Prognostic and Diagnostic Significance of circRNA Expression in Esophageal Cancer: A Meta-analysis

Background and Aims Circular RNA (circRNA) demonstrates potential biological application in various solid tumors. We intended to evaluate the diagnostic, prognostic, and clinicopathological value of circRNA for esophageal cancer (EC). Methods We screened relative studies from Pubmed, Embase, Web of Science, and Cochrane Library. The diagnostic role of circRNAs was testified by pooled sensitivity and specificity. Pooled odds ratio (OR) and pooled hazard ratio (HR) were computed to appraise the clinicopathological and prognostic value, respectively. Results There were total 15 articles suitable with our included criteria, in which 7 for diagnosis, 8 for prognosis, and 9 for clinicopathological features. The pooled sensitivity and specificity were 0.77 and 0.80, respectively, while the AUC was 0.85. Patients with aberrant expression of circRNAs had a 2.92-fold increased risk of developing EC. The proportion of EC patients with normal circRNA expression only accounted for 29%. Upregulated expression of oncogenic circRNA was correlated with poor clinicopathological features, including lymph node metastasis, tumor size, and T classification, while downregulation of tumor-suppressor circRNA was contributed to worse TNM stage. As for prognosis, upregulated expression of circRNA carried out a diverse survival outcome, with a pooled HR of 2.76 for tumor promoter and that of 0.21 for tumor suppressor. High expression of oncogenic circRNA in both plasma and tumor tissue would lead to a shorter survival duration. Conclusion circRNAs might be a promising biomarker for diagnosis, prognosis, and clinicopathological features of EC.

With the development of sequencing technology, circRNA was found to be of remarkable importance in various dis-eases, including cardiovascular disease, diabetes, Alzheimer's disease, and cancer [5,7,9,10]. Due to its high conservatism and stability, circRNA might become a promising biomarker in cancer diagnosis and therapy.
Esophageal cancer (EC) is a common malignancy of the digestive tract with poor prognosis. In all malignancies, EC ranks among the top ten both in morbidity and mortality [11]. Adenocarcinoma and squamous cell carcinoma are two most frequent histologic types of EC. The former occurs mainly in developed countries, and the latter occurs mostly in Eastern Asia [12,13]. Regrettably, EC is often diagnosed at middle-advanced stage, which results in the omitting of optimal therapeutic opportunity for patients. The primary reasons include the delayed emergence of initial symptoms, discomfort caused by endoscopy, and the nonspecific and insensitive tumor markers [14]. Recently, the role of epigenetics in esophageal cancer is gradually being discovered. The occurrence and progression of malignant tumors are usually first accompanied by changes in the microenvironment and signaling pathways, such as the regulation of the vascular network of EC by miR-126 and miR-377 [15]. Many studies discovered abnormal expression of circRNA in EC, which may provide crucial reference for diagnosis and treatment. We incorporated relevant studies for a meta-analysis, in order to summarize the correlation between circRNA expression and diagnosis, prognosis, and clinical characteristics of EC.

Search Strategy.
Our research was carried out on the basis of the preferred reporting items for systematic reviews and meta-analyses (PRISMA) checklist (Supplementary file (available here)) [16]. We searched for studies from four online databases, including Pubmed, Embase, Web of Science, and Cochrane Library, by using the following terms: (1) ("esophageal carcinoma" or "esophageal cancer" or "esophageal tumor" or "esophageal neoplasm" and (2) ("circular RNA" or "circRNA"). The deadline for searching was June 17th, 2020. Two researchers (HL and JPY) evaluated Articles for clinicopathological feature (n = 9) Articles for prognosis (n = 8) Articles for diagnosis (n = 7) Identification Screening Eligibility Included Figure 1: The flowchart of research selection.

Study Selection.
Studies that met the following eligibility were included into our meta-analysis: (1) patients were diagnosed as EC by positive histology, (2) the studies were performed to estimate the diagnostic or prognostic  3 Gastroenterology Research and Practice efficiency of circRNA for EC or to identify the relationship between the expression of circRNA and clinicopathologic features, and (3) cohort or case-control researches. The excluded criteria were listed as the following: (1) articles that were not published in English; (2) review, meta-analysis, letter, and animal studies; and (3) with incomplete information.

