Circulating MicroRNA Biomarkers for Lung Cancer Detection in East Asian Populations

Background: Lung cancer (LC) is the leading cause of cancer-related death in Eastern Asia. The prognosis of LC highly depends on tumor stages and early detection could substantially reduce LC mortality. Accumulating evidence suggested that circulating miRNAs in plasma or serum may have applications in early LC detection. We thus conducted a systematic literature review on the diagnostic value of miRNAs markers for LC in East Asian populations. Methods: PubMed and ISI Web of Knowledge were searched to retrieve relevant articles published up to 17 September 2018. Information on study design, population characteristics, investigated miRNAs and diagnostic accuracy (including sensitivity, specificity and area under the curve (AUC)) were independently extracted by two reviewers. Results: Overall, 46 studies that evaluated a total of 88 miRNA markers for LC diagnosis in East Asian populations were identified. Sixteen of the 46 studies have incorporated individual miRNA markers as panels (with 2–20 markers). Three promising miRNA panels with ≥90% sensitivity and ≥90% specificity were discovered, two of which were externally validated. Diagnostic performance of circulating miRNAs in East Asian populations was comparable to previously summarized performance in Western populations. Forty-four miRNAs were reported in both populations. No major differences in diagnostic performance by ethnicity of the same miRNA was observed. Conclusions: Circulating miRNAs or miRNA panels, possibly in combination with other promising molecular markers including epigenetic and genetic markers, may be promising candidates for noninvasive LC early detection. However, large studies with samples collected prospectively in true screening settings are required to validate the promising markers or marker panels.


Introduction
Lung cancer (LC) is the leading cause of cancer mortality in Eastern Asia, with 950,015 cases and 815,635 deaths estimated in 2018 [1]. Although advances in therapy have led to improvements in survival of LC patients [2], the 5-year survival rate remains very low, mainly due to late diagnosis of disease [3]. Traditional screening methods such as chest radiography and sputum cytology have limited clinical applications as they display low sensitivity and specificity in detection of LC [4]. In recent years, Low-dose CT has been recommended for LC screening in high-risk smokers. However, potential hazards of CT screening, including radiation exposure, high false-positive rates, overdiagnosis and Our literature search initially yielded 1543 articles using the search terms mentioned above, 551 from PubMed and 992 from Web of Science ( Figure 1). After removal of 340 duplicate articles, titles and abstracts of 1203 articles were carefully reviewed. A total of 1097 articles were excluded according to the above described exclusion criteria. The remaining 106 articles were selected for full text reading, of which 60 articles were removed: 24 using disease controls, 16 not reporting sensitivity, specificity or AUC values, and 20 reporting studies in non-East Asian countries. In the end, 46 studies that evaluated diagnostic performance of circulating miRNAs for LC in East Asian populations and published between 2011 and 2018 (Tables 1 and 2) were eligible for this systematic review [8][9][10]. Of the included 46 studies, 45 were from China and one was from Japan [29].

Study Quality and Characteristics
QUADAS-2 assessment was completed by two reviewers (H.Y. and Z.G.) independently. Any inconsistencies were discussed and resolved between the investigators. High applicability concerns were found in the patient selection domain in 10 (22%) of the included studies. Unclear risk of bias were found in the patient selection domain and the index test domain in 24 (52%) and 6 (13%) of the included studies, respectively. No risk of bias or applicability concern was found in the reference standard domain, and the flow and timing domain. Details of QUADAS-2 results of the 46 studies are displayed in Figures S1 and S2.
All 46 studies were case-control studies in which blood samples were collected after disease diagnosis. Of the 46 studies, 43 evaluated individual miRNAs, eight of which conducted independent validation (Table 1). Sixteen studies assessed miRNA panels, 13 of which carried out independent validation ( Table 2). Detailed information on each study, including the number of cases and controls, mean or median age, male proportion, specimen type, histological subtype, tumor stage, and diagnostic indicators, are summarized in Tables 1 and 2. Table 1 additionally shows the p-value for testing the difference of each individual miRNA between cases and controls or the statistical significance of AUC values (indicated in the footnotes of Table 1).

