Abnormally Expressed Circular RNAs as Novel NonInvasive Biomarkers for Hepatocellular Carcinoma: A Pair-wise Meta-Analysis

Background: In the recent literature, dysregulated circular RNAs (circRNAs) have been extensively investigated in hepatocellular carcinoma (HCC). This study strives to evaluate the diagnostic as well as the predictive value of abnormally expressed circRNAs in HCC. Methods: Eligible studies were sourced from PubMed, EMBASE, and CNKI online databases. Data on patients’ clinical characteristics, including diagnostic efficacy and overall survival (OS) were extracted. The diagnostic and prognostic parameters were respectively synthesized using the bivariate meta-analysis model and multivariate Cox hazard regression analysis based on STATA 12.0. The trim and fill approach was employed to evaluate the impacts of publication bias. Results: A sum of 21 eligible types of research was incorporated. The pooled sensitivity, specificity and area under the curve (AUC) of abnormally expressed circRNAs in distinguishing HCC from non-cancer controls were estimated to be 0.78 (95% confidence interval (CI): 0.69–0.85), 0.80 (95% CI: 0.74–0.86) and 0.86, respectively. Survival analyses expressed that the down-regulated circRNA expression signature correlated perfectly with HCC survival (hazard ratio (HR) = 0.42, 95% CI: 0.19–0.91, p = 0.028; I2 = 92.7%; p = 0.000), whereas the HCC cases with high circRNA levels had significantly poorer prognoses than those of patients with low circRNA levels (HR = 2.22, 95% CI: 1.50–3.30, p = 0.000; I2 = 91%; p = 0.000). Moreover, abnormally expressed circRNAs were intimately linked with tumor size, differentiation grade, microvascular invasion, metastasis, TNM stage, and serum AFP level in patients with HCC. Stratified analysis based on sample type, control source, and expression status also yielded robust results. Conclusions: Abnormally expressed circRNA signatures show immense potential as novel non-invasive biomarker(s) in complementing HCC diagnosis and prognosis.

3 Background Hepatocellular carcinoma (HCC), which is a prevalent digestive system cancer, continues to be the prime cause of cancer deaths worldwide [1]. In China, the incidence of HCC was shown to have increased remarkably over the past decades, which has resulted in great health and economic burdens worldwide [2]. Although the technological advances for HCC treatment in recent years have vastly improved the clinical outcomes of patients with GC, the 5-year survival rate is very low [3]. It was discovered that for individuals with HCC, the average life span is 3-9 months [4]. The sensitivity and specificity of the currently used blood biomarkers such as alpha fetal protein (AFP), Protein Induced by Vitamin K Absence or Antagonist-II (PIVKA-II), and alpha-fetoprotein heterogeneity are not satisfactory for HCC detection [5]. For prognosis monitoring, no biomarkers were well developed. Thus, it was essential for us to develop and examine the novel efficient biomarkers for HCC.
Non-coding RNAs perform crucial roles in cancer biology, providing objectives for cancer intervention. As a new class of endogenous noncoding RNAs, circular RNAs (circRNAs) are a series of functional non-coding transcripts initiated from either joining exons, introns or both. [6]. Unlike linear RNAs, circRNAs actualize covalently closed continuous loop structures, characterized by stability, abundance and specific expression in different tissues as well as cells during maturation [6,7]. CircRNAs act as key regulators in a broad scope of biological processes, including the initiation and progression of several types of cancer [8,9]. CircRNAs are aberrantly expressed in cancer tissues, especially in HCC, suggesting that these molecules could be novel biomarkers for HCC diagnosis and prognosis [10-30]. Whether or not circRNAs are of clinical value for the diagnosis of HCC must be clarified. Herein, we carried out this meta-analysis, aiming to evaluate the diagnostic and prognostic utilities of circRNAs expression signature in HCC.

Selection Criteria
Studies were in compliance with the following criteria: (1) Studies were limited to those which evaluated the diagnostic or prognostic or clinicopathological features of circRNA(s) in HCC patients; (2) the TP (true positive), FP (false positive), FN (false negative), and TN (true negative) values for diagnosis, or estimated HR (hazard ratio) values with 95% CIs for survival, were either available among studies or could be extracted indirectly; (3) cases were definitely diagnosed with pathological evidence; and (4) the specimens were obtained prior to any radiotherapy or chemotherapy treatments. Irrelevant papers were excluded according to the following criteria: (1) Studies with insufficient data to form the 2 × 2 table for diagnosis, or the HRs with 95% CIs for survival were unavailable; (2) studies were rated as low quality; and (3) basic studies, reviews, meta-analyses, comments, letters or case reports, etc. were also excluded.

Data Extraction
The baseline contents were collected independently by two trained authors . The basic   information covered comprised facts like the author's first name, year of publication,   design of the research, ethnicity, sample capacity, pathologic data of the population   study, circRNA signature, test methods, sensitivity, specificity, cut-off value setting, HR values with 95% CIs for survival, follow-up time, etc. Any disagreements which appeared during data summarization were resolved by group consensus, or the articles' authors were reached out to.

