Yu-Rong Wang,Xiao-Qian Hao,Qian Xiang,Yu-Sheng Liu,Jie-Yan Wu

1College of traditional Chinese Medicine,Jinan University,Guangzhou 510632,China.

Abstract

Key words: Diagnosis,MicroRNA-155,Biomarker,Breast cancer

Background

Breast cancer (BC) is the leading cause of cancerous death among females with the highest diagnostic frequency.About 63,410 cases of female breast carcinoma in situ are expected to be diagnosed in 2017 in America,and it alone is anticipated to account for 30% of all new cancer diagnoses in women [1].There are many diagnostic tools for detection of BC,such as imaging examination and biopsy,although there are still rooms for improvement [2,3].In recently years,the prediction on the recurrence or progression of BC and the detection of it on the early stage is more and more commonly conducted using tumorous bio-markings of tissues and serum.Nonetheless,their use does not allow an accurate prediction of treatment resistance or relapse[4,5].The occurrence and development of BC is a multistep,multistage complicated process involving multiple factors,among which genetic factors are considered to play a crucial role [6,7].Consequently,it is in urgent need to develop new and economical ways to diagnose BC and predict prognosis,so as to provide precise and personalized treatment for patients.

It is widely accepted that cancer therapy has entered a new stage since 1993 when microRNAs were discovered in studies onCaenorhabditiselegans [8],and new opportunities of diagnosis and therapy have been introduced due to exploded information nowadays.miRNAs are small and noncoding RNA molecules with length ranged from 20 to 25 nucleotides [9].The binding area of miRNAs is mainly the 3’ untranslated region of mRNAs and the function of it is degrading target mRNAs or inhibiting translation to down-regulate target proteins [10,11].They are known to be dominated by genes which participate in the following biological processes,including migration,apoptosis,differentiation,stress reaction,cell cycle as well as an inflammatory reaction [12-14].Hence,it can be seen that miRNAs exert an essential role in regulating almost entire signal pathways within normal and dysregulated cells.miRNAs are closely related to the processes that cancers develop and progress [15-17].A more interesting thing is that,miRNAs have been proved to exist in both tissues and body fluid such as plasma [18,19].The latter provides a possibility that miRNAs might be used as not only bio-markers for non-invasive diagnosis and prognosis but also targets for therapy[20].Taken the fact that miRNAs exhibit high stability in the blood circulation system and tissue samples fixed by formalin and coated by paraffin in consideration,it is recognized by more and more researchers that miRNAs are becoming a powerful tool for diagnosing cancer as well as predicting the prognosis and reaction to cancer treatments.Several studies have reported the potential of using miRNAs as novel diagnostic markers for BC [21-26].MicroRNA-155 (miR-155) is one of the most frequently studied oncomiRNAs.In recent years,downstream targets of miR-155 have widely been investigated in several signal pathways,which may provide evidence for its potential clinical use at a molecule level.miR-155 prevents SOCS1 and stimulates activity of the JAK2/STAT3 pathway,which lead to tumor growth [27].A finding indicates that miR-155 has a pivotal role in tumor angiogenesis by downregulation of Von Hippel-Lindau,and provides a basis for miR-155-expressing tumors to embody an aggressive malignant phenotype [28].In contrast,FOXP3 activation ofMIR155HGexpression in normal epithelia may prevent the transformation of the healthy breast epithelium to a cancerous phenotype [29].In BC tissue,the overexpression of miR-155 was observed[30].Recently,several studies have shown that miR-155 is detected at a high level in the serum of patients with BC [18,31,32].However,the value of miR-155 in the application for BC diagnosis varied in the different studies.Other studies reported no significant association between serum miR-155 expression and clinicopathologic features such as hormone receptors,histologic grade and lymph node metastasis [33,34].Thus,the relationship among miR-155,BC diagnosis and other factors needed to be clarified beyond the limits of these single studies.We conducted a metaanalysis to evaluate the diagnostic values of miRNA-155 for detecting BC.

