Clinicopathological Significance of EBV-Infected Gastric Carcinomas: A Meta-Analysis

Background and objectives: The present study aims to elucidate the clinicopathologic significance of Epstein–Barr virus (EBV) infection in gastric carcinomas (GCs) through a meta-analysis. Materials and Methods: Sixty-one eligible studies were included in the present meta-analysis. The included patients, with and without EBV infection, were 2063 and 17,684, respectively. We investigated the clinicopathologic characteristics and various biomarkers, including programmed death-ligand 1 (PD-L1) expression and tumor-infiltrating lymphocytes (TILs). Results: The estimated EBV-infected rate of GCs was 0.113 (95% confidence interval (CI): 0.088–0.143). The EBV infection rates in GC cells were 0.138 (95% CI: 0.096–0.194), 0.103 (95% CI: 0.077–0.137), 0.080 (95% CI: 0.061–0.106), and 0.042 (95% CI: 0.016–0.106) in the population of Asia, America, Europe, and Africa, respectively. There was a significant difference between EBV-infected and noninfected GCs in the male: female ratio, but not other clinicopathological characteristics. EBV infection rates were higher in GC with lymphoid stroma (0.573, 95% CI: 0.428–0.706) than other histologic types of GCs. There were significant differences in high AT-rich interactive domain-containing protein 1A (ARID1A) and PD-L1 expressions, and high CD8+ TILs between EBV-infected and noninfected GCs. Conclusions: Our results showed that EBV infection of GCs was frequently found in male patients and GCs with lymphoid stroma. EBV infection was significantly correlated with ARID1A and PD-L1 expressions and CD8+ TILs in GCs.


Introduction
The Epstein-Barr virus (EBV) is a ubiquitous human herpesvirus associated with several lymphoid and epithelial malignancies, including Burkitt's lymphoma, Hodgkin's lymphoma, nasal NK/T cell lymphoma, and a subset of gastric carcinomas (GCs) [1][2][3][4][5][6]. In 1990, Burke et al. first detected the EBV genomes in a small group of GCs using a polymerase chain reaction [1]. Shibata et al. demonstrated that EBV genomes were uniformly present in GC cells, resembling lymphoepithelioma cells [4]. After that, EBV involvement was detected not only in lymphoepithelioma-like GCs but also in a subset of ordinary GCs [4,7].

Data Extraction
Data associated with clinicopathological characteristics based on EBV positivity in GCs were extracted from each of the eligible studies . Two independent authors obtained all the data. The data extracted were the author's information, study location, number of patients analyzed, EBV-positive rates, and clinicopathological characteristics by EBV infection. Additional information on immunohistochemical stains is shown in Table S2.

Statistical Analyses
The meta-analysis was performed using the Comprehensive Meta-Analysis software package 2.0 (Biostat, Englewood, NJ, USA). The EBV positivity rate was investigated in GCs. In addition, a subgroup analysis based on study location and histologic subtypes of GCs was performed. The correlations between EBV infection and clinicopathological characteristics were evaluated in GCs. In the present study, the following were included in the evaluated clinicopathological characteristics: age, sex, tumor size, tumor differentiation, histologic type, lymphatic, vascular, and perineural invasions, lymph node metastasis, and pTNM stages. Furthermore, the correlations between EBV positivity and p53, ARID1A, human epidermal growth factor receptor (HER2), and PD-L1 expressions, tumor-infiltrating lymphocytes (TILs), and microsatellite instability (MSI) in GCs were analyzed. We checked the heterogeneity between the studies by Q and I 2 statistics, expressed as p-values. Additionally, we conducted a sensitivity analysis to assess the heterogeneity of the eligible studies and the impact of each study on the combined effects. In the meta-analysis, as the eligible studies used various populations, a random-effect model (rather than a fixed-effect model) was determined to be more suitable. The statistical difference between subgroups was evaluated by a metaregression test. We used Begg's funnel plot and Egger's test to assess the publication bias; if significant publication bias was found, the fail-safe N and trim-fill tests were additionally used to confirm the degree of publication bias. The results were considered statistically significant at p < 0.05.

Selection and Characteristics of the Studies
In this study, 1301 relevant articles were found from the PubMed database and reviewed for a meta-analysis. Of these, 405 articles had no or a lack of sufficient information for the meta-analysis. A further 346 were excluded due to nonoriginal articles. Among the remaining articles, 489 reports were excluded for the following reasons: nonhuman studies (n = 238), articles on other diseases (n = 185), in a language other than English (n = 40), and duplication (n = 26); see Figure 1. Finally, 61 eligible articles were selected and included for the meta-analysis (Table 1). These studies included 19,747 GC patients with and without EBV infection (2063 and 17,684, respectively).
Medicina 2020, 56, x FOR PEER REVIEW 3 of 13 significant publication bias was found, the fail-safe N and trim-fill tests were additionally used to confirm the degree of publication bias. The results were considered statistically significant at p < 0.05.

