Prognostic and clinicopathological value of circPVT1 in human cancers: A meta‐analysis

Abstract Background Circular RNA PVT1 (circPVT1) is significantly upregulated in various human cancers and is related to poor clinical outcome of cancer patients. However, the prognostic and clinicopathological value of circPVT1 in diverse human cancers remains controversial and inconclusive. Aim The objective of our study is to evaluate the prognostic and clinicopathological role of circPVT1 for cancer patients. Methods and results PubMed, Embase, Web of Science, and Cochrane Library were searched for eligible studies by October 1, 2020. The correlation between circPVT1 expression, and overall survival (OS) and clinical parameters was assessed by pooled hazard ratios (HRs) and odds ratios (ORs) with 95% confidence intervals (CIs). Subgroup analyses, heterogeneity, and publication bias were conducted to further enhance reliability. Twelve studies (1282 patients) were selected for this meta‐analysis, including 11 on prognosis and 10 on clinicopathological parameters. Elevated expression of circPVT1 was associated with a worse OS in cancer patients (HR, 2.009; 95% CI, 1.667‐2.408, 1.892; P < .001). For clinicopathological value, upregulation of circPVT1 was closely related to poor clinical parameters lymph node metastasis (OR = 2.019; 95% CI, 1.026‐3.976; P = .042; I 2 = 77.5%; PH = 0.000), late clinical stage (OR = 3.594; 95% CI, 1.828‐7.065; P < .001; I 2 = 71.7%; PH = 0.001), distant metastasis (OR = 4.598; 95% CI, 1.411‐14.988; P = .011; I 2 = 78.1%; PH = 0.001), and chemoresistant (OR = 6.400; 95% CI, 2.107‐19.441; P = .001; I 2 = 49.6%; PH = 0.159). Conclusion High expression of circPVT1 is correlated with unfavorable prognosis of cancer patients, indicating that circPVT1 can function as a potential prognostic biomarker in human cancer.

With the development of experimental technologies, such as highthroughput RNA sequencing, an increasing number of circRNAs has been detected in the past decades. [4][5][6] Notably, numerous works have indicated that circRNAs play pivotal roles in various physiological processes and diverse human diseases, including human cancer. [7][8][9] It has been found that a great number of circRNAs are dysregulated in various human cancers and play an important role in tumor progression. 10 Furthermore, the promising roles of several circRNAs as prognostic and diagnostic biomarkers in human cancer have been reported in recent studies. For example, it has been found that circHIPK3 can serve as a diagnostic biomarker for osteosarcoma, and downregulation of circHIPK3 is associated with poor OS of osteosarcoma patients. 11 Circular RNA PVT1 (circPVT1), also named hsa_circ_0001821, is located on chromosome 8q24 and is derived from the PVT1 gene locus. 12 Very recent works have proved that circPVT1 is evidently upregulated in multiple cancer types and serves as an oncogenic noncoding RNA during cancer development. CircPVT1 has been found to promote proliferation, metastasis, and chemoresistance in several human cancers. 13 For example, circPVT1 can facilitate cell proliferation and inhibit apoptosis via sponging miR-497 in non-small cell lung cancer (NSCLC) in both vivo and vitro. 14 It has also been found that circPVT1 has a great diagnostic accuracy for some cancer types, such as NSCLC, gastric cancer (GC), and oral squamous cell carcinoma (OSCC). [15][16][17][18][19] Moreover, emerging evidence has shown that the upregulation of circPVT1 is correlated with poor OS in several types of human cancer, such as lung adenocarcinoma (LAD), breast cancer (BCa), and GC. 15,20,21 Besides, the clinicopathological value of circPVT1 in human cancers has also been identified in multiple studies. 18 However, the certain prognostic and clinicopathological value of circPVT1 in human cancers is still inclusive and some studies provided contradictory outcomes. For instance, Chen et al 15  Similarly, it was found that elevated expression of circPVT1 was related to tumor size in NSCLC and hepatocellular carcinoma (HCC); in contradiction to these studies, studies of Yan et al 23 and Bian et al 20 indicated that there was no evident relationship between circPVT1 expression and tumor size. Therefore, we conducted the following meta-analysis to assess the prognostic and clinicopathological value of circPVT1 in human cancers, thus highlighting the potential role of circPVT1 as clinical biomarker and novel therapy target for cancer.

| Data search strategy
The present study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA). 24 Relevant studies published in English before October 1, 2020 were searched in PubMed, Embase, Web of Science, and Cochrane Library online databases. The key words were as follows: (a) "circular RNA PVT1"or "circPVT1" or "hsa_circ_0001821"; (b) "cancer" or "tumor" or "neoplasm" or "carcinoma" or "sarcoma".
Two researchers (ZJL and XZT) conducted the literature search independently, and reference lists were also searched for possible relevant studies. Each retrieved study was comprehensively checked and identified by the two researchers. Any discrepancies were resolved by the corresponding author (LL).

| Inclusion and exclusion criteria
Studies in accordance with the following criteria were included: (a) patients with diagnosis of human cancer; (b) studies examining the correlation between prognosis or clinicopathological features of cancer patients; (c) studies published in English; (d) cohort studies that patients divided into "high circPVT1" and "low circPVT1" group.
On the contrary, these studies were excluded: (a) duplication; (b) studies irrelevant to circPVT1 or human cancer; (c) reviews, case reports, letters, and editorials; (d) studies without available data, or studies that we could not contact the authors/investigators for further clarification or data.

