Prognostic value of microRNAs in hepatocellular carcinoma: a meta-analysis

Background Numerous articles reported that dysregulated expression levels of miRNAs correlated with survival time of HCC patients. However, there has not been a comprehensive meta-analysis to evaluate the accurate prognostic value of miRNAs in HCC. Design Meta-analysis. Materials and Methods Studies, published in English, estimating expression levels of miRNAs with any survival curves in HCC were identified up until 15 April, 2017 by performing online searches in PubMed, EMBASE, Web of Science and Cochrane Database of Systematic Reviews by two independent authors. The pooled hazard ratios (HR) with 95% confidence intervals (CI) were used to estimate the correlation between miRNA expression and overall survival (OS). Results 54 relevant articles about 16 miRNAs, with 6464 patients, were ultimately included. HCC patients with high expression of tissue miR-9 (HR = 2.35, 95% CI = 1.46–3.76), miR-21 (HR = 1.76, 95% CI = 1.29–2.41), miR-34c (HR = 1.64, 95% CI = 1.05–2.57), miR-155 (HR = 2.84, 95% CI = 1.46–5.51), miR-221 (HR = 1.76, 95% CI = 1.02–3.04) or low expression of tissue miR-22 (HR = 2.29, 95% CI = 1.63–3.21), miR-29c (HR = 1.35, 95% CI = 1.10–1.65), miR-34a (HR = 1.84, 95% CI = 1.30–2.59), miR-199a (HR = 2.78, 95% CI = 1.89–4.08), miR-200a (HR = 2.64, 95% CI = 1.86–3.77), miR-203 (HR = 2.20, 95% CI = 1.61–3.00) have significantly poor OS (P < 0.05). Likewise, HCC patients with high expression of blood miR-21 (HR = 1.73, 95% CI = 1.07–2.80), miR-192 (HR = 2.42, 95% CI = 1.15–5.10), miR-224 (HR = 1.56, 95% CI = 1.14–2.12) or low expression of blood miR-148a (HR = 2.26, 95% CI = 1.11–4.59) have significantly short OS (P < 0.05). Conclusions In conclusion, tissue miR-9, miR-21, miR-22, miR-29c, miR-34a, miR-34c, miR-155, miR-199a, miR-200a, miR-203, miR-221 and blood miR-21, miR-148a, miR-192, miR-224 demonstrate significantly prognostic value. Among them, tissue miR-9, miR-22, miR-155, miR-199a, miR-200a, miR-203 and blood miR-148a, miR-192 are potential prognostic candidates for predicting OS in HCC.


INTRODUCTION
Numerous studies reported expression levels of tissue  or blood  miRNAs were related with prognosis of HCC patients.HCC is one of the most common tumors, over 700,000 new cases are reported yearly, and HCC is considered as the third primary etiology of tumor-associated mortality rate globally [222][223][224].In spite of enormous process in diagnosis and comprehensive therapy over the last few decades, HCC patients still have poor prognosis, primarily due to its high rate of recurrence [225] and metastasis [226].miRNAs, a cluster of endogenous short noncoding single strand RNAs, serve as significant posttranscriptional regulatory factor of genetic expression via interacting with the 3′-UTR of the targeted mRNAs [227].Conspicuously, due to widespread RNAase in the blood environment, circulating miRNAs displayed predominant stability.As a noninvasive detection method, circulating miRNA (blood) demonstrated more potential value as diagnostic and prognostic biomarkers than tissue miRNAs.Studies [228,229] conducted in preclinical models and cancer patients proved that malignant tumor influences expression levels of miRNAs in the blood and that certain serum miRNAs are correlated with particular cancers.Though the way requires more validation, the finding possibly discloses the avenue to a creative method of detecting cancers via measurement of serum or plasma miRNAs.
Thus far, substantial investigations have discovered that miRNAs are involved and play a crucial role in the carcinogenesis of HCC [230,231] while some miRNAs are up-regulated and others down-regulated in HCC.For example, Wong et al. [232] gained contrasting results that identifiable difference in miRNA expression pattern could not be discovered between primary HCC and venous metastases.However, comparing venous metastases to primary HCC, a prominent universal decrease of miRNA expression levels was assayed.Their results indicated that miRNA abnormality relatively early occured in liver carcinogenesis and the later universally decreased miRNA aggravated the preexisting miRNA abnormity to further accelerate HCC metastasis.
Nevertheless, there has not been a synthetic metaanalysis to assess precise prognostic value of miRNAs in HCC.As a consequence, it is of vital significance to develop a meta-analysis with an aim to evaluate it.

