The diagnostic role of DNA methylation in sporadic endometrial cancer: a systematic review and meta-analysis

Background Although increasing numbers of methylated genes have been identified as biomarkers for endometrial cancer, the results have been inconsistent. We therefore carried out a systematic review and meta-analysis to evaluate the diagnostic accuracy of methylated genes as markers for sporadic endometrial cancer. Results A total of 22 studies including 1930 participants (sporadic endometrial cancer patients and normal individuals) met our eligibility criteria. The pooled sensitivity and specificity were 0.93 (95% confidence interval: 0.91−0.94) and 0.48 (95% confidence interval: 0.46–0.50), respectively. The area under the summary receiver operating characteristic curve was 0.8834. The presence of DNA methylation was significantly associated with lymph node metastasis of endometrial cancer (pooled odds ratio: 0.28, 95% confidence interval: 0.15–0.52, p < 0.001). Materials and Methods We searched the relevant literature systematically using the PubMed and Web of Science databases up to April 2017. Diagnostic accuracy variables were pooled and analyzed using Meta-DiSc software. Sensitivity analysis and publication bias were evaluated using Review Manager. Conclusions This meta-analysis suggests that the detection of DNA methylation is associated with lymph node metastasis, with high sensitivity but relatively low specificity for the diagnosis of sporadic endometrial cancer.


INTRODUCTION
Endometrial cancer (EC) is one of the three main tumors originating in the female genital system.Its incidence is higher than that of cervical cancer in many countries, and is located in the top of gynecologic malignant tumors [1,2].EC patients often present with abnormal vaginal bleeding; although many patients are diagnosed when the disease is still confined to the uterus, about 30% are diagnosed with EC in its later stage.Most clinical trials for the treatment of advanced and recurrent EC have shown limited benefits, and the mortality rate has increased dramatically over the past few years [3,4].
Abnormal genetic and epigenetic alterations have been widely recognized as being associated with EC [5,6].However, genetic markers have not yet proven reliable for identifying the entire spectrum of the disease, especially sporadic EC.Aberrant DNA methylation is one of the most widely studied epigenetic modifications with a critical role in EC [7,8].Diagnostic markers based on gene methylation have recently been developed and have shown considerable promise for the detection of EC, while aberrant promoter methylation has been found Meta-Analysis www.impactjournals.com/oncotarget to be an early and widespread alteration in endometrial tumorigenesis [9].However, although specific gene methylation patterns have been widely used for the diagnosis of many different cancers, including EC, there is currently a lack of effective diagnostic biomarkers for EC, and novel, accurate markers are urgently needed.
Previous studies have investigated the use of aberrant gene methylation in tissue samples as potential diagnostic biomarkers for EC, with encouraging but variable results.Further studies are therefore needed to describe the associations between DNA methylation in different tumor suppressor genes and the clinicopathologic features of sporadic EC.Moreover, no systematic review or metaanalysis has yet been conducted to evaluate the diagnostic accuracy of the existing studies.We therefore conducted a comprehensive, systematic review and meta-analysis of eligible studies to resolve the inconsistent and ambiguous findings, and to clarify the diagnostic value of DNA methylation in EC.Furthermore, we aimed to identify the diagnostic accuracy of gene methylation markers to predict other clinical pathological outcomes of EC.

Study characteristics
The workflow of the systematic literature search is displayed in Figure 1.The primary search of PubMed and Web of Science identified 701 articles, of which 194 were duplicate articles.A total of 120 studies were initially obtained after filtering the titles, abstracts, and full texts.Of these, 98 studies were excluded based on the eligibility criteria: 11 studies were excluded based on the use of in vitro/ex vivo cell lines and human xenografts; 41 studies did not have healthy normal controls; 11 were not casecontrol studies; it was not possible to extract or calculate the diagnostic sensitivity and specificity of the methylation biomarkers in 30 studies; and five studies had small sample sizes (n ≤ 10).All the included studies focused on DNA methylation/hypermethylation in tissues.The systematic literature search thus finally yielded 22 studies including 1930 participants (1418 patients and 512 normal individuals).None of the patients had received preoperative chemotherapy, radiotherapy, or hormone therapy.The included studies were published between 2001 and 2016 and originated from nine regions (China, Czech Republic, Hong Kong, Japan, Netherland, Russia, Slovak Republic, Taiwan, and USA) (Table 1).The sample sizes of these studies ranged from 19-155 (median 88).Fourteen studies evaluated the diagnostic value of methylation of a single gene (Sasaki et al., 2001(1) [10]; Sasaki et al., 2001(2) [11]; Saito et al., 2003 [12]; Sasaki et al., 2003 [13]; Li et al., 2005 [14]; Pijnenborg et al., 2007 [17]; Yanokura et al., 2007 [18]; Tse et al., 2009 [20]; Varley et al., 2009 [21]; Yi et al., 2011 [22]; Kovalenko et al., 2013 [25]; Yang et al., 2013 [27]; Chmelarova et al., 2014 [28]; Dong et al., 2015 [30]), five studies evaluated multiple genes (Banno et al., 2006 [15]; Suehiro et al., 2008 [19]; Fiolka et al., 2013 [24]; Visnovsky et al., 2013 [26]; Sheng et al., 2016 [31]), and the other three studies evaluated both single and combined genes (Shih et al., 2006 [16]; Zhang et al., 2011 [23]; Chen et al., 2015 [29]).Seventeen studies measured the methylation patterns of the genes using methylationspecific polymerase chain reaction (MSP), three studies used combined bisulfite restriction analysis (COBRA), one used quantitative MSP (qMSP), and one used MSP and COBRA for two genes, respectively.Details of the DNA methylation biomarkers and their diagnostic powers are shown in Supplementary Table 1.All 22 selected publications were evaluated and checked by two reviewers.High levels of methodological quality (more than five stars) were observed according to the NOS scale.