Materials and methods

Search strategy

The resources of literature included databases such as VIP database,WanFang,China National Knowledge Infrastructure,Cochrane Library,PubMed,Web of Science,Embase with search terms of “(microRNA-155 OR miR-155 OR mir155) AND (breast cancer OR breast neoplasm OR breast tumor OR breast carcinoma OR breast tumour) AND (diagnosis OR ROC curve OR sensitivity OR specificity OR diagnostic OR diagnose)”.The search was updated up until April 10,2020.

The criterions that qualified studies included in the meta-analyses need to meet are as follows: (1) studies in respect to the diagnosis capacity of miRNA-155 for BC;(2) BC was diagnosed on the basis of histopathologic evidence,which are extensively accepted as golden standards;(3) studies which contained enough data to construct 2 × 2 tables,including false negative;true negative,false positive as well as true positive;and (4) healthy participants or patients suffering from benign diseases made up the control group.Exclusion criteria were as follows: (1)reviews or letters;(2) repetitive publication;(3)literature that mainly concerning survival rate or prognosis of BC;and (4) insufficient data.If the content of a study arouses controversy,it would be discussed until agreement was reached.All studies were qualified as long as the criterions mentioned above were met.

Data extraction

Related data was collected using a standard sheet.After 1 author extracting data,the other 2 authors checked the results by themselves.The information was collected as follows: values of a diagnostic 4-fold contingency table(true positives,false positives,false negatives,true negatives),detection methods,sample source,sample capacity,race,country,edition year,first author.The authors of the selected articles were also connected to the great extent and disputes were settled by discussion among authors.

Quality assessment

The Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) was applied to assessing the quality of chosen literatures [35].In QUADAS-2 tool,patients through the study and timing of the index tests and reference standard (flow and timing),reference standard,index test as well as patient selection were discussed by four key domains and each of which contained 2 sections including applicability and risk of bias.The options of “yes”,“no”,or “unclear” were used to answer each question with a score of 1,0,0,respectively.

Statistical analysis

The application value of miR-155 for diagnosing BC was assessed through calculating corresponding 95%CI,diagnostic odds ratio (DOR),pooled specificity and the pooled sensitivity.The property of diagnostic tests was evaluated by DOR and summary receiver operator characteristics (SROC) and the heterogeneity was evaluated by I2statistic and Q test.If I2＞ 50% orP＜0.05 which indicated heterogeneity,a model of random effect was used.Conversely,ifP≥ 0.05 and I2≤ 50%,a model of fixed effect was used.The source of heterogeneity was investigated by meta-regression and sensitivity analyses.The funnel was used to analyze publication bias.The application value of miR-155 diagnosing BC was described by Fagan’s nomogram.We tested the potential for publication bias by using Deeks’ funnel plots [36].All these analyses were carried out by Meta-Disc statistical software [37] and Stata 12.0(Stata Corporation,College Station,TX,USA).

Results

Study selection and quality assessment

The flow path of this study is shown in Figure 1.Totally,714 studies were selected after the above-mentioned databases were searched,of which 263 were duplicates.From the remaining 451 publications,109 studies concerning cancer except BC,136 non-diagnostic studies,86 studies without content concerning miR-155 and 87 non-clinical studies were eliminated from the analyses.Thirty-three articles remained and their entire characters were screened.After that,13 reviews or letters,5 without complete data and 3 data duplication were eliminated.Finally,11 eligible studies were selected for the meta-analysis [34,38-47].The main basic information of the chosen articles was shown in Table 1.The assessing tool of QUADAS-2 was utilized to evaluate the 11 chosen articles.On all accounts,the majority of these articles attained relatively high scores.Quality assessment was conducted with a bar graph in Figure 2.

Data analysis

The specificity and sensitivity of miR-155 were analyzed using forest plot and the results are shown in Figure 3.The negative likelihood ratio (NLR),positive likelihood ratio (PLR),specificity,sensitivity were 0.17(95 % CI,0.09-0.31),4.75(95 % CI,2.98-7.57),0.82(95 % CI,0.72-0.89) and 0.86 (95 % CI,0.76-0.93),separately,which proved the advantages of miR-155 as the indicator of BC (Figure 3,4).On the other hand,the I2value of NLR,PLR,specificity and sensitivity were 95.85% (P= 0.00),90.25% (P= 0.00),89.00% (95 %CI,84.41-93.59;P= 0.00),94.93 % (95 % CI,93.24-96.61;P= 0.00),which indicated that the data in the current study was significantly heterogeneous.Consequently,the model of random effects were chosen for further analyses.

