MicroRNAs with prognostic significance in bladder cancer: a systematic review and meta-analysis

The aim of this study was to systematically review articles that investigated the prognostic significance of different microRNAs in bladder cancer (BC). We systematically searched PubMed, Web of Science, and Embase to identify relevant studies until March 2016. After screening, 26 studies that involved 2753 patients were included. Results suggested that many miRs expression aberration may predict prognosis in patients with BC. There are six miRs (miR-21, miR-143, miR-155, miR-200, miR-214, and miR-222) were reported by at least two studies, and we performed meta-analysis in the corresponding studies. Accordingly, we found that high miR-21 expression was associated with poor overall survival [OS; hazard ratio (HR) = 3.94, 95% CI 2.08–7.44]. High miR-143 expression was associated with poor progression-free survival (PFS; HR = 3.78, 95% CI 1.61–8.89). High miR-155 expression was associated with poor PFS (HR = 8.10, 95% CI 2.92–22.48). High miR-222 expression was associated with poor OS (HR = 3.39, 95% CI 1.10–10.41). Meanwhile, low miR-214 expression was correlated with poor RFS(HR = 0.34, 95% CI 0.22–0.53). Our comprehensive systematic review concluded that microRNAs, particularly miR-21, miR-143, miR-155, miR-214, and miR-222, could serve as meticulous follow-up markers for early detection of progression or recurrence and even useful therapeutic targets for the treatment in patients with BC.

Study Characteristics. All of the included studies were recently published (2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). They reported the prognostic significance of 37 different miRs in BC patients with varying tumor stages. Thirteen studies originated from China, two from Korea, two from Greece, two from Spain, one from Sweden, one from Denmark, one from Israel, one from Germany, one from France, one from the United States, and one from the United Kingdom. Most studies applied the quantitative real-time polymerase chain reaction to measure the miR expression. Four studies applied in situ hybridization. The miR expressions were mainly detected in the tissue samples. However, two and four studies measured miRs in serum and urine cell-free, respectively. The HR was adjusted for corresponding covariates in 18 studies using the Cox regression multivariate analysis. Table 1 summarizes the detailed information of the twenty-six included studies.
Two studies addressed the relationship between miR-214 expression and survival outcomes in BC, of which two contained RFS 18, 24 data, and one contained OS 18 data. A meta-analysis was conducted on the miR-214 expression and RFS relationship. We found that low miR-214 expression was predicted poor RFS (a fixed-effect model, HR, 0.34; 95% CI: 0.22-0.53; p < 0.001; I 2 = 49.3%, p = 0.160). In addition, Wang et al. 18 also found that a shorter OS in BC patients with an decreased level of miR-214 (HR, 0.28; 95% CI: 0.16-0.50; p < 0.001) (Fig. 3E).
For miR-222, two studies reported OS 22,27 , one reported CSS 27 , one reported RFS 27 , and one reported PFS 27 . We performed a meta-analysis on the miR-222 expression and OS relationship. The results suggested that increased miR-222 expression tended to occur in patients with poor OS (a random-effect model, HR, 3.39; 95% CI: 1.10-10.41; p = 0.033; I 2 = 82.4, p = 0.017). Additionally, Puerta-Gil et al. 27  Considering that few studies were included in this meta-analysis, the publication bias evaluated by Begg's and Egger's test was not considered necessary.

