http://orcid.org/0000-0001-6328-6526
Figures
Abstract
Midkine (MK) has been reported as the potential novel diagnostic biomarker for cancer in several studies, but their results were controversial. Therefore, we performed a diagnostic meta-analysis to assess the diagnostic value of serum MK in cancer patients. A systematic electronic and manual search was performed for relevant literatures through several databases up to June 1, 2017. The quality of the studies included in the meta-analysis was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. All analyses were conducted using stata12.0 software. Ten studies collectively included 1119 cancer patients and 1441 controls met the eligible criteria. The summary estimates were: sensitivity 0.78 (95% CI = 0.68–0.85), specificity 0.83 (95% CI = 0.72–0.90), positive likelihood ratio 4,54 (95% CI = 2.64–7.80), negative likelihood 0.27 (95% CI = 0.18–0.40), diagnostic odds ratio 16.79 (95% CI = 7.17–39.33), and area under the curve 0.87 (95% CI = 0.84–0.89). Publication bias was suggested by Deeks’ funnel plot asymmetry test (P = 0.92). According to our results, serum MK has greater diagnostic value in diagnosing cancer, however, more reliable studies in larger cohort should be conducted to evaluate the diagnostic accuracy of serum MK.
Citation: Jing X, Cui X, Liang H, Hao C, Han C (2017) Diagnostic accuracy of ELISA for detecting serum Midkine in cancer patients. PLoS ONE 12(7): e0180511. https://doi.org/10.1371/journal.pone.0180511
Editor: Hiroyoshi Ariga, Hokkaido Daigaku, JAPAN
Received: April 21, 2017; Accepted: June 18, 2017; Published: July 7, 2017
Copyright: © 2017 Jing et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This work was supported by the National Natural Science Foundation of China (81670270) and Scientific Research Project of Shanxi Provincial Department of Health (201601070).
Competing interests: The authors have declared that no competing interests exist.
Introduction
Cancer is among the leading cause of morbidity and mortality across the globe, accounting for 17.5 million cancer cases worldwide and 8.7 million deaths in 2015, making it a global health problem with a heavy burden on the society[1–3]. The increasing morbidity and mortality from cancer might attribute to earlier diagnosis and treatment[4, 5]. Most of the cancer patients could have a high chance of cure if diagnosed early and treated adequately. Growing evidences suggested that the five-year survival rate could be improved significantly with diagnosis in the early stage of cancer such as gastric cancer, lung cancer, and renal cancer[6–8]. Unfortunately, the majority of cancer patients are diagnosed in advanced stages where early metastasis to lymph nodes, adjacent tissue or organs occurs, following with a reduced overall five-year survival rate. Therefore, it is urgent to find accurate and efficient methods for cancer diagnosis at its early stage to improve the survival status for cancer patients and simultaneously reduce the cancer burden.
Biopsy is currently considered the golden standard for diagnosing solid tumor, often combined with ultrasound and imageological examination[9]. However, the invasive nature of biopsy and low sensitivity of ultrasound and imageological examination restrict their application in the early diagnosis of tumor[10, 11]. Therefore, exploring more reliable non-invasive detection of new or recurrent tumor is the need of the hour.
Midkine (MK), a heparin-binding growth factor, was a highly expressed factor during the early differentiation stage in embryonal carcinoma and was weak or undetectable in normal adult tissues[12–14]. Mounting evidence has indicated that MK plays an essential role in various biological activities in malignant tumors such as proliferation, migration, antiapoptosis, angiogenesis, invasion, and metastasis[15]. Increased MK mRNA and protein expressions have been reported in many early stage human cancers tissue, such as breast cancer, gastrointestinal cancer, and lung cancer[16–19]. Due to the secretory protein properties, the detection of MK in serum or plasma could serve as a diagnostic marker and allow the early detection of malignancy. Nowadays, the diagnostic value of MK for the early detection of cancer has been successfully detected in various tumors, such as duodenal, colon, pancreatic, lung, esophageal, tumors and hepatocellular carcinoma[19–21].
Despite many studies have demonstrated the potential of serum or plasma MK as a novel marker for tumor diagnosis, the previous studies have been limited by relatively small sample size recruited in the individual studies, and still no comprehensive conclusion have been reached for the diagnostic value of MK in detection cancers. Therefore, the aim of the present study was to review and assess the overall diagnostic test accuracy of serum MK for cancer diagnosis.