Direction of Dysregulation of Circulating miRNAs
Of the 46 studies, 45 studies described the direction of dysregulation of miRNAs in blood, and only one study did not report information on miRNA dysregulation (Table S1). Among the 22 miRNAs reported in ≥2 studies, inconsistent directions were reported for six miRNAs (Table S3), whereas consistent direction of dysregulation was observed for most of the markers. For example, up-regulation of miR-24, miR-223, miR-155 and miR-221, and down-regulation of miR-126 were consistently reported in all studies reporting on these miRNAs regardless of histological subtype, stage or sample type (Table S1). In addition, miR-21, the most frequently reported miRNA, was also up-regulated in both plasma and serum samples in all included studies, except one study reporting down-regulation of miR-21 in exosome [8].

Comparison of miRNAs Profiles for LC Detection between Eastern Asian and Western Populations
We previously performed a systematic review of literature on circulating miRNAs for lung cancer detection in Western populations [12]. A further update of this review until September 17th, 2018 identified a total number of 110 miRNAs, 34 of which were reported in ≥2 studies. We plotted Venn diagrams to illustrate the overlap of identified miRNAs between East Asian and Western populations. Forty-four miRNAs were reported in both populations ( Figure 3A), which accounted for 50% of miRNAs reported in the East Asian populations, and 40% of miRNAs reported in the Western populations. Restricting miRNAs reported ≥2 times in each type of population, there were 12 overlapping miRNAs ( Figure 3B), which account for 55% of miRNAs reported in East Asian and 35% of miRNAs reported in Western populations. Sixteen miRNAs were evaluated individually in both East Asian and Western populations, AUC values of which are shown in Figure 4. For some miRNAs showing differences in diagnostic performance were observed between East Asian and Western studies, but such differences were also seen within each type of study populations. Detailed information of the 16 miRNAs is provided in Table S4. Of the 22 miRNAs reported at least twice, miR-21 was most frequently reported (13 studies), followed by miR-145, miR-20a, miR-24, miR-223, miR-155, miR-25, miR-152, miR-125a-5p, miR-126, miR-221 and miR-93 (all three to five studies) ( Table 3). Of note, higher frequency of reporting of the investigated markers does not go along with higher values of diagnostic performance parameters. For example, the median sensitivity and specificity of miR-21 were 69% (46.3-91%) and 71.9% (60-92%) respectively.

Direction of Dysregulation of Circulating miRNAs
Of the 46 studies, 45 studies described the direction of dysregulation of miRNAs in blood, and only one study did not report information on miRNA dysregulation (Table S1). Among the 22 miRNAs reported in ≥2 studies, inconsistent directions were reported for six miRNAs (Table S3), whereas consistent direction of dysregulation was observed for most of the markers. For example, up-regulation of miR-24, miR-223, miR-155 and miR-221, and down-regulation of miR-126 were consistently reported in all studies reporting on these miRNAs regardless of histological subtype, stage or sample type (Table S1). In addition, miR-21, the most frequently reported miRNA, was also up-regulated in both plasma and serum samples in all included studies, except one study reporting down-regulation of miR-21 in exosome [8].

Comparison of miRNAs Profiles for LC Detection between Eastern Asian and Western Populations
We previously performed a systematic review of literature on circulating miRNAs for lung cancer detection in Western populations [12]. A further update of this review until September 17th, 2018 identified a total number of 110 miRNAs, 34 of which were reported in ≥2 studies. We plotted Venn diagrams to illustrate the overlap of identified miRNAs between East Asian and Western populations. Forty-four miRNAs were reported in both populations ( Figure 3A), which accounted for 50% of miRNAs reported in the East Asian populations, and 40% of miRNAs reported in the Western populations. Restricting miRNAs reported ≥2 times in each type of population, there were 12 overlapping miRNAs ( Figure 3B), which account for 55% of miRNAs reported in East Asian and 35% of miRNAs reported in Western populations. Sixteen miRNAs were evaluated individually in both East Asian and Western populations, AUC values of which are shown in Figure 4. For some miRNAs showing differences in diagnostic performance were observed between East Asian and Western studies, but such differences were also seen within each type of study populations. Detailed information of the 16 miRNAs is provided in Table S4.