Study selection
Figure 1 showed a flow chart with details about the study selection process.

Study frequency
Tables 1 (tissue) and 2 (blood) showed the frequency of researches evaluating prognostic value of miRNAs, including miRNA name, number of investigations assessing prognostic value, and reference.

Study characteristics
Supplementary Table 1 comprehensively presented the characteristics and details (names of miRNAs, information about the included articles, detected samples, sample size, stage, cut-off value, detection methods, follow-up, survival outcome with HR and 95% CI) of studies with Kaplan-Meier survival curves (K-M curves) in HCC.If the survival outcome was not furnished directly and merely as K-M curves, we used the software Engauge Digitizer version 4.1 [233] to extract the data from K-M curves.Additionally, if both the univariate and multivariate outcomes were covered, we just chose the latter in that the confounding factors were corrected.

Meta-analysis
Table 3 presented a summary of the HR estimated from pooled analysis for the included miRNAs.A total of 16 miRNAs were screened by our present study.

Publication bias
For the purpose of evaluating publication bias on OS of HCC patients with high tissue miR-21 level, we employed the Begg's funnel plot (Figure 2B).Accordingly, the P value was 0.21, suggesting nonexistent publication bias.

Sensitivity analysis
The sensitivity analysis did not manifest variances among the outcomes in terms of the exclusion of any single research (Figure 2C) in the estimation on OS of HCC patients with high tissue miR-21 level, indicating that no individual investigation significantly affected the merged HR with 95% CI.

Publication bias
For the sake of estimating publication bias on OS of HCC patients with high blood miR-122 level, we employed the Begg's funnel plot (Figure 3B).Consequently, the P value was 0.56, suggesting nonexistent publication bias.

Sensitivity analysis
The sensitivity analysis did not manifest variances among the outcomes in terms of the exclusion of any single research (Figure 3C) in the estimation on OS of HCC patients with high blood miR-122 level, indicating that no individual investigation significantly affected the merged HR with 95% CI.

Meta-regression
We employed the meta-regression to seek source of heterogeneity (I 2 = 84.8%) on OS of HCC patients with high blood miR-122 level.The details were shown in Table 4, and source of heterogeneity was

Status quo
Numerous articles reported that dysregulated expression levels of miRNAs correlated with survival time of HCC patients .Nevertheless, there has not been a comprehensive meta-analysis to assess the accurate prognostic value of miRNAs in HCC.Therefore, it was conducted to estimate the relationship between dysregulated miRNA level and survival time of HCC patients.

Molecular mechanisms for included miRNAs
For included miRNAs in the current study, a summary of miRNAs with changed levels, their possible targets and pathways enrolled in the present study has been presented in Table 5.From the data of the table, these potential targets and pathways may be involved with survival outcome of HCC patients.

Strengths of the meta-analysis
There are a few strengths in this study, which are as follows: (1) nearly all articles estimating associations between miRNA level and survival result of HCC patients are shown in the current meta-analysis; (2) the number about HCC patients of all researches included in this study are more than or equal to 30, which makes the metaanalyses more convincing; (3) the Begg's funnel plot and sensitivity analysis were used for miR-21 and miR-122, which excluded publication bias and excessive influence of individual study; (4) we employed meta-regression to seek source of heterogeneity, which indicated that months of follow-up were significantly associated with it; (5) studies merely proposing HR or 95% CI without K-M curves were excluded.

Limitations
Simultaneously, there are also limitations for the current work: (1) only English articles were included by us, which possibly excluded some studies written in other languages; (2) not all the articles assessing associations between miRNA level and survival time were included

Implications for future clinical and scientific research
With expression profiles shown in Tables 1 and 2, we can conveniently find relevant article about a single miRNA.
Thus, the present study tendency for miRNAs in HCC can be easily obtained by basic researchers.Meanwhile, combined detection of multi-miRNAs can greatly increase the predict level for HCC patients.Besides, for clinical doctors, combined use of tissue and blood from HCC patients can bring about synergistic effect to estimation of prognosis.

