Clinicopathological and Prognostic Significance of CD24 Overexpression in Patients with Gastric Cancer: A Meta-Analysis

Objective The prognostic significance of CD24 expression for survival in patients with gastric cancer remains controversial. We conducted a meta-analysis to investigate the impact of CD24 expression on clinicopathological features and survival outcomes in gastric cancer. Methods A comprehensive literature search of the electronic databases PubMed, Embase, Web of Science and China National Knowledge Infrastructure (CNKI; up to April 8, 2014) was performed for relevant studies using multiple search strategies. Correlations between CD24 expression and clinicopathological features and overall survival (OS) were analyzed. Results A total of 1,041 patients with gastric cancer from 9 studies were included. The pooled odds ratios (ORs) indicated CD24 expression was associated with tumor depth (OR = 0.45, 95% confidence interval [CI]  = 0.32–0.63; P<0.00001), status of lymph nodes (OR = 0.40, 95% CI = 0.25–0.64; P = 0.0001) and tumor node metastasis (TNM) stage (OR = 0.56, 95% CI = 0.41–0.77; P = 0.0003). The pooled hazard ratio (HR) for OS showed overexpression of CD24 reduced OS in gastric cancer (HR = 1.99, 95% CI = 1.29–3.07, P = 0.002). Whereas, combined ORs showed that CD24 expression had no correlation with tumor differentiation or Lauren classifications. Conclusion CD24 overexpression in patients with gastric cancer indicated worse survival outcomes and was associated with common clinicopathological poor prognostic factors.


Introduction
Gastric cancer (GC) is the fourth most common cancer and the second leading cause of cancer-related death worldwide [1]. Although having undergone radical resection and postoperative adjuvant therapy, most patients with GC will die of recurrence and metastasis. Several clinicopathological parameters such as tumor size, histological type, tumor differentiation, depth of tumor invasion, regional lymph node involvement, distant metastasis and tumor stage, have been reported as important prognostic factors for GC [2][3][4][5][6]. However, the advance in treatment of GC was relatively small in the past few decades. Understanding the molecular mechanisms that lead to the development and progression of GC remains an important challenge in translational research.
Using immunohistochemistry, a series of studies has revealed that CD24 was expressed in a variety of human malignancies, such as nasopharyngeal carcinoma, non-small-cell lung cancer, breast cancer, hepatocellular carcinoma, pancreatic cancer, colorectal cancer, renal cell carcinoma, bladder carcinoma, ovarian cancer, prostate cancer and intrahepatic cholangiocarcinoma [16,17]. Expression of CD24 might facilitate the interactions of cancer cells with endothelial cells and platelets, which promotes the dissemination of CD24-expressing cancer cells [18,19]. Previous studies have shown that the expression of CD24 is correlated with tumor progression and a poor prognosis in various carcinomas [20].
Although evidence exists that CD24 is an important factor implicated in clinicopathological features [21][22][23][24][25][26][27][28][29] and the prognosis of GC [22,23], some conflicting results have been reported. A study on CD24 in GC reported that positive CD24 expression tended to indicate worse survival outcomes, but the difference was not statistically significant [27]. Moreover, two other recent studies on a panel of tumor markers demonstrated that CD24 expression was not an independent prognostic factor for patients with GC [25,28]. Whether discrepancy in these data was due to limited sample sizes or genuine heterogeneity is still unknown. To address the controversial issues, a meta-analysis was carried out to determine the association between CD24 and clinicopathological parameters as well as the significance of CD24 expression in the prediction of clinical outcomes in GC.

Search Strategy
A comprehensive literature search of the electronic databases PubMed, Embase, Web of Science and China National Knowledge Infrastructure (CNKI) was performed up to April 8, 2014. Studies were selected using the following search terms: 'gastric or stomach' and 'cancer or neoplasm or carcinoma' and 'CD24'. The references of articles and reviews were also manually searched for additional studies. The eligible reports were identified by two reviewers (J-X.W. and Y-Y.Z.) and controversial studies were adjudicated by a third reviewer (H-X.A.).

