Tumor Biol. DOI 10.1007/s13277-014-2777-0
RESEARCH ARTICLE
MicroRNAs used as novel biomarkers for detecting cancer metastasis Chunshan Han & Haixiang Yu & Lening Zhang & Xiaoli Li & Yonggang Feng & Hua Xin
Received: 23 September 2014 / Accepted: 23 October 2014 # International Society of Oncology and BioMarkers (ISOBM) 2014
Abstract The low survival rates of cancers are primarily due to late diagnosis and metastasis. Discriminating the metastasis is a crucial factor for prognosis and improving the survival rate of cancer patients. MicroRNAs (miRNAs) can regulate the expression of hundreds of downstream genes, which has a broad effect on the regulation of the whole cell cycle. Accumulating studies have found that the aberrant expression of miRNAs is associated with cancer genesis. The aim of this study is to evaluate the diagnostic value of miRNAs in detecting cancer metastasis. Medline, PubMed, Embase, and CNKI were searched for relevant articles. Sensitivity, specificity, positive and negative likelihood ratio (PLR, NLR) and diagnostic odds ratio (DOR), the summary receiver operator characteristic (SROC) curve and the calculated AUC (area under the SROC curve) were applied to explore the diagnostic accuracy of miRNAs in metastasis. Seven hundred seventyone metastatic cancer patients and 552 non-metastatic cancer controls from 14 articles were involved in our meta-analysis. A sensitivity of 0.75 (95 % confidence interval (CI), 0.72– 0.79) and a specificity of 0.80 (95 % CI, 0.76–0.84) were observed from metastatic patients and non-metastatic controls in the combined analysis. And the AUC was 0.83 (95 % CI, Electronic supplementary material The online version of this article (doi:10.1007/s13277-014-2777-0) contains supplementary material, which is available to authorized users. C. Han : H. Yu : L. Zhang : Y. Feng : H. Xin (*) Department of Thoracic Surgery, China-Japan Union Hospital of Jilin University, No. 126, Xiantai Street, Changchun 130033, Jilin, China e-mail: [email protected] H. Xin e-mail: [email protected] X. Li Pharmaceutical Department, China-Japan Union Hospital of Jilin University, Changchun 130033, Jilin, China
0.79–0.86). In addition, results from subgroup analyses suggested that a higher diagnostic value for metastasis was acquired in tissue sample other than blood sample (sensitivity, 0.82 versus 0.73; specificity, 0.84 versus 0.79; PLR, 5.0 versus 3.5; NLR, 0.22 versus 0.34; DOR, 23 versus 10; AUC, 0.88 versus 0.80). In summary, this meta-analysis proved the relatively high diagnostic value of miRNA in metastasis, which might be applied as a novel screening tool to detect metastasis along with other biomarkers. We also illustrated that tissue-based miRNAs may have a better diagnostic accuracy than blood-based miRNAs. Keywords microRNA . Tissue . Blood . Metastasis . Cancer . Screening . Meta-analysis
Introduction Cancer has become a major health problem worldwide through years. It is reported that in the USA, one in four deaths is caused by cancer [1]. With the aged tendency of population, deteriorative environment, and unhealthy lifestyles, the cancer incidence and mortality has still been in a rising tendency. Although the prediction and treatment on cancer have achieved a great progress, which retards the mortality rate, the 5-year survival rate of some kinds of cancer can be relatively low. Some studies showed that lung cancer had the 5-year survival rate of 13 % [2], while hepatocellular carcinoma had a low rate of 5-year survival being approximately 5 % [3]. The survival rates of cancers are due to the following factors: diagnosis stages and metastasis. It was reported that gastric cancer had 5 % 5-year survival rate in the advanced stage (III and IV), but it can achieve nearly 90 % if patients are diagnosed in the early stage (I and II) [4]. Hence, it is an efficient method to reduce cancer mortality by detecting cancer in the early stage. Metastasis is another factors
Tumor Biol.