Meta-analysis of diagnostic value
We assessed the risk of bias for each study.The detailed evaluation criteria and results for each item are shown in Supplementary Table 2 and Supplementary Figure 1, and the risk of bias is summarized in Figure 2. The risk of bias was high or unclear in most included studies.Four studies stated that the sequences of participants were generated randomly.The diagnostic values for all the assessed methylation biomarkers were reported in 64% of studies, indicating no selective reporting.Eight studies were reported to be free of other biases and were defined as low risk.
Taken together, we considered that the detection method and gene combination contributed to the heterogeneity source.Further, well-designed studies with different detection methods are therefore needed to clarify the diagnostic role of gene methylation in sporadic EC.

Sensitivity analysis and publication bias
Sensitivity analysis was conducted to evaluate the stability of the results, by removing one study at a time.However, this did not significantly affect the pooled OR or inconsistency index, indicating the stability of our meta-analysis.The funnel plot was relatively symmetrical and no single study fell outside the funnel (Figure 4), suggesting that there were no significant publication biases in this meta-analysis of gene methylation in patients with sporadic EC.

DISCUSSION
Surgical resection or biopsy specimens are used for diagnosing numerous cancers, and are considered to be the gold standard for clinical applications.Pathological results from these tissues can also provide important information in relation to clinical decision making.DNA methylation has previously demonstrated potential use as a biomarker for EC, showing distinctly different patterns between EC and normal tissues [32].Alterations in DNA methylation in EC reflect transcription changes during endometrial carcinogenesis.According to a recent review, numerous studies have assessed abnormal promoter methylation of different genes in EC [33].Although methylation studies in relation to EC remain at the preclinical stage, many results offer the potential for clinical use as diagnostic, prognostic, and therapeutic-response biomarkers, as well as targets of epigenetic therapies.However, although aberrant methylated regions of P16 and RASSF1A demonstrated sensitivity and specificity for the detection of EC [34,35], aberrant methylation statuses of other genes, such as ER, PR, MLH1, MGMT, APC, and CDH1, have shown variable diagnostic accuracies for EC [33].
In the present study, we evaluated the diagnostic accuracy of methylated markers for sporadic EC, based on previously published studies.Overall, the pooled sensitivity and specificity of DNA methylation for EC diagnosis were 0.93 and 0.48, respectively, and the area under the SROC curve (AUSROC) was 0.88 in tissue samples.The presence of DNA methylation had a relatively high diagnostic ability (area under curve [AUC] = 0.8834) for the risk of sporadic EC (Figure 5), but a relatively low specificity (0.48) and high sensitivity (0.93).In terms of sensitivity, DNA methylation is chemically stable and has shown high sensitivity in other tumors [36].The present evidence demonstrated that methylated genes  were superior to protein biomarkers such as CA-125, either alone or in combination, which had diagnostic sensitivities of only 40%-80% [37,38].A clinical diagnosis of EC currently relies on a combination of ultrasound, magnetic resonance imaging, and CA-125, though none of these alone is completely satisfactory.CA-125 is often used as a biomarker for ovarian cancer, but has also been suggested as a good prognostic marker for EC [37].The presence of circulating tumor DNA (ctDNA) has been considered to be highly specific for certain cancers, mainly because somatic mutations identified in tumor DNA are absent from normal DNA, whereas gene methylation may occur in normal as well as cancer DNA [39].Our study confirmed that the detection of DNA methylation in clinical samples had a relatively low diagnostic specificity for EC (pooled specificity: 0.48, 95%CI: 0.46-0.50),suggesting that more specific methylated genes are required.
The current study included gene methylation or hypermethylation alterations, which could lead to aberrant silencing of tumor suppressor genes in most cases.Consistent with these types of gene alterations, the most common detection method for methylation was MSP, which has various advantages including relatively high sensitivity [40].In our meta-analysis, MLH1, CDH1, and RASSF1A gene methylation were detected in more than one study.MLH1 was identified as the cause of abnormal DNA mismatch repair and microsatellite instability (MSI), and its dysfunction might be associated with Lynch syndrome (also known as hereditary nonpolyposis colorectal cancer syndrome, HNPCC) [41].Although MSI has been detected in numerous patients with EC, its mutation frequency is extremely low, suggesting that it may arise as a result of aberrant methylation of promoter regions in MSI-positive EC patients [15].Furthermore, the diagnostic value of aberrant DNA methylation for EC was not significantly influenced by the technique used, except possibly for qMSP, suggesting that this kind of assay may show greater specificity.A combination of gene methylation markers also showed potentially higher diagnostic accuracy in sporadic EC.
This preliminary study clearly indicated that detecting gene methylation patterns could offer a diagnostic test for sporadic EC with high sensitivity, but low specificity.The presence of DNA methylation in EC patients may predict positive lymph node metastasis and unfavorable survival.In general, there is lack of diagnostic and prognostic biomarkers for EC, and methylation markers may thus be useful for distinguishing between EC and non-malignant causes in women with abnormal vaginal bleeding.Further meta-analyses need to be conducted to address the use of accurate methylated targets and suitable detection techniques, while more prospective studies utilizing consistent and standardized methodologies are urgently required to resolve these problems.
This study had several limitations.Firstly, selection bias might have occurred due to enrichment of studies reporting positive results, and the relatively small sample sizes in some of the selected literature may also have led to bias.Furthermore, the included studies were mostly from East Asia, and the conclusions may therefore not be universally applicable.The use of different MSP primers and/or equipment, and the lack of a well-accepted methylated gene in sporadic EC might also have been potential sources of bias.

