The Prognostic Value and Regulatory Mechanisms of microRNA-145 in Various Tumors: A Systematic Review and Meta-analysis of 50 Studies

As an important type of noncoding RNA, miRNA comprises a class of small endogenous RNAs of approximately 22 nt in length that play a crucial role in the regulation of gene expression at the posttranscriptional level (1, 2). Previous studies have shown that many miRNAs are aberrantly expressed in tumor tissues and play crucial roles in the growth, differentiation, angiogenesis, metastasis, and drug resistance of various tumors (3, 4). Moreover, some miRNAs are significantly associated with the prognosis of patients with various tumors and are potential prognostic predictors and candidate treatment targets (5). Therefore, the recognition of the clinical significance and regulatory mechanisms of miRNAs may assist with the diagnosis, prognosis prediction, and treatment of malignant tumors.

miR-145 is derived from chromosome 5q32 and contains two mature subtypes of miR-145-5p and miR-145-3p (6). On the basis of the deep sequencing data referred to in miRBase (7), miR-145-5p is much more abundantly expressed than miR-145-3p. Substantial data obtained from previous studies have demonstrated that miR-145 is downregulated in various tumors and corresponding cell lines to be considered as tumor suppressors (8–14). On the contrary, several studies have found that miR-145 is upregulated in tumor tissues and functions as an oncogene (15, 16). For example, Naito and colleagues (15) showed that miR-145 was upregulated in patients with gastric cancer with more advanced tumor stages or with scirrhous type histology, and highly expressed miR-145 was significantly associated with poor prognosis in patients with gastric cancer. To date, although the correlation between miR-145 expression and the prognosis of patients with various tumors has been investigated by numerous studies, the conclusions are not completely consistent. Therefore, we aimed to conduct a meta-analysis based on all eligible evidence to evaluate the association between miR-145 expression and the prognosis of patients with malignant tumors. In addition, in response to the need for comprehensive recognition of miR-145, known regulatory mechanisms of miR-145 will be illustrated in this study via systematically reviewing previous studies.

A systematic literature search was conducted using online databases including MEDLINE, Embase, PubMed, Google Scholar, and China Biology Medicine disc. The keywords used in the searches were “miR-145 or miRNA-145 or microRNA-145 (all fields).” The categories of diseases and research types were not limited so that the maximum number of studies was identified. In addition, the reference lists of relevant reviews, meta-analyses, and original studies were manually screened to acquire more studies. The language was not restricted.

In this study, two primary outcomes, overall survival (OS) and progression-free survival (PFS), were selected to calculate the association between miR-145 expression and survival outcome of patients with multiple tumors. OS was measured from the time at which the baseline blood or tissue sample was obtained to the date of death from any cause or the date of last follow-up. PFS was recorded as the time between the baseline blood and tissue sampling for miRNA analysis and documentation of the first tumor progression, based on clinical or radiological findings. In different studies, OS was also expressed as disease-specific survival (DSS; ref. 17) and cancer-specific survival (CSS; ref. 15), while PFS was also described as recurrence-free interval (RFI; ref. 8), disease-free survival (DFS; refs. 9, 10, 18–26), biochemical-free survival (BFS; ref. 27), metastasis-free survival (MFS; ref. 28), relapse-free survival (RFS; ref. 29), and time to relapse (TTR; ref. 11).

The following criteria were used to help select eligible literature: (i) published studies that could be retrieved from the abovementioned online databases; (ii) the expression of miR-145 was measured in the tumor tissue, peripheral blood, or body liquid; (iii) the association between miR-145 expression level and survival outcome was analyzed; and (iv) HR and 95% confidence interval (CI) were reported or enough information was provided to calculate such parameters. Studies were excluded if they included the following items: (i) the patients did not suffer from malignant tumors; (ii) the study was only conducted on an animal model or tumor cell lines; and (iii) no data could be extracted or the studies were published as abstracts, reviews, conference reports, letters, or editorials.