acting on cancer mortality. Metastasis is the spread cancer cells from one organ to another, which often occurs in malignant tumor tissues. The metastasis can be divided into four types based on routes: transcoelomic, lymphatic spread, hematogenous spread, and transplantation. Study showed that the 5-year survival rate of patients with lymph node metastasis declined drastically to 50–65 %, compared with the rate of approximately 80–95 % in patients with positive lymph nodes [5]. In another study, it showed that the 5-year survival rate of cervical cancer patients without lymph node metastasis can reach 100 % compared to that of 63.4 % in patients with lymph node metastasis at early stage (I/II) [6]. Therefore, discriminating the metastasis is a crucial factor for prognosis and improving the survival rate of cancer patients, and the application of a suitable therapeutic method is grounded in the identification of metastasis [7]. The traditional method to detect cancer metastasis is by imaging diagnosis. Doctors examine the samples taken from patients with microscope and other adjuvant techniques such as FISH. The normal existence of various type of cell can be tested from the tissue sample, the lack of which suggests the metastasis of tumor cells. Despite the high accuracy of this method, it hurts patients for its invasive nature. And it can be found only after a period of metastasis, which cannot provide efficient information for treating cancer in advance. The high expense and complex procedures are also obstacles for the clinical application of imaging diagnosis. However, there are other special ways to detect the corresponding cancer metastasis. For example, a subfraction of TA-4 and a tumorassociated antigen, called squamous cell carcinoma antigen, is widely used as the marker for squamous cell carcinoma. However, because of the non-organ specificity for cervix, the normal initial squamous cell carcinoma antigen is unable to exclude the lymph node metastasis. Hence, a more sensitive, non-invasive, and efficient strategy for detecting metastasis is needed. MicroRNAs, which are found to participate in tumorgenesis, provide a new road for detecting metastasis. MicroRNAs (miRNAs) are small non-protein-coding RNA molecules with a length of 18 to 24 nucleotides. A single miRNA can regulate the expression of hundreds of downstream genes, which has a broad effect on the regulation of the whole cell cycle. It is revealed that miRNAs are involved in cellular proliferation, apoptosis, differentiation, metabolism, and other cellular processes [8]. MiRNAs play important roles in transcription or posttranscription level by specific pairing with the 3′ UTR of the target messenger RNA (mRNA), leading mRNA activation inhibition or degradation of target mRNA. Accumulating studies have found that the aberrant expression of miRNAs is associated with cancer genesis. MiRNAs are found to be stable and abundant in body liquid, such as serum, plasma, and sputum, indicating it can be served as biomarker for cancer diagnosis. As a novel, cost-efficient, and convenient biomarker, miRNAs are widely used in detecting metastasis of cancers. However, there are inconsistent results revealed from
studies, leading us to conduct this meta-analysis. The aim of this study is to evaluate the diagnostic value of miRNAs in detecting cancer metastasis.
Methods Search strategy We conducted a literature search to identify relevant studies on miRNAs detecting metastasis from databases including Medline, PubMed, Embase, and CNKI. The articles were searched until July 2014, with no language restriction. The following keywords were used: (“neoplasm metastasis”[Mesh] OR neoplasm metastases OR metastasis OR metastases OR lymphatic metastasis OR lymphatic metastases OR lymph node metastasis) AND (“microRNAs”[MeSH Terms] OR “microRNAs” OR “miRNA”) AND (“ROC curve”[Mesh] OR “diagnosis”[Mesh] OR “sensitivity and specificity”[Mesh]). More studies were added by reviewing the reference lists of articles. This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidance (Supplement S1). Inclusion and exclusion criteria The eligible studies have to fulfill the following criteria: (1) studies related to miRNAs in detecting cancer metastasis, (2) histopathology used for detection of cancer, (3) studies providing necessary data for evaluating diagnostic value of miRNAs for detecting metastasis, and (4) studies with sufficient validity. The publications were removed based on the following criteria: (1) duplicate publications, (2) studies with insufficient data, and (3) studies in the form of letters, reviews, editorials, meeting abstracts, and case reports. Data extraction Two reviewers extracted data independently from the eligible studies. We collected relevant information including country, sample size, cancer site, metastasis site, the types of miRNAs, and specimen. Statistical analysis The STATA 12.0 (Stata Corp, College Station, TX, USA) software was performed to analyze the extracted data, with two-sided P value. We calculated the pooled parameters for assessing the diagnostic value, including sensitivity, specificity, positive and negative likelihood ratio (PLR, NLR), and diagnostic odds ratio (DOR). The PLR and NLR present the clinical utility, and the DOR indicates the accuracy with the
Tumor Biol.
