’Original article The role of circulating microRNAs as novel biomarkers in diagnosing colorectal cancer: a meta-analysis Theodore Rokkas, Fotis Kothonas, Androniki Rokka, Georgios Koukoulis and Emmanuel Symvoulakis Objectives Several microRNAs (miRNAs) have been identified as potential circulating biomarkers in a number of different cancers including colorectal cancer (CRC). This study aims to assess the diagnostic performance of circulating miRNAs in detecting CRC through meta-analysis of all eligible relevant studies. Materials and methods An extensive literature search was performed and studies that estimated the diagnostic accuracy of miRNAs in CRC were identified. Data from the eligible studies were collected and pooled; sensitivity, specificity, positive and negative likelihood ratios, diagnostic odds ratios, weighted symmetric summary ROC curve and the area under the curve (AUC) were calculated. Heterogeneity was evaluated using the Q-test and I2-statistics. In addition, subgroup analyses and metaregression analyses were carried out to explore the potential sources of significant heterogeneity. Results A total of 16 studies were included in the meta-analysis according to the inclusion criteria. The overall analysis showed that circulating miRNAs have a relatively good diagnostic performance in CRC, with a sensitivity of 78%, a specificity of 79% and an AUC of 0.87. Subgroup analyses showed that a single miRNA-21 test, as opposed to a panel miRNAs test, significantly improved the diagnostic accuracy with 83.4% sensitivity, 91.6% specificity, and AUC increasing to 0.94. Conclusion Circulating miRNAs, especially miRNA-21, are promising diagnostic biomarkers in CRC. However, more prospective studies are required to further explore their diagnostic role and their usefulness in clinical practice. Eur J Gastroenterol Hepatol 27:819–825 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Introduction
Colorectal cancer (CRC) accounts for over 9% of all cancer incidences worldwide and is the 4th most common cause of cancer-related mortality [1]. CRC is curable if detected at an early stage. The 5-year survival rate for CRC patients is related to the stage of the disease, ranging from more than 93% for stage I to less than 8% for stage IV [2]. Therefore, the development of a screening test, by which the cancer can be diagnosed at an early stage, is necessary. Although population-based screening combining faecal occult blood tests and colonoscopy has been implemented with significant success, there is a trend towards the use of biomarkers and novel screening tools to improve detection performance [3,4]. MicroRNAs (miRNAs) are a class of small noncoding RNA molecules, 19–25 nucleotides in length, that have been shown to regulate many cellular processes [5,6]. The first report suggesting miRNA involvement in cancer was published just over a decade ago and soon after the first study on the association of tissue miRNAs with CRC was European Journal of Gastroenterology & Hepatology 2015, 27:819–825 Keywords: biomarkers, colorectal cancer, diagnosis, meta-analysis, microRNA Gastroenterology Clinic, Henry Dunant Hospital Center, Athens, Greece Correspondence to Theodore Rokkas, MD, PhD, FACG, AGAF, FEBG, Gastroenterology Clinic, Henry Dunant Hospital Center, 107 Messogion Ave., Athens 11526, Greece Tel/fax: 30 210 6431334; e-mail: [email protected] Received 19 January 2015 Accepted 5 March 2015
published [7]. Since then, a variety of candidate miRNA biomarkers in the tissues of CRC patients have been reported [8]. Apart from tissue, specific miRNAs have also been found in the blood or its components in a number of different cancers [9] and it is of importance that in addition to being highly abundant in circulation, miRNAs show remarkable stability in both plasma and serum [10]. These properties make circulating miRNAs ideal novel tumour biomarkers for noninvasive detection in CRC and indeed several miRNAs have been studied. However, the published literature in this field is diverse and, so far, none of the circulating miRNAs have been incorporated into clinical practice. This prompted us to further investigate the diagnostic role of circulating miRNAs in CRC by carrying out a meta-analysis, in which we pooled the results of all existing relevant studies, aiming to precisely define their diagnostic accuracy. Materials and methods Data identification and extraction
Extensive English-language, computer-aided medical literature searches of the PubMed/MEDLINE and Embase databases for human studies were carried out to find relevant publications on the diagnostic role of miRNA in diagnosing CRC, compared with established diagnostic methods, using the following keywords or MeSH terms: Colorectal cancer, microRNA, diagnosis. In detail, the following search items were used: (‘colorectal neoplasms’[MeSH Terms] OR (‘colorectal’[All Fields]
0954-691X Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
DOI: 10.1097/MEG.0000000000000363
819
820
July 2015 • Volume 27 • Number 7
European Journal of Gastroenterology & Hepatology
AND ‘neoplasms’[All Fields]) OR ‘colorectal neoplasms’ [All Fields] OR (‘colorectal’[All Fields] AND ‘cancer’ [All Fields]) OR ‘colorectal cancer’[All Fields]) AND (‘micrornas’[MeSH Terms] OR ‘micrornas’[All Fields] OR ‘microrna’[All Fields]) AND (‘diagnosis’[Subheading] OR ‘diagnosis’[All Fields] OR ‘diagnosis’[MeSH Terms]). No beginning date limit was used. The search was updated to the end of January 2015. To expand our search, we screened the articles that remained after the selection process for potentially suitable references. Two authors (F.K. and G.K.) extracted data from each study independently using a predefined form, and disagreements were resolved by discussion with a third author (T.R.) and consensus. The work was carried out in accordance with the PRISMA statement [11]. Selection criteria
Inclusion and exclusion criteria were delineated before the commencement of the literature search. Thus, eligible studies were included in this meta-analysis if they fulfilled the following criteria: (a) they were published as full articles, (b) they were written in the English language and (c) they provided sufficient data for the authors to construct a 2 × 2 contingency table to calculate sensitivity, specificity, positive likelihood ratios (PLRs) and negative likelihood ratios (NLRs), diagnostic odds ratios (DORs), receiver operating characteristic (ROC) and area under the curve (AUC). When two articles reported the same study, the publication that was more informative was selected. In studies reporting data according to the level of bowel preparation, data referring to the best level were taken into account. The methodological quality of each study was assessed by QUADAS (quality assessment for studies of diagnostic accuracy), which is an evidence-based quality assessment tool for use in systematic reviews of diagnostic accuracy studies [12]. The 14 items (Table 1) were assessed in all included articles, each of which was assessed as ‘yes’, ‘no’ or ‘unclear’. Accordingly, they were scored with 1 for ‘yes’, 0 for ‘unclear’ and − 1 for ‘no’ with a maximum score of 14. Statistical analysis
Agreement on the selection of studies between the two reviewers was evaluated using the k coefficient. Sensitivity,
specificity, NLRs and NLRs (PLR = sensitivity/1–specificity, NLR = 1–sensitivity/specificity) and DORs (DOR = PLR /NLR) of miRNAs [with corresponding 95% confidence intervals (CIs)] were calculated from the original numbers given in the included studies by constructing 2 × 2 contingency tables. PLR indicates how much the odds of disease increase when the test is positive, whereas NLR indicates how much the odds of CRC decrease when the miRNA test is negative. PLRs and NLRs are considered to be more clinically meaningful than sensitivity or specificity with PLR of more than 10 or NLR less than 0.1, indicating high accuracy [13]. Pooled results with corresponding 95% CIs were derived using the fixed-effects model (Mantel and Haenszel method) [14], unless significant heterogeneity was present, in which case the randomeffects model was used (DerSimonian and Laird method) [15]. Forest plots were constructed for visual display of individual studies and pooled results. In addition, the results of the individual studies were presented in a ROC space to show the distribution of sensitivities and specificities. A weighted symmetric summary ROC (sROC) curve was computed using the Moses–Shapiro–Littenberg method [16] and the AUC was calculated, with perfect tests having an AUC of 1 and poor tests having an AUC close to 0.5 [17]. To assess whether heterogeneity exists among the eligible studies, we first calculated the correlation coefficient and the P value between the logit of sensitivity and logit of 1 − specificity using the Spearman test to exclude the threshold effect. In this case, the Spearman correlation coefficient was significant when the P value was less than 0.05. When heterogeneity was not explained by the threshold effect, further evaluation was performed by calculating the Cochran Q-test and the inconsistency index (I2) [18,19]. Heterogeneity was considered to be present if the Q-test provided a P value of less than 0.1 [10] and I2 of more than 50% and in case of significant heterogeneity, we continued our attempts to explain this by exploring the study characteristics using metaregression. In addition, Deeks’ funnel plot asymmetry test was used to evaluate the potential publication bias [20] and a P value of less than 0.1 was considered to be representative of a significant publication bias. The statistical analyses were carried out using statistical package for the social sciences software version 14.0 (SPSS Inc., Chicago, Illinois, USA), the Comprehensive Meta-Analysis
Table 1. The QUADAS tool Yes 1. 2. 3. 4.