Results
Included studies
A detailed flowchart of our literature research is presented in Fig 1. The initial search returned a total of 361 articles involving MK and cancer, and 26 of them were excluded due to duplications. The remaining 335 articles were subjected to the next stage of evaluation. Among the remaining 312 excluded articles, 178 were reviews, letters, meta-analyses or not related to the research topic, 36 were not from serum, 87 were not related to cancer diagnose and 11 articles did not contain sufficient and useful data. Finally, 11 articles were included in this meta-analysis[21–30].
Study characteristics and quality assessment
The main characteristics of the included studies were summarized and listed in Table 1 by order of publication year, ranging from 2003 to 2017. A total of 2826 subjects (1128 Cancer patients and 1574 controls) from 11 studies were included in the meta-analysis. Among the 11 studies, 4 were conducted in China, 2 in Japan, 1 in Germany, 1 in Poland, 1in Egypt, 1 in Australia, 1 in Turkey. In addition, different types of cancer were recorded including esophageal squamous cell carcinoma (n = 1), pediatric embryonal tumor (n = 1), colorectal cancer (n = 1), hepatocellular carcinoma (n = 3), thyroid cancer (n = 1), non-small cell lung cancer (n = 2), mesothelioma (n = 1), head and neck squamous cell carcinoma (n = 1). All studies used the ELISA to measure the levels MK in serum or plasma. The QUADAS-2 scores of the included studies were shown in Table 1. Overall, the majority of studies presented a moderate-high quality.
Diagnostic accuracy and threshold analysis
In this meta-analysis, the threshold effect, an inverse correlation between the sensitivity and specificity, was conducted to explore whether it was existed in the study, which was shown in an ROC plane (Fig 2). The shape of ROC curve showed a nontypical shoulder-arm appearance and the spearmen correlation coefficient value of sensitivity and 1-specificity was -0.1455 with a P value was 0.6696 suggesting that there was no heterogeneity from the threshold effect. The forest plot of sensitivity and specificity for MK assays was shown in Fig 3. Heterogeneity in sensitivity and specificity was detected in 11 studies (I2 = 88.20 and 93.01, respectively), suggesting that significant heterogeneity was existed in sensitivity and specificity from non-threshold effect. Therefore, the random effects model was employed in this study. The pooled sensitivity, specificity, PLR, NLR, and DOR with their 95% confidence intervals (CIs) were 0.78 (95% CI: 0.68–0.85, Fig 3A), 0.83 (95% CI: 0.72–0.90, Fig 3B), 4.54 (95% CI: 2.64–7.80 Fig 3C), 0.27 (95% CI: 0.18–0.40, Fig 3D), and 16.79 (95% CI: 7.17–39.33, Fig 4), respectively. The SROC curve for the 11 included studies is shown in Fig 5. The AUC of MK tests was 0.87 (95% CI = 0.84–0.89), thereby implying a relatively high diagnostic accuracy.
Forest plot of sensitivity(A), specificity(B), positive likelihood(C),negative likelihood(D) with their 95%CIs.
Meta-regression and sensitivity analyses
Meta-regression analysis was performed to explore the potential source of heterogeneity in sensitivity and specificity base on the covariates including ethnicity, simple size, number of case and control, year of publication. The results in Fig 6 showed that sample size (P<0.01) had an effect on sensitivity, while the ethnicity (P<0.01) contributed to interstudy heterogeneity for specificity.
Additionally, sensitivity analyses was conducted and 2 outlier was found by influence analysis (Fig 7C) and outlier detection (Fig 7D). After the one outlier excluded from the test, sensitivity (from 0.78 to 0.74) and specificity (from 0.83 to 0.81), AUC (from 0.87 to 0.83) showed minimal changes, indicating the outlier did not have a substantial effect on the overall analysis. Combined with goodness of fit (Fig 7A) and bivariate normality analyses (Fig 7B), we confirmed that the random-effect bivariate model was moderately robust for this meta-analysis.
Graphical depiction of residual-based goodness-of-fit(A), bivariate normality(B), influence(C) and outlier detection(D).
Publication bias
Fagan’s Nomogram for likelihood ratio in Fig 8 showed that when a pre-test probability of 20% was specified, the post-test probability positivity would raise to 53% with a PLR of 5, and the post-test probability negativity would decreased to 6% with a NLR was 0.27. These results suggest a moderate value for MK in the diagnosis of cancer in serum. In addition, we conducted Deeks’ test to assess the publication bias and found no significant publication bias in our study with P value 0.92 (Fig 9).