combination of sensitivity and specificity. We also constructed the summary receiver operator characteristic (SROC) curve (based on the sensitivity and specificity of each study) and calculated the AUC (area under the SROC curve) to explore the diagnostic accuracy of miRNAs in metastasis. Considering studies involving in more than one miRNA, the studied miRNAs were integrated into a comprehensive factor by regression analysis. The I2 test was conducted; with I2 > 50 % indicates the existence of heterogeneity. We used subgroup analyses to find the potential sources of heterogeneity. The influence of single on overall result was analyzed by goodness of fit, bivariate normality, Cook’s distance, and outlier detection. The publication bias was tested by Deek’s funnel plot asymmetry test.
Databases Search Medline, PubMed, Embase, CNKI, etc.
(N = 1013) Duplicate Publications (N = 117) Abstract and Keyword Review Two reviewers review abstract and keyword
(N = 896) Unrelated Studies (N = 853) Full-text and Data Accessment Two reviewers assess full-tex and data
(N = 43) Data Unavailable
Results Data selection and study characteristics A flow diagram (Fig. 1) illustrated the selection process and the reasons for exclusion. Of all the 1013 retrieved articles, we selected 14 distinguished articles according to our inclusion and exclusion criteria [9–22]. In the 14 available articles, 34 studies were conducted, with a total of 771 metastatic cancer patients and 552 non-metastatic cancer controls enrolled in our meta-analysis. Characteristics of all identified studies were listed in Table 1 in the order of publication year. All the articles were published between 2011 and 2014. Asian population is the focus of all studies. Specimens were from tissue (n=10), serum (n=17), or plasma (n=7). Specimens from serum and plasma were combined as blood samples (n=24). Pooled diagnostic accuracy of miRNA profiles in cancer metastasis The diagnostic performance of miRNAs in differentiating cancer with metastasis from those without metastasis was presented in Table 2. In the analysis for combined specimens, the I 2 value of sensitivity was 40.12 %; no statistically significant heterogeneity was identified between studies corresponding to sensitivity. The fixed effect model was applied with no heterogeneity observed, whereas the I 2 value of specificity was 62.70 %, which indicated the existence of significant heterogeneity, and the random effect model was adopted. Similar results were also acquired from stratified analysis by tissue and blood sample. Based on the models mentioned above, the evaluation for the sensitivity and specificity of miRNAs in diagnosing metastasis was illustrated in Fig. 2. A sensitivity of 0.75 (95 % confidence interval (CI), 0.72–0.79) and a specificity of 0.80
(N = 29) Final Inclusion of Articles (N = 14)
Fig. 1 A flow chart of study identification, inclusion, and exclusion