To make the management of literature more convenient, all identified citations were imported into an EndNote library (Thomson Corporation). After removing duplicated studies, two independent investigators (Y. Zhang and J. Tang) carefully screened the relevant studies by reading the titles and abstracts. Then, the entire text of potential eligible studies was evaluated to confirm the final inclusion. Any discrepancies during the study selection were resolved by discussion with the corresponding author (M. Xu) for consensus.

The relevant information was extracted from all included studies by two independent authors (P. Wang and L. Li). The following data elements were sought and recorded: (i) first author, publication year, and nationality of study population; (ii) miR-145 subtype, tumor type, sample type, and miR-145 assay method; and (iii) sample size, period of follow up, cut-off value, HR, and corresponding 95% CI. When a study reported the survival results of both univariate and multivariate analyses, only the latter was extracted because it is more accurate as it accounts for confounding factors. If a study only reported the survival results using Kaplan–Meier curves, then the statistical variables were read from the graphical survival plots with the Engauge Digitizer 4.1 software program, and then the HR value and 95% CI were calculated via the method reported by Tierney and colleagues (30). Regarding other missing information, e-mails were sent to corresponding authors requesting useful data. Finally, the extracted data forms were crosschecked between the abovementioned two reviewers, and any disagreements during the process of data extraction were resolved by discussions with a third author (M. Zhang).

In this study, the quality of included studies was assessed by two independent investigators (P. Wang and L. Li) using the Newcastle-Ottawa Scale (NOS; ref. 31). This is an acknowledged tool for assessing the quality of nonrandomized studies via the judgment of three main study characteristics as follows: selection and definition of the study groups, comparability of the groups, and ascertainment of outcomes. Then, a possible score of 0–9 was assigned to each study. A study with a NOS score greater than 6 was considered to be high quality.

The RevMan software version 5.3 (Cochrane Collaboration) and STATA software version 14.1 (StataCorp) were used to perform statistical analyses in this study. The pooled HR and corresponding 95% CI values were used to evaluate the prognostic value of miR-145 for various malignant neoplasms. The statistical significance of the outcomes was determined by the Z-test and P values less than 0.05 were considered statistically significant. miR-145 is a known tumor suppressor in most tumors; therefore, low expression of miR-145 was thought as a risk factor and HR greater than one indicated a poor prognosis. The heterogeneity among studies was assessed using Cochran Q test and Higgins I-square (I²) statistic (32, 33). If a significant heterogeneity was observed, namely P < 0.05 and/or I² > 50%, the random-effects model (DerSimonian and Laird method; ref. 34) was used. Alternatively, the fixed-effects model (Mantel–Haenszel method; ref. 35) was applied to calculate the pooled HR and 95% CI of survival outcomes. To decrease the heterogeneity among studies and recognize the prognostic value of miR-145 in greater detail, subgroup analyses were conducted on the basis of multiple criteria such as miR-145 subtype, tumor type, sample type, and patient ethnicity. In addition, Begg test (rank correlation test; ref. 36) and Egger test (weighted linear regression test; ref. 37) were employed to evaluate the potential publication bias (P < 0.05 was considered statistically significant). Furthermore, one-way sensitivity analysis was performed to identify studies that had a crucial influence on the pooled HR by removing one study at a time.

A total of 1,262 studies were identified from online databases MEDLINE, Embase, PubMed, Google Scholar, and China Biology Medicine disc. Another five studies (17, 38–41) were acquired through manually screening the reference lists of relevant reviews and meta-analyses. After removing 54 duplicated publications, the remaining 1,213 studies were evaluated by carefully reading the titles and abstracts, after which 1,059 studies were excluded because of the following reasons: not tumor studies, unpublished, withdrawn articles, letters, abstracts, reviews, or meta-analyses. Next, the entire text of the remaining 154 studies was assessed. Among them, 106 were removed because survival analyses were not performed in 92 studies, and HR could not be extracted or calculated from the other 14 studies. Finally, 48 articles containing 50 studies, which were published between 2010 and 2017, were included in the meta-analysis for this study. A detailed flowchart illustrating the process of literature selection is shown in Fig. 1.