Discussion
Cancer continues to be a major public health challenge and an important cause of mortality and morbidity worldwide[31]. Early diagnosis and treatment are the key point for the clinical course and the outcome patients with cancer[32, 33]. Cancer-specific variations in secreted proteins in the blood are promising noninvasive biomarkers for identifying and monitoring patients with cancer. MK, as one of the most extensively studied aberrant secreted proteins, was discovered to be increased in the early stages of numerous tumors[16]. In 1999, Konishi N et al revealed that MK is highly expressed in precancerous lesions and the early stage of prostate carcinoma development[34]. Furthermore, abnormal high expression of MK could promote cell proliferation, angiogenesis, invasion, migration and metastasis[35]. Meanwhile, as a feasible and non-invasive alternative, MK was promising to be a biomarker for cancer. However, the diagnosis accuracy of serum MK has been reported to be variable. Hence, we conducted a comprehensive meta-analysis to evaluate the diagnostic accuracy of MK for cancers.
In the present meta-analysis, we performed a diagnostic meta-analysis to assess the accuracy of serum MK as a biomarker for cancer. The pooled sensitivity and specificity of serum MK were 0.78 (95% CI = 0.68–0.85) and 0.83 (95% CI = 0.72–0.90) respectively, indicating that 78% of cancer patients had high serum MK levels and that 83% of the non-cancer patients had low serum MK levels. The pooled DOR of 16.79 (95% CI = 7.17–39.33) suggested that the overall accuracy of serum MK for the diagnosis of cancer is credible. The likelihood ratio, including PLR and NLR, could also reflect the diagnostic accuracy. When PLR>10 or NLR<0.1, the likelihood of diagnosis or exclusion of a disease increased remarkably. Nevertheless, in our meta-analysis, a pooled PLR of 4.54 (95% CI = 2.64–7.80) and NLR 0.27 (95% CI = 0.18–0.40) suggested that MK may not be powerful enough to confirm or exclude the potential patient with cancers. However, an AUC of 0.87 (95% CI = 0.84–0.89) means a high ability for cancer detection. Therefore, serum MK was an effective biomarker for cancer diagnosis.
Heterogeneity is a potential problem that can influence the incorporation effect and the interpretation of the results[36]. The shape of ROC curve and the spearmen analysis suggested that heterogeneity could not be explained by a threshold effect. Meta-regression analysis was performed to explore the potential source of heterogeneity in sensitivity and specificity base on the covariates including ethnicity, simple size, number of case and control, year of publication. The results indicated that sample size had an effect on sensitivity, while the ethnicity contributed to interstudy heterogeneity for specificity. Due to there were insufficient eligible studies to fully elucide the source of the heterogeneity, the possible reason for the heterogeneity need to be investigated in future studies.
The current study has several limitations. First, despite extensive literature search were performed, the limited number of included studies and cancer types may restrict our ability to evaluate the accuracy of serum MK. Second, we could not determine the ideal cut-off value for serum MK test, due to inconsistent cut-off values were adopted in each study. Third, the majority of these studies are retrospective analyses on historical cohorts, which may limit the conclusion due to selection bias. Fourth, only articles published in English or Chinese were enrolled in our meta-analysis, and studies published in other languages were not included, which may cause inevitable bias.
Despite these limitations, the present evidence indicates that serum MK is potential to be a diagnosis biomarker for cancer, because this non-invasive method has good overall diagnostic performance. However, large-scale and comprehensive studies must be performed in the future to validate this finding.
Materials and methods
Search strategy and selection criteria
This meta-analysis was performed in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines[37] Studies were systematically searched from PubMed, Google Scholar, EMBASE, Chinese National Knowledge Infrastructure (CNKI), Wan Fang Database up to June 1, 2017 with the following MeSH and key words:(”midkine” or “MK” or “MDK”), and “serum” and (“cancer” or “carcinoma” or “tumor neoplasm”) and (“diagnosis” or “ROC curve”) without language restriction or publication date restrictions. In addition, references of all articles in these eligible studies were also read to identify additional relevant literature.
All studies should meet the following criteria: (1) studies appraising MK for cancer diagnosis; (2) the levels of MK in serum or plasma were determined; (3) sufficient information were provided or can be used to calculate the sensitivity and specificity (including true positive-TP, false positive-FP, false negative-FN and true negative-TN). While the exclusion criteria were list as follows: (1) duplications or overlapping studies; (2) studies without adequate data to construct the 2×2 table; (3) reviews, letters, case report and conference abstracts.