(95 % CI, 0.76–0.84) were observed from metastatic patients and non-metastatic controls in the combined analysis. And the AUC was 0.83 (95 % CI, 0.79–0.86), corresponding SROC curve was constructed and presented in Fig. 3. A relatively high diagnostic accuracy was thus acquired from the result. Subgroup and meta-regression analyses Subgroup analyses were conducted by a type of specimen which have been mentioned to be related to the accuracy of diagnosis by previous research [10]. The results of subgroups were also presented in Table 2. A higher diagnostic value for metastasis was observed in tissue sample other than blood sample (sensitivity, 0.82 versus 0.73; specificity, 0.84 versus 0.79; PLR, 5.0 versus 3.5; NLR, 0.22 versus 0.34; DOR, 23 versus 10; AUC, 0.88 versus 0.80), indicating that cancer tissue might be a better sample for miRNA assays in metastasis diagnosis. The SROC curve was illustrated in Fig. 4. Sensitivity analyses and publication bias The influence of a single study on the overall results was assessed by influence analysis. Goodness of fit, bivariate normality, Cook’s distance, and outlier detection were applied. From the illustrated results, a study from Chen et al. [9] was
Tumor Biol. Table 1 Characteristics of all identified studies
Included study
Wu et al. [19] Chen et al. [10] Huang et al. [15] Wang and Gu [18] Zhao et al. [21] Chen et al. [9]
Chen et al. [12] Chen et al. [13] Kim et al. [16] Zhao et al. [22] Feng et al. [14] Chen et al. [11] Lu et al. [17]
Yin et al. [20]
Country
Sample size
Cancer site
Metastasis site
MicroRNA
Specimen
Lymph node Lymph node
miR-212, −195
Tissue
miR-21
Serum
Tissue
Serum
M (+)
M (−)
China
22
22
Gastric
China
40
25
Cervical
China
13
31
Cervical
Lymph node
China
38
36
Colorectal
Liver
miR-100, −125b, −143, −145, −199a-5p, let7c miR-29a
China
100
22
Breast
Bone
miR-10b
Serum
China
40
40
Cervical
Lymph node
Tissue
40
40
Cervical
Lymph node
China
25
35
Breast
Lymph node
miR-1246, −20a, −2392, −3147, −3162-5p, −4484 miR-1246, −20a, −2392, −3147, −3162-5p, −4484 miR-10b, −373,
Plasma
China
14
37
Cervical
Lymph node
miR-143
Tissue
Korea
16
15
Gastric
Lymph node
Serum
China
40
40
Cervical
Lymph node
miR-21, −27a, −106b, −146a, −148a, −223, −433 miR-20a, −203
China
48
50
Colorectal
Liver
miR-29a
Serum
China
36
36
Gastric
Various
miR-122, −192
Plasma
China
120
30
Nasopharyngeal
miR-9
Plasma
63
21
Nasopharyngeal
miR-9
Plasma
116
72
Colorectal
Lymph node Lymph node Liver
miR-126, −141, −21
Serum
China
identified as an outlier. Yet, a similar accuracy for diagnosis was acquired with the outlier removed from the analyses. The result of influence test suggested that our meta-analysis was robust. The publication bias of eligible articles was assessed by Deek’s funnel plot and linear regression test. A P value of 0.17 was appraised. It was suggested that no statistically significant publication bias was observed across enrolled studies.
Serum
Serum
Discussion In this meta-analysis, we investigate the diagnostic value of miRNAs in metastasis diagnosis. Seven hundred seventy-one metastatic patients and 552 non-metastatic controls from 34 studies were enrolled in our analysis. We observed a relatively high diagnostic value for majority of miRNAs, which indicated the validity of miRNAs as biomarkers.