Was the spectrum of patients representative of the patients who will receive the test in practice? Were selection criteria clearly described? Is the reference standard likely to correctly classify the target condition? Is the time period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests? 5. Did the whole sample or a random selection of the sample receive verification using a reference standard of diagnosis? 6. Did patients receive the same reference standard regardless of the index test result? 7. Was the reference standard independent of the index test (i.e. the index test did not form part of the reference standard)? 8. Was the execution of the index test described in sufficient detail to permit replication of the test? 9. Was the execution of the reference standard described in sufficient detail to permit its replication? 10. Were the index test results interpreted without knowledge of the results of the reference standard? 11. Were the reference standard results interpreted without knowledge of the results of the index test? 12. Were the same clinical data available when test results were interpreted as would be available when the test is used in practice? 13. Were uninterpretable//intermediate test results reported? 14. Were withdrawals from the study explained?
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Item No Unclear
Circulating miRNAs and colorectal cancer Rokkas et al.
751 potentially eligible studies initially generated by the literature searches 603 rejected (title suggested article not appropriate) 148 abstracts retrieved
59 studies with extractable data
89 excluded (duplications, editorials, review articles)
43 excluded as not fulfilling the inclusion criteria 16 studies meta-analyzed Fig. 1. Flow diagram of the studies identified in this meta-analysis.
software version 2 (Biostat Inc., Englewood, New Jersey, USA), Stata version 12 (Stata Corporation, College Station, Texas, USA) and Meta-DiSc statistical software version 1.4 [Unit of Clinical Biostatistics, Ramon and Cajal Hospital, Madrid, Spain [21]. Results Descriptive assessment and study characteristics
A flow chart describing the process of study selection is shown in Fig. 1. Out of 751 titles initially generated by the literature searches, 16 studies remained eligible for metaanalysis [22–37]. Initial agreement between the reviewers for the selection of relevant articles was high (k = 0.995, 95% CI 0.989–1.00). The main characteristics of the 16 studies eligible for meta-analysis are shown in Table 2. They were carried out in different parts of the world and included a total of 1230 CRC patients and 810 controls.
www.eurojgh.com
821
data showed a PLR of 3.66 (2.62–5.10) and an NLR of 0.27 (0.21–0.36) (Fig. 2). As shown in the likelihood matrix (Fig. 3), the summary point of PLR combined with LRN lies in the right lower quadrant (PLR < 10, LRN > 0.1). This is in agreement with the modest pooled DOR magnitude of 14.25 (8.63–23.52) (Fig. 4). The corresponding ROC plot with sROC is also shown in Fig. 3. The AUC under the sROC was 0.860. There was no publication bias as judged by Deeks’ test (P = 0.384). However, significant heterogeneity (P < 0.001) was found in all parameters of diagnostic performance (i.e. sensitivity, specificity, PLR, NLR and DOR) (Fig. 2). Thus, apart from use of the random-effects model, metaregression and subgroup analyses were carried out to explore the reasons for the significant heterogeneity observed. First, we checked whether there was any significant threshold effect and found that this was not the case as the Spearman correlation coefficient was 0.117 and the P value was 0.666. Then, we attempted to explain the heterogeneity by exploring study characteristics, that is, age, ethnicity (Asians vs. Whites), QUADAS score and blood specimen (plasma vs. serum) through metaregression analyses, and no significant results were found (Table 3). Subsequently, we carried out subgroup analysis comparing the group of three studies [26–28] in which a single circulating miRNA (i.e. miRNA-21) was used with the group of 13 studies [22–25,29–37] that used a panel of various circulating miRNAs (Table 4). We found that the single miRNA-21 group performed better than the panel miRNAs group in all diagnostic value parameters, that is sensitivity 83.4% (78.7–87.5), specificity 91.6% (85.5– 95.7), PLR 9.88 (5.60–17.42), NLR 0.18 (0.14–0.24), DORs 56.91 (28.00–115.68) and AUC 0.947. Most importantly, in this group, heterogeneity became nonsignificant in all diagnostic accuracy parameters in comparison with the panel miRNA group.
Diagnostic performance of microRNAs
Discussion
The pooled data (random-effects model) showed circulating miRNA sensitivity of 78% [95% CI (75–80%)] and specificity of 79% (76–82%). Accordingly, the pooled
The development of suitable noninvasive biomarkers is important for the diagnosis of CRC. The recent discovery that noncoding miRNAs are stable in body fluids, such as
Table 2. Characteristics of the studies included References
Country
Ethnicity
Total number (patients/controls)
Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Kanaan et al. [26] Wang and Zhang [27] Toiyama et al. [28] Zhang et al. [29] Giráldez et al. [30] Wang et al. [31] Luo et al. [32]
China China China China USA China Japan China Spain China Germany
Asian Asian Asian Asian White Asian Asian Asian White Asian White
140 (90/50) 159 (100/59) 140 (103/37) 148 (90/58) 40 (20/20) 71(32/29) 239 (186/53) 164 (78/86) 95 (42/53) 49 (22/27) 224 (80/144)
Plasma Plasma Plasma Plasma Plasma Serum Serum Plasma Plasma Plasma Plasma
Liu et al. [33] Hofsli et al. [34] Brunet Vega et al. [35] Zanutto et al. [36] Du et al. [37]
China Norway Spain Italy China
Asian White White White Asian
280 50 56 58 98
Serum Serum Serum Plasma Plasma
(200/80) (40/10) (30/26) (29/29) (49/49)
Blood specimen
mRNA profile miRNA-17-3p, -92 miRNA-29a, -92a miRNA-21, -221, -222 miRNA-601, -760, -29a, -92a miRNA-21 miRNA-21 miRNA-21 miRNA-200c, 18a miRNA-19a, -19b, -15b miRNA-409-3p, -7, -93 miRNA -18a, -20a, -21, -29a, -92a, -106b, -133a, -143, -145, -181b, -342-3p, -532-3p miRNA-21, -31, -92a,- 18a, -106a Panel of 21 miRNAsa miRNA-16-5p, let7a-5p, 103-5p miRNA-378, -21 miRNA-21, 92a
a
miRNA-423-5p, -210, -720, -320a, -378, -92a, -29a, -155, -106a, -143, -103,-199a-3p, -151-5p, -107, -191, -423-3p, -221, -let7d, -382, -409-3p, -652.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Assay method qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR
822
July 2015 • Volume 27 • Number 7
European Journal of Gastroenterology & Hepatology
(a)
(c)
Sensitivity (95% CI) Ng et al. [22] 0.89 0.83 Huang et al. [23] 0.86 Pu et al. [24] 0.83 Wang et al. [25] Wang and Zhang [27] 0.88 Kanaan et al. [26] 0.90 0.82 Wang et al. [31] Luo et al. [32] 0.54 Zhang et al. [29] 0.85 Giráldez et al. [30] 0.79 Hofsli et al. [34] 0.88 Brunet Vega et al. [35] 0.67 Liu et al. [33] 0.65 Toiyama et al. [28] 0.83 0.59 Zanutto et al. [36] Du et al. [37] 0.76
Positive LR (95% CI)
(0.81 − 0.95) (0.74 − 0.90) (0.78 − 0.92) (0.74 − 0.90) (0.71 − 0.96) (0.68 − 0.99) (0.60 − 0.95) (0.42 − 0.65) (0.75 − 0.92) (0.63 − 0.90) (0.73 − 0.96) (0.47 − 0.83) (0.58 − 0.72) (0.77 − 0.88) (0.39 − 0.76) (0.61 − 0.87)
Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31] Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
Pooled positive LR = 3.66 (2.62 − 5.10) Cochran Q = 82.61, d.f. = 15 (P = 0.0000) Inconsistency (I 2) = 81.8% τ 2 = 0.3365
Pooled sensitivity 0.78 (0.75−0.80) 2 = 74.6, d.f. = 15 (P = 0.0000) Inconsistency (I 2) = 79.9%
0
0.2
0.4 0.6 Sensitivity
0.8
1
Random-effects model
1 Positive LR
0.01
(b)
100.0
Random-effects model
Specificity (95% CI) (d) Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31] Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
Negative LR (95% CI)
0.70 (0.55 − 0.82) 0.85 (0.73 − 0.93) 0.41 (0.25 − 0.58) 0.93 (0.83 − 0.98) 0.74 (0.58 − 0.87) 0.90 (0.68 − 0.90) 0.89 (0.71 − 0.98) 0.82 (0.75 − 0.88) 0.76 (0.65 − 0.84) 0.79 (0.66 − 0.89) 0.90 (0.55 − 1.00) 0.58 (0.37 − 0.77) 0.85 (0.75 − 0.92) 0.91 (0.79 − 0.97) 0.59 (0.39 − 0.76) 0.82 (0.68 − 0.91)
Ng et al. [22] Huang et al. [23 Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31 Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
0.2
0.4 0.6 Specificity
0.8
1
0.16 (0.09 − 0.29) 0.20 (0.13 − 0.31) 0.34 (0.18 − 0.63) 0.18 (0.11 − 0.29) 0.17 (0.07 − 0.43) 0.11 (0.03 − 0.42) 0.20 (0.08 − 0.50) 0.56 (0.44 − 0.72) 0.20 (0.12 − 0.35) 0.27 (0.15 − 0.49) 0.14 (0.06 − 0.32) 0.58 (0.32 − 1.06) 0.41 (0.33 − 0.51) 0.19 (0.14 − 0.26) 0.71 (0.42 − 1.20) 0.30 (0.18 − 0.50)
Pooled negative LR = 0.27 (0.21-0.36) Cochran Q = 77.92, d.f. = 15 (0.0000) Inconsistency (I 2) = 80.8% τ 2 = 0.2270
Pooled specificity 0.79(0.76-0.82) 2 = 63.71, d.f. = 15 (P = 0.0000) Inconsistency (I 2) = 76.5%
0
2.96 (1.93 − 4.55) 5.44 (2.96 − 9.99) 1.45 (1.10 − 1.92) 12.08 (4.67 − 31.2) 3.41 (1.97 − 5.92) 9.00 (2.40 − 33.79) 7.36 (2.49 − 21.79) 2.98 (2.98 − 4.45) 3.47 (2.36 − 5.09) 3.79 (2.19 − 6.56) 8.75 (1.36 − 56.38) 1.58 (0.94 − 2.64) 4.33 (2.55 − 7.37) 8.78 (3.80 − 20.26) 1.42 (0.83 − 2.41) 4.11 (2.23 − 7.58)
Random-effects model
0.01
1 Negative LR
100.0
Random-effects model
Fig. 2. Forest plots showing (a) individual and pooled (95% CIs) sensitivities. (b) Individual and pooled (95% CIs) specificities. (c) Individual and pooled (95% CIs) positive likelihood ratios (LR). (d) Individual and pooled (95% CIs) negative LR. CI, confidence interval.