The SROC curve was plotted and the area under the summary receiver operating characteristic (AUC) was calculated to be 0.90 (95% CI,0.88-0.93),indicating an overall high diagnostic accuracy as a diagnostic test(Figure 5).Moreover,the pooled DORalso was calculated 28.63 (95% CI,11.22-73.04) (Figure 6).

Analysis diagnostic threshold effect

The threshold effect and the nonthreshold effect are 2 important causes of heterogeneity in diagnostic tests.The threshold effect is often evaluated by the Spearman correlation coefficient of specificity and sensitivity[37].The existence of the threshold effect was denied since the corresponding Spearman correlation coefficient was -0.345 with aP-value of 0.328.Additionally,in our study,the value of I2＞ 50%indicated that the heterogeneity by the nonthreshold effect also existed among these studies [46].

Meta-regression analysis

In order to illustrate the source of heterogeneity,we conducted the meta-regression by exploring study characteristics,which include year,country,ethnicity and specimen.It was revealed by the results that these factors were not significantly heterogeneous (Table 2).

Table1 Summary of studies using miR-155 as a biomarker of BC and study quality assessment

Sensitivity analysis

The random-effect bivariate model was robust for the calculation of the pooled estimates by executing the goodness of fit and bivariate normality analyses (Figure 7a,b).One outlier study was assessed by outlier detection (Figure 7d) and influence analysis (Figure 7c).

But the overall parameters did not demonstrate evidential changes when the outlier study was eliminated,which indicated that data of individual articles could not affect the integral pooled estimated value.

Publication bias

The publication bias among the chosen articles was assessed using Deek’s funnel plot in meta-analysis.The existence of significant publication bias was denied due to the symmetry of the funnel plot and aP-value of 0.14(Figure 8).

Clinical utility and index test

The application value of miR-155 on diagnosing BC was described by Fagan’s nomogram (Figure 9).When 50% was selected as the pretest probability,the data indicated that,if the result of a diagnostic test was positive,the post-test probability that the individual suffered BC would be approximately 83% (red line).If the result was negative,the post-test probability that the participant was affected with BC would be approximately 14% (blue line).That suggesting that miR-155 was a promising indicator for the diagnosis of BC.

Discussion

Recently,in the wake of development in early-stage diagnosing and treating technologies,the death rate of breast carcinoma has been decreased.Yet,there are still several thousands of women suffering from BC each year in the world.Contemporarily,histological evaluation of biopsy is the gold diagnosis standard for BC.Whereas,it fails to diagnose cancer at an early stage.Mammography x-ray photography is extensively applied to BC screening and detection,but its usage is restricted in irradiation [49].Although some biomarkers such as progesterone receptor and estrogen receptor are helpful in the treatment of BC,they are not sensitive and accurate enough to predict the prognosis of BC.It has been verified continually that miRNAs that exert an essential role in regulating vast biological processes participate in pathogenic mechanisms [50].miRNAs have high stability and are highly conserved in tissues and cells with excellent convenience of detection and extraction [51],and can tolerate repeated freezing and thawing which make them excellent candidate biomarkers for detecting early-stage BC,predicting prognosis and selecting optimal treatment [53,53].

Table2 Univariate meta-regression of miRNA-155 for the sensitivity and specificity of miRNAs for diagnosis of BC

A number of studies have built a pivotal role of miR-155 in breast carcinogenesis [27,54].Changs et al.[55]confirmed that the mutated BRCA1(breast carcinoma1,early onset) would trigger miR-155-5p’s overexpression.Neilsen and co-workers indicated that MIR155HGmay act as a target in the downstream signaling pathway of mutated p53,since the expressions of both are related to invasiveness inside the living body[56].Correspondingly,miR-155 has been observed to be overexpressed in BC [31,40,45,57].