Discussion
We performed comprehensive systematic review and meta-analysis of the current literature on BC in response to the need for independent prognostic molecular markers that are readily assayable before, during, and/or after BC treatment. A total of 26 studies involving 2753 patients were analyzed to evaluate the relationship between the miRs and the BC prognosis in our study. Accordingly, 37 different miRs involved in the survival outcomes of patients with BC were compared. This study aimed to identify the miRs associated with BC prognosis, which could be further validated in future studies and eventually evaluated before the treatment, thereby improving BC management.
This study found that high miR-452 28 , miR-452* 28 , miR-21 16 Our study systematically assessed the role of 37 different miRs in BC prognosis. However, most of the miRs were investigated only by a single study, and only six miRs (miR-21, miR-143, miR-155, miR-200, miR-214, and miR-222) were identified by at least two studies. Therefore, we conducted the meta-analysis on these six miRs to determine a pooled conclusion. We found that high miR-21 expression was associated with poor OS. High miR-143 expression was associated with poor PFS. High miR-155 expression was associated with poor PFS. Meanwhile, high miR-222 expression was associated with poor OS. Low miR-214 expression was correlated with poor RFS, and MiR-200 expression was not related to the RFS in the pooled analysis. Pignot et al. 23 reported that an elevated miR-200 level was associated with a shorter OS. But it was important to note that the study was small (n = 72), had limited power, could lead to a premature result, and was not adjusted for other associated variables (covariates) that could also affect the survival outcome. Therefore, more multicenter prospective studies with large scale and long-term follow-up are needed to obtain a more persuasive conclusion.
MiR-21 is one of the most extensively studied cancer-related miRs; it is an abnormal expression in most cancers functioning as oncogene [40][41][42][43] . Zhou et al. 44 published a meta-analysis evaluating the prognostic value of miR-21 in various cancers. They pooled 63 published studies and found that the increased miR-21 expression predicted worse OS in cancers. This finding is consistent with our results in BC. Some researchers explored the possible miR-21 mechanism in BC. Zhang et al. 16 suggested that miR-21 through the maspin expression down-regulation up-regulated the VEGF-C expression, thereby increasing tumor growth, migration, and invasion in BC. Additionally, Zaravinos et al. 25 demonstrated that miR-21 repressed the tumor suppressors, PTEN and PDCD4, to enhance angiogenesis, tumor cell proliferation, EMT, and metastasis activation in BC. MiR-143 is located in chromosome 5q33, which is a well-known fragile site in human genome and highly co-expressed with miR-145 in BC 45,46 . Many studies have demonstrated the tumor suppressor role of miR-143 in BC. MiR-143 can suppress cell proliferation and migration and promote apoptosis in BC by inhibiting PI3K/Akt and MAPK signaling and targeting AKT 47 , KRAS 48 , ERK5 49 , and PAI-1 45 . Intriguingly, miR-143 has been found to be up-regulated in aggressive BC, and that the case of patients with evaluated miR-143 is associated with worse prognosis. This finding seemed to be contrary to the tumor suppressor functions of miR-143. Many studies have also observed this interesting outcome 21,23,27,29 . However, the definite mechanism elucidating it remains unclear and needs further studies. A large number of studies confirmed the oncogene function of miR-155 in various cancers, including renal, thyroid, hematological, pancreatic, and bladder cancers [50][51][52][53][54] . High miR-155 expression has recently been correlated with BC recurrence and progression. Some studies might provide potential evidence linking the miR-155 expression and BC. Zhang et al. 14 suggested that miR-155 overexpression promoted tumor cell growth via Wnt/β-catenin signaling activation, which is also a vital pathway in the BC tumorigenesis. In addition, Wang et al. 19 reported that the oncogenic properties of miR-155 can be attributed to its antiapoptotic function through a blockade of caspase-3 activity or suppression of proapoptotic genes, such as TP53BP1, and promoted cell proliferation by down-regulating the SOCS1 gene, or activated PKB signaling via downregulation of tumor suppressors, including PTEN, PDCD4, and SHIP1. The role of miR-214 acting as an oncogene or a tumor suppressor is quite distinctive in different cancer types. MiR-214, serving as an oncogene, has been found in many human cancers, such as nasopharyngeal, gastric, and ovarian cancers as well as malignant melanomas [55][56][57][58] . Nevertheless, miR-214, serving as a tumor suppressor, has been found in several other cancers, such as cervical, breast, liver, esophagus, and bladder cancers 17, 59-62 . The functional discrepancies of miR-214 in different cancer types may be derived from its varied target genes or distinction among tissue types and cellular environments. Wang et al. 17 showed that miR-214 decreased cellular proliferation, migration, and invasion and simultaneously increased apoptosis, suggesting that miR-214 functions a tumor suppressor in BC. The tumor-suppressive role of miR-214 might explain the unfavorable prognosis of BC patients with low miR-214 expression. High miR-222 expression has been observed in many human cancers, such as glioblastoma, prostate, colorectal, and pancreatic cancers [63][64][65][66] , suggesting that miR-222 might play a vital role in carcinogenesis. Puerta-Gil et al. 27 recently found that high miR-222 expression is correlated with more advanced tumor stage and grade in BC, indicating its important role in tumorigenesis and metastasis. Moreover, Zhang et al. 22 obtained the same conclusion in Asian patients with BC. Calderaro et al. 67 explored the possible molecular mechanism of miR-222 in BC. They found that miR-222 decreased the tumor suppressor, PTEN, which is considered to enhance angiogenesis, tumor cell proliferation,  Aside from the abovementioned miRs, the current study also systematically investigated the relationship between other miRs and BC prognosis. Figure 2 and Table 3 summarize the relation with prognosis and the possible role of other miRs in BC progression.
Accordingly, several limitations should be pointed out in the interpretation of the results of the current study. First, 26 involving 2753 patients were included in the systematic review. However, most of them investigated diverse miRs. Only six miRs (miR-21, miR-143, miR-155, miR-200, miR-214, and miR-222) were assessed by at least two studies that used different survival outcome assessments. Therefore, only two or three records were eligible in most of the meta-analyses in the current study. Furthermore, more multicenter prospective studies should be performed to verify our results and make a more mature conclusion. The publication bias was not evaluated because only few studies were included in this meta-analysis. The lack of analysis might have affected the interpretation of the results and make them less reliable. Second, error may have occurred because of inaccurate readings even if three independent reviewers extracted the data and the HR extrapolation method from a Kaplan-Meier graph was previously validated 68 . Therefore, the extrapolated HRs might be less dependable compared with the reported statistics. Third, the marked heterogeneity of studies was observed in some analyses. The heterogeneity of the pooled analysis might have been caused by several differences among the studies, including the baseline characteristics of the patients (i.e., study size, age, gender, ethnicity, and tumor stage), detecting methods, detecting sample, cut-off value, HR source, adjusted covariates, and follow-up duration. Finally, positive results were more likely to be published in most studies, whereas studies with negative results were often rejected or less assessable, which could compromise the validity of such analyses 69 . Urine markers have recently become more desirable than tissue or serum markers because urine is more convenient and less invasive to collect. Additionally, urine markers can be evaluated before, during, and/or after surgery and then monitored throughout a patient's life. More well-conducted and appropriately designed studies are thus needed to establish the prognostic value of the miR urine levels. Some studies have already developed a combined expression signature of multiple miRs, which require a powerful validation strategy. More importantly, independent validation studies are needed to validate these results and evaluate the performance and prognostic power of the signature. Pignot et al. 23 used three miRs to develop a molecular signature for BC with promising results. In the future, developing a new molecular signature using diverse miRs and then identifying their efficacy may also be useful.
In conclusion, our comprehensive systematic review and meta-analysis suggested that miRs, particularly miR-21, miR-143, miR-155, miR-214, and miR-222, could potentially serve as risk stratification markers and even therapeutic targets in BC despite the abovementioned limitations. However, more large scale, multicenter prospective studies with standardized methods and long-term follow up are needed to verify our results.