(a)
(b) 100
Sensitivity 1
SROC curve Symmetric SROC AUC = 0.8605 SE(AUC) = 0.0232 Q∗ = 0.7913 SE(Q∗) = 0.0224
0−9
Positive likelihood ratio
0−8 0−7 0−6 10
0−5 0−4 0−3 0−2 0−1
1 0.01
0 0.1
0.3 0.5 0.7 0.9 Negative likelihood ratio
0.99
0
0−2
0−4 0−6 1−specificity
0−8
1
Fig. 3. (a) Likelihood matrix showing individual (solid dots) and pooled (diamond shape) values of positive likelihood ratios combined with negative likelihood ratios. (b) Summary receiver operating characteristic curves (SROC) with area under the curve (AUC) of miRNAs describing the diagnostic performance. Every dot stands for a study.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
95% CI
0.009 − 0.297 − 0.105 − 0.555
0.1253 0.8005 0.3162 0.9817
0.9454 0.7196 0.7464 0.5856
1.01 0.74 0.90 0.57
0.12–4.55 0.12–4.55 0.44–1.84 0.06–5.29
plasma and serum, created an opportunity to develop such novel diagnostic biomarkers. However, their diagnostic accuracy is inconsistent or even contradictory in the literature. The pooled data in this meta-analysis showed circulating miRNA sensitivity of 78% and specificity of 79%, with corresponding PLR 3.66 and NLR 0.27. As shown in the likelihood matrix (Fig. 3), the summary point of PLR combined with LRN lies in the right lower quadrant (PLR < 10, LRN >0.1), indicating that the circulating miRNAs may not have sufficient power either to confirm or to exclude cancer. This is in agreement with the modest pooled DOR magnitude of 14.25, which expresses the strength of the association between circulating miRNAs and CRC. Accordingly, the AUC was 0.860, indicating a relatively good accuracy. However, these results are overshadowed by significant heterogeneity, which was found in all parameters of diagnostic performance (i.e. sensitivity, specificity, PLR, NLR and DOR). There was no obvious explanation for this significant heterogeneity s there was no significant threshold effect and in addition no significant results were found in meta-regression analyses exploring study characteristics, that is, age, ethnicity (Asians vs. Whites), QUADAS score and blood specimen (plasma vs. serum). The subsequent subgroup analysis, that is the single miRNA versus panel of various miRNAs, yielded interesting findings in that the single miRNA-21 group showed good results in all parameters of diagnostic performance, that is sensitivity 83.4%, specificity 91.6%, PLR 9.88, NLR 0.18, DOR 56.91 and AUC 0.947 and, in Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Table 4. Subgroup analysis (microRNA panel vs. microRNA-21)
CI, confidence interval; Cte, constant coefficient; RDOR, relative diagnostic odds ratio; S, statistic S.
0.947
CI, confidence interval; DOR, diagnostic odds ratio; LR, likelihood ratio.
RDOR
3.55 (2.54–4.96) (Q = 60.59, d.f. = 12, P < 0.001, I2 = 80.2%) 9.88 (5.6 0–17.42) (Q = 0.27, d.f. = 2, P = 0.874, I2 = 0%)
P value
78.6 (75.4–81.5) (Q = 50.65, d.f. = 12, P < 0.001, I2 = 76.3%) 91.6 (85.5–95.7) (Q = 0.36, d.f. = 2, P = 0.854, I2 = 0%)
SE
77.1 (74.3–79.7) (Q = 64.1, d.f. = 12, P < 0.001, I2 = 81.3%) 83.4 (78.7–87.5) (Q = 0.76, d.f. = 2, P = 0.683, I2 = 0%)
Cte S Age Ethnicity QUADAS score Specimen source (plasma vs. serum)
Coefficient
Positive LR (95% CI) (heterogeneity)
Variables
Specificity (95% CI) (heterogeneity)
Table 3. Results of the multivariable meta-regression model, for the parameters taken into account, with backward regression analysis (inverse variance weights)
Sensitivity (95% CI) (heterogeneity)
Negative LR (95% CI) heterogeneity
Fig. 4. Forest plots showing individual and pooled (95% CIs) diagnostic odds ratios. CI, confidence interval.
13.72 (8.50–22.13) (Q = 37. 79, d.f. = 12, P < 0.001, I2 = 65.9%) 56.91 (28.00–115.68) (Q = 0.36, d.f. = 2, P = 0.834, I2 = 0%)
Random-effects model Pooled diagnostic odds ratio = 14.25 (8.64 − 23.52) Cochran Q = 59.40; d.f. = 15(P = 0.0000) Inconsistency (I 2) = 74.7% τ 2 = 0.7272
0.27 (0.20–0.36) (Q = 56.60, d.f. = 12, P < 0.001, I2 = 78.8%) 0.18 (0.14–0.24) (Q = 0.67, d.f. = 2, P = 0.716, I2 = 0%)
DOR (95% CI) heterogeneity
0.01 1 100.0 Diagnostic odds ratio
18.67 (7.64 − 45.61) 27.12 (11.24 − 65.45) 4.33 (1.82 − 10.29) 67.50 (21.22 − 214.69) 20.30 (5.70 − 72.32) 81.00 (10.26 − 639.34) 36.00 (7.15 − 181.35) 5.27 (2.86 − 9.72) 17.02 (7.74 − 37.42) 14.00 (5.19 − 37.75) 63.00 (6.52 − 608.93) 2.73 (0.92 − 8.09) 10.52 (5.34 − 20.75) 46.20 (17.05 − 125.17) 2.01 (0.71 − 5.71) 13.70 (5.18 − 36.26)
miRNA panel (13 studies) miRNA-21 (three studies)
Diagnostic OR (95% CI) Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31] Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
0.858
www.eurojgh.com
AUC
Circulating miRNAs and colorectal cancer Rokkas et al.
