer OS or DFS in gastric cancer or colorectal cancer. In summary, our study indicated that high survivin expression was related to poor prognosis in GI cancer. Well-designed studies with large sample and more convincing data are needed to confirm our conclusion.


Introduction
Gastrointestinal (GI) cancer, including gastric cancer and colorectal cancer, is highly aggressive malignancy that constitutes a major risk to public health worldwide, with nearly 2.4 million new cancer patients and 1.4 million deaths in 2012 [1].The Combination of surgery, chemotherapy and radiotherapy remains the standard regimens for GI cancer patients.However, not all patients derive a benefit from these strategies [2].Thus, novel therapeutic strategy is urgently needed.It is of great clinical value to identify applicable prognostic biomarkers that may not only predict the effectiveness of treatment above but also derive the potential therapeutic drug.
Survivin is the most extensively investigated member of inhibitor of apoptosis proteins family, and it can enhance proliferation, promote angiogenesis and inhibit apoptosis [3,4].High survivin expression levels were detectable in various types of malignancies including colorectal cancer [5], glioma [6], breast cancer [7] and gastric cancer [8].Recently, high survivin expression was found to play a active role in the failure of chemotherapy and radiotherapy [9,10] and predict the poor prognoses in patients with GI cancer in numerous studies [11][12][13][14].
However, the results remain controversial and inconsistent.For this reason, we performed an up-to-date systematic analysis to evaluate the potential impact of survivin in the clinical outcome of patients with GI cancer.

Literature search
The flow chart of the literature search was showed in Figure 1.Totally, 532 potentially related articles were selected, and thirty studies were finally included.The majority of removing articles were studies with animals or cell lines or reviews.

Characteristics of the study
Among the thirty studies, nineteen studies were on colorectal cancer, and eleven studies were on gastric cancer, 3622 patients with GI cancer were finally collected.The fundamental feature of the included studies was shown in Table 1.In addition, OS were extracted from twenty three studies and DFS was extracted from eight studies.

Rate of high survivin expression
The prevalence of high survivin expression in these studies ranged from 20.4% to 88.3%, and it partly reflected the heterogeneity in the criteria for its high expression levels.In the meta-analysis, the prevalence of high survivin expression was 0.57 (95% CI: 0.51-0.63, Figure 2).Subgroup analysis was stratified by colorectal cancer and gastric cancer.The rates of high survivin expression in colorectal cancer and gastric cancer were 0.54 (95%CI: 0.47-0.62)and 0.62 (95%CI: 0.52-0.71),respectively (Supplementary Figures S1-S2).

Association of survivin with GI cancer prognoses
The pooled OS was utilized to summarize the overall effect of high survivin expression for the twenty three studies.The analysis indicated that the pooled HR was 1.57 (95% CI: 1.42-1.74,Figure 3A) in GI cancer patients.However, the results showed high heterogeneity (I 2 ＜ = 57.4%;P 0.001).In the eight studies reporting DFS, the analysis showed that high survivin expression also contributed to poor DFS (1.38, 95%CI: 1.21-1.58,Figure 3B).In addition we carried out the meta-analysis of high survivin expression and survival in colorectal cancer and gastric cancer.Our results showed that high survivin expression was related to poor OS (HR 1.61, 95% CI: 1.43-1.81,Figure 4A) in colorectal cancer as well as gastric cancer (HR: 1.43, 95% CI: 1.14-1.79,Figure 4B).Further, two pooled results showed that high survivin expression also contributed to shorter DFS (HR: 1.33, 95% CI: 1.16-1.53,

Sensitivity and Publication bias
The sensitivity analysis indicated that the results were stable and were not affected when removing one study each time.No obvious asymmetry was found in the chart of the funnel

Discussion
The meta-analysis was conducted to assess the potential role of high survivin expression in clinical outcomes in GI cancers patients.Our analysis showed that the overall incidence of high survivin expression was 0.57 in GI cancers patients.Besides, we found that high expression of survivin was the prognostic factor for both OS and DFS.On account of the heterogeneity, we performed the subgroup analysis on gastric cancer and colorectal cancer.
Our data also indicated that high survivin expression was significantly related to poor OS and DFS.As far as we know, this was the first meta-analysis to investigate the association of high survivin expression and the clinical outcomes in GI cancers patients.

Survivin as an prognostic biomarker in GI cancer
The current study revealed that high survivin expression was related to poor clinical outcome.Further investigations are necessary to clarify the function of survivin as a biomarker for prognosis in GI cancer patients.The molecular mechanisms involved in the process of survivin contributing to the clinical outcome remains uncertain.Ye et al [50] reported that survivin knockdown by siRNA resulted in the inhibition of colorectal cancer cell invasion and migration by mediating the AKT and the ERK pathways.Fei et al [51] found that survivin transfected by shRNA could activate the extrinsic and intrinsic apoptotic process in SW-620 colon cells.Pandey et al's study [52] showed that survivin inhibition induced enhanced cell death in gastric cells with the 5-fluorouracil (5-FU) treatment.Wang et al [53]revealed that a survivin promoter-regulated oncolytic adenoviral vector has broad-spectrum antitumour properties, and survivin has been closely correlated with an elevated proliferative capacity, an enhanced metastatic capacity, and chemotherapy and radiotherapy resistance in cancer cells.In addition to being a prognostic biomarker, our results are of great value in view of the emergence of new drugs targeting survivin.Currently, several clinical trials targeting survivin at different phases are being developed which will likely benefit patients with certain conditions (Table 2).