Data Extraction and Quality Assessment.
Two researchers (HL and JPY) extracted the following information from each study independently: (1) first author, country, edition year, cancer and circRNA type, the number of samples, sample species, experimental method, and regulated signature of circRNA; (2) the follow-up duration of EC patients; (3) diagnostic specificity and sensitivity, the area under the receiver operating characteristic (ROC) curve (AUC), the value of true positive (TP), false negative (FN), true negative (TN), and false positive (FP); and (4) clinicopathological features including age, gender, smoking, drinking, TNM stage, T classification, lymph node metastasis, distant metastasis, tumor size, and differentiation. If the parameter of TP, TN, FP, and FN was not offered, we assessed it according to sample size, specificity, sensitivity, and AUC.
Two independent researchers (HL and JPY) performed quality assessment of the included studies by using the Newcastle-Ottawa Scale (NOS) [17]. A score no less than 6 was conferred with high quality for a study.

Statistical
Analysis. Stata 15.0 was utilized to develop related statistical analysis. By combining the number of TP, TN, FP, and FN, the pooled specificity, sensitivity, diagnostic odds ratio (DOR), and negative and positive likelihood ratio (NLR and PLR) were calculated. Summary receiver operator characteristic (sROC) curve with AUC (the area under sROC) was plotted to evaluate the diagnostic value of circRNA. Pooled odds ratios (ORs) with 95% confidence intervals (CI) were utilized to assess the relationship between the expression of circRNA and clinicopathologic features. In addition, we estimated the prognostic value of circRNA for overall survival (OS) via using pooled hazard ratios (HRs). Subgroup analysis was performed to determine whether the aberrant expression of circRNA in plasma or tumor tissue had an impact on prognosis. I 2 value and chi-squared test were used to evaluate heterogeneity. A <50% I 2 value or a <0.10 p value was considered of no conspicuous heterogeneity, so a Wang (1)  Wang (2)   Gastroenterology Research and Practice fixed-effect model was applicable. Otherwise, a randomeffect model should be adopted [18]. Potential source of heterogeneity was investigated via sensitivity analyses. In addition, funnel plots and Begg and Egger's tests were established to estimate publication bias.

Search
Results. The flowchart of study selection was plotted in Figure 1. A total of 242 articles were retrieved from online databases, in which 15 were suitable for being incorporated in the meta-analysis. There were seven [19][20][21][22][23][24][25]  As for prognostic analysis, species included both plasma and tumor sample. What is more, the involved studies were of high quality (Table 3).  (Figure 4). Moreover, the pooled PLR was 3.92 (95% CI 2.51-6.12), and pooled DLR was 0.29 (95% CI 0.22-0.37) ( Figure 5). Aforementioned outcomes demonstrated that circRNA could be a precise biomarker for EC diagnosis. Table 4 reveals the relation between clinicopathological features and circRNA. High expression of tumor-promoter circRNA was contributed to poor

Overall Survival (OS).
With no significant heterogeneity (I 2 = 0%), fixed-effect models were applied to estimate the role of circRNA in OS prognosis. Upregulated tumorpromoter circRNA was correlated with worse OS (HR 2.76, 95% CI 2.09-3.63, Figure 6(a)) for EC patients. Oppositely, upregulation of tumor-suppressor circRNA notably carried out more favorable OS probability (HR 0.21, 95% CI 0.08-0.57, Figure 6(b)). Furthermore, for tumor-promoter cir-cRNA, subgroup analysis declared that the high expression both in plasma (HR 2.52, 95% CI 1.56-4.09) and tissue (HR 2.88, 95% CI 2.06-4.02) carried out worse prognosis (Figure 7). Figure 8 demonstrated that there was no publi-cation bias in our meta-analysis. We also performed further qualitative analysis by using Begg's test and Egger's test to evaluate the publication bias, and the results supported the conclusion that there was no publication bias (Begg's test: p = 0:076; Egger's test: p = 0:107; Figures 9 and 10). Additionally, sensitivity analysis showed that the outcomes of metaanalysis were invariable when removed the studies one by one, which concluded that the pooled outcomes were stable ( Figure 11). What is more, no evidence of publication bias was implied by developing Deeks' funnel plot asymmetry test (p = 0:44; Figure 12).