| Data extraction and quality assessment
Two researchers (ZJL and LW) extracted the following data from eligible studies: first author name; publication year; country; type of cancer; sample size; detection method; cut-off value; prognostic parameters of circPVT1 including HRs, corresponding 95% CIs and P value; clinicopathological characteristics with age, gender, histological grade, tumor size, T stage, lymph node metastasis, clinical stage, distal metastasis, and chemoresistant. If HRs and 95% CIs were not directly provided, they were extracted from Kaplan-Meier curves according to Tierney's method. 25 Excel was used to record all the available information. The guideline of Newcastle-Ottawa Scale (NOS) was adopted to evaluate the quality of eligible studies. Three perspectives including selection, comparability, and exposure were considered. The NOS scores were ranged from 0 (lowest score) to 9 (highest score). A study with a score ≥ 7 was defined as a study with high quality. 26 The quality of each eligible study was evaluated by two independent researchers. Any discrepancies were resolved by the corresponding author (LL).

| Statistical analysis
Statistical analysis was performed using STATA 12.0 software. The correlation between circPVT1 expression and clinicopathological characteristics was evaluated by ORs and 95% CIs. The HR and 95% CI were directly extracted from the original publications. If the study only provided Kaplan-Meier survival curve, HR and 95% CI were synthesized by Engauge Digitizer 11.2 software (http://markummitchell. github.io/engauge-digitizer/) according to Tierney's method. 25 Higgin's I 2 statistics and Cochran's Q test were used to identify the heterogeneity of the included studies. If I 2 ≤ 50% and P > .05, a fixedeffect model test was utilized to analyze the results. Otherwise, a random-effect model test was applied for analysis. 27 HR > 1 indicates the poor prognosis for high circPVT1expression level, and P < .05 was considered to be statistically significant. Publication bias was qualitatively determined using funnel plot analysis, and quantified by Egger's test and Begg's test. Subgroup analysis and sensitivity analysis were conducted to evaluate the source of heterogeneity and stability of the results.

| Search results
One hundred and fifty-one relevant studies were retrieved after searching database key words. Eighty-nine studies were excluded based on duplication criteria. After title, abstract, and article type screening, 29 studies were further excluded. Then, based on criteria, 21 articles were excluded for: not focused on circPVT1 or cancer, not reported relevant results, not studied research studies, and no sufficient data. In summary, 12 eligible studies were included in the metaanalysis ( Figure 1).  (Table 1).

| Overall survival
The combined analysis of published data from 11 analyses indicated that upregulation of circPVT1 was an indicator for poor OS in patients with human cancer (pooled HR, 2.009; 95% CI, 1.667-2.408; P < .001). A fixed-effects model was selected since no heterogeneity was reported (I 2 = 0.0%, P H = 0.670, Figure 2).

| Subgroup meta-analyses based on sample size
In subgroups divided by sample size, the results showed that in both

| Clinicopathological parameters
The correlation between circPVT1 and the clinicopathological parameters is shown in

| Publication bias and sensitivity analysis
We conducted funnel plot analysis, Egger's, and Begg's tests to assess potential publication bias for OS. The funnel plot of OS revealed no evident asymmetry, indicative of no obvious evidence of publication bias for OS ( Figure 4). Meanwhile, Begg's test: P = .213 and Egger's test: P = .073 also indicated there was no evident publication bias in publications reported OS ( Figure 5). The sensitivity analysis was conducted to evaluate the stability of pooled results. After excluding each individual study, there was no significant change on the results, indicating the pooled result was reliable and stable ( Figure 6).

| DISCUSSIONS
CircPVT1, as an oncogenic circRNA, has been found to be upregulated in various human cancers, such as GC, CRC, BCa, and osteosarcoma. 15 indicating that targeting circPVT1 may be a promising therapeutic strategy to overcome chemoresistance and improve prognosis of cancer patients. For example, exosomal circPVT1 has been found to induce cisplatin resistance by stimulating YAP1 via sponging miR-30a-5p in GC in both vivo and vitro. 34 Via sponging miR-134p to upregulate the expression of ZEB1, circPVT1 can also promote GC cells resistance to paclitaxel. 35 In osteosarcoma, circPVT1 has been found to confer multidrug resistance by upregulating ABCB1 expression. 16 Similarly, circPVT1, which is upregulated in LAD tissues, can promote chemoresistance by inducing ABCC1 via sponging miR-145-5p in LAD A549/DR cells in vitro. 21 Furthermore, emerging evidence suggests that circPVT1 can function as a promising clinical biomarker for human cancer. The diagnostic significance of circRNA in several cancer types, including osteosarcoma, GC, and OSCC, has been validated in some studies. 16,17,36 Notably, several studies have  F I G U R E 4 Funnel plot of publication bias based on overall survival may contribute to selection bias and publication bias. Secondly, the number of studies in some major tumor types, such as bone neoplasms, was limited, which may influence the results of subgroup meta-analysis based on cancer types. Also, there was only one study researching the prognostic value of circRNA in some cancer types, such as HNSCC, PTD, and HCC, so it was difficult to identify the correlation between circPVT1 expression and clinical outcome in those cancers. Thirdly, several studies only provided the survival curve, so the indirectly extracted HRs and 95% CIs may cause some errors and lead to potential source of bias. Fourthly, the cutoff value for distinguishing high or low circPVT1 expression levels was inconsistent, which may account for potential heterogeneity. Finally, the sample size of patients in this meta-analysis was relatively small, so more studies and large samples would be necessary to better explore the correlation between circPVT1 and prognosis and clinical characteristics in patients of various cancer types.

| CONCLUSIONS
In conclusion, upregulation of circPVT1 is significantly correlated with worse OS in human cancers. This study indicates that