Among the 50 included studies, 46 studies investigated the prognostic value of miR-145-5p for malignant tumors and only 4 studies focused on miR-145-3p. A total of 6,875 patients suffering from 18 different tumors were included in the meta-analysis, with the sample size in each study ranging from 20 to 1,141 patients (median 74.5). Quantitative real-time PCR was the method used most often to measure the expression of miR-145 (43/50, 86%). A total of 17 of 50 studies (34%) measured the expression of miR-145 in frozen tissues, 18 (36%) in formalin-fixed, paraffin-embedded tissues (FFPE), 6 (12%) in plasma or serum, while the remaining 8 studies (16%) did not specifically report the tissue type used. The cut-off value that stratified patients into high and low expression groups varied among the different studies, with the median value being the most widely used value (28/50, 56%). For the prognosis, 25 studies reported a correlation between miR-145 expression and OS, 14 reported PFS, and the remaining 11 reported both OS and PFS. The length of follow-up ranged from 24 to 310 months with a median of 69.5 months. With respect to HR, 18 studies reported the HR and 95% CI directly, whereas the remaining 32 studies reflected the survival outcomes using survival curves, and thus the HR and 95% CI could be calculated. With respect to the quality of included studies, most included studies (43/50, 86%) were high quality with the NOS score greater than 6. Other detailed information of enrolled studies is listed in Table 1.

A total of 36 (9, 12–15, 17, 21, 22, 24–26, 29, 39, 40, 42–62) of the 50 studies that included 5,074 patients evaluated the relationship between miR-145 expression and OS of patients suffering from various tumors. Among them, two studies investigated the expression level and prognostic value of miR-145 in blood samples, and the results of these two studies were similar with most tissue studies. Therefore, a pooled analysis containing blood and tissue studies were performed. Owing to a significant heterogeneity existing among studies (I² = 62%, P < 0.001), a random model was employed to calculated the pooled HR and 95% CI of OS. The result showed that low miR-145 expression was associated with poor OS in multiple tumors, with a pooled HR of 1.70 (95% CI, 1.46–1.99; P < 0.001; Fig. 2). To decrease the heterogeneity among studies, subgroup analyses were performed on the basis of 7 criteria: miR-145 subtype, tumor type, sample type, HR resource, patient ethnicity, miR-145 assay method, and cut-off value (Table 2; Supplementary Figs. S1A–S7A). The results showed that low expression of miR-145 was significantly associated with worse OS in subgroup analyses of miR-145 subtype, HR resource, patient ethnicity, miR-145 assay method, and cut-off value. However, the results from subgroup analysis of tumor type suggested that the downregulation of miR-145 was obviously associated with poor OS in colorectal cancer (HR = 2.17; 95% CI, 1.52–3.08; P < 0.001), ovarian cancer (HR = 2.15; 95% CI, 1.29–3.59; P = 0.003), glioma (HR = 1.65; 95% CI, 1.30–2.10; P < 0.001), and osteosarcoma (HR = 2.28; 95% CI, 1.50–3.47; P < 0.001), but not in lung cancer (HR = 1.54; 95% CI, 0.70–3.36; P = 0.28), cervical cancer (HR = 1.32; 95% CI, 0.64–2.68; P = 0.45), esophageal cancer (HR = 0.97; 95% CI, 0.30–0.09; P = 0.95), and breast cancer (HR = 1.19; 95% CI, 0.79–1.81; P = 0.41). The study conducted by Naito and colleagues (15) found that miR-145 was upregulated in scirrhous type gastric cancer and the downregulation of miR-145 was significantly associated with better OS in gastric cancer (HR = 0.50; 95% CI, 0.26–2.98; P = 0.44; ref. 15). The pathologic type and survival result from this previous study were totally different from other studies focusing on the association between miR-145 expression and OS in gastric cancer (46, 60). Therefore, further subgroup analysis for gastric cancer was performed after omitting the study conducted by Naito and colleagues, and the result without heterogeneity (I² = 0; P = 0.55) suggested that the low miR-145 expression also indicated worse OS in gastric cancer (HR = 1.78; 95% CI, 1.35–2.36; P < 0.001). For the subgroup analysis of sample type, the downregulated miR-145 in frozen tissues and serum was significantly associated with poor OS (frozen tissues: HR = 1.81; 95% CI, 1.39–2.35; P < 0.001; serum: HR = 1.74; 95% CI, 1.21–2.49; P = 0.003), whereas the association between the downregulation of miR-145 in FFPE tissues and OS was not statistically significant (HR = 1.35; 95% CI, 0.99–1.84; P = 0.06).