Study Selection
We collected all eligible articles about the relationship between CD24 and clinicopathological features and clinical outcomes in GC in this meta-analysis. Studies meeting the following inclusion criteria were included: (1) CD24 expression was evaluated in the primary GC tissues; (2) CD24 expression was examined by immunohistochemistry (IHC); (3) studies revealed the relationship between CD24 expression and GC clinicopathological parameters or prognosis; (4) studies regarding the prognosis provided sufficient information to estimate hazard ratios (HRs) for overall survival (OS) and 95% confidenceintervals (CIs); (5) if there were multiple articles based on similar patients, only the largest or most recently published article was included. The exclusion criteria used in this meta-analysis were: (1) letters, reviews, case reports, conference abstracts, editorials and expert opinion; and (2) patients who had received previous chemotherapy or radiotherapy.

Data Extraction
Two investigators (J-X.W. and H-X.A.) independently extracted data from eligible studies. Disagreements were resolved by discussion and consensus. Two investigators reviewed all studies that met the inclusion and exclusion criteria. The following information was recorded for each study: name of the first author, year of publication, sample source, number of cases, clinicopathological parameters, Lauren classification for gastric adenocarcinomas [30], tumor node metastasis (TNM) stage, immunohistochemical technique, definition of a positive CD24, and survival of patients. If the HR or standard errors (SEs) were not reported in the included studies, we calculated or estimated the HR from available data or Kaplan-Meier curves using the methods reported by Tierney et al. [31].

Assessment of Study Quality
Two authors (J-X.W. and Y-Y.Z.) independently assessed the quality of all studies on the basis of a 9-score system; i.e., the Newcastle-Ottawa Scale (NOS) [32]. In this scoring system, each study included in the meta-analysis was judged on three broad perspectives: the selection of the study cases (four items, one score for each item), the comparability of the study populations (one item, up to two scores) and the ascertainment of either the exposure or outcome of interest (three items, one score for each item). Scores of items were added up and used to compare study quality in a quantitative manner. A higher score indicated the individual study was of higher quality. Discrepancies in the score were resolved through discussion between the authors.

Statistical Methods
The meta-analysis was performed using the Stata 12.0 (Stata Corporation, College Station, TX, USA) and Review Manager 5.2 (Cochrane Collaboration, Oxford, UK) programs. Comparisons of CD24 expression in groups with different clinicopathologic features were performed by pooled estimates of odds ratios (ORs), as well as the 95% CI. Statistical significance was determined at a50.05. In consideration of the possibility of heterogeneity among the studies, a statistical test for heterogeneity was examined by the Chi-square-based Q-test, and the significance level was fixed at P,0.10. The inconsistency index I 2 was also calculated to evaluate the variation caused by heterogeneity. A high value of I 2 indicated a higher probability of the existence of heterogeneity. The DerSimonian and Laird random-effects model [33] was used if substantial heterogeneity was detected (Q-statistic: P,0.10; I 2 .50%). Otherwise, a fixed-effects model of Mantel-Haenszel [34] was applied in the absence of between-study heterogeneity (Q-statistic: P.0.10; I 2 ,50%). Publication bias was assessed by a Begg's rank correlation test and Egger's regression asymmetry test [35,36].

Search Results
Twenty-nine articles were identified using the search strategy described above. Twenty of them were excluded due to being irrelevant to the current analysis, non-original articles (review) or having insufficient primary outcome data. There were nine studies finally included in the current meta-analysis (Fig. 1).

Study Characteristics
Nine publications from 2004 to 2014 were eligible for meta-analysis. Their characteristics are summarized in Table 1. A total of 1,041 patients from China, Korea, Turkey and Japan were enrolled, including 707 males and 334 females. CD24 overexpression was found in 521 patients (50.0%). Immunohistochemistry was the only method used to evaluate the expression of CD24 in GC specimens. The definition of CD24 positive staining varied among the studies. Six researches defined CD24 expression by different percentages of positive cells, whereas only one study determined CD24 expression by both staining intensity score and percentage of positive cells.

Qualitative Assessment
The study quality was assessed using the Newcastle-Ottawa quality assessment scale, generating scores ranging from 5 to 8 (with a mean of 6.56), with a higher value indicating better methodology. The results of quality assessment are shown in Table 1.  Quantitative Synthesis

Sensitivity Analyses and Publication Bias
A sensitivity analysis, in which one study was removed at a time, was performed to evaluate results stability. The corresponding pooled ORs and HRs were not significantly altered, suggesting stability of our results. Egger's and Begg's tests indicated no publication bias among these studies regarding the HR for OS with P values of 0.272 and 0.734 respectively. The funnel plots were largely symmetric (Fig. 4), which also suggested no evidence of publication bias in this meta-analysis.