Recently,only 1 meta-analysis on the application value of miR-155 on the diagnosis of breast carcinoma was published.Dr.Wang and Dr.Hou [58] had evaluated 3 BC studies and showed that miR-155 is highly sensitive 79% (95% CI,72%-84%) and specific 85% (95% CI,75%-92%) to diagnosis of breast carcinoma thus possessing application value (58).Nevertheless,this meta-analysis contained merely 3 works of literature in which 2 contained comparatively few sample capacities,which might contribute to the instability of results and decreased importance of metaanalysis.Additionally,a similar study reported by Dr.Hou had also performed in the same way to analysis the diagnosis ability of miR-155 in several various malignant tumors,like pancreatic ductal adenocarcinoma,non-small cell lung cancer,and BC.This study was intended to assess the role of miR-155 in the diagnosis of multiple tumors,not only of BC,although they used subgroup analysis with a few sample sizes [59].

It was revealed by the results of a meta-analysis that the pooled specificity and sensitivity were 0.82 (95%CI,0.72-0.89) and 0.86 (95% CI,0.76-0.93),respectively,which indicated favorable application value of miR-155 on diagnosing BC.The property of diagnostic tests was represented by DOR and area under SROC (AUC).The interval of DOR is 0 to infinity and the discrimination of test is getting better along with the increase of DOR values [60].Ideally,the SROC curve is located beside the upper and left corner so that the result of the test would be desirable [61].The AUC and DOR of miR-155 were 0.90 (95% CI,0.88-0.93) and 28.63 (95% CI,11.22-73.04),respectively.Additionally,likelihood ratios (NLR,PLR) are clinic measurement criteria of diagnosis accuracy which exhibit higher significance and usefulness,presenting miR-155 has favorable test performance.Furthermore,it was shown by the results of the meta-analysis that the value of pooled PLR was 4.75,which suggested that the probability of a BC patient being diagnosed as positive was five folds to that of a normal woman.The pooled NLR of 0.17 (95% CI,0.09-0.31),which presents a cancer probability of 17% on the basis of normal miR-155 level in peripheral blood.In spite of imperfect NLR and PLR values (PLR ＞ 10,NLR ＜ 0.1) [62],miRNA-155 is still satisfactory for diagnosing BC.

Heterogeneity should be taken into account as it significantly affects the diagnostic value of miR-155.Since heterogeneity was obviously revealed by test results,the sources were investigated in the current study.Primarily,although the threshold effect is considered as a dominating source of heterogeneity among studies of test accuracy [37],the Spearman correlation coefficient turned out to be -0.345 (P=0.328),suggesting that threshold effect was excluded from the factors causing heterogeneity in the current study.Secondly,meta-regression analysis was utilized to figure out whether heterogeneity was caused by any single article.It was unfortunate that no source was of heterogeneity found.Other than those factors,one outlier study was excluded that could interpret interstudy heterogeneity.Consequently,there was a minimal change of heterogeneity,demonstrating that our results were not driven by the outlier.

Although this meta-analysis suggested a prognostic role of miRNAs in diagnosing the outcome of BC patients,several limitations of this meta-analysis must be pointed out.Initially,only 5 articles on the basis of Caucasian and 5 on the basis of Asians were included,indicating probably insufficient sample capacity in statistics.Secondly,heterogeneity existed in the analysis and could not be explained by the results of meta-regression and sensitivity analyses.Meanwhile,although the diagnosis values of miR-155 were moderate,based on the high accuracy criterions (NLR ＜0.1,PLR ＞ 10) [62],the performance of miR-155 was not as satisfactory as expectation.

The application value of miR-155 as an innovative bio-marker for diagnosing and expecting prognosis of breast carcinoma was evaluated by meta-analysis.In conclusion,miR-155 could be utilized as a complementary method of initial cancer and noninvasive screening biomarkers for cancer detection in clinical practice.Additionally,large appropriately designed prospective studies,both clinical trials and observational cohorts employing standard methods,are peremptorily needed to confirm the conclusion of this study.