Data extraction and quality assessment
Data extraction from eligible studies was performed independently by 2 investigators, and disagreements were resolved by deliberation with a third investigator. From each study, the following study characteristics were extracted: (1) basic information of articles (the first author, year of publication, country of publication); (2) research object’s general features (ethnicity, sample size, mean age of subjects, type of cancer); (3) data used for our final meta-analysis (detection method, cut-off value, sensitivity, specificity, TP, FP, FN, and TN); (4) other research features (sample source, study design, etc).
The Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) instrument, a quality assessment tool, was used to assess the methodological quality of individual studies by the same two independent investigators[38]. The QUADAS-2 tool consisted of four domains: Patient selection, index test, reference standard, and flow and timing. Each of the assessment has seven questions, which should be answered with “yes”, “no”, or “unclear”. An answer of “yes” gets 1 score, while others get 0, the highest score is seven[39].
Statistical analysis
All the analyses of diagnostic accuracy were performed using STATA 12.0 software. The bivariate meta-analysis model was used to calculate the pooled results of sensitivity, specificity, positive and negative likelihood ratios (PLR and NLR), diagnostic odds ratio (DOR) with their corresponding 95% confidence intervals (CIs)[40, 41]. The summary receiver operator characteristic (SROC) curve and the area under the SROC curve (AUC) were generated to evaluate the accuracy of cancer diagnosis. An AUC of 1.0 indicates perfect diagnostic accuracy whereas an AUC of 0.5 indicates poor diagnostic accuracy[42]. Additionally, the Spearman correlation coefficient was used to evaluate the diagnostic threshold effects, and P<0.05 indicates significant heterogeneity. The statistical heterogeneity from non-threshold was assessed by the Q value and I2 statistic. A probability value of P<0.05 and I2≥50% indicated the existence of significant heterogeneity among individual studies[43]. Furthermore, meta-regression and subgroup analyses were conducted to explore the potential sources of heterogeneity. The potential publication bias was estimated by Deeks’ funnel plot and a probability value of P<0.05 indicates significant publication bias[41]. Finally, the Fagan’s nomogram was graphed to estimate the post-test probabilities.
Supporting information
S1 Table. PRISMA checklist.
PRISMA checklist identifying how and where each element of the PRISMA process has been addressed in this paper.
https://doi.org/10.1371/journal.pone.0180511.s001
(PDF)
References
- 1. Global Burden of Disease Cancer C, Fitzmaurice C, Allen C, Barber RM, Barregard L, Bhutta ZA, et al. Global, Regional, and National Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disability, and Disability-Adjusted Life-years for 32 Cancer Groups, 1990 to 2015: A Systematic Analysis for the Global Burden of Disease Study. JAMA oncology. 2016. pmid:27918777.
- 2. Navarro SA, Carrillo E, Grinan-Lison C, Martin A, Peran M, Marchal JA, et al. Cancer suicide gene therapy: a patent review. Expert opinion on therapeutic patents. 2016;26(9):1095–104. pmid:27424657.
- 3. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA: a cancer journal for clinicians. 2013;63(1):11–30. pmid:23335087.
- 4. Bartels CL, Tsongalis GJ. MicroRNAs: novel biomarkers for human cancer. Clinical chemistry. 2009;55(4):623–31. pmid:19246618.
- 5. Shen J, Stass SA, Jiang F. MicroRNAs as potential biomarkers in human solid tumors. Cancer letters. 2013;329(2):125–36. pmid:23196059; PubMed Central PMCID: PMC3552101.
- 6. Wang J, Yu JC, Kang WM, Ma ZQ. Treatment strategy for early gastric cancer. Surgical oncology. 2012;21(2):119–23. pmid:21256735.
- 7. Motzer RJ, Mazumdar M, Bacik J, Berg W, Amsterdam A, Ferrara J. Survival and prognostic stratification of 670 patients with advanced renal cell carcinoma. Journal of clinical oncology: official journal of the American Society of Clinical Oncology. 1999;17(8):2530–40. pmid:10561319.
- 8. Iwamoto H, Kanda Y, Sejima T, Osaki M, Okada F, Takenaka A. Serum miR-210 as a potential biomarker of early clear cell renal cell carcinoma. International journal of oncology. 2014;44(1):53–8. pmid:24212760.