Tumor Biol. 1.0
Table 2 Diagnostic performance of miRNAs in differentiating cancer with metastasis from those without metastasis Analysis
Combine
Tissue
12 9 18 1
Blood
28
20 13 21
5 7 19
2 23 2 27
6
34 0.75 [0.72–0.79] 40.12 (P=0.01) 0.80 [0.76–0.84]
10 0.82 [0.72–0.88] 42.27 (P=0.06) 0.84 [0.70–0.92]
24 0.73 [0.70–0.76] 26.46 (P=0.012) 0.79 [0.74–0.83]
I2 (%) PLR (95 % CI) NLR (95 % CI) DOR (95 % CI) AUC (95 % CI)
62.70 (P
RESEARCH ARTICLE
MicroRNAs used as novel biomarkers for detecting cancer metastasis Chunshan Han & Haixiang Yu & Lening Zhang & Xiaoli Li & Yonggang Feng & Hua Xin
Received: 23 September 2014 / Accepted: 23 October 2014 # International Society of Oncology and BioMarkers (ISOBM) 2014
Abstract The low survival rates of cancers are primarily due to late diagnosis and metastasis. Discriminating the metastasis is a crucial factor for prognosis and improving the survival rate of cancer patients. MicroRNAs (miRNAs) can regulate the expression of hundreds of downstream genes, which has a broad effect on the regulation of the whole cell cycle. Accumulating studies have found that the aberrant expression of miRNAs is associated with cancer genesis. The aim of this study is to evaluate the diagnostic value of miRNAs in detecting cancer metastasis. Medline, PubMed, Embase, and CNKI were searched for relevant articles. Sensitivity, specificity, positive and negative likelihood ratio (PLR, NLR) and diagnostic odds ratio (DOR), the summary receiver operator characteristic (SROC) curve and the calculated AUC (area under the SROC curve) were applied to explore the diagnostic accuracy of miRNAs in metastasis. Seven hundred seventyone metastatic cancer patients and 552 non-metastatic cancer controls from 14 articles were involved in our meta-analysis. A sensitivity of 0.75 (95 % confidence interval (CI), 0.72– 0.79) and a specificity of 0.80 (95 % CI, 0.76–0.84) were observed from metastatic patients and non-metastatic controls in the combined analysis. And the AUC was 0.83 (95 % CI, Electronic supplementary material The online version of this article (doi:10.1007/s13277-014-2777-0) contains supplementary material, which is available to authorized users. C. Han : H. Yu : L. Zhang : Y. Feng : H. Xin (*) Department of Thoracic Surgery, China-Japan Union Hospital of Jilin University, No. 126, Xiantai Street, Changchun 130033, Jilin, China e-mail: [email protected] H. Xin e-mail: [email protected] X. Li Pharmaceutical Department, China-Japan Union Hospital of Jilin University, Changchun 130033, Jilin, China
0.79–0.86). In addition, results from subgroup analyses suggested that a higher diagnostic value for metastasis was acquired in tissue sample other than blood sample (sensitivity, 0.82 versus 0.73; specificity, 0.84 versus 0.79; PLR, 5.0 versus 3.5; NLR, 0.22 versus 0.34; DOR, 23 versus 10; AUC, 0.88 versus 0.80). In summary, this meta-analysis proved the relatively high diagnostic value of miRNA in metastasis, which might be applied as a novel screening tool to detect metastasis along with other biomarkers. We also illustrated that tissue-based miRNAs may have a better diagnostic accuracy than blood-based miRNAs. Keywords microRNA . Tissue . Blood . Metastasis . Cancer . Screening . Meta-analysis
Introduction Cancer has become a major health problem worldwide through years. It is reported that in the USA, one in four deaths is caused by cancer [1]. With the aged tendency of population, deteriorative environment, and unhealthy lifestyles, the cancer incidence and mortality has still been in a rising tendency. Although the prediction and treatment on cancer have achieved a great progress, which retards the mortality rate, the 5-year survival rate of some kinds of cancer can be relatively low. Some studies showed that lung cancer had the 5-year survival rate of 13 % [2], while hepatocellular carcinoma had a low rate of 5-year survival being approximately 5 % [3]. The survival rates of cancers are due to the following factors: diagnosis stages and metastasis. It was reported that gastric cancer had 5 % 5-year survival rate in the advanced stage (III and IV), but it can achieve nearly 90 % if patients are diagnosed in the early stage (I and II) [4]. Hence, it is an efficient method to reduce cancer mortality by detecting cancer in the early stage. Metastasis is another factors
Tumor Biol.