In this study, PFS was analyzed along with disease-free survival (DFS), recurrence-free interval (FRI), biochemical-free survival (BFS), metastasis-free survival (MFS), time to relapse (TTR), and relapse-free survival (RFS), because all these indices were used to indicate the tumor recurrence or deterioration after surgery or treatment. A total of 26 studies (8–11, 16, 18–29, 38, 41, 49, 52, 53, 56, 61, 63) encompassing 1,971 patients with carcinoma evaluated the correlation between miR-145 expression and PFS. The random-effects model was employed to estimate the pooled HR owing to an obvious heterogeneity among studies (I² = 70%, P < 0.001). Results showed that downregulated miR-145 significantly predicted unfavorable PFS in various cancers, with a HR of 1.39 (95% CI, 1.16–1.67; P < 0.001; Fig. 3). Similar to OS, subgroup analyses of PFS were also performed. The results showed that the predictive value of miR-145 on PFS in various cancers was not altered when patients were stratified on the basis of HR resource and miR-145 assay method (Table 2; Supplementary Figs. S1B–S7B). Nevertheless, the results from the subgroup analysis of tumor type showed that the low miR-145 expression was only significantly associated with worse PFS in lung cancer (HR = 1.97; 95% CI, 1.25–3.09; P = 0.003). The subgroup analysis based on patient ethnicity found that the downregulated miR-145 was only associated with poor PFS in Asian patients (HR = 1.50; 95% CI, 1.23–1.82; P < 0.001) and not with the European (HR = 1.41; 95% CI, 0.87–2.29; P = 0.16) or American patients (HR = 0.80; 95% CI, 0.32–2.02; P = 0.64). For the subgroup analysis of the cut-off value, the results showed that the PFS of patients in high and low miR-145 expression groups that were stratified using the median value was comparable (HR = 1.14; 95% CI, 0.86–1.51; P = 0.37). The subgroup analysis found that the low miR-145 expression in all types of tissues was not associated with the PFS (frozen tissue: HR = 1.51; 95% CI, 0.96–2.37; P = 0.08; FFPE tissue: HR = 1.18; 95% CI, 0.94–1.48; P = 0.15; serum: HR = 1.26; 95% CI, 0.86–1.86; P = 0.23).

In this study, the funnel plots of Begg and Egger tests were employed to evaluate the publication bias of all included studies. No obvious asymmetry was observed in the funnel plots of Begg (OS, P = 0.17, Fig. 4A; PFS, P = 0.63, Fig. 4C) and the P values of the Egger tests were all higher than 0.05 (OS, P = 0.41, Fig. 4B; PFS, P = 0.33, Fig. 4D). Therefore, significant publication bias did not exist in this meta-analysis.

Sensitivity analysis of OS and PFS was performed to investigate the influence of each individual study on the pooled HRs (Fig. 5A and B). The result showed that the pooled results were not significantly altered by sequentially omitting any single dataset, demonstrating that the results of this meta-analysis were robust.