823
824
July 2015 • Volume 27 • Number 7
European Journal of Gastroenterology & Hepatology
this respect, performed much better than the panel of the various miRNAs group. Most importantly, in the single miRNA-21 group, heterogeneity became nonsignificant in all diagnostic parameters in comparison with the panel miRNA group, where heterogeneity was significant. All the above might mean that miRNA-21 is a good candidate as an accurate noninvasive novel biomarker for CRC diagnosis, provided that these results are confirmed by further prospective studies. It must be noted that the diagnostic performance of miRNA-21 in this meta-analysis is the best so far in comparison with other blood markers studied, such as carcinoembryonic antigen [38], carbohydrate antigen 19-9 [30] and miRNA-92a [39] In this meta-analysis, we included 16 eligible studies, which is the largest so far in comparison with previous similar studies [40–42], and undoubtedly, this represents an advantage. However, there are several limitations in this meta-analysis. Thus, because of the fact that the clinical value of miRNA in CRC has only been explored recently, small sample size studies are included in our meta-analysis, and as a result, small-study effects are inescapable. Therefore, it is necessary to strengthen our conclusion by further validations of miRNA-21 in large prospective studies. Furthermore, there might still be some more biases in our meta-analysis. For example, in terms of the design of eligible studies, only two had a clearly prospective design as opposed to the remaining studies. Moreover, no studies unequivocally mentioned whether a blind design was used. Future studies should address these points; in addition, other parameters, which might potentially influence the results, should be addressed. These parameters might be related to patients, such as TNM classification (I/II/III/IV) for CRC, or technical aspects in terms of the methods of RNA extraction (miRNsasy vs. TRIzol LS) and the method of miRNA detection [qRT-PCR (Taqman) vs. qRT-PCR (SYBR green) vs. miCURY vs. Agilent microarray]. In conclusion, circulating miRNAs are promising noninvasive diagnostic biomarkers in CRC. In particular, miRNA-21 is a good candidate as an accurate novel biomarker. However, more prospective well-designed studies are required to further explore its diagnostic role and its incorporation into clinical practice.
6
7
8
9
10
11
12
13
14
15 16
17 18 19 20
21
22
23
Acknowledgements 24
Conflicts of interest
There are no conflicts of interest. 25
References 1 Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011; 61:69–90. 2 O’Connell JB, Maggard MA, Ko CY. Colon cancer survival rates with the new American Joint Committee on Cancer sixth edition staging. J Natl Cancer Inst 2004; 96:1420–1425. 3 Zauber AG, Lansdorp-Vogelaar I, Knudsen AB, Wilschut J, van Ballegooijen M, Kuntz KM. Evaluating test strategies for colorectal cancer screening: a decision analysis for the U.S. Preventive Services Task Force. Ann Intern Med 2008; 149:659–669. 4 Wu WK, Law PT, Lee CW, Cho CH, Fan D, Wu K, et al. MicroRNA in colorectal cancer: from bench top to bedside. Carcinogenesis 2011; 32:247–253. 5 Bartel DP. MicroRNAs: target recognition and regulatory functions. Cell 2009; 136:215–233.
26
27
28
29
30
Chiang HR, Schoenfeld LW, Ruby JG, Auyeung VC, Spies N, Baek D, et al. Mammalian microRNAs: experimental evaluation of novel and previously annotated genes. Genes Dev 2010; 24:992–1009. Michael MZ, O' Connor SM, van Holst Pellekaan NG, Young GP, James RJ. Reduced accumulation of specific microRNAs in colorectal neoplasia. Mol Cancer Res 2003; 1:882–891. Ma Y, Zhang P, Yang J, Liu Z, Yang Z, Qin H. Candidate microRNA biomarkers in human colorectal cancer: systematic review profiling studies and experimental validation. Int J Cancer 2012; 130:2077–2087. Healy NA, Heneghan HM, Miller N, Osborne CK, Schiff R, Kerin MJ. Systemic miRNAs 609 as potential biomarkers for malignancy. Int J Cancer 2012; 131:2215–2222. Chen X, Ba Y, Ma L, Cai X, Yin Y, Wang K, et al. Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Res 2008; 18:997–1006. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health car-interventions: explanation and elaboration. J Clin Epidemiol 2009; 62:e1–34. Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003; 3:25. Rosenfeld RM, Shiffman RN. Clinical practice guidelines: a manual for developing evidence-based guidelines to facilitate performance measurement and quality improvement. Otolaryngol Head Neck Surg 2006; 135:S1–S28. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 719–748. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177–188. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med 1993; 12:1293–1316. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982; 143:29–36. Cochran WG. The combination of estimates from different experiments. Biometrics 1954; 8:101–129. Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med 2002; 21:1539–1558. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005; 58:882–893. Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Methodol 2006; 6:31. Ng EK, Chong WW, Jin H, Lam EK, Shin VY, Yu J, et al. Differential expression of microRNAs in plasma of patients with colorectal cancer: a potential marker for colorectal cancer screening. Gut 2009; 58:1375–1381. Huang Z, Huang D, Ni S, Peng Z, Sheng W, Du X. Plasma microRNAs are promising novel biomarkers for early detection 684 of colorectal cancer. Int J Cancer 2010; 127:118–126. Pu XX, Huang GL, Guo HQ, Guo CC, Li H, Ye S, et al. Circulating miR-221 directly amplified from plasma is a potential diagnostic and prognostic marker of colorectal cancer and is correlated with p53 expression. J Gastroenterol Hepatol 2010; 25:1674–1680. Wang Q, Huang Z, Ni S, Xiao X, Xu Q, Wang L, et al. Plasma miR-601 and miR-760 are novel biomarkers for the early detection of colorectal cancer. PLoS One 2012; 7:e44398. Kanaan Z, Rai SN, Eichenberger MR, Roberts H, Keskey B, Pan J, Galandiuk S. Plasma miR-21: a potential diagnostic marker of colorectal cancer. Ann Surg 2012; 256:544–551. Wang B, Zhang Q. The expression and clinical significance of circulating microRNA-21 in serum of five solid tumors. J Cancer Res Clin Oncol 2012; 138:1659–1666. Toiyama Y, Takahashi M, Hur K, Nagasaka T, Tanaka K, Inoue Y, et al. Serum miR-21 as a diagnostic and prognostic biomarker in colorectal cancer. J Natl Cancer Int 2013; 105:849–859. Zhang GJ, Zhou T, Liu ZL, Tian HP, Xia SS. Plasma miR-200c and miR-18a as potential biomarkers for the detection of colorectal carcinoma. Mol Clin Oncol 2013; 1:379–384. Giráldez MD, Lozano JJ, Ramírez G, Hijona E, Bujanda L, Castells A, Gironella M. Circulating microRNAs as biomarkers of colorectal cancer:
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Circulating miRNAs and colorectal cancer Rokkas et al.