Discussion
Our systematic literature review identified 46 studies that evaluated a total of 88 miRNA markers for LC diagnosis in East Asian populations. Sixteen of the 46 studies have incorporated individual miRNA markers as panels (with 2-20 markers). Three promising miRNA panels with ≥90% sensitivity and ≥90% specificity were discovered, two of which were verified externally [33,54]. Diagnostic performance of circulating miRNA in East Asian populations was comparable to diagnostic performance in Western populations. Forty-four miRNAs were reported in both populations. No major differences in diagnostic performance by ethnicity of the same miRNA was observed.
In general, the performance of the investigated miRNAs and miRNA panels for detecting LC in East Asian populations appears promising, and in most cases the sum of the sensitivity and specificity by far exceeded 100% (Figure 2A). There were 17 individual miRNAs and 11 miRNA panels for which both sensitivity and specificity above 80% were reported ( Figure 2B). Several miRNA panels even showed rather good diagnostic efficiency. For instance, Fan et al. [33] used a panel composed of miR-20a-5p, miR-16-5p and miR-15b-5p in serum to discriminate NSCLC cases from healthy controls, and the sensitivity and specificity reached 94% and 94%, respectively, in the validation set. Chen et al. [54] used a 10-miRNA panel in serum, and reported 93% sensitivity and 90% specificity in the validation set. Several miRNA panels were reported to be useful for the detection of early stage LC,

Discussion
Our systematic literature review identified 46 studies that evaluated a total of 88 miRNA markers for LC diagnosis in East Asian populations. Sixteen of the 46 studies have incorporated individual miRNA markers as panels (with 2-20 markers). Three promising miRNA panels with ≥90% sensitivity and ≥90% specificity were discovered, two of which were verified externally [33,54]. Diagnostic performance of circulating miRNA in East Asian populations was comparable to diagnostic performance in Western populations. Forty-four miRNAs were reported in both populations. No major differences in diagnostic performance by ethnicity of the same miRNA was observed.
In general, the performance of the investigated miRNAs and miRNA panels for detecting LC in East Asian populations appears promising, and in most cases the sum of the sensitivity and specificity by far exceeded 100% (Figure 2A). There were 17 individual miRNAs and 11 miRNA panels for which both sensitivity and specificity above 80% were reported ( Figure 2B). Several miRNA panels even showed rather good diagnostic efficiency. For instance, Fan et al. [33] used a panel composed of miR-20a-5p, miR-16-5p and miR-15b-5p in serum to discriminate NSCLC cases from healthy controls, and the sensitivity and specificity reached 94% and 94%, respectively, in the validation set. Chen et al. [54] used a 10-miRNA panel in serum, and reported 93% sensitivity and 90% specificity in the validation set. Several miRNA panels were reported to be useful for the detection of early stage LC, with verified AUC values over 0.85 [8,10,48]. However, most studies evaluated miRNAs or miRNA panels in samples of stage I-IV LC cases, the diagnostic efficiency of the reported markers or panels for early stage LC needs to be validated in true screening settings. Differences in the subgroup analyses for histology and stage of LC were rather small and relevant data were not sufficient to obtain robust results.
Aberrant expression of specific circulating miRNAs may provide important information for distinguishing LC histological subtypes or stages. However, subgroup analyses with respect to LC histological subtypes or stages have only been performed in a small proportion of the included studies [17,26,28,30,34,39,45,48,50,53]. Even though histology-specific analysis identified variations of diagnostic indicators of miRNA in different subtypes, no histology-specific miRNA was discovered due to the overlap in profiles. Consistent findings were also observed in studies conducted in the Western populations. Stage-specific analyses in both the Eastern Asian and Western populations showed that diagnostic efficacy was better in advanced stage compared to early stage; however, the differences with respect to AUC were quite small.
Over the past decade, studies focusing on miRNA and LC have been emerging rapidly and have identified a number of LC-specific miRNA expression profiles. However, these miRNA expression profiles are not consistent among different studies [58][59][60][61]. In this systematic review, the degree of overlap of LC-specific miRNAs reported by various studies was found to be low and inconsistent directions of dysregulation of miRNAs was observed. Similar findings were also noted among the Western studies [12]. To be used as biomarkers for LC screening, miRNAs should show consistent degrees and direction of dysregulation in different settings [62,63]. Possible sources of inconsistency of results across studies could include the use of different biospecimen or different analytical platforms. Ideally, future studies should involve different types of biospecimen and different types of analytical platforms in the same study population in order to further elucidate the role of these factors in miRNA profiling.
Further causes of the heterogeneity of reported miRNA biomarkers may be the differences in the characteristics of study populations. For the included studies in both East Asian and Western populations, we extracted demographic information such as age, sex and country, as these factors may affect the identification of miRNA markers [11,64,65]. Cigarette smoking could also induce the dysregulation of some circulating miRNAs, which might be associated with smoking-related LC [47,66,67]. However, information on smoking status was incomprehensive in most studied populations, especially in controls. Sample sizes of the majority of both East Asian and Western studies were rather small, which might have resulted in substantial random variation of LC-specific miRNAs identified among studies. In addition, miRNA profiles may change in different phases of cancer [68]. Most studies in both East Asian and Western populations however recruited LC cases with various stage proportions.
The use of different miRNA detection protocols could also affect the identification of LC-specific miRNAs. Pre-analytical factors such as speed and duration of centrifugation may affect the amount of cell debris remaining in the sample supernatant (especially in plasma samples) and cause miRNA contaminations [69][70][71][72]. Only a fraction of included studies in both East Asian and Western populations have applied a second high-speed centrifugation step to reduce the remaining cell debris (Table S5). Moreover, reports on most studies did not specifically address potential hemolysis, which could result in an increase of multiple miRNA levels [70,71,73,74]. Analytical factors such as differences in miRNA extraction and quantification methods could also contribute to the variability of identified LC-specific miRNAs [75]. The extraction methods in the included studies were diverse (Table S5), only few studies in both East Asian and Western populations have used miRNeasy kit which has been suggested to have better extraction efficiency compared to other kits [76,77]. In addition, a very important yet unresolved issue is the normalization of miRNA expression data. Circulating miR-16 is usually used as endogenous control for normalization (Table S5), however, it shows an unstable expression in the circulation of cancer patients as well as in hemolytic samples [71,78,79]. Several new bioinformatics tools such as miRNA ratios and differentially expressed miRNA pairs have been developed to build up miRNA panels to reduce the analytical bias [80][81][82]. These attempts so far have shown promising results but still require further validation.
The herein described circulating miRNAs have several advantages over some other markers of cancer detection: (i) miRNAs are extremely stable in cell-free fluids and in less than ideal sample handling conditions [79,83,84], (ii) they can be measured repeatedly over time in a non-invasive manner [85], (iii) they can be used to predict cancer in high-risk populations years in advance [80,82,86] and (iv) the cost of analysis is relative low. Still, other blood-based markers have been proposed for LC detection, such as cytokeratin 19 fragment 21-1 (CYFRA21-1), carcinoembryonic antigen (CEA), tissue polypeptide specific antigen (TPS) or neuron-specific enolase (NSE). However, these markers are primarily utilized for monitoring of disease progression and tend to show suboptimal diagnostic value for LC with sensitivity values usually under 50% [81,87]. Sanfiorenzo et al. [88] developed a signature of 11 plasma miRNAs to discriminate healthy individuals from NSCLC patients which yielded 85% sensitivity and 82.9% specificity. This estimate of sensitivity was compared favorably with an estimated sensitivity of blood-based biomarkers, such as CYFRA 21-1, TPS and CEA, which ranged from 34% to 53% in the same study population. Recently, DNA methylation markers have showed good diagnostic efficacy for LC. Zhang et al. [89] used a combination of F2RL3 methylation in whole blood and smoking exposure to predict LC incidence with an AUC value of 0.86 for participants ≥65 years. Since miRNA markers and DNA methylation markers both shown good diagnostic performance for LC, the combination of the two may offer an improvement of diagnostic efficiency in LC.