acting on cancer mortality. Metastasis is the spread cancer cells from one organ to another, which often occurs in malignant tumor tissues. The metastasis can be divided into four types based on routes: transcoelomic, lymphatic spread, hematogenous spread, and transplantation. Study showed that the 5-year survival rate of patients with lymph node metastasis declined drastically to 50–65 %, compared with the rate of approximately 80–95 % in patients with positive lymph nodes [5]. In another study, it showed that the 5-year survival rate of cervical cancer patients without lymph node metastasis can reach 100 % compared to that of 63.4 % in patients with lymph node metastasis at early stage (I/II) [6]. Therefore, discriminating the metastasis is a crucial factor for prognosis and improving the survival rate of cancer patients, and the application of a suitable therapeutic method is grounded in the identification of metastasis [7]. The traditional method to detect cancer metastasis is by imaging diagnosis. Doctors examine the samples taken from patients with microscope and other adjuvant techniques such as FISH. The normal existence of various type of cell can be tested from the tissue sample, the lack of which suggests the metastasis of tumor cells. Despite the high accuracy of this method, it hurts patients for its invasive nature. And it can be found only after a period of metastasis, which cannot provide efficient information for treating cancer in advance. The high expense and complex procedures are also obstacles for the clinical application of imaging diagnosis. However, there are other special ways to detect the corresponding cancer metastasis. For example, a subfraction of TA-4 and a tumorassociated antigen, called squamous cell carcinoma antigen, is widely used as the marker for squamous cell carcinoma. However, because of the non-organ specificity for cervix, the normal initial squamous cell carcinoma antigen is unable to exclude the lymph node metastasis. Hence, a more sensitive, non-invasive, and efficient strategy for detecting metastasis is needed. MicroRNAs, which are found to participate in tumorgenesis, provide a new road for detecting metastasis. MicroRNAs (miRNAs) are small non-protein-coding RNA molecules with a length of 18 to 24 nucleotides. A single miRNA can regulate the expression of hundreds of downstream genes, which has a broad effect on the regulation of the whole cell cycle. It is revealed that miRNAs are involved in cellular proliferation, apoptosis, differentiation, metabolism, and other cellular processes [8]. MiRNAs play important roles in transcription or posttranscription level by specific pairing with the 3′ UTR of the target messenger RNA (mRNA), leading mRNA activation inhibition or degradation of target mRNA. Accumulating studies have found that the aberrant expression of miRNAs is associated with cancer genesis. MiRNAs are found to be stable and abundant in body liquid, such as serum, plasma, and sputum, indicating it can be served as biomarker for cancer diagnosis. As a novel, cost-efficient, and convenient biomarker, miRNAs are widely used in detecting metastasis of cancers. However, there are inconsistent results revealed from
studies, leading us to conduct this meta-analysis. The aim of this study is to evaluate the diagnostic value of miRNAs in detecting cancer metastasis.
Methods Search strategy We conducted a literature search to identify relevant studies on miRNAs detecting metastasis from databases including Medline, PubMed, Embase, and CNKI. The articles were searched until July 2014, with no language restriction. The following keywords were used: (“neoplasm metastasis”[Mesh] OR neoplasm metastases OR metastasis OR metastases OR lymphatic metastasis OR lymphatic metastases OR lymph node metastasis) AND (“microRNAs”[MeSH Terms] OR “microRNAs” OR “miRNA”) AND (“ROC curve”[Mesh] OR “diagnosis”[Mesh] OR “sensitivity and specificity”[Mesh]). More studies were added by reviewing the reference lists of articles. This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidance (Supplement S1). Inclusion and exclusion criteria The eligible studies have to fulfill the following criteria: (1) studies related to miRNAs in detecting cancer metastasis, (2) histopathology used for detection of cancer, (3) studies providing necessary data for evaluating diagnostic value of miRNAs for detecting metastasis, and (4) studies with sufficient validity. The publications were removed based on the following criteria: (1) duplicate publications, (2) studies with insufficient data, and (3) studies in the form of letters, reviews, editorials, meeting abstracts, and case reports. Data extraction Two reviewers extracted data independently from the eligible studies. We collected relevant information including country, sample size, cancer site, metastasis site, the types of miRNAs, and specimen. Statistical analysis The STATA 12.0 (Stata Corp, College Station, TX, USA) software was performed to analyze the extracted data, with two-sided P value. We calculated the pooled parameters for assessing the diagnostic value, including sensitivity, specificity, positive and negative likelihood ratio (PLR, NLR), and diagnostic odds ratio (DOR). The PLR and NLR present the clinical utility, and the DOR indicates the accuracy with the
Tumor Biol.