Methods
Search Strategy. This meta-analysis was conducted following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 70 . A systematic literature search was performed in the electronic databases PubMed, Web of Science, and Embase on 15 March 2016 using the following search strategy: (microRNA or miRNA or miR) and (bladder cancer or bladder tumor or bladder carcinoma or bladder neoplasm or urothelial cancer or urinary tract cancer) and (prognosis or prognostic or survival or outcome or mortality). We also manually searched the reference lists of the relevant literature.

Selection Criteria.
The studies were included based on the following criteria: (1) the association of miRs with the prognosis significance in BC should be described; (2) the studies defined the miR cut-off and clearly described the miR measurement; and (3) the studies correlated survival outcomes with a single miR expression. The exclusion criteria were as follows: (1) non-English papers; (2) case reports, letters, commentaries, meeting records, or review articles; (3) sample number fewer than 30 patients; (4) focused on animal models or cancer cells; (5) concerned genetic variation of an miR; (6) calculated HRs based on multiple miRs; and (7) the study lacked sufficient data for obtaining HR and 95% CI. All evaluations were independently performed by three individual researchers to ensure the accurate inclusion of studies. The discrepancies were resolved by discussion. We only retrieved the most informative and recently studied one for further analyses of duplicate studies.
This study was based on published literature. Therefore, ethical approval from ethics committees was not needed.
Data Extraction. Three investigators independently extracted data from eligible studies using a predefined form. The discrepancies in data extraction were resolved by discussion. The following data were extracted: surname of the first author, publication year, investigated miRs, origin of the studied population, study design, tumor stage, sample size, gender, follow-up time, detecting method, detected sample, cutoff value, and effect estimates, namely, HR of miR expression for OS, CSS, DFS, RFS or PFS, as well as their 95% CI (Table 1). We calculated HRs and their 95% CI based on the methods reported by Tierney et al. 68 if the HR and 95% CI were not directly available. Quality Assessment. The quality of the included studies was evaluated using the Newcastle-Ottawa scale, as recommended by the Cochrane Non-randomized Studies Methods Working Group 71 . Each study can be assessed by eight methodology items with a score ranging from 0 to 9. The high scores indicated high quality. We considered studies with scores of 6 or more as high quality for the meta-analysis. Only studies with high quality were included in the further analysis to assure the quality of this meta-analysis.
Statistical Analysis. Pooled HR with 95% CI was used to evaluate the association of the miR expression with the BC prognosis. An observed HR > 1 implied an unfavorable prognosis for the group with an elevated miR expression. Conversely, an observed HR < 1 implied a favorable prognosis for the group with an elevated miR expression. A heterogeneity test of the pooled HR was conducted using Cochran's Q test and Higgins I-squared statistic 72 . A p value of less than 0.05 was considered significant. A random-effect model was used miR-29c* Tumor suppressor miR-29c* acts through downregulation of DNA methyltransferases as well as upregulation of demethylating genes to keep the normal methylation pattern. 37 miR-9 Proto-oncogene miR-9 directly targeted the CDH-1 gene encoding E-cadherin, a regulator of EMT, considered to be an important initiating step for tumor metastasis. 23 Table 3. Summary of possible role and potential mechanism of miRs entered this study in bladder cancer. when between-study heterogeneity was observed (Cochran's Q test, p < 0.05); otherwise, a fixed-effect model was used. All statistical analyses were performed using Stata 12.0 software (StatCorp, College Station, TX, USA), and p < 0.05 was considered statistically significant.