Consistent with other mammalian miRNAs, miR-145 was first transcribed from its parental gene and termed as pri-miRNA, then the pri-miRNA was cleaved into hairpin intermediates (pre-miRNAs) by the nuclear RNase III Drosha and further processed to mature miRNAs by cytosolic Dicer, another RNase-III related enzyme (Fig. 6; refs. 64, 65). Therefore, the expression level of miR-145 might be regulated in the transcription and maturation process by various factors. During the transcription step, Sachdeva and colleagues (66) first discovered that P53, a well-demonstrated central tumor suppressor, could induce miR-145 transcription by directly interacting with its promoter. From then on, other molecules including EWS-FLI-1 (67), FOXO (68), TP53 (69), EGFR (70), PPARγ (71), DNMT3b (72), AR (73), and DDX3 (74) could also stimulate the transcription of miR-145 by enhancing promoter activity. On the contrary, DNA methylation at the miR-145 promoter region (75), C/EBP-b (76), DCLK1 (77), RREB1 (78), and ZEB2 (79) caused the downregulation of miR-145 by repressing the activity of the miR-145 promoter. With regard to the maturation-process, P53 (80) and BRCA1 (81) could increase the expression of miR-145 by directly interacting with the Drosha complex, whereas p70S6K (82), methyltransferase BCDIN3D (83), and TARBP2 (84) restrained the maturation of miR-145 via inhibition of Dicer activity. Moreover, recent studies have shown that the function of miR-145 could also be repressed by various competing endogenous RNAs (ceRNA) including transcribed pseudogenes (e.g., OCT4-pg4; ref. 85), lncRNAs [e.g., lncRNA RoR (86–88), TUG1 (89), MALAT1 (90), UCA1 (91), and CRNDE (92)], and circular RNAs [e.g., circRNA_001569 (93) and circBIRC6 (94)]. The ceRNAs identified thus far mainly regulate the function of miR-145-5p instead of miR-145-3p.

After processing with Dicer, pre-miR-145 generated two mature subtypes named miR-145-5p and miR-145-3p. Then, an RNA-induced silencing complex that silenced the expression of target transcripts by either facilitating corresponding mRNA degradation or blocking its translation was formed (64). By systematically reviewing previous studies, the transcripts that could be targeted by miR-145-5p, miR-145-3p, or both miR-145-5p and miR-145-3p in different malignant tumors were summarized (Table 3). Generally, most of them focused on miR-145-5p and it was found to silence many genes, which participated in almost every aspect of tumor activities, including tumor growth, metastasis, differentiation, angiogenesis, and drug resistance. Among these genes, some played important roles in only one aspect of tumor behavior. Minami and colleagues (95) revealed that miR-145-5p perturbed the Warburg effect by silencing KLF4 in bladder cancer cells, resulting in significant cell growth inhibition. Eades and colleagues (96) found that the small GTPase ADP-ribosylation factor 6, a target of miR-145-5p in the triple-negative breast cancer, promoted cell invasion by regulating E-cadherin localization and impacting cell–cell adhesion. Gao and colleagues (97) suggested overexpression of miR-145-5p sensitized breast cancer cells to doxorubicin in vitro and enhanced the doxorubicin chemotherapy in vivo via inhibition of the multidrug resistance-associated protein 1. Yet, some targeted genes of miR-145-5p possess multiple functions in an individual tumor type. For instance, overexpression of miR-145-5p suppressed esophageal squamous cell carcinoma cell proliferation and invasion via targeting c-Myc (98), and the knockdown of miR-145-5p responsively increased both the mRNA and protein levels of Ets1, and thus promoted the metastasis and angiogenesis of gastric cancer cells (99). In addition, some genes can be regulated by miR-145-5p in multiple tumors. For example, FSCN1, one of the most frequently reported target genes of miR-145-5p, was involved in bladder cancer (100), esophageal cancer (101), hepatocellular carcinoma (HCC; ref. 102), lung cancer (103), nasopharyngeal cancer (104), and prostate cancer (105). Compared with miR-145-5p, the biological functions of miR-145-3p, which were derived from the antisense of miR-145-5p, have been reported in few studies. It was reported that three genes, HMGA2 (45), Ang-2 (106), and HIF-2α (107), could be regulated by miR-145-3p in ovarian cancer, pancreatic cancer, and neuroblastoma, respectively, to inhibit tumor growth or metastasis. Furthermore, previous studies found that MDTH (108) and UHRF1 (109) could be coregulated by both miR-145-5p and miR-145-3p in lung cancer and bladder cancer, respectively.