combination of sensitivity and specificity. We also constructed the summary receiver operator characteristic (SROC) curve (based on the sensitivity and specificity of each study) and calculated the AUC (area under the SROC curve) to explore the diagnostic accuracy of miRNAs in metastasis. Considering studies involving in more than one miRNA, the studied miRNAs were integrated into a comprehensive factor by regression analysis. The I2 test was conducted; with I2 > 50 % indicates the existence of heterogeneity. We used subgroup analyses to find the potential sources of heterogeneity. The influence of single on overall result was analyzed by goodness of fit, bivariate normality, Cook’s distance, and outlier detection. The publication bias was tested by Deek’s funnel plot asymmetry test.
Databases Search Medline, PubMed, Embase, CNKI, etc.
(N = 1013) Duplicate Publications (N = 117) Abstract and Keyword Review Two reviewers review abstract and keyword
(N = 896) Unrelated Studies (N = 853) Full-text and Data Accessment Two reviewers assess full-tex and data
(N = 43) Data Unavailable
Results Data selection and study characteristics A flow diagram (Fig. 1) illustrated the selection process and the reasons for exclusion. Of all the 1013 retrieved articles, we selected 14 distinguished articles according to our inclusion and exclusion criteria [9–22]. In the 14 available articles, 34 studies were conducted, with a total of 771 metastatic cancer patients and 552 non-metastatic cancer controls enrolled in our meta-analysis. Characteristics of all identified studies were listed in Table 1 in the order of publication year. All the articles were published between 2011 and 2014. Asian population is the focus of all studies. Specimens were from tissue (n=10), serum (n=17), or plasma (n=7). Specimens from serum and plasma were combined as blood samples (n=24). Pooled diagnostic accuracy of miRNA profiles in cancer metastasis The diagnostic performance of miRNAs in differentiating cancer with metastasis from those without metastasis was presented in Table 2. In the analysis for combined specimens, the I 2 value of sensitivity was 40.12 %; no statistically significant heterogeneity was identified between studies corresponding to sensitivity. The fixed effect model was applied with no heterogeneity observed, whereas the I 2 value of specificity was 62.70 %, which indicated the existence of significant heterogeneity, and the random effect model was adopted. Similar results were also acquired from stratified analysis by tissue and blood sample. Based on the models mentioned above, the evaluation for the sensitivity and specificity of miRNAs in diagnosing metastasis was illustrated in Fig. 2. A sensitivity of 0.75 (95 % confidence interval (CI), 0.72–0.79) and a specificity of 0.80
(N = 29) Final Inclusion of Articles (N = 14)
Fig. 1 A flow chart of study identification, inclusion, and exclusion