Selection and Characteristics of the Studies
In this study, 1301 relevant articles were found from the PubMed database and reviewed for a meta-analysis. Of these, 405 articles had no or a lack of sufficient information for the meta-analysis. A further 346 were excluded due to nonoriginal articles. Among the remaining articles, 489 reports were excluded for the following reasons: nonhuman studies (n = 238), articles on other diseases (n = 185), in a language other than English (n = 40), and duplication (n = 26); see Figure 1. Finally, 61 eligible articles were selected and included for the meta-analysis (Table 1). These studies included 19,747 GC patients with and without EBV infection (2,063 and 17,684, respectively).

Epstein-Barr virus (EBV) Infected Rates of Gastric Carcinomas (GCs)
First, we investigated and analyzed the EBV-positive rates of GCs. The estimated EBV-positive rate was 0.113 (95% CI: 0.088-0.143) in overall GC cases. In the subgroup analysis based on study location, the EBV infected rate was the highest in Asia, compared to that in other regions. The EBV infected rate in the Asia region was 0.138 (95% CI: 0.096-0.194). In other areas, the EBV infected rates were 0.103, 0.080, and 0.042 in America, Europe, and Africa, respectively ( Table 2).

Correlations Between Epstein-Barr virus (EBV) Infection and Clinicopathological Characteristics in Gastric Carcinomas (GCs)
The clinicopathological characteristics, according to EBV positivity, were investigated in GCs. The male patients showed a significantly higher estimation rate in the EBV-positive group than in the EBV-negative group (0.824 vs. 0.639; p < 0.001 in a metaregression test). Other clinicopathological characteristics, including age, tumor size, tumor differentiation, lymphatic, vascular, and perineural invasions, pT stage, lymph node metastasis, and pTNM stage, had no significant differences between EBV-infected and noninfected GCs (Table 3). Next, the EBV-positive rates by histologic type of GC were investigated ( Table 4). The EBV-positive rate of GC with lymphoid stroma was 0.573 (95% CI: 0.428-0.706). This GC with lymphoid stroma showed higher EBV-positive rates compared to other tumor subtypes such as tubular adenocarcinoma (0.174), poorly cohesive carcinoma (0.078), papillary carcinoma (0.022), mucinous carcinoma (0.053), and undifferentiated carcinoma (0.111).