- 9. Li H, Zhang Q, Chen L, Min L, Wang X, Liu F, et al. [Role of diagnostic laparoscopy in the treatment plan of gastric cancer]. Zhonghua wei chang wai ke za zhi = Chinese journal of gastrointestinal surgery. 2017;20(2):195–9. pmid:28226355.
- 10. Shin JH. Ultrasonographic imaging of papillary thyroid carcinoma variants. Ultrasonography. 2017. pmid:28222584.
- 11. Tanaka IB 3rd, Komura J, Tanaka S. Pathology of Serially Sacrificed Female B6C3F1 Mice Chronically Exposed to Very Low-Dose-Rate Gamma Rays. Radiation research. 2017. pmid:28218887.
- 12. Kadomatsu K, Tomomura M, Muramatsu T. cDNA cloning and sequencing of a new gene intensely expressed in early differentiation stages of embryonal carcinoma cells and in mid-gestation period of mouse embryogenesis. Biochemical and biophysical research communications. 1988;151(3):1312–8. pmid:3355557.
- 13. Tomomura M, Kadomatsu K, Nakamoto M, Muramatsu H, Kondoh H, Imagawa K, et al. A retinoic acid responsive gene, MK, produces a secreted protein with heparin binding activity. Biochemical and biophysical research communications. 1990;171(2):603–9. pmid:2403350.
- 14. Muramatsu T, Kadomatsu K. Midkine: an emerging target of drug development for treatment of multiple diseases. British journal of pharmacology. 2014;171(4):811–3. pmid:24460672; PubMed Central PMCID: PMC3925019.
- 15. Muramatsu T. Midkine, a heparin-binding cytokine with multiple roles in development, repair and diseases. Proceedings of the Japan Academy Series B, Physical and biological sciences. 2010;86(4):410–25. PubMed Central PMCID: PMC3417803. pmid:20431264
- 16. Jones DR. Measuring midkine: the utility of midkine as a biomarker in cancer and other diseases. British journal of pharmacology. 2014;171(12):2925–39. 10.1111/bph.12601. PubMed Central PMCID: PMC4055197. pmid:24460734
- 17. Garver RI Jr., Radford DM, Donis-Keller H, Wick MR, Milner PG. Midkine and pleiotrophin expression in normal and malignant breast tissue. Cancer. 1994;74(5):1584–90. pmid:7520350.
- 18. Aridome K, Tsutsui J, Takao S, Kadomatsu K, Ozawa M, Aikou T, et al. Increased midkine gene expression in human gastrointestinal cancers. Japanese journal of cancer research: Gann. 1995;86(7):655–61. pmid:7559083.
- 19. Garver RI Jr., Chan CS, Milner PG. Reciprocal expression of pleiotrophin and midkine in normal versus malignant lung tissues. American journal of respiratory cell and molecular biology. 1993;9(5):463–6. pmid:8217186.
- 20. Ikematsu S, Yano A, Aridome K, Kikuchi M, Kumai H, Nagano H, et al. Serum midkine levels are increased in patients with various types of carcinomas. British journal of cancer. 2000;83(6):701–6. pmid:10952771; PubMed Central PMCID: PMC2363529.
- 21. Zhu WW, Guo JJ, Guo L, Jia HL, Zhu M, Zhang JB, et al. Evaluation of midkine as a diagnostic serum biomarker in hepatocellular carcinoma. Clinical cancer research: an official journal of the American Association for Cancer Research. 2013;19(14):3944–54. pmid:23719264
- 22. Shimada H, Nabeya Y, Tagawa M, Okazumi S, Matsubara H, Kadomatsu K, et al. Preoperative serum midkine concentration is a prognostic marker for esophageal squamous cell carcinoma. Cancer science. 2003;94(7):628–32. pmid:12841873.
- 23. Lucas S, Reindl T, Henze G, Kurtz A, Sakuma S, Driever PH. Increased midkine serum levels in pediatric embryonal tumor patients. Journal of pediatric hematology/oncology. 2009;31(10):713–7. pmid:19727009.
- 24. Krzystek-Korpacka M, Diakowska D, Neubauer K, Gamian A. Circulating midkine in malignant and non-malignant colorectal diseases. Cytokine. 2013;64(1):158–64. pmid:23899719.
- 25. Meng Z, Tan J, Zhang G, Tian W, Fu Q, Li W, et al. Evaluation of serum midkine as a biomarker in differentiated thyroid cancer. Life sciences. 2015;130:18–24. pmid:25817231.