31
32
33
34
35
36
results from a genome-wide profiling and validation study. Clin Gastroenterol Hepatol 2013; 11:681–688. Wang S, Xiang J, Li Z, Lu S, Hu J, Gao X, et al. A plasma microRNA panel for early detection of colorectal cancer. Int J Cancer 2015; 136:152–161. Luo X, Stock C, Burwinkel B, Brenner H. Identification an evaluation of plasma microRNAs for early detection of colorectal cancer. PLoS One 2013; 14:5e62880. Liu GH, Zhou ZG, Chen R, Wang MJ, Zhou B, Li Y, Sun XF. Serum miR-21 and miR-92a as biomarkers in the diagnosis and prognosis of colorectal cancer. Tumour Biol 2013; 34:2175–2181. Hofsli E, Sjursen W, Prestvik WS, Johansen J, Rye M, Tranø G, et al. Identification of serum microRNA profiles in colon cancer. Br J Cancer 2013; 108:1712–1719. Brunet Vega A, Pericay C, Moya I, Ferrer A, Dotor E, Pisa A, et al. microRNA expression profile in stage III colorectal cancer: circulating miR-18a and miR-29a as promising biomarkers. Oncol Rep 2013; 30:320–326. Zanutto S, Pizzamiglio S, Ghilotti M, Bertan C, Ravagnani F, Perrone F, et al. Circulating miR-378 in plasma: a reliable, haemolysis-
www.eurojgh.com
37
38
39
40
41
42
825
independent biomarker for colorectal cancer. Br J Cancer, 2014; 110:1001–1007. Du M, Liu S, Gu D, Wang Q, Zhu L, Kang M, et al. Clinical potential role of circulating microRNAs in early diagnosis of colorectal cancer patients. Carcinogenesis 2014; 35:2723–2730. Thomson DM, Krupey J, Freedman SO, Gold P. The radioimmunoassay of circulating carcinoembryonic antigen of the human digestive system. Proc Natl Acad Sci U S A 1969; 64:161–167. Hundt S, Haug U, Brenner H. Blood markers for early detection of colorectal cancer: a systematic review. Cancer Epidemiol Biomarkers Prev 2007; 16:1935–1953. Yang X, Zeng Z, Hou Y, Yuan T, Gao C, Jia W, et al. MicroRNA-92a as a potential biomarker in diagnosis of colorectal cancer: a systematic review and meta-analysis. PLoS One 2014; 9:e88745. Zhou XJ, Dong ZG, Yang YM, Du LT, Zhang X, Wang CX. Limited diagnostic value of microRNAs for detecting colorectal cancer: a metaanalysis. Asian Pac J Cancer Prev 2013; 14:4699–4704. Jiang JX, Zhang N, Liu ZM, Wang YY. Detection of microRNA-21 expression as a potential screening biomarker for colorectal cancer: a meta-analysis. Asian Pac J Cancer Prev 2014; 15:7583–7588.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Introduction
Colorectal cancer (CRC) accounts for over 9% of all cancer incidences worldwide and is the 4th most common cause of cancer-related mortality [1]. CRC is curable if detected at an early stage. The 5-year survival rate for CRC patients is related to the stage of the disease, ranging from more than 93% for stage I to less than 8% for stage IV [2]. Therefore, the development of a screening test, by which the cancer can be diagnosed at an early stage, is necessary. Although population-based screening combining faecal occult blood tests and colonoscopy has been implemented with significant success, there is a trend towards the use of biomarkers and novel screening tools to improve detection performance [3,4]. MicroRNAs (miRNAs) are a class of small noncoding RNA molecules, 19–25 nucleotides in length, that have been shown to regulate many cellular processes [5,6]. The first report suggesting miRNA involvement in cancer was published just over a decade ago and soon after the first study on the association of tissue miRNAs with CRC was European Journal of Gastroenterology & Hepatology 2015, 27:819–825 Keywords: biomarkers, colorectal cancer, diagnosis, meta-analysis, microRNA Gastroenterology Clinic, Henry Dunant Hospital Center, Athens, Greece Correspondence to Theodore Rokkas, MD, PhD, FACG, AGAF, FEBG, Gastroenterology Clinic, Henry Dunant Hospital Center, 107 Messogion Ave., Athens 11526, Greece Tel/fax: 30 210 6431334; e-mail: [email protected] Received 19 January 2015 Accepted 5 March 2015
published [7]. Since then, a variety of candidate miRNA biomarkers in the tissues of CRC patients have been reported [8]. Apart from tissue, specific miRNAs have also been found in the blood or its components in a number of different cancers [9] and it is of importance that in addition to being highly abundant in circulation, miRNAs show remarkable stability in both plasma and serum [10]. These properties make circulating miRNAs ideal novel tumour biomarkers for noninvasive detection in CRC and indeed several miRNAs have been studied. However, the published literature in this field is diverse and, so far, none of the circulating miRNAs have been incorporated into clinical practice. This prompted us to further investigate the diagnostic role of circulating miRNAs in CRC by carrying out a meta-analysis, in which we pooled the results of all existing relevant studies, aiming to precisely define their diagnostic accuracy. Materials and methods Data identification and extraction
Extensive English-language, computer-aided medical literature searches of the PubMed/MEDLINE and Embase databases for human studies were carried out to find relevant publications on the diagnostic role of miRNA in diagnosing CRC, compared with established diagnostic methods, using the following keywords or MeSH terms: Colorectal cancer, microRNA, diagnosis. In detail, the following search items were used: (‘colorectal neoplasms’[MeSH Terms] OR (‘colorectal’[All Fields]
0954-691X Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
DOI: 10.1097/MEG.0000000000000363
819
820
July 2015 • Volume 27 • Number 7
European Journal of Gastroenterology & Hepatology
AND ‘neoplasms’[All Fields]) OR ‘colorectal neoplasms’ [All Fields] OR (‘colorectal’[All Fields] AND ‘cancer’ [All Fields]) OR ‘colorectal cancer’[All Fields]) AND (‘micrornas’[MeSH Terms] OR ‘micrornas’[All Fields] OR ‘microrna’[All Fields]) AND (‘diagnosis’[Subheading] OR ‘diagnosis’[All Fields] OR ‘diagnosis’[MeSH Terms]). No beginning date limit was used. The search was updated to the end of January 2015. To expand our search, we screened the articles that remained after the selection process for potentially suitable references. Two authors (F.K. and G.K.) extracted data from each study independently using a predefined form, and disagreements were resolved by discussion with a third author (T.R.) and consensus. The work was carried out in accordance with the PRISMA statement [11]. Selection criteria
Inclusion and exclusion criteria were delineated before the commencement of the literature search. Thus, eligible studies were included in this meta-analysis if they fulfilled the following criteria: (a) they were published as full articles, (b) they were written in the English language and (c) they provided sufficient data for the authors to construct a 2 × 2 contingency table to calculate sensitivity, specificity, positive likelihood ratios (PLRs) and negative likelihood ratios (NLRs), diagnostic odds ratios (DORs), receiver operating characteristic (ROC) and area under the curve (AUC). When two articles reported the same study, the publication that was more informative was selected. In studies reporting data according to the level of bowel preparation, data referring to the best level were taken into account. The methodological quality of each study was assessed by QUADAS (quality assessment for studies of diagnostic accuracy), which is an evidence-based quality assessment tool for use in systematic reviews of diagnostic accuracy studies [12]. The 14 items (Table 1) were assessed in all included articles, each of which was assessed as ‘yes’, ‘no’ or ‘unclear’. Accordingly, they were scored with 1 for ‘yes’, 0 for ‘unclear’ and − 1 for ‘no’ with a maximum score of 14. Statistical analysis
Agreement on the selection of studies between the two reviewers was evaluated using the k coefficient. Sensitivity,
specificity, NLRs and NLRs (PLR = sensitivity/1–specificity, NLR = 1–sensitivity/specificity) and DORs (DOR = PLR /NLR) of miRNAs [with corresponding 95% confidence intervals (CIs)] were calculated from the original numbers given in the included studies by constructing 2 × 2 contingency tables. PLR indicates how much the odds of disease increase when the test is positive, whereas NLR indicates how much the odds of CRC decrease when the miRNA test is negative. PLRs and NLRs are considered to be more clinically meaningful than sensitivity or specificity with PLR of more than 10 or NLR less than 0.1, indicating high accuracy [13]. Pooled results with corresponding 95% CIs were derived using the fixed-effects model (Mantel and Haenszel method) [14], unless significant heterogeneity was present, in which case the randomeffects model was used (DerSimonian and Laird method) [15]. Forest plots were constructed for visual display of individual studies and pooled results. In addition, the results of the individual studies were presented in a ROC space to show the distribution of sensitivities and specificities. A weighted symmetric summary ROC (sROC) curve was computed using the Moses–Shapiro–Littenberg method [16] and the AUC was calculated, with perfect tests having an AUC of 1 and poor tests having an AUC close to 0.5 [17]. To assess whether heterogeneity exists among the eligible studies, we first calculated the correlation coefficient and the P value between the logit of sensitivity and logit of 1 − specificity using the Spearman test to exclude the threshold effect. In this case, the Spearman correlation coefficient was significant when the P value was less than 0.05. When heterogeneity was not explained by the threshold effect, further evaluation was performed by calculating the Cochran Q-test and the inconsistency index (I2) [18,19]. Heterogeneity was considered to be present if the Q-test provided a P value of less than 0.1 [10] and I2 of more than 50% and in case of significant heterogeneity, we continued our attempts to explain this by exploring the study characteristics using metaregression. In addition, Deeks’ funnel plot asymmetry test was used to evaluate the potential publication bias [20] and a P value of less than 0.1 was considered to be representative of a significant publication bias. The statistical analyses were carried out using statistical package for the social sciences software version 14.0 (SPSS Inc., Chicago, Illinois, USA), the Comprehensive Meta-Analysis
Table 1. The QUADAS tool Yes 1. 2. 3. 4.
Was the spectrum of patients representative of the patients who will receive the test in practice? Were selection criteria clearly described? Is the reference standard likely to correctly classify the target condition? Is the time period between reference standard and index test short enough to be reasonably sure that the target condition did not change between the two tests? 5. Did the whole sample or a random selection of the sample receive verification using a reference standard of diagnosis? 6. Did patients receive the same reference standard regardless of the index test result? 7. Was the reference standard independent of the index test (i.e. the index test did not form part of the reference standard)? 8. Was the execution of the index test described in sufficient detail to permit replication of the test? 9. Was the execution of the reference standard described in sufficient detail to permit its replication? 10. Were the index test results interpreted without knowledge of the results of the reference standard? 11. Were the reference standard results interpreted without knowledge of the results of the index test? 12. Were the same clinical data available when test results were interpreted as would be available when the test is used in practice? 13. Were uninterpretable//intermediate test results reported? 14. Were withdrawals from the study explained?
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Item No Unclear
Circulating miRNAs and colorectal cancer Rokkas et al.
