The diagnostic value of EBV-DNA and EBV-related antibodies detection for nasopharyngeal carcinoma: a meta-analysis

Numerous individual studies have investigated the diagnostic value of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection for nasopharyngeal carcinoma (NPC), but the conclusions remain controversial. This meta-analysis aimed to determine the value of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection in the diagnosis of NPC. PROSPERO registration number: CRD42019145532. PubMed, EMBASE, Cochrane Library, and Chinese data libraries (Wanfang, CNKI, and CBM) were searched up to January 2019. The pooled sensitivity, specificity, and positive likelihood, negative likelihood, and diagnostic odds ratios were conducted in this meta-analysis. Summary receiver operating characteristic curves evaluated the test-performance global summary. Publication bias was examined by Deek’s funnel plot asymmetry test. Forty-seven studies with 8382 NPC patients (NPC group) and 15,089 individuals without NPC (Control group) were included in this meta-analysis. The sensitivity, specificity, positive likelihood (+ LR), negative likelihood (-LR), DOR and AUC of EBV-DNA in diagnosis of NPC were: 0.76 (95% CI 0.73–0.77), 0.96 (95% CI 0.95–0.97), 14.66 (95% CI 9.97–21.55), 0.19 (95% CI 0.13–0.28), 84 (95% CI 50.45–139.88), 0.96 (SE: 0.001), and 0.55 (95% CI 0.54–0.57), 0.96 (95% CI 0.96–0.97), 12.91 (95% CI 9.55–17.45), 0.35 (95% CI 0.29–0.43), 39.57 (95% CI 26.44–59.23), 0.94 (SE: 0.002) for the EA-IgA, and 0.85 (95% CI 0.84–0.85), 0.89 (95% CI 0.88–0.89), 6.73 (95% CI5.38–8.43), 0.17 (95% CI 0.12–0.23), 43.03 (95% CI 31.51–58.76), 0.93 (SE: 0.007) for the VCA-IgA, and 0.86 (95% CI 0.85–0.88), 0.87 (95% CI 0.88–0.90), 7.55 (95% CI 5.79–9.87), 0.16 (95% CI 0.13–0.19), 50.95 (95% CI 34.35–75.57), 0.94 (SE: 0.008) for the EBNA1-IgA, and 0.70 (95% CI 0.69–0.71), 0.94 (95% CI 0.94–0.95), 9.84 (95% CI 8.40–11.54), 0.25 (95% CI 0.21–0.31), 40.59 (95% CI 32.09–51.35), 0.95 (SE: 0.005) for the Rta-IgG. The EBV-DNA had larger AUC compared with other EBV-based antibodies (P < 0.05), while the difference between EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG was not statistically significant (P > 0.05). EBV-DNA, VCA-IgA, EBNA1-IgA and Rta-IgG detection have high accuracy in early diagnosis NPC. In addition, EBV-DNA detection has the higher diagnosis accuracy in NPC. On the other hand, EA-IgA is suitable for the diagnosis but not NPC screening.


Background
Nasopharyngeal carcinoma (NPC) is the most common malignant tumor in head and neck surgery and it is highly prevalent in southern China and Southeast Asia [1]. Unfortunately, early-stage patients with NPC are asymptomatic. More than 70% newly diagnosed NPC are local-advanced or distant metastasis, and the extent of NPC at diagnosis is the most important factor affecting survival rate [2]. Despite radiotherapy and chemoradiotherapy in widespread use as the primary treatment for NPC, the overall prognosis remains poor [3]. Therefore, the use of ideal NPC early diagnosis markers is crucial. Clinical information, laboratory exams and biomedical informatics are significance component in cancer patients [4]. NPC is related to Epstein-Barr virus (EBV) infection which can promote the development of NPC [5]. The detection of specific Epstein-Barr virus DNA and antibodies are important means for the early diagnosis of NPC [6]. In addition, EBV-based antibodies detection has the advantages of rapid, convenient, and low cost. Numerous individual studies have investigated the diagnostic value of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection for nasopharyngeal carcinoma, but variable sensitivities and specificities were reported. Currently, there is no consensus which is a better test for early diagnosis of NPC. This metaanalysis aimed to determine the value of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection in the diagnosis of NPC and to provide an important basis for NPC screening and early diagnosis. This meta-analysis followed the PRISMA Diagnostic Test Accuracy reporting guidelines [7].