Search strategy
We conducted a systematic literature search to identify studies assessing DNA methylation changes as   Two authors (YF) and (YW) consulted, and screened the titles and abstracts of the identified papers independently.All the studies were evaluated and discussed by the authors until a consensus of criteria was reached.The Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) statement [42] was applied as the template for the searching process.

Eligibility criteria
Duplicate articles were removed after combining the retrieved publications from the two databases.Initial screening was conducted by reviewing the titles and abstracts.Only full-text reports of original studies were included, and meeting abstracts, reviews, and editorials were excluded.Articles not focusing on DNA methylation changes in tissues and not in the context of sporadic EC detection/diagnosis were also excluded.After the firstround examination, we conducted a full-text review of the remaining articles and excluded studies based on the following exclusion criteria: studies without healthy normal individuals or controls (e.g., studies with only paired samples or benign disease were not considered); studies that were not case-control studies; studies for which sensitivity and specificity values for the diagnosis of sporadic EC were not reported or could not be calculated from the published data; studies with small sample sizes (n ≤ 10); and studies based on cell lines/animals rather than human clinical samples.

Data extraction and statistical analysis
Two investigators independently reviewed and evaluated the eligible studies according to the Newcastle-Ottawa scale (NOS) [43], and studies awarded five or more stars were considered as high-quality.The following data were then extracted by the two authors using a standardized form: first author, year of publication, region, sample size, detection techniques, biomarkers, alteration type, diagnostic sensitivity and specificity values, and risk of bias according to criteria from the Cochrane Collaboration tool (Cochrane handbook for systematic reviews of interventions version 5.1.0.).The following risk-of-bias items were evaluated using standardized methods: random sequencing generation, allocation concealment, blinding of patients and study personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases.Any disagreement was resolved by further discussion.If methylation values were not definitively reported, the information was extracted from available tables or figures as far as possible.The diagnostic value or accuracy of the methylated genes for sporadic EC was evaluated by analyzing diagnostic variables such as sensitivity, specificity, positive likelihood ratios (PLR) and negative likelihood ratios (NLR), diagnostic ratios (DOR), and summary receiver operating characteristic curves (SROC) using Meta-DiSc software [44].A PLR >5.0 and NLR <0.2 were considered clinically significant.The DOR represented the increased risk of EC in patients with the methylated gene compared with those without.Sensitivity and publication bias were analyzed using Review Manager 5.3, and publication bias was presented using funnel plots.Substantial heterogeneity was considered to exist when I 2 was >50%.In the event of heterogeneity among studies, the results were pooled using a random effect model; otherwise a fixed effect model was adopted.

Figure 2 :
Figure 2: Risk of bias graph (reviewers' judgements about each risk of bias item presented as percentages across enrolled studies).

Figure 3 :Figure 4 :
Figure 3: Diagnostic accuracy of forest plots.(A) Forest plots of pooled sensitivity.(B) Forest plots of pooled specificity.(C) Forest plots of pooled positive likelihood ratio.(D) Forest plots of pooled negative likelihood ratio.

Table 3 : Meta-analysis of the association DNA methylation with clinicopathological features
Abbreviation: The implication of No. of studies and patients containing multiple and different genes detected in the same study; P = p value; OR=odds ratio; I 2 = inconsistency index; 95% CI = 95% confidence intervals; BMI = body mass index; others = non-endometrioid; Invasion = myometrial invasion.www.impactjournals.com/oncotarget