Although miRNAs only encompass 19–23 nucleotides and do not have the ability to encode proteins, they are involved in multiple cellular pathways and play crucial roles in various diseases (1). Identifying aberrantly expressed miRNAs and illustrating their underlying mechanisms may assist with early diagnosis, prognosis evaluation, and treatment development of numerous diseases (110, 111). Among thousands of known cancer-related miRNAs, miRNA-145 is considered an important one whose biological function and clinical significance have been investigated by many studies. To date, only two other meta-analyses have assessed the association between miR-145 expression and the prognosis of patients with malignant tumors (112, 113). Zhang and colleagues (112) utilized four prostate cancer studies and evaluated the predicted value of miR-145 for the DFS. In this study, the author defined the pooled HR < 1 indicated poor prognosis for the groups with lower miR-145 expression and was considered statistically significant if the 95% CI did not overlap 1. The pooled result indicated that the low expression of miR-145 in prostate cancer tissues predicted poor DFS (HR = 0.48; 95% CI, 0.30–0.75; P = 0.001). However, the result of their study cannot be applied to other cancers due to the heterogeneity among different tumors and the small sample size. Yang and colleagues (113) reported that overexpression of miR-145 was significantly associated with favorable OS in various carcinomas (HR = 0.47; 95% CI, 0.31–0.72; P < 0.01), but not with DFS (HR = 0.87; 95% CI, 0.51–1.47; P = 0.569). However, the conclusion of their study was not robust enough because it was published 3 years ago and only included 18 studies. In this study, a comprehensive literature search was performed to collect all relevant evidence available from previous studies, and we performed more detailed subgroup analyses to recognize the prognostic value of miR-145 in greater detail.

A total of 50 studies that included 6,875 patients evaluated the association between miR-145 expression and the prognosis of patients with malignant tumors. The pooled results from 36 studies suggested that low expression of miR-145 significantly predicted a poor OS, and this result was not altered when the patients were stratified into different subgroups based on HR resource, patient ethnicity, miR-145 assay method, and cut-off value. The subgroup analysis based on tumor type suggested that the downregulation of miR-145 was obviously associated with poor OS in colorectal cancer, ovarian cancer, gastric cancer, glioma, and osteosarcoma. In addition, the subgroup analysis based on tissue type found the prognosis value of frozen tissue and peripheral blood was better than that of FFPE tissues, which was probably caused by higher speed of RNA degradation in FFPE tissues than that in the other two types of tissues (114).

Consistent with OS, the pooled results showed that the downregulation of miR-145 was significantly associated with worse PFS in patients with various cancers. This result did not change when the patients were assigned to different subgroups based on miR-145 subtype, HR resource, and miR-145 assay method. However, the subgroup analysis based on tumor types showed that the downregulation of miR-145 was only associated with poor PFS in patients suffering from lung cancer, and not those with prostate cancer, colorectal cancer, esophageal cancer, breast cancer, and glioma. Meanwhile, the subgroup analysis based on patient ethnicity found that the downregulation of miR-145 was only associated with poor PFS among Asian patients, but not with the European and American patients. This discrepancy might arise from differences in environment or genetic background. Nevertheless, the subgroup analyses indicated that the aberrantly expressed miR-145 in all types of tissues (e.g., blood, frozen, and FFPE tissues) was not significantly associated with PFS. In addition, there was no significant difference in PFS between high and low miR-145 expressed groups, which was classified by median values.