Discussion
CD24 is a cell surface molecule that plays an important role in the migration and adhesion of cells [19]. The correlation between CD24 expression and GC has been studied by a number of researchers. Some of them showed overexpression of CD24 is associated with several clinicopathologic features and a poor prognosis [21][22][23][24][25][26][27][28][29]. However, the clinical relevance of CD24 as a prognostic factor in GC remains controversial. For procuring a reasonable conclusion, we combined 9 eligible studies including 1,041 cases to perform this meta-analysis. The depth of tumor invasion, lymph node metastasis and distant metastases are the major prognostic factors in GC. It has been reported that CD24 expression increases the proliferation of tumor cells, induces cell motility and invasion and promotes cell spreading [11]. CD24 was identified as a ligand to P-selectin, which is a surface molecule expressed by activated endothelial cells and platelets. Tumor cells with increased CD24 expression acquired the capacity to form stabilized platelet-tumor thrombi and then adhered to endothelial cells at the distal metastatic sites, which protected tumor cells from destruction and promoted tissue penetration and tumor extravasation [11,16,22,37]. On the other hand, by a subtractive technique, CD24 was revealed as one of the metastasis-associated genes and confirmed to be overexpressed in the metastatic phenotype [23]. A previous meta-analysis investigating the relationship between CD24 expression and prognostic parameters in different carcinomas suggested CD24 expression was associated with lymph node metastasis in breast cancer and advanced clinical stages in urothelial carcinomas. Limited by insufficient cases (151 patients were included), the expression of CD24 was found to be not associated with lymph  node metastasis in GC [20]. In the present meta-analysis, we found CD24 expression was related to tumor depth, status of lymph nodes and TNM staging in an expanded sample. In line with basic studies, the results of our research supported that CD24 was involved in cell-cell and cell-matrix interactions. Although the mechanism of cancer progression caused by expressional alterations remains unknown, CD24 overexpression may be considered as a marker of GC that indicates invasiveness. Further functional investigation of CD24 activity in GC may clarify the importance of CD24 in the progression of the disease.
In previous studies, only one of four studies suggested statistically significant HRs for OS of elevated CD24 expression [23] and the remaining three studies showed a trend for reduced survival outcomes [22,27,28]. Sample size, as a strong predictor for epidemiological studies, may play an important role in this controversy. In the current meta-analysis, we pooled the data from these studies together, and demonstrated a remarkable association between CD24 expression and OS of patients with GC.
According to the Lauren classification system [30], gastric adenocarcinomas were classified into a diffuse or intestinal type. The diffuse type, which occurs more commonly in younger women, has a diffuse infiltrative growth pattern and is usually a poorly differentiated adenocarcinoma. In contrast, the intestinal type occurs more frequently in elderly men and is characterized by well-defined glandular structures and is always associated with atrophic gastritis and intestinal metaplasia in adjacent mucosa [38]. The diffuse type of GC has a familial tendency, relates to decreased E-cadherin, and has a less favorable prognosis, whereas the intestinal type is associated with environmental etiology, microsatellite instability and adenomatous polyposis coli (APC) mutation [39]. These findings indicated that the diffuse and intestinal typesof GC have distinct molecular pathways. Previous studies revealed there was a potential relationship between Lauren classifications and CD24 expression. However, our meta-analysis of these pooled patients, showed that overexpression of CD24 was not associated with the diffuse or intestinal types of GC.
Efforts were made to conduct a comprehensive analysis, but some limitations need to be acknowledged. First, the survival analysis was not performed by multivariate analyses in most reported studies. We, therefore, calculated the HR for OS from available data or Kaplan-Meier curves. Second, it was clear that the two types of GC might have different biological behaviors. Although our study demonstrated CD24 expression was not associated with Lauren classifications, whether CD24 played a different role in intestinal-type and diffuse-type GC is still unknown. Limited to insufficient information to estimate the relationship between CD24 and clinicopathologic features in different types of GC, more efforts are needed in the future. Third, the different concentrations of antibodies and the variable definitions of CD24 expression used in the included studies might have influenced the results of our meta-analysis. Fourth, most of the included studies were carried out in East Asia and the results may be different in Western countries.
Our meta-analysis indicated that overexpression of CD24 in GC was not only associated with tumor exterior expansion, lymph node metastasis and advanced TNM stage, but also was a biomarker for poor prognosis. Further clinical studies may be performed to clarify the precise prognostic significance of CD24 in GC.
Supporting Information S1