Discussion
circRNA might be a novel tumor biomarker. Its predictive value in diagnosis and prognosis for malignancy has been gradually explored. Several circRNAs have been certified to be associated with the development and progression of various tumors, such as ciRs-7 [34]. The predictive role of circRNA in different malignancies, including lung cancer, colorectal cancer, and laryngeal cancer, has also been reported recently [35][36][37]. Niu et al. [38] conducted a meta-analysis to investigate the diagnostic role of circRNA in EC, which illustrated that circRNA had a favorable biological value for EC diagnosis. However, the number of studies included was small, and the relation between cir-cRNA expression and prognosis or clinicopathological characteristics was not investigated. To our knowledge, this is the first meta-analysis involving the relationship between circRNA expression and diagnosis, prognosis, and clinicopathological characteristics of EC. In our analysis, sensitivity and specificity of circRNA for diagnosis were 0.77 and 0.80, respectively, and the AUC was 0.85. In addition, the overall DOR was 13.71, while incorporated PLR and NLR were 3.92 and 0.29, respectively. In other words, patients with the aberrant expression of cir-cRNA were 3.92 times more likely to develop EC compared with the general population, and the proportion of patients with the normal circRNA expression only accounted for 29%. Upregulation of oncogenic circRNA was obviously associated with lymph node metastasis, tumor size, and T classification. Upregulation of downregulation of tumor-suppressor circRNA contributed to poor TNM stage. As for prognostic value, the abnormal expression of circRNA was closely associated with poor OS. Of course, just as the downregulation of miR-20b, miR-27a, and miR-181a leads to the upregulation of drug-resistant genes in gastric cancer to affect the sensitive of chemotherapy, we also expect circRNA to serve for the precise and individualized treatment of EC [39]. This is an attractive challenge that requires more clinical trials.
We investigated the diagnostic value of circRNA for EC. Due to the anomalous expression of circRNA in plasma, it is easy to obtain samples for testing when a person was suspected of suffering from EC. Meanwhile, stable structures and conservative sequences guarantee that cir-cRNA is not prone to denature. The expression of cir-cRNA from preoperative plasma or postoperative tumor tissue is a powerful supplement to the assessment of    Gastroenterology Research and Practice patient's prognosis. Our meta-analysis comprised 15 studies involving 1032 patients, which strongly manifested the function of circRNA in diagnosis, prognosis, and clinicopathological relevance for EC. It is expected that more investigations will be performed to further confirm our results, especially on tumor-suppressor circRNA.
Our analysis was developed based on PRISMA guidelines strictly and was accomplished by independent researchers utilizing appropriate retrieval strategies. We screened the studies in compliance with the rigorous inclusion and exclusion criteria. For statistical analyses, we applied precise and appropriate statistical methods, and the statistical outcomes were analyzed and interpreted carefully. Nevertheless, there were still some limitations in our study. First, the number of studies included was relatively small, especially the studies on tumor-suppressor circRNA. In order to further ascertain Wang (2)     Heterogeneity between groups: p = .
Overall (I-squared = 0.0%, p = 0.681) the results, more studies are necessary to perform in the future. Second, all studies were from China, indicating that studies on other races were needed. In addition, some studies did not provide clear sensitivity, specificity, or HR. We extracted indispensable data from supplied ROC curves and KM curves, which may lead to potential bias. Finally, we analyzed the prognostic role of circRNA by using HR, which was provided in each research via univariate analysis. The tests performed may be statistically significant but biologically less relevant if placed into a more complex context. As a result, an HR obtained from multivariate will be more credible. The prognostic role of circRNA after adjusting for other prognostic factors remains to be further explored.

Conclusion
In summary, our meta-analysis declared that the expression of circRNA in plasma had a certain value in the   Figure 7: Subgroup analysis for verifying the relationship between tumor-promoter circRNA expression and overall survival prognosis. circRNAs, circular RNAs.   Upper CI limit Estimate Meta-analysis estimated, given named study is omitted Figure 11: Sensitivity analysis of circRNAs for esophageal cancer. differential diagnosis of EC. Meanwhile, the aberrant expression of circRNA both in malignancy tissue and plasma indicated worse prognosis. circRNA might be a promising biomarker, and further researches are needed to verify its role in EC.

Data Availability
The data used to support the findings of this study are included within the article.

Conflicts of Interest
The authors declare that they have no conflict of interest.