Meta‐analysis of the diagnostic value of exosomal miR‐21 as a biomarker for the prediction of cancer

Abstract Background Early diagnosis of cancer is still the most effective method to increase survival and therapeutically effective patient management. Accumulating studies had exploited exosomes as an indicator for the diagnosis and prognosis of cancer. In addition to exosomes, exosome‐derived miRs are widely investigated as a novel biomarker for diagnosis in cancer patients. The aim of this study was to clarify the diagnostic value of ex‐miR‐21 in cancer. Methods Databases were searched for eligible studies up to June, 2021. Studies included in this meta‐analysis were reviewed and selected independently by two authors. The data of sensitivity, specificity, diagnostic odds ratio (DOR), and summary receiver operating characteristic curves (SROC) of exosomal miR‐21 as a diagnostic biomarker were extracted and calculated. Quality assessment was conducted by using the QUADAS‐2 tool. Results A total of 26 studies were included in the systematic analysis and meta‐analysis. The pooled results of sensitivity, specificity, PLR/NLR, DOR, and area under the curve were 76% (95%CI, 0.70–0.81), 82% (0.77–0.87), 4.3 (3.1–6.0), 0.29 (0.22–0.38), 15 (8–26), and 0.86 (0.83–0.89), respectively. Sensitivity analysis and Deeks' funnel plot indicated that results remained unchanged and had no publication bias. For the subgroup analysis, it was showed that ex‐miR‐21 had a superior diagnostic accuracy on identifying PC. Conclusion Exosomal microRNA‐21 can serve as an effective and widely used diagnostic biomarker for cancer, especially in PC. The using field of exosomes and exosome‐derived miR can further extend the prognosis and therapeutic management. Standardized isolation of exosomes and miRNA‐21 should be developed.


| INTRODUC TI ON
Cancer and its high incidence have become a serious threat to human health and bring a great challenge to treatment planning and research progress. 1 Timely diagnosis and therapeutic monitoring remain one of the major challenges during the treatment of disease.
Lacking a reliable detection method, many cases of cancer were misdiagnosed or delayed diagnosed and missed the best treatment time eventually. 2 The five-year survival rate was poor in several types of cancer as the disease is hardly detectable in its early stage and is often being found out until the late stages with limited treatment options.
Accumulating studies have demonstrated the positive correlation between exosomes and the biological process of cancer, such as tumor growth, metastasis, and chemoresistance. [3][4][5] Applications of exosome-shuttled proteins and nucleic acids in cancer diagnosis and prognosis have also caught attention in the detection of several types of cancers. 6,7 Given its high stability in the extracellular environment, less interference to serum and detectable in various body liquids (plasma, serum, urine, CSF, etc.), exosomes are advantaged as a novel biomarker for tumor diagnosis.
Exosomal miRNAs, small noncoding RNAs of approximately 18-25 nucleotides length, are one of the study subjects with the significant expression level of RNAs and are suggested as diagnostic and prognostic indicators for various types of cancer. 8,9 Among those microRNAs, exosomal mir-21 is the only significantly overexpressed miRNA in a wide range of solid cancer. Current studies on the diagnosed accuracy of ex-mir-21disperse in a single type of cancer, such as lung, 10 breast, 11 and ovarian cancers, 12 merely discussed the overall application of ex-mir-21 in different types of cancer. In this study, we perform a comprehensive analysis of the diagnostic performance of ex-mir-21 in cancer diagnosis.

| Literature search and search strategy
Eligible studies were identified in electronic databases that are recommended by the handbook and by cross-checking the reference lists of relevant papers up to March 2021. The Cochrane Library, Embase via OVID, PubMed, ScienceDirect, and Web of Science databases were systematically searched using search terms: ("exosomes" OR "exosomal" OR "exocrine"), ("microRNA-21" OR "microR-21" OR "mir-21"), ("biomarker" OR "marker, biological"), and ("neo-plasms" OR "cancer" OR "carcinoma" OR "tumor") as well as their abbre-

| Inclusion and exclusion criteria
Two investigators (FL and HM) independently extracted data from the eligible papers complying with the inclusion and exclusion criteria. Disagreements were subsequently reviewed and resolved through discussion. Studies were included in the meta-analysis if they met the following inclusion: (1) studies investigating the role of ex-mir-21 in the diagnosis of cancer; (2) study patients with any type of carcinoma should be confirmed by the gold standard; (3) studies reported the detailed clinical data that can be used to calculate diagnostic accuracy data, including and the number of true-positive (TP), false-positive (FP), true-negative (TN), and false-negative (FN) cases, 2×2 diagnostic table; (4) articles written in English. Animal experiments were excluded, and only original articles were considered.
Other publications, including letters, reviews, case reports, or editorial articles, were excluded.

| Quality assessment
The quality and potential bias of the studies were assessed using the QUADAS-2 tool in the Review Manager. The risk of bias for each study was ranked as "low," "unclear," and "high." The same two investigators assessed the study quality independently, and discrepancies were resolved by discussion. The risk of bias level was ranked according to four different areas: (i) patient selection, (ii) index test, (iii) reference standard, and (iv) flow of patients through the study and timing of the index tests and reference standard (flow and timing).