(95 % CI, 0.76–0.84) were observed from metastatic patients and non-metastatic controls in the combined analysis. And the AUC was 0.83 (95 % CI, 0.79–0.86), corresponding SROC curve was constructed and presented in Fig. 3. A relatively high diagnostic accuracy was thus acquired from the result. Subgroup and meta-regression analyses Subgroup analyses were conducted by a type of specimen which have been mentioned to be related to the accuracy of diagnosis by previous research [10]. The results of subgroups were also presented in Table 2. A higher diagnostic value for metastasis was observed in tissue sample other than blood sample (sensitivity, 0.82 versus 0.73; specificity, 0.84 versus 0.79; PLR, 5.0 versus 3.5; NLR, 0.22 versus 0.34; DOR, 23 versus 10; AUC, 0.88 versus 0.80), indicating that cancer tissue might be a better sample for miRNA assays in metastasis diagnosis. The SROC curve was illustrated in Fig. 4. Sensitivity analyses and publication bias The influence of a single study on the overall results was assessed by influence analysis. Goodness of fit, bivariate normality, Cook’s distance, and outlier detection were applied. From the illustrated results, a study from Chen et al. [9] was
Tumor Biol. Table 1 Characteristics of all identified studies
Included study
Wu et al. [19] Chen et al. [10] Huang et al. [15] Wang and Gu [18] Zhao et al. [21] Chen et al. [9]
Chen et al. [12] Chen et al. [13] Kim et al. [16] Zhao et al. [22] Feng et al. [14] Chen et al. [11] Lu et al. [17]
Yin et al. [20]
Country
Sample size
Cancer site
Metastasis site
MicroRNA
Specimen
Lymph node Lymph node
miR-212, −195
Tissue
miR-21
Serum
Tissue
Serum
M (+)
M (−)
China
22
22
Gastric
China
40
25
Cervical
China
13
31
Cervical
Lymph node
China
38
36
Colorectal
Liver
miR-100, −125b, −143, −145, −199a-5p, let7c miR-29a
China
100
22
Breast
Bone
miR-10b
Serum
China
40
40
Cervical
Lymph node
Tissue
40
40
Cervical
Lymph node
China
25
35
Breast
Lymph node
miR-1246, −20a, −2392, −3147, −3162-5p, −4484 miR-1246, −20a, −2392, −3147, −3162-5p, −4484 miR-10b, −373,
Plasma
China
14
37
Cervical
Lymph node
miR-143
Tissue
Korea
16
15
Gastric
Lymph node
Serum
China
40
40
Cervical
Lymph node
miR-21, −27a, −106b, −146a, −148a, −223, −433 miR-20a, −203
China
48
50
Colorectal
Liver
miR-29a
Serum
China
36
36
Gastric
Various
miR-122, −192
Plasma
China
120
30
Nasopharyngeal
miR-9
Plasma
63
21
Nasopharyngeal
miR-9
Plasma
116
72
Colorectal
Lymph node Lymph node Liver
miR-126, −141, −21
Serum
China
identified as an outlier. Yet, a similar accuracy for diagnosis was acquired with the outlier removed from the analyses. The result of influence test suggested that our meta-analysis was robust. The publication bias of eligible articles was assessed by Deek’s funnel plot and linear regression test. A P value of 0.17 was appraised. It was suggested that no statistically significant publication bias was observed across enrolled studies.
Serum
Serum
Discussion In this meta-analysis, we investigate the diagnostic value of miRNAs in metastasis diagnosis. Seven hundred seventy-one metastatic patients and 552 non-metastatic controls from 34 studies were enrolled in our analysis. We observed a relatively high diagnostic value for majority of miRNAs, which indicated the validity of miRNAs as biomarkers.
Tumor Biol. 1.0
Table 2 Diagnostic performance of miRNAs in differentiating cancer with metastasis from those without metastasis Analysis
Combine
Tissue
12 9 18 1
Blood
28
20 13 21
5 7 19
2 23 2 27
6
34 0.75 [0.72–0.79] 40.12 (P=0.01) 0.80 [0.76–0.84]
10 0.82 [0.72–0.88] 42.27 (P=0.06) 0.84 [0.70–0.92]
24 0.73 [0.70–0.76] 26.46 (P=0.012) 0.79 [0.74–0.83]
I2 (%) PLR (95 % CI) NLR (95 % CI) DOR (95 % CI) AUC (95 % CI)
62.70 (P