751 potentially eligible studies initially generated by the literature searches 603 rejected (title suggested article not appropriate) 148 abstracts retrieved
59 studies with extractable data
89 excluded (duplications, editorials, review articles)
43 excluded as not fulfilling the inclusion criteria 16 studies meta-analyzed Fig. 1. Flow diagram of the studies identified in this meta-analysis.
software version 2 (Biostat Inc., Englewood, New Jersey, USA), Stata version 12 (Stata Corporation, College Station, Texas, USA) and Meta-DiSc statistical software version 1.4 [Unit of Clinical Biostatistics, Ramon and Cajal Hospital, Madrid, Spain [21]. Results Descriptive assessment and study characteristics
A flow chart describing the process of study selection is shown in Fig. 1. Out of 751 titles initially generated by the literature searches, 16 studies remained eligible for metaanalysis [22–37]. Initial agreement between the reviewers for the selection of relevant articles was high (k = 0.995, 95% CI 0.989–1.00). The main characteristics of the 16 studies eligible for meta-analysis are shown in Table 2. They were carried out in different parts of the world and included a total of 1230 CRC patients and 810 controls.
www.eurojgh.com
821
data showed a PLR of 3.66 (2.62–5.10) and an NLR of 0.27 (0.21–0.36) (Fig. 2). As shown in the likelihood matrix (Fig. 3), the summary point of PLR combined with LRN lies in the right lower quadrant (PLR < 10, LRN > 0.1). This is in agreement with the modest pooled DOR magnitude of 14.25 (8.63–23.52) (Fig. 4). The corresponding ROC plot with sROC is also shown in Fig. 3. The AUC under the sROC was 0.860. There was no publication bias as judged by Deeks’ test (P = 0.384). However, significant heterogeneity (P < 0.001) was found in all parameters of diagnostic performance (i.e. sensitivity, specificity, PLR, NLR and DOR) (Fig. 2). Thus, apart from use of the random-effects model, metaregression and subgroup analyses were carried out to explore the reasons for the significant heterogeneity observed. First, we checked whether there was any significant threshold effect and found that this was not the case as the Spearman correlation coefficient was 0.117 and the P value was 0.666. Then, we attempted to explain the heterogeneity by exploring study characteristics, that is, age, ethnicity (Asians vs. Whites), QUADAS score and blood specimen (plasma vs. serum) through metaregression analyses, and no significant results were found (Table 3). Subsequently, we carried out subgroup analysis comparing the group of three studies [26–28] in which a single circulating miRNA (i.e. miRNA-21) was used with the group of 13 studies [22–25,29–37] that used a panel of various circulating miRNAs (Table 4). We found that the single miRNA-21 group performed better than the panel miRNAs group in all diagnostic value parameters, that is sensitivity 83.4% (78.7–87.5), specificity 91.6% (85.5– 95.7), PLR 9.88 (5.60–17.42), NLR 0.18 (0.14–0.24), DORs 56.91 (28.00–115.68) and AUC 0.947. Most importantly, in this group, heterogeneity became nonsignificant in all diagnostic accuracy parameters in comparison with the panel miRNA group.
Diagnostic performance of microRNAs
Discussion
The pooled data (random-effects model) showed circulating miRNA sensitivity of 78% [95% CI (75–80%)] and specificity of 79% (76–82%). Accordingly, the pooled
The development of suitable noninvasive biomarkers is important for the diagnosis of CRC. The recent discovery that noncoding miRNAs are stable in body fluids, such as
Table 2. Characteristics of the studies included References
Country
Ethnicity
Total number (patients/controls)
Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Kanaan et al. [26] Wang and Zhang [27] Toiyama et al. [28] Zhang et al. [29] Giráldez et al. [30] Wang et al. [31] Luo et al. [32]
China China China China USA China Japan China Spain China Germany
Asian Asian Asian Asian White Asian Asian Asian White Asian White
140 (90/50) 159 (100/59) 140 (103/37) 148 (90/58) 40 (20/20) 71(32/29) 239 (186/53) 164 (78/86) 95 (42/53) 49 (22/27) 224 (80/144)
Plasma Plasma Plasma Plasma Plasma Serum Serum Plasma Plasma Plasma Plasma
Liu et al. [33] Hofsli et al. [34] Brunet Vega et al. [35] Zanutto et al. [36] Du et al. [37]
China Norway Spain Italy China
Asian White White White Asian
280 50 56 58 98
Serum Serum Serum Plasma Plasma
(200/80) (40/10) (30/26) (29/29) (49/49)
Blood specimen
mRNA profile miRNA-17-3p, -92 miRNA-29a, -92a miRNA-21, -221, -222 miRNA-601, -760, -29a, -92a miRNA-21 miRNA-21 miRNA-21 miRNA-200c, 18a miRNA-19a, -19b, -15b miRNA-409-3p, -7, -93 miRNA -18a, -20a, -21, -29a, -92a, -106b, -133a, -143, -145, -181b, -342-3p, -532-3p miRNA-21, -31, -92a,- 18a, -106a Panel of 21 miRNAsa miRNA-16-5p, let7a-5p, 103-5p miRNA-378, -21 miRNA-21, 92a
a
miRNA-423-5p, -210, -720, -320a, -378, -92a, -29a, -155, -106a, -143, -103,-199a-3p, -151-5p, -107, -191, -423-3p, -221, -let7d, -382, -409-3p, -652.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Assay method qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR qRT-PCR
822
July 2015 • Volume 27 • Number 7
European Journal of Gastroenterology & Hepatology
(a)
(c)
Sensitivity (95% CI) Ng et al. [22] 0.89 0.83 Huang et al. [23] 0.86 Pu et al. [24] 0.83 Wang et al. [25] Wang and Zhang [27] 0.88 Kanaan et al. [26] 0.90 0.82 Wang et al. [31] Luo et al. [32] 0.54 Zhang et al. [29] 0.85 Giráldez et al. [30] 0.79 Hofsli et al. [34] 0.88 Brunet Vega et al. [35] 0.67 Liu et al. [33] 0.65 Toiyama et al. [28] 0.83 0.59 Zanutto et al. [36] Du et al. [37] 0.76
Positive LR (95% CI)
(0.81 − 0.95) (0.74 − 0.90) (0.78 − 0.92) (0.74 − 0.90) (0.71 − 0.96) (0.68 − 0.99) (0.60 − 0.95) (0.42 − 0.65) (0.75 − 0.92) (0.63 − 0.90) (0.73 − 0.96) (0.47 − 0.83) (0.58 − 0.72) (0.77 − 0.88) (0.39 − 0.76) (0.61 − 0.87)
Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31] Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
Pooled positive LR = 3.66 (2.62 − 5.10) Cochran Q = 82.61, d.f. = 15 (P = 0.0000) Inconsistency (I 2) = 81.8% τ 2 = 0.3365
Pooled sensitivity 0.78 (0.75−0.80) 2 = 74.6, d.f. = 15 (P = 0.0000) Inconsistency (I 2) = 79.9%
0
0.2
0.4 0.6 Sensitivity
0.8
1
Random-effects model
1 Positive LR
0.01
(b)
100.0
Random-effects model
Specificity (95% CI) (d) Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31] Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
Negative LR (95% CI)
0.70 (0.55 − 0.82) 0.85 (0.73 − 0.93) 0.41 (0.25 − 0.58) 0.93 (0.83 − 0.98) 0.74 (0.58 − 0.87) 0.90 (0.68 − 0.90) 0.89 (0.71 − 0.98) 0.82 (0.75 − 0.88) 0.76 (0.65 − 0.84) 0.79 (0.66 − 0.89) 0.90 (0.55 − 1.00) 0.58 (0.37 − 0.77) 0.85 (0.75 − 0.92) 0.91 (0.79 − 0.97) 0.59 (0.39 − 0.76) 0.82 (0.68 − 0.91)
Ng et al. [22] Huang et al. [23 Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31 Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
0.2
0.4 0.6 Specificity
0.8
1
0.16 (0.09 − 0.29) 0.20 (0.13 − 0.31) 0.34 (0.18 − 0.63) 0.18 (0.11 − 0.29) 0.17 (0.07 − 0.43) 0.11 (0.03 − 0.42) 0.20 (0.08 − 0.50) 0.56 (0.44 − 0.72) 0.20 (0.12 − 0.35) 0.27 (0.15 − 0.49) 0.14 (0.06 − 0.32) 0.58 (0.32 − 1.06) 0.41 (0.33 − 0.51) 0.19 (0.14 − 0.26) 0.71 (0.42 − 1.20) 0.30 (0.18 − 0.50)
Pooled negative LR = 0.27 (0.21-0.36) Cochran Q = 77.92, d.f. = 15 (0.0000) Inconsistency (I 2) = 80.8% τ 2 = 0.2270
Pooled specificity 0.79(0.76-0.82) 2 = 63.71, d.f. = 15 (P = 0.0000) Inconsistency (I 2) = 76.5%
0
2.96 (1.93 − 4.55) 5.44 (2.96 − 9.99) 1.45 (1.10 − 1.92) 12.08 (4.67 − 31.2) 3.41 (1.97 − 5.92) 9.00 (2.40 − 33.79) 7.36 (2.49 − 21.79) 2.98 (2.98 − 4.45) 3.47 (2.36 − 5.09) 3.79 (2.19 − 6.56) 8.75 (1.36 − 56.38) 1.58 (0.94 − 2.64) 4.33 (2.55 − 7.37) 8.78 (3.80 − 20.26) 1.42 (0.83 − 2.41) 4.11 (2.23 − 7.58)
Random-effects model
0.01
1 Negative LR
100.0
Random-effects model
Fig. 2. Forest plots showing (a) individual and pooled (95% CIs) sensitivities. (b) Individual and pooled (95% CIs) specificities. (c) Individual and pooled (95% CIs) positive likelihood ratios (LR). (d) Individual and pooled (95% CIs) negative LR. CI, confidence interval.