| Statistical analysis
TP, TN, FP, and FN rates in the included studies were the primary data to assess the sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratios (DORs), and SROC.
The diagnostic performance of ex-mir-21 was determined by calculating those values with 95% CIs. The DOR was used to reflect the relationship between diagnostic tests and disease where higher numbers would indicate improved performance in diagnosing patients with/ without cancer. The DOR summarized the diagnostic accuracy of the index test as a single number that describes how many times higher the odds were of obtaining a test positive result in a diseased, rather than a non-diseased person. PLR/NLR was a ratio of the probability that a test result is correct to the probability that the test is incorrect. The larger the ratio of PLR, the greater is the probability of being a true positive when the result is positive. The smaller the ratio of NLR, the greater is the probability of being a true negative when the result is negative.
Summary receiver operating characteristic (SROC) curves were generated to estimate the effect of sensitivity and specificity based on TP and FP rates. The area under the curve (AUC) of the SROC was also calculated to assess the performance of ex-mir-21 in diagnosing cancer.
A prediction interval (the 95% prediction contour line in Figure 3) was used to consider the potential effect of the biomarker in the cancer diagnosis when it is applied within an individual setting. 13 The heterogeneity between studies was assessed using the χ 2 test and the inconsistency index (I 2 ). An I 2 > 50% with p < 0.05 from the χ 2 test is indicative of significant heterogeneity. In this case, a randomeffects model was chosen to pool the data of sensitivity, specificity, and AUC. Otherwise, a fixed-effects model was used. The threshold effect was considered a possible cause of heterogeneity in diagnostic accuracy analysis. Spearman correlation was used to analyze the logit of sensitivity and the logit of (1-specificity) and to verify the ex-

| Study evaluation
In total, 803 studies were identified in the search of multiple databases and cross-checking for reference lists. After removing duplicates, 309 studies contained and went through a screening process based on titles and abstracts. One hundred fifty-three studies were identified for full-text review, and eighty-six of them were excluded because of nondiagnostic study, review articles, or diagnostic accuracy data not included or calculable in the publication. Finally, 26 studies met the inclusion criteria and were included in the metaanalysis and review. 10-12,14-36 A detailed flow diagram of the study selection is shown in Figure 1 Table 1 summarizes the characteristics of the 26 included studies in our meta-analysis. There were 30 trials in the 26 articles with the publication years ranged from 2008 to 2020. The investigated cancer types were divided into four groups: digestive cancer (n = 13), breast cancer (n = 2), lung cancer (n = 3), and other types of cancer (n = 11), including colon cancer (CC), ovarian cancer (OC), hepatocellular carcinoma (HCC), glioblastoma, renal cell carcinoma (RCC), intraductal papillary mucinous neoplasm (IPMN), and lymphoma.

| Quality assessment
Regarding the quality evaluation of the included studies, a majority of included studies had a low risk of bias. The risk of bias in the patient selection domain was considered having a relatively higher amount of bias in five of twenty-seven included studies. These studies did not clarify if their studies avoid inappropriate exclusions.
Several studies that did not mention its interpretation way of blindness were assessed as unclear or high risk of bias in the "index test" and "reference standard" category. Our ratings of the risk of bias and applicability concerns in each study are displayed in Figure 2.

| The output results of this meta-analysis
The pooled sensitivity and specificity of the ex-mir-21 diagnostic ac-  Outlier detection was conducted for the investigation of heterogeneity. Most appropriate is meta-analysis with low risk of bias, although it also suggested that the source of heterogeneity may come from the studies of Jin et al., 18 Lai et al., 20 Matsuzaki et al., 22 Pan et al., 12 Que et al., 26 Taylor et al., 29  To account for the potential sources of heterogeneity, sensitivity and subgroup analyses were also conducted. The heterogeneity was still unchanged, and no significant difference was detected in the sensitivity analysis by omitting each of the included studies (data not shown). During the subgroup analysis, the pooled sensitivity, specificity, PLR, NLR, DOR, and AUC for each subgroup were calculated and are presented in Table 2.
There was an obvious difference between the pooled data in ethnicity analysis that the diagnostic accuracy of ex-mir-21 was superior in studies with Caucasian-based compared with Asian-based (sen: 84% vs 73%, spc: 91% vs 80%, AUC: 95% vs 83%), implicating that the ethnicity may be a potential factor impacting on heterogeneity and the diagnostic accuracy.
The cancer types divided into four cancer types revealed a less effective performance of ex-mir-21 in diagnosing NSCLC (sen: 70%, spc: 78%) and other types of cancer (sen: 71%, spc: 87%). According to a subdivided analysis, our results suggested that ex-mir-21 has an excellent diagnostic performance in the diagnosis of PC (sen:85%, spc:84%, AUC: 91%). On the other side, there was no significant difference between the pooled data in sample sources; the diagnostic accuracy was slightly better in the group of serum-based studies revealing that ex-mir-21 showed the versatility to detect cancer from various bodily fluids of humans. According to Deeks' funnel plot asymmetry test, no publication bias was detected among the studies (p = 0.703, Figure S4).

| DISCUSS ION
Exosomes have increasingly come to the front as important sources of reliable biomarkers for cancer diagnosis and prognosis. 37         converted the hepatic stellate cells (HSCs) to cancer-associated fibroblasts (CAFs), which further secrete exosomal-mir-21 to promote HCC progression. 50 Therefore, more and more attention has been paid to the role of miR-21 in prevention and therapeutic strategies.
The study of exosomes is an active area of research. The outstanding performance of ex-mir-21 is significantly higher than that of CA19-9 and CEA, whose diagnostic abilities were specific for certain types of cancers. Its high stability in circulation and microenvironment and reproducible detection also advantage its further application in the clinical setting. However, to utilize exosomal miR-

| Limitation
Several points should be a concern before clinical application.

| CON CLUS ION
A favorable and preferred choice of ex-mir-21 as an effective biomarker in diagnosing cancer is recommended and shows strong exploring the potential prognostic role of exosomes will contribute to the management of the therapeutic strategies.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.