Conclusions
Our systematic review suggests a number of circulating miRNAs to be promising candidates for noninvasive LC detection in East Asian populations. The heterogeneity of reported LC-specific miRNA profiles in published studies needs to be addressed and protocols for the standardization of miRNA analysis procedures need to be put into place. Larger prospective studies, the improvement of miRNA detection technologies, the minimization of pre-analytical or analytical variability as well as the development of new analysis methods will be crucial to further reduce the bias in measurement and analysis and improve the diagnostic performance. Most importantly, however, will be the rigorous independent validation of identified promising miRNA algorithms or their combination with other biomarkers in prospective screening cohorts.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/11/3/415/s1, Figure S1: Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies, Figure S2: Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study, Table S1: Summary of studies reporting significant associations of miRNAs with lung cancer in East Asian populations, Table S2: Diagnostic performance of miRNAs and miRNA panels according to lung cancer stage in East Asian populations, Table S3: MiRNAs for which opposite directions of dysregulation were reported in lung cancer blood samples, Table S4: East Asian studies versus Western studies of miRNAs for lung cancer detection, Table S5: Protocols of blood miRNA detection.
Author Contributions: H.B., Y.Z. and H.Y. designed the study. H.Y. and Z.G. extracted the data. H.Y. wrote the manuscript. Y.Z., K.C. and H.B. revised the manuscript. All authors read and approved the final manuscript.