Strengths and limitations of our study
Our results had several important implications.First, our analysis had been enhanced by the involvement of over 3,000 patients in 30 centres.Second, the relationship between high survivin expression and prognosis in GI cancers persisted and remained statistically significant by subgroup analysis.Third, all of the analyses were performed by the fixed effects model or random effects model.Models was applied by the different heterogeneous, which contributed to the validity and reliability of the statistical results.Our meta analysis had also some limitations as follows.First, our study was a literature-based analysis.It was possible that there may be some degree of publication bias in this area of research.To address this issue, Begg's test was applied.Second, there was statistical heterogeneity among the patients contributing the basic feature of the studies.Accordingly, we further performed a subgroup analysis according to the stratified types of cancers.Third, there was clearly a multitude of confounding factors (test method，definition of high expression, difference between extracting data from Kaplan-Meier curves and raw data ,etc ) that made meta-analysis inaccurate.

Identification and Selection of Studies
EMBASE, PubMed, Medline, Cochrane library data and the Web of Science were screened for articles exploring the expression of survivin in GI cancer by two independent investigators.We identified the articles with the following terms: (1)"survivin"or "BIRC5" and ( 2) "digestive tract" or "gastrointestinal" or "gastric" or "stomach" or "colorectal" or "colon" or "rectal" and (3) "cancer" or "carcinoma" or "tumor" or "neoplasm".We also collected the reference from the original articles and review papers.Paper published in English was selected only and the ending date was January 31, 2017.
Eligibility criteria were as follows: (1) expression of survivin was conducted using immunohistochemistry or qRT-PCR;(2) association of survivin expression with survival in GI caner; and (3) the hazard ratio (HR) for the overall survival (OS) or disease free survival (DFS) related to survivin expression was reported or could be calculated from the paper.
The following studies were excluded :(1) articles about animals or cell lines;(2)studies lacking sufficient data on survival;(3)overlapping samples or duplication of previous publications; and (4)the exclusion of letters to the editor, articles and reviews that were not published in English.

Data Collection and Synthesis
Data were collected by two independent researchers using predefined variables abstraction forms.The following information was collected: gender, age, tumor type, first author, study location, published year, number of patients, cutoff for considering survivin as being highly expressed and HR with corresponding 95% confidence interval(CI) for OS or DFS.When both multivariate analysis and univariate analysis were used to calculate HR, the data taken from multivariate analysis were selected.When HR was not provided directly, they were estimated using technique published by Tierney [15].
The overall HR was pooled by the random effects model or the fixed effects model, which was determined by the heterogeneity calculated by Chi-square test [16].When the heterogeneity was high, the former model was used.Otherwise, the fixed effects model was considered.Pooled HR > 1 indicated that high survivin expression contributed to poor OS or DFS in GI cancer patients.

Quality assessment
The Newcastle Ottawa Scale was applied to evaluate the quality of the studies recruited.The scale is an eight-item system (with a maximum of nine stars) to evaluate a study in three aspects: selection, comparability, and outcome.Any disagreements were negotiated between two investigators and reached an agreement finally.

Statistical Analysis
Data was pooled and analyzed by STATA 10.0 (Stata Corporation, Texas, USA).For the rate of high survivin expression, we combined the incidences and 95%CI.The relationship between survivin expression and the survival such as OS and DFS were estimated using HR.In our study, the HR with corresponding 95%CI was preferably extracted from the article and calculated by the technique mentioned above.The publication bias was visually checked by funnel plots, and was further assessed using Begg's test [17,18].The sensitivity test was evaluated by deleting one study each time.The statistical significance was defined as P＜0.05 and all statistical tests were two-sided.

Conclusions
Based on our study, it was found that high survivin expression was associated with poor prognosis in terms of OS and DFS, which indicates that developing the strategies targeting survivin could be a promising therapeutic treatment for GI cancer patients.
Supplementary Materials: Supplementary materials can be found at www.mdpi.com/link.

Figure 1 .
Figure 1.Flow chart representing the process of study selection

Figure 2 .
Figure 2. Meta-analysis of the prevalence of high survivin expression.

PreprintsFigure 3 :
Figure 3: Prognostic value of high survivin expression in GI cancer patients.A. meta-analysis of high survivin expression and OS in GI cancer patients; B. meta-analysis of survivin expression and DFS in GI cancer patients.

Figure 5 :
Figure 5: Sensitivity analysis and publication bias for the association between high survivin expression and survival.A. sensitivity analysis of the association between high survivin expression and OS; B. sensitivity analysis of the association between high survivin expression and DFS;C.funnel plot for the association between high survivin expression and OS; D. funnel plot for the association between high survivin expression and DFS.

Table 2 .
Studies targeting survivin in cancer