(a)
(b) 100
Sensitivity 1
SROC curve Symmetric SROC AUC = 0.8605 SE(AUC) = 0.0232 Q∗ = 0.7913 SE(Q∗) = 0.0224
0−9
Positive likelihood ratio
0−8 0−7 0−6 10
0−5 0−4 0−3 0−2 0−1
1 0.01
0 0.1
0.3 0.5 0.7 0.9 Negative likelihood ratio
0.99
0
0−2
0−4 0−6 1−specificity
0−8
1
Fig. 3. (a) Likelihood matrix showing individual (solid dots) and pooled (diamond shape) values of positive likelihood ratios combined with negative likelihood ratios. (b) Summary receiver operating characteristic curves (SROC) with area under the curve (AUC) of miRNAs describing the diagnostic performance. Every dot stands for a study.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
95% CI
0.009 − 0.297 − 0.105 − 0.555
0.1253 0.8005 0.3162 0.9817
0.9454 0.7196 0.7464 0.5856
1.01 0.74 0.90 0.57
0.12–4.55 0.12–4.55 0.44–1.84 0.06–5.29
plasma and serum, created an opportunity to develop such novel diagnostic biomarkers. However, their diagnostic accuracy is inconsistent or even contradictory in the literature. The pooled data in this meta-analysis showed circulating miRNA sensitivity of 78% and specificity of 79%, with corresponding PLR 3.66 and NLR 0.27. As shown in the likelihood matrix (Fig. 3), the summary point of PLR combined with LRN lies in the right lower quadrant (PLR < 10, LRN >0.1), indicating that the circulating miRNAs may not have sufficient power either to confirm or to exclude cancer. This is in agreement with the modest pooled DOR magnitude of 14.25, which expresses the strength of the association between circulating miRNAs and CRC. Accordingly, the AUC was 0.860, indicating a relatively good accuracy. However, these results are overshadowed by significant heterogeneity, which was found in all parameters of diagnostic performance (i.e. sensitivity, specificity, PLR, NLR and DOR). There was no obvious explanation for this significant heterogeneity s there was no significant threshold effect and in addition no significant results were found in meta-regression analyses exploring study characteristics, that is, age, ethnicity (Asians vs. Whites), QUADAS score and blood specimen (plasma vs. serum). The subsequent subgroup analysis, that is the single miRNA versus panel of various miRNAs, yielded interesting findings in that the single miRNA-21 group showed good results in all parameters of diagnostic performance, that is sensitivity 83.4%, specificity 91.6%, PLR 9.88, NLR 0.18, DOR 56.91 and AUC 0.947 and, in Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Table 4. Subgroup analysis (microRNA panel vs. microRNA-21)
CI, confidence interval; Cte, constant coefficient; RDOR, relative diagnostic odds ratio; S, statistic S.
0.947
CI, confidence interval; DOR, diagnostic odds ratio; LR, likelihood ratio.
RDOR
3.55 (2.54–4.96) (Q = 60.59, d.f. = 12, P < 0.001, I2 = 80.2%) 9.88 (5.6 0–17.42) (Q = 0.27, d.f. = 2, P = 0.874, I2 = 0%)
P value
78.6 (75.4–81.5) (Q = 50.65, d.f. = 12, P < 0.001, I2 = 76.3%) 91.6 (85.5–95.7) (Q = 0.36, d.f. = 2, P = 0.854, I2 = 0%)
SE
77.1 (74.3–79.7) (Q = 64.1, d.f. = 12, P < 0.001, I2 = 81.3%) 83.4 (78.7–87.5) (Q = 0.76, d.f. = 2, P = 0.683, I2 = 0%)
Cte S Age Ethnicity QUADAS score Specimen source (plasma vs. serum)
Coefficient
Positive LR (95% CI) (heterogeneity)
Variables
Specificity (95% CI) (heterogeneity)
Table 3. Results of the multivariable meta-regression model, for the parameters taken into account, with backward regression analysis (inverse variance weights)
Sensitivity (95% CI) (heterogeneity)
Negative LR (95% CI) heterogeneity
Fig. 4. Forest plots showing individual and pooled (95% CIs) diagnostic odds ratios. CI, confidence interval.
13.72 (8.50–22.13) (Q = 37. 79, d.f. = 12, P < 0.001, I2 = 65.9%) 56.91 (28.00–115.68) (Q = 0.36, d.f. = 2, P = 0.834, I2 = 0%)
Random-effects model Pooled diagnostic odds ratio = 14.25 (8.64 − 23.52) Cochran Q = 59.40; d.f. = 15(P = 0.0000) Inconsistency (I 2) = 74.7% τ 2 = 0.7272
0.27 (0.20–0.36) (Q = 56.60, d.f. = 12, P < 0.001, I2 = 78.8%) 0.18 (0.14–0.24) (Q = 0.67, d.f. = 2, P = 0.716, I2 = 0%)
DOR (95% CI) heterogeneity
0.01 1 100.0 Diagnostic odds ratio
18.67 (7.64 − 45.61) 27.12 (11.24 − 65.45) 4.33 (1.82 − 10.29) 67.50 (21.22 − 214.69) 20.30 (5.70 − 72.32) 81.00 (10.26 − 639.34) 36.00 (7.15 − 181.35) 5.27 (2.86 − 9.72) 17.02 (7.74 − 37.42) 14.00 (5.19 − 37.75) 63.00 (6.52 − 608.93) 2.73 (0.92 − 8.09) 10.52 (5.34 − 20.75) 46.20 (17.05 − 125.17) 2.01 (0.71 − 5.71) 13.70 (5.18 − 36.26)
miRNA panel (13 studies) miRNA-21 (three studies)
Diagnostic OR (95% CI) Ng et al. [22] Huang et al. [23] Pu et al. [24] Wang et al. [25] Wang and Zhang [27] Kanaan et al. [26] Wang et al. [31] Luo et al. [32] Zhang et al. [29] Giráldez et al. [30] Hofsli et al. [34] Brunet Vega et al. [35] Liu et al. [33] Toiyama et al. [28] Zanutto et al. [36] Du et al. [37]
0.858
www.eurojgh.com
AUC
Circulating miRNAs and colorectal cancer Rokkas et al.
823
824
July 2015 • Volume 27 • Number 7
European Journal of Gastroenterology & Hepatology
this respect, performed much better than the panel of the various miRNAs group. Most importantly, in the single miRNA-21 group, heterogeneity became nonsignificant in all diagnostic parameters in comparison with the panel miRNA group, where heterogeneity was significant. All the above might mean that miRNA-21 is a good candidate as an accurate noninvasive novel biomarker for CRC diagnosis, provided that these results are confirmed by further prospective studies. It must be noted that the diagnostic performance of miRNA-21 in this meta-analysis is the best so far in comparison with other blood markers studied, such as carcinoembryonic antigen [38], carbohydrate antigen 19-9 [30] and miRNA-92a [39] In this meta-analysis, we included 16 eligible studies, which is the largest so far in comparison with previous similar studies [40–42], and undoubtedly, this represents an advantage. However, there are several limitations in this meta-analysis. Thus, because of the fact that the clinical value of miRNA in CRC has only been explored recently, small sample size studies are included in our meta-analysis, and as a result, small-study effects are inescapable. Therefore, it is necessary to strengthen our conclusion by further validations of miRNA-21 in large prospective studies. Furthermore, there might still be some more biases in our meta-analysis. For example, in terms of the design of eligible studies, only two had a clearly prospective design as opposed to the remaining studies. Moreover, no studies unequivocally mentioned whether a blind design was used. Future studies should address these points; in addition, other parameters, which might potentially influence the results, should be addressed. These parameters might be related to patients, such as TNM classification (I/II/III/IV) for CRC, or technical aspects in terms of the methods of RNA extraction (miRNsasy vs. TRIzol LS) and the method of miRNA detection [qRT-PCR (Taqman) vs. qRT-PCR (SYBR green) vs. miCURY vs. Agilent microarray]. In conclusion, circulating miRNAs are promising noninvasive diagnostic biomarkers in CRC. In particular, miRNA-21 is a good candidate as an accurate novel biomarker. However, more prospective well-designed studies are required to further explore its diagnostic role and its incorporation into clinical practice.