Discussion
In other epithelial malignancies, the prevalence of EBV positivity was found to be 26.37%, 33.44%, and 45.37% in breast, cervical, and oral squamous cell carcinomas, respectively [68][69][70]. The range of EBV positivity reported was variable in GC tissues . However, Chen et al. reported that non-neoplastic gastric tissue did not detect EBV positivity [71]. A TCGA study stated that the incidence of EBVaGCs was 9% [2]. Previous meta-analyses have reported the range as 2-20% and 6-33% [72,73]. In addition, the clinicopathological features of EBV positivity in GCs were variable, according to reports [72,73]. Therefore, the impact of variable EBV positivity on the controversy of clinicopathological implications of EBV in GCs needs to be elucidated. The present study includes a detailed meta-analysis of the clinicopathological implications of EBV positivity in GCs.
In the present study, the estimated EBV positive rate was 11.3%. EBV positive rates ranged from 1.2% to 89.2% in the individual eligible studies . In previous meta-analyses, EBV positive rates have been reported as 7.5% and 12.6% in 2010 and 2019, respectively [74,75]. Various factors, including the eligible studies, may have affected the differences of EBV positivity between meta-analyses. In Murphy's report, a subgroup analysis based on study location was performed, and the estimated EBV positive rates in America, Asia, and Europe were 9.88%, 8.28%, and 8.70%, respectively [72]. In the current study, the positive rate was highest in Asia at 13.8%. However, there were no significant differences between study locations in the metaregression test. Lee et al. reported that locations with a high prevalence of GCs had low EBV positivity [76]. They showed only odds ratios according to study locations, but not the estimated rates. As the criteria of the odds ratio were not described, interpretation of the odds ratio was not possible. They described that the EBV-positive rate of Asians was 8.4% through simple estimation using the raw data of each study. A meta-analysis did not obtain this result. Moreover, the estimated EBV-positive rates of Caucasian and Hispanic patients did not differ from Asians. In another meta-analysis, there was no significant difference in EBV-positive rates between study locations [75].
In addition, EBV positivity rates can differ according to the histologic type of GC. The highest EBV-positive rate was found in GC with lymphoid stroma at 57.3%. The implications of study location and ethnicity on EBV positivity may be less important when compared to the subtype of GC. Furthermore, the impact of studied years can contribute to varying EBV-positive rates. Additionally, we investigated EBV positivity in tubular adenocarcinoma according to study years. Based on 2017 data, EBV-positive rates were 0.113 (95% CI: 0.063-0.195) and 0.375 (95% CI: 0.132-0.703) after 2017 and before 2017, respectively, with a significant difference between subgroups (p = 0.012 in a metaregression test; data not shown). The possible causes are different methodologies and different histologic subtypes of the included cases. The cellular component can affect EBV positivity. In GCs, TILs can show EBV positivity [71]. Of course, the use of a PCR method with microdissection is possible for a more detailed examination; however, this limitation cannot be solved by microdissection due to intratumoral and peritumoral lymphocytes. Although PCR methods are more sensitive than in situ hybridization (ISH) methods, EBV positivity should be elucidated by evaluating cellular fractions, such as in ISH [71]. However, a definitive cause for the difference of EBV positivity by study years could not be found.
In previous studies, EBV positivity has been significantly correlated with some clinicopathological characteristics, sex, and tumor location [22,26,53]. In the present study, there was a significant correlation between EBV positivity and the patient's sex; however, EBV positivity was not correlated with lymphovascular invasion or pTNM stage. The clinicopathological significance of EBV infection is different by reports [24,25,[74][75][76]. Huang et al. reported that EBV infection in GCs was correlated with high pTNM stages and lymphatic tumor invasion, as opposed to our results [24,25]. Lee et al. reported that EBV positivity was higher in younger patients than in older patients [76]. Li et al. reported a correlation between EBV positivity and lymph node metastasis [74]. However, other meta-analyses showed no correlation between EBV positivity and lymph node metastasis, in agreement with our result [75,76]. For the evaluation of correlation with lymph node metastasis, Li's meta-analysis and our meta-analysis included 5 and 40 datasets, respectively. Moreover, they analyzed their data using odds ratios, unlike our analysis. These discrepancies could be involved in the difference of results between the meta-analyses.
Although the molecular characteristics of GCs have been studied [2], previous meta-analyses have not dealt with their correlation with various molecular markers [75]. In our results, CD8+ TILs and PD-L1 expressions of the tumor and immune cells were more frequently found in EBVaGCs than in non-EBVaGCs. Abundant TILs are one of the histologic features in GCs with EBV infection [77][78][79]. In the TCGA report, PD-L1 gene amplification was elevated in EBVaGCs [2]. Furthermore, PD-L1 immunohistochemical expression in tumor cells was more frequently found in EBVaGCs than in non-EBVaGCs [28]. However, the impact of TILs in GCs is not yet fully understood. In addition, further evaluation of the tumor-infiltrating and peritumoral lymphocytes will be needed in GC with lymphoid stroma, which was significantly associated with high EBV positivity. In GCLS, EBV-positive tumors had more PI3K/AKT pathway mutations than EBV-negative tumors [80]. In addition, because EBVaGCs are significantly correlated with high TILs, new immunotherapeutic strategies associated with T-cells are challenging for the treatment of advanced EBVaGCs [81,82]. ARID1A expression was higher in EBVaGCs than in non-EBVaGCs. In the previous meta-analysis, correlations between EBV positivity and molecular markers, such as p53 and CpG island methylator phenotype, were found [76].
This study has some limitations. First, a subgroup analysis based on EBV detection methods could not be performed due to the methods used in the eligible studies. Second, the impact of study years on EBV positivity could not be fully investigated based on subtypes of GCs. We evaluated only tubular adenocarcinomas among the various GC subtypes. Third, the eligible studies used different antibodies and evaluation criteria for immunohistochemistry. However, subgroup analysis based on antibody and evaluation criteria could not be performed due to insufficient information.

Conclusions
Taken together, our results show that the EBV positivity of GCs is frequently found in male patients and GC with lymphoid stroma. Although EBV positivity was highest in Asians, there was no significant difference between study locations. EBV positivity is significantly correlated with ARID1A and PD-L1 expressions, as well as CD8+ TILs in GCs.

Conflicts of Interest:
The authors declare no conflict of interest.