In response to the need for comprehensive recognition of miR-145, the regulatory mechanisms of miR-145-5p/3p were elucidated in this study. Normally, miR-145 was downregulated in tumor tissues. Chivukula and colleagues (115) found that miR-145 was not expressed in colonic epithelial cells, but highly expressed in mesenchymal cells, such as fibroblasts and smooth muscle cells, this result was further validated by Kent and colleagues (116). Hence, they considered that the downregulation of miR-145 in colorectal tumor tissues was the depletion of mesenchymal cells in tumors relative to adjacent normal tissues. In addition, the underlying molecular mechanisms of the downregulation of miR-145 were also investigated in multiple previous studies. In tumor tissues and cell lines, the factors that prompted the transcription and maturation of miR-145 were downregulated, whereas the molecules that repressed the transcription and maturation of miR-145 were upregulated. These contributed to the downregulation of miR-145. Furthermore, the function of miR-145-5p could also be restrained by various ceRNAs in the cytoplasm. OCT4-pg4, a pseudogene of OCT4, holds the common binding sequence with OCT4 for miR-145-5p, and thus functions as a natural miR-145-5p sponge to protect the OCT4 transcript from being inhibited by miR-145 (85). Long noncoding RNAs (lncRNAs) are an important class of noncoding RNA and partly serve as molecular sponges to competitively inhibit miRNAs. Previous studies have demonstrated that the function of miR-145-5p could be restrained by multiple lncRNAs in various diseases, such as lncRNA-ROR in lung cancer (117), pancreatic cancer (87), endometrial cancer (86), and colorectal cancer (88); lncRNA-TUG1 in gastric cancer (118) and bladder cancer (89); and lncRNA-MALAT1 in cervical cancer (90). In addition, the biological function of miR-145-5p could also be inhibited by circular RNAs (circRNA), which is a novel class ceRNA shaped by a covalently closed loop without 5′-3′ polarity (119). Xie and colleagues (93) found that circ-001569 promoted the proliferation and invasion of colorectal cancer cell by sequestering miR-145-5p. Yu and colleagues (94) found that circBIRC6 directly interacts with miR-145-5p and miR-34a to modulate target genes that maintain human pluripotency and differentiation. The demonstrated target genes of miR-145-5p/3p and their biological functions in different neoplasms were also illustrated in this study. It was shown that numerous oncogenes that are involved in almost every aspect of tumor activity can be regulated by miR-145-5p/3p.

Although this study provides important information in recognition of the clinical value and regulatory mechanism of miR-145, many limitations in the study should be noted. First, despite 50 relevant studies being utilized, the number of studies belonging to each type of tumor was not sufficient and the sample sizes in most of the studies were small, with these factors compromising the statistical power of the meta-analysis. Second, owing to relevant studies being less than two, the prognostic value of miR-145 for some important solid tumors such as HCC, pancreatic cancer, and renal cell cancer could not be evaluated during this meta-analysis; therefore, more well-designed clinical studies with larger sample sizes for these tumors are imperative. Third, the cut-off value in each study varied, with a golden standard regarding the cut-off value should be verified to better evaluate the prognostic value of miR-145. Fourth, owing to several eligible studies not providing the survival data directly, corresponding HRs and 95% CIs were calculated from survival curves, which might cause several micro statistical errors. Fifth, although no significant publication bias was identified in this meta-analysis, potential publication bias might exist owing to desirable results being published more easily, resulting in over estimation of survival outcomes. Finally, this study only collected the regulatory mechanisms and biological functions of miR-145-5p/3p reported thus far, and more novel biological mechanisms containing miR-145-5p/3p might be unveiled in the future.

In conclusion, based on all eligible evidence, our study demonstrated that the downregulation of miR-145 was significantly associated with the poor prognosis of patients with various malignant tumors. The subgroup analyses indicated that low expression of miR-145 significantly predicted worse OS in patients with colorectal cancer, ovarian cancer, glioma, and osteosarcoma. Low expression of miR-145 was also significantly associated with PFS in patients with lung cancer and those of Asian descent. In addition, via comprehensively review previous studies, we found that miR-145 is involved in multiple tumor activities by targeting numerous genes, and the expression level of miR-145 could also be regulated by multiple factors.

No potential conflicts of interest were disclosed.

This study was supported by grants from the National Natural Science Foundation of China (grant no. 71673193) and the Key Technology Research and Development Program of the Sichuan Province (grant nos. 2015SZ0131 and 2017FZ0082).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.