6
7
8
9
10
11
12
13
14
15 16
17 18 19 20
21
22
23
Acknowledgements 24
Conflicts of interest
There are no conflicts of interest. 25
References 1 Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011; 61:69–90. 2 O’Connell JB, Maggard MA, Ko CY. Colon cancer survival rates with the new American Joint Committee on Cancer sixth edition staging. J Natl Cancer Inst 2004; 96:1420–1425. 3 Zauber AG, Lansdorp-Vogelaar I, Knudsen AB, Wilschut J, van Ballegooijen M, Kuntz KM. Evaluating test strategies for colorectal cancer screening: a decision analysis for the U.S. Preventive Services Task Force. Ann Intern Med 2008; 149:659–669. 4 Wu WK, Law PT, Lee CW, Cho CH, Fan D, Wu K, et al. MicroRNA in colorectal cancer: from bench top to bedside. Carcinogenesis 2011; 32:247–253. 5 Bartel DP. MicroRNAs: target recognition and regulatory functions. Cell 2009; 136:215–233.
26
27
28
29
30
Chiang HR, Schoenfeld LW, Ruby JG, Auyeung VC, Spies N, Baek D, et al. Mammalian microRNAs: experimental evaluation of novel and previously annotated genes. Genes Dev 2010; 24:992–1009. Michael MZ, O' Connor SM, van Holst Pellekaan NG, Young GP, James RJ. Reduced accumulation of specific microRNAs in colorectal neoplasia. Mol Cancer Res 2003; 1:882–891. Ma Y, Zhang P, Yang J, Liu Z, Yang Z, Qin H. Candidate microRNA biomarkers in human colorectal cancer: systematic review profiling studies and experimental validation. Int J Cancer 2012; 130:2077–2087. Healy NA, Heneghan HM, Miller N, Osborne CK, Schiff R, Kerin MJ. Systemic miRNAs 609 as potential biomarkers for malignancy. Int J Cancer 2012; 131:2215–2222. Chen X, Ba Y, Ma L, Cai X, Yin Y, Wang K, et al. Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Res 2008; 18:997–1006. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JP, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health car-interventions: explanation and elaboration. J Clin Epidemiol 2009; 62:e1–34. Whiting P, Rutjes AW, Reitsma JB, Bossuyt PM, Kleijnen J. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003; 3:25. Rosenfeld RM, Shiffman RN. Clinical practice guidelines: a manual for developing evidence-based guidelines to facilitate performance measurement and quality improvement. Otolaryngol Head Neck Surg 2006; 135:S1–S28. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 719–748. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177–188. Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med 1993; 12:1293–1316. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982; 143:29–36. Cochran WG. The combination of estimates from different experiments. Biometrics 1954; 8:101–129. Higgins JP, Thompson SG. Quantifying heterogeneity in a metaanalysis. Stat Med 2002; 21:1539–1558. Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol 2005; 58:882–893. Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Methodol 2006; 6:31. Ng EK, Chong WW, Jin H, Lam EK, Shin VY, Yu J, et al. Differential expression of microRNAs in plasma of patients with colorectal cancer: a potential marker for colorectal cancer screening. Gut 2009; 58:1375–1381. Huang Z, Huang D, Ni S, Peng Z, Sheng W, Du X. Plasma microRNAs are promising novel biomarkers for early detection 684 of colorectal cancer. Int J Cancer 2010; 127:118–126. Pu XX, Huang GL, Guo HQ, Guo CC, Li H, Ye S, et al. Circulating miR-221 directly amplified from plasma is a potential diagnostic and prognostic marker of colorectal cancer and is correlated with p53 expression. J Gastroenterol Hepatol 2010; 25:1674–1680. Wang Q, Huang Z, Ni S, Xiao X, Xu Q, Wang L, et al. Plasma miR-601 and miR-760 are novel biomarkers for the early detection of colorectal cancer. PLoS One 2012; 7:e44398. Kanaan Z, Rai SN, Eichenberger MR, Roberts H, Keskey B, Pan J, Galandiuk S. Plasma miR-21: a potential diagnostic marker of colorectal cancer. Ann Surg 2012; 256:544–551. Wang B, Zhang Q. The expression and clinical significance of circulating microRNA-21 in serum of five solid tumors. J Cancer Res Clin Oncol 2012; 138:1659–1666. Toiyama Y, Takahashi M, Hur K, Nagasaka T, Tanaka K, Inoue Y, et al. Serum miR-21 as a diagnostic and prognostic biomarker in colorectal cancer. J Natl Cancer Int 2013; 105:849–859. Zhang GJ, Zhou T, Liu ZL, Tian HP, Xia SS. Plasma miR-200c and miR-18a as potential biomarkers for the detection of colorectal carcinoma. Mol Clin Oncol 2013; 1:379–384. Giráldez MD, Lozano JJ, Ramírez G, Hijona E, Bujanda L, Castells A, Gironella M. Circulating microRNAs as biomarkers of colorectal cancer:
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Circulating miRNAs and colorectal cancer Rokkas et al.
31
32
33
34
35
36
results from a genome-wide profiling and validation study. Clin Gastroenterol Hepatol 2013; 11:681–688. Wang S, Xiang J, Li Z, Lu S, Hu J, Gao X, et al. A plasma microRNA panel for early detection of colorectal cancer. Int J Cancer 2015; 136:152–161. Luo X, Stock C, Burwinkel B, Brenner H. Identification an evaluation of plasma microRNAs for early detection of colorectal cancer. PLoS One 2013; 14:5e62880. Liu GH, Zhou ZG, Chen R, Wang MJ, Zhou B, Li Y, Sun XF. Serum miR-21 and miR-92a as biomarkers in the diagnosis and prognosis of colorectal cancer. Tumour Biol 2013; 34:2175–2181. Hofsli E, Sjursen W, Prestvik WS, Johansen J, Rye M, Tranø G, et al. Identification of serum microRNA profiles in colon cancer. Br J Cancer 2013; 108:1712–1719. Brunet Vega A, Pericay C, Moya I, Ferrer A, Dotor E, Pisa A, et al. microRNA expression profile in stage III colorectal cancer: circulating miR-18a and miR-29a as promising biomarkers. Oncol Rep 2013; 30:320–326. Zanutto S, Pizzamiglio S, Ghilotti M, Bertan C, Ravagnani F, Perrone F, et al. Circulating miR-378 in plasma: a reliable, haemolysis-
www.eurojgh.com
37
38
39
40
41
42
825
independent biomarker for colorectal cancer. Br J Cancer, 2014; 110:1001–1007. Du M, Liu S, Gu D, Wang Q, Zhu L, Kang M, et al. Clinical potential role of circulating microRNAs in early diagnosis of colorectal cancer patients. Carcinogenesis 2014; 35:2723–2730. Thomson DM, Krupey J, Freedman SO, Gold P. The radioimmunoassay of circulating carcinoembryonic antigen of the human digestive system. Proc Natl Acad Sci U S A 1969; 64:161–167. Hundt S, Haug U, Brenner H. Blood markers for early detection of colorectal cancer: a systematic review. Cancer Epidemiol Biomarkers Prev 2007; 16:1935–1953. Yang X, Zeng Z, Hou Y, Yuan T, Gao C, Jia W, et al. MicroRNA-92a as a potential biomarker in diagnosis of colorectal cancer: a systematic review and meta-analysis. PLoS One 2014; 9:e88745. Zhou XJ, Dong ZG, Yang YM, Du LT, Zhang X, Wang CX. Limited diagnostic value of microRNAs for detecting colorectal cancer: a metaanalysis. Asian Pac J Cancer Prev 2013; 14:4699–4704. Jiang JX, Zhang N, Liu ZM, Wang YY. Detection of microRNA-21 expression as a potential screening biomarker for colorectal cancer: a meta-analysis. Asian Pac J Cancer Prev 2014; 15:7583–7588.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