Search strategy, inclusion criteria and exclusion criteria
The details were presented in Table 6.Two authors (Yue Zhang and Chao Wei) independently performed this comprehensive online search.

Quality assessment
Yue Zhang and Chao Wei confirmed all eligible investigations that analyzed the prognostic value of miRNAs in HCC, and Yue-Hua Jiang reassessed uncertain data.

Statistical analysis
All analyses were conducted using Stata version 13.0 (StataCorp, College Station, Texas, USA).The relative effect sizes for HR were characterized as moderate (protective [0.51-0.75]or contributory [1.35-1.99])and large (≤ 0.50 or ≥ 2).The HR was considered significant at the P < 0.05 level if the 95% CI did not include the value 1.If the P values from OS and other survival results about corresponding miRNAs were inconsistent, the HR from OS was considered to the main reference standard.Because different types of samples (tissue, plasma and serum) from HCC patients at different disease stages, cut-off values and miRNA methods were used in individual studies, random-effects models (DerSimonian-Laird method) were more appropriate than fixed-models (Mantel-Haenszel method) for most of the analyses.Consequently, the random-effects models were used in the current meta-analysis.Source of heterogeneity was explored through meta-regression.Publication bias was estimated using the Begg's funnel plot.A two-tailed P value < 0.05 was considered significant.Sensitivity analysis (influence analysis) was carried out to test how powerful the combined effect size was to removal of individual investigation.If the point assessment was out of the 95% CI of the pooled effect size after it was removed from the analysis, an individual study was doubted to have excessive influence.

Figure 1 :
Figure 1: Flow diagram of literature search and selection.

Figure 2 :
Figure 2: (A) Forest plot of the analyses about high expression of tissue miR-21 and OS, DFS or OS (multivariate analysis); (B) Publication bias of the analysis about high expression of tissue miR-21 and OS and (C) Sensitivity analysis of the study about high expression of tissue miR-21 and OS.

Figure 3 :
Figure 3: (A) Forest plot of the analyses about high expression of blood miR-122 and OS or DFS; (B) Publication bias of the analysis about high expression of blood miR-122 and OS and (C) Sensitivity analysis of the study about high expression of blood miR-122 and OS.

Figure 4 :
Figure 4: Forest plot of the analyses about high expression of tissue miR-9 or low expression of tissue miR-22, 29a, 29c and OS, DFS/RFS or RFS.

Figure 6 :
Figure 6: Forest plot of the analyses about high expression of tissue miR-221 or low expression of tissue miR-199a, 200a, 203 and OS, RFS or RFS/MFS/DFS.

Figure 7 :
Figure 7: Forest plot of the analyses about high expression of blood miR-21, 29a, 192, 224 or low expression of blood miR-148a and RFS/DFS, OS or PFS/DFS.

Table 1 : Frequency of studies estimating prognostic value of tissue miRNA expression in hepatocellular carcinoma
Highlighted studies were included in the present meta-analysis; N: Number of studies estimating prognostic value; R: References.

Table 2 : Frequency of studies estimating prognostic value of blood miRNA expression in hepatocellular carcinoma
Highlighted studies were included in the present meta-analysis; N: Number of studies estimating prognostic value; R: References.

Table 3 : Summary of the HR for miRNA expression in hepatocellular carcinoma
HR: hazard ratios; CI: confidence intervals; OS: overall survival; DFS: disease-free survival; RFS: recurrence-free survival; PFS: progression-free survival; MFS: metastasis-free survival; m Multivariate analysis.

Table 3 and
showed the details.

Table 5 : Summary of miRNAs with altered expression, their potential targets and pathways entered this study
the present study, which might neglect some potential miRNAs; (3) a few variables emerged among the included investigations, including different kinds of samples from HCC patients at different stages, cut-off values and detection methods, and only random-effects models were employed for all meta-analyses; (4) although overall studies included 54 relevant articles and 6464 patients in the present study, the number of articles and patients may be not enough for 16 miRNAs focused on. in

Table 6 : Information of search methods and criteria of inclusion and exclusion Methods Information
(2)Frequency of studies estimating prognostic value of miRNAs ≤ 2 in tissue;(3) Studies which cannot be merged; (4) If more than one article had been published on the identical study cohort, only the most comprehensive study was selected for the present meta-analysis