Study Quality Grading
The quality of the studies for diagnosis was assessed by the Quality Assessment for Studies of Diagnostic Accuracy II (QUADAS II) checklist [31]. The tool comprised of two domains that included "risk of bias" and "applicability concerns", which contained 7 queries concerning patient selection, index tests, reference standards, flow, and timing.
The answer of risk for bias could be rated as "no" (0 score), "yes" (1 score), or "unclear" (0 score). The study quality for the case-control study was judged in line with the Newcastle-Ottawa Quality Assessment Scale (NOS) checklist [32], in which the assessment focuses on a total of 8 items categorized in terms of study selection, comparability, and outcome, with a maximum judgment score of "9". An answer of "yes" receives a score of "1"; otherwise, no sores were awarded.

Statistical Analysis
Statistical evaluation was carried out established on the Stata 12.0 program (Stata Corporation, College Station, TX, USA). Heterogeneity among studies was assessed by employing Chi 2 (Chi-square) and Inconsistency I 2 (I-square) tests. Either p < 0.05 in the Chi 2 test or I 2 > 50% were both regarded as studies with significant heterogeneity. The diagnostic parameters were synthesized using the bivariate meta-analysis model, and HRs with 95% CIs were combined using multivariate Cox hazard regression analysis. The random effect model was chosen when significant heterogeneity appeared in the pooled effect size. Sensitivity analysis was performed to trace the underlying outlier studies included in the pooled effects. Due to publication, bias was recognized by Deek's funnel plot, Begg's and Egger's tests, and P <0.05 was situated to denote statistically significant differences. If at any instance bias was noticed, the trim and fill approaches were employed to evaluate the impacts of bias on the overall joint effects [33].

Methodological Quality Assessment
The quality and bias of all diagnostic studies were independently appraised by two authors in compliance with the QUADAS-II criteria, whereby studies were assessed for patient selection, index test, reference standard, flow and timing [31]. As reported by Figure 2, all included 8 publications for diagnosis were judged as low risk for applicability concerns, and 3 studies were assessed with bias in patient selection, or index test, or reference standard, and received rated QUADAS scores equal to 3 points. Evaluation of the quality of all case-control studies was enabled by applying the NOS checklist [32]. As shown in Table   3, all the included prognostic studies received rated NOS scores higher or equal to 6, and thus they were all included in the final synthesis.

Investigations of Heterogeneity
In the overall diagnostic meta-analysis, the Chi 2 and I 2 tests revealed significant substantial heterogeneity among pooled effects (Q = 49.403, df = 2.00, p = 0.000; I 2 = 95.95, 95% CI: 92.85-99.05). In line with the diagnostic effects, clear heterogeneity was also observed in the pooled prognostic effects for both the elevated (p = 0.000; I 2 = 91%) and down-regulated circRNA profiles (p = 0.000; I 2 = 92.7%). Thus, all weights were synthesized using a random effect model.

Overall Diagnostic Performance
The summary receiver-operating characteristic (SROC) curve was employed to assess the diagnostic efficacy of circRNA profiling in distinguishing HCC from non-tumorous controls.
The pooled sensitivity ( Figure

Clinicopathological Association
Evaluation of the link between circRNA expression and clinicopathological elements in HCC also produced robust results. As shown in Table 4 (Table 5).

Sensitivity Analysis
Influence analysis was performed in both the diagnostic and prognostic effect sizes. As exemplified by Figure 5, one study [30] was identified as the outlier in the pooled prognostic effects of down-regulated circRNAs in HCC. After elimination of the outlier data and re-analysis of the effect, the I 2 dropped from 92.3% to 90%, indicating that included heterogeneous studies were a substantial cause of study heterogeneity. No outliers were detected in other pooled effects ( Figure 5).

Publication Bias
Publication bias was judged using different methods for different pooled effects. As shown in Figure 6A, no clear publication bias was detected in the combined diagnostic effects (Deek's funnel plot, p = 0.446), nor in the analysis of down-regulated circRNA profiling (Egger's test, p = 0.606, Figure 6B). Nevertheless, the funnel plot expressed evidence of a publication bias in the effects of up-regulated circRNA profiling (Egger's test, p = 0.001, Figure 6C), and the trim and fill technique was employed to detect the outcome of bias [33]. As indicated in Figure 6D, the filled funnel plots identified 5 imputed studies, but the effect was slightly altered before and after adjustment (variance = 0.187, p = 0.005 vs. variance = 0.287, p = 0.000).

Discussion
As stated earlier, hepatocellular carcinoma, (HCC) is among the causes of cancer death in digestive system tumors [1][2][3]. There are numerous reviewed biomarkers for HCC, which include AFP, PIVKA-II, and the ratio of lens culinaris agglutinin-reactive alpha-fetoprotein to total AFP (AFP-L3/AFP) [34]. However, these biomarkers retain several limitations on their overall diagnostic efficacies [5,34]. In this respect, ideal noninvasive biomarkers are Our study still retains many limitations. The overriding problem is the substantial heterogeneity which appeared among studies. The sensitivity analysis identified one study [30] as the outlier in the pooled prognostic effects of down-regulated circRNAs in HCC. Our analysis further confirmed the impacts of heterogeneous studies on the generation of heterogeneity among combined effects. Additionally, biases from publications appeared in one of our pooled prognostic analyses. Nevertheless, our further assessment through the use of a nonparametric trim and fill approach affirmed that the joint precision is not subjected to the unprinted negative studies. Consequently, the accuracy of all the pooled effects was shown to be relatively reliable.
In summary, our study shows evidence that abnormally expressed circRNAs may perform a critical role in HCC progression and could serve as diagnostic and prognostic biomarkers for cases of HCC. Future in-depth research is required to further evaluate the utilities of