- 26. Shaheen KY, Abdel-Mageed AI, Safwat E, AlBreedy AM. The value of serum midkine level in diagnosis of hepatocellular carcinoma. International journal of hepatology. 2015;2015:146389. pmid:25737783; PubMed Central PMCID: PMC4337261.
- 27. Vongsuvanh R, van der Poorten D, Iseli T, Strasser SI, McCaughan GW, George J. Midkine Increases Diagnostic Yield in AFP Negative and NASH-Related Hepatocellular Carcinoma. PloS one. 2016;11(5):e0155800. pmid:27219517; PubMed Central PMCID: PMC4878793.
- 28. Xia X, Lu JJ, Zhang SS, Su CH, Luo HH. Midkine is a serum and urinary biomarker for the detection and prognosis of non-small cell lung cancer. Oncotarget. 2016;7(52):87462–72. pmid:27974680.
- 29. Yamashita T, Shimada H, Tanaka S, Araki K, Tomifuji M, Mizokami D, et al. Serum midkine as a biomarker for malignancy, prognosis, and chemosensitivity in head and neck squamous cell carcinoma. Cancer medicine. 2016;5(3):415–25. pmid:26798989; PubMed Central PMCID: PMC4799940.
- 30. Sun C, Wang L, Liu M, Zhao W, Piao Z, ZHANG ML. The diagnosis value of non-small cell lung cancer ofSerum medium-term factors combined carcinoembryonic antigen Guangdong Medical Journal. 2015;(05):768–9.
- 31.
Bray F, Soerjomataram I. The Changing Global Burden of Cancer: Transitions in Human Development and Implications for Cancer Prevention and Control. In: Gelband H, Jha P, Sankaranarayanan R, Horton S, editors. Cancer: Disease Control Priorities, Third Edition (Volume 3). Washington (DC)2015.
- 32. Ma R, Xu H, Wu J, Sharma A, Bai S, Dun B, et al. Identification of serum proteins and multivariate models for diagnosis and therapeutic monitoring of lung cancer. Oncotarget. 2017. pmid:28121629.
- 33. Cui X, Jing X, Long C, Tian J, Zhu J. Long noncoding RNA MEG3, a potential novel biomarker to predict the clinical outcome of cancer patients: a meta-analysis. Oncotarget. 2017. pmid:28157702.
- 34. Konishi N, Nakamura M, Nakaoka S, Hiasa Y, Cho M, Uemura H, et al. Immunohistochemical analysis of midkine expression in human prostate carcinoma. Oncology. 1999;57(3):253–7. pmid:10545795.
- 35. Rawnaq T, Dietrich L, Wolters-Eisfeld G, Uzunoglu FG, Vashist YK, Bachmann K, et al. The multifunctional growth factor midkine promotes proliferation and migration in pancreatic cancer. Molecular cancer research: MCR. 2014;12(5):670–80. pmid:24567526.
- 36. Coory MD. Comment on: Heterogeneity in meta-analysis should be expected and appropriately quantified. International journal of epidemiology. 2010;39(3):932; author reply 3. pmid:19349478.
- 37. Moher D, Liberati A, Tetzlaff J, Altman DG, Group P. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. International journal of surgery. 2010;8(5):336–41. pmid:20171303.
- 38. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Annals of internal medicine. 2011;155(8):529–36. pmid:22007046.
- 39. Zhang Y, Zhong Q, Chen X, Fang J, Huang Z. Diagnostic value of microRNAs in discriminating malignant thyroid nodules from benign ones on fine-needle aspiration samples. Tumour biology: the journal of the International Society for Oncodevelopmental Biology and Medicine. 2014;35(9):9343–53. pmid:24943685.
- 40. Wan C, Shen Y, Yang T, Wang T, Chen L, Wen F. Diagnostic value of microRNA for pancreatic cancer: a meta-analysis. Arch Med Sci. 2012;8(5):749–55. pmid:23185182; PubMed Central PMCID: PMC3506232.
- 41. 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. Journal of clinical epidemiology. 2005;58(9):882–93. pmid:16085191.
- 42. Harbord RM, Deeks JJ, Egger M, Whiting P, Sterne JA. A unification of models for meta-analysis of diagnostic accuracy studies. Biostatistics. 2007;8(2):239–51. pmid:16698768.
- 43. Dinnes J, Deeks J, Kirby J, Roderick P. A methodological review of how heterogeneity has been examined in systematic reviews of diagnostic test accuracy. Health technology assessment. 2005;9(12):1–113, iii. pmid:15774235.