Data sources and literature search strategy
Literature review was separately conducted by two investigators that queried online databases, including Pub-Med, EMBASE, Cochrane Library, and Chinese data libraries (WanFang, CNKI, and CBM), and the search concluded in January 2019, using the following keywords: nasopharyngeal carcinoma, Epstein-Barr virus, capsid antigen-IgA, early antigen antibody, nuclear antigen antibody, BRLF1 transcription activator IgG, EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG.

Study selection
Inclusion criteria 1. Studies that assessed the performance of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection for untreated NPC identification; 2. All patients included in the study were diagnosed using a reference test (such as needle biopsy or postoperative tissue specimens with pathological confirmation); 3. Studies that used a pre-specified threshold; 4. Studies that clearly stated the number of true positive, false positive, false negative, and true negative results in the diagnosis of NPC or these values could be calculated from the data; 5. Studies that provided a clear definition of the control sources (healthy individual or non-NPC patients); 6. In cases of multiple reports describing the same population, the most recent or most complete report was selected.

Exclusion criteria
1. Reported results were insufficient for construction of the 2 × 2 table; 2. Studies that failed to clearly define the control types; 3. The NPC group contained other tumors; 4. Basic research, review articles, comments, letters, case reports, abstracts in conference, responding letters and experimental animal studies.

Study quality assessment and data extraction
Study quality assessment was conducted using the diagnostic accuracy (QUADAS) II checklist [8]. Studies considered of high quality were eligible for this metaanalysis. Data on study characteristics, the first author, year of publication, country of origin, article language, sample size, control sources (healthy individuals or non-NPC patients), detection method, sample types and cutoff value were extracted from the selected studies by one author and checked by another author. If agreement cannot be reached, a third reviewer will be consulted. Any disagreements were discussed until consensus was reached.

Statistical analysis
Standard methods recommended for meta-analyses of diagnostic test evaluations were used to perform this meta-analysis [9]. Review Manager version 5.3, Meta-DiSc statistical software version 1.4 and Stata version 14.0 (STATA Corporation, College Station, TX, USA) were used in this meta-analysis. The Cochrane Q test and inconsistency index (I 2 ) were used to estimate the heterogeneity within studies [10]. Heterogeneity was considered statistically significant when P < 0.05 or I 2 > 50%. If statistically significant heterogeneity existed, metaanalysis was performed using the random effects model, otherwise, a fixed effect model was used.
The accuracy indexes of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG was pooled by meta-analysis, such as sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR) and AUC. The likelihood ratios (PLR and NLR) are clinically meaningful for the measurement of diagnostic accuracy; PLR > 10 and NLR < 0.1 are considered high [11]. The DOR is a single indicator of test accuracy that combines the data from sensitivity and specificity into a single metric. The summary receiver operating characteristic (SROC) curve was used to evaluate the global summary of test performance.
Sensitivity analyses were performed to explore the sources of heterogeneity of the included studies by removing each included study consecutively. The heterogeneity was investigated by meta-regression according to different covariates, including publication year (Year ≥ 2011 or < 2011), NPC size (NPC ≥ 100 or NPC < 100), control sources (Control sources from healthy serum or from healthy persons and non-NPC patients), detection method, and article language (English or Chinese). Publication bias was examined by Deek's funnel plot asymmetry test. All P values were two sides and P < 0.05 was regarded as statistically significant.

Article search and study quality
In this meta-analysis, 47 publications on the role of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG concentrations in the diagnosis of NPC that met the criteria for inclusion were included in the analysis . Figure 1 shows a flowchart of the study selection process. The 47 studies included 8382 patients with NPC (NPC group) and 15,089 patients without NPC (Control group). The main features of enrolled studies are summarized in Table 1. Article quality was judged in terms of the QUADAS II recommendations. The proportions of studies with low, high, or unclear risk of bias and applicability concerns are displayed in Fig. 2.

Diagnostic accuracy
The pooled sensitivity, specificity, PLR, NLR, DOR and AUC for the value of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG in the diagnosis of NPC are displayed in Table 2 and the the diagnostic characteristics of included studies are in Tables 3, 4, 5, 6, 7. The EBV-DNA had larger areas under the summary receiver operator curve when compared with EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG (P < 0.05), while EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG were no statistically different from each other (P > 0.05) in Table 8. The summary receiver operator curve of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection for NPC were showed in Fig. 3. Additional file 1 showed the Forest plots of sensitivity, specificity, PLR, NLR, DOR for acoustic analysis of EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG.

Sensitivity analysis and meta-regression
The sensitivity analysis showed that the results were not affected by the exclusion of any individual trial. As meta-regression result indicated that publication year, NPC or control size, control sources, detection method, cutoff value, and article language are not the DOR heterogeneity of EBV-DNA, VCA-IgA, EBNA1-IgA and Rta-IgG, whereas detection method was possible DOR heterogeneity sources for the EA-IgA (P < 0.0095).

Discussion
EBV infection plays a critical role in the progression of nasopharyngeal carcinoma, as body can produce lots of EBV-related antigens at the early stage, which can be used for NPC screening and EBV-DNA, EA-IgA, VCA-IgA and EBNA1-IgA are usually involved [59,60]. EBV-DNA in circulation may be released from cancer cells during the process of apoptosis or generated from viral replication and different EBV antigens are expressed at different stages of infection [61]. Circulating EBV-DNA has been shown to correlate with the stage of NPC, recurrence rate and screening for NPC [62]. In this study, a meta-analysis was conducted to assess the diagnostic significance of five EBV-based markers for patients with NPC. This study showed that the EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG detection were effective method for NPC diagnosis.
Previous meta-analyses have been published on the value of some EBV-based markers in the detection of NPC. For the EBV-DNA, Han et al.conducted a metaanalysis based on 18 studies involving 1492 NPC cases and 2461 health controls in Asians, in which the pooled sensitivity and specificity of EBV-DNA detection for NPC were 0.73 (95% CI 0.71-0.75) and 0.89 (95% CI 0.88-0.90) [63]. Furthermore, Han et al.found that the accuracy of NPC detection was lower by serum (0.81) than that by plasma (0.86), with SROCs being 0.91 and 0.97, respectively. The heterogeneity across studies showed significant difference in the Han's study, and Han et al.did not evaluate the threshold effect and publication bias [63]. The Han's study should also perform sensitivity analysis and meta-regression to explore the sources of heterogeneity. Li et al. conducted another important meta-analysis on the diagnosis value of VCA-IgA detection for NPC based on 4671 patients with NPC and 7663 patients without NPC [64], which was correlated with higher pooled sensitivity 0.91 (95% CI 0.90-0.92) and specificity 0.92 (95% CI 0.92-0.93), with SROC 0.98. But the Li's study existed language bias   [65]. Threshold effect, publication bias as well as complicated control types presented in the Cui's study, which may contribute to heterogeneity and affect the accuracy of pooled results. Additionally, previous meta-analyses could not reach a conclusive result as to the most favorable choice for NPC diagnosis.
To our knowledge, this is the first meta-analysis to determine the usefulness of EA-IgA and EBNA1-IgA and compare the accuracy EBV-DNA, EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgA in diagnosis of NPC. In this study, the highest sensitivity was EBNA1-IgA (0.86), and the specificity of EBV-DNA (0.96) and EA-IgA (0.96) were highest. Besides, the sensitivity of EA-IgA (0.55) was lowest, and screening for NPC using only EA-IgA may lead to misdiagnosis, but the specificity was high, which indicated that EA-IgA was suitable for the diagnosis but not screening of NPC. EA-IgA or EBV-DNA detection combined with other  indicators may also improve the sensitivity and specificity for the serological diagnosis of NPC [44,51]. The likelihood ratios (PLR and NLR) calculated from the sensitivity and specificity indicate the discriminatory properties of negative and positive test results. The pooled PLR of EBV-DNA, VCA-IgA and EA-IgA were above 5 and NLR below 0.2, which given strong diagnosis evidence especially for EBV-DNA and EA-IgA with PLR above 10 [11]. Furthermore, a summary receiver operating characteristic (SROC) curve was also conducted to describe the relationship between sensitivity and specificity. AUC can summarize the inherent capacity of a test to discriminate the participant with disease from those without it [66,67]. The AUC of EBV-DNA and other antibodies were more than 90%, indicating a very high level of overall accuracy. The EBV-DNA (AUC = 0.96) had slightly larger AUC compared with EA-IgA, VCA-IgA, EBNA1-IgA and Rta-IgG (P < 0.05). Additionally, the pooled DOR of EBV-DNA (84.00) that differed 33.05-44.43 was higher than other EBV-based antibodies. These results indicated that EBV-DNA detection had higher accuracy in diagnosis of NPC. In addition, a meta-analysis included 8128 NPC cases showed that pre-EBV-DNA levels can also be a prognostic indictor for patients with NPC [68]. A recent prospective screening study involving 20,174 participants showed that plasma EBV-DNA detection was useful in screening for early asymptomatic nasopharyngeal carcinoma screening, with 97.1% sensitivity and 98.6% specificity [2]. But only 309/1112 had detectable Epstein-Barr virus DNA in plasma at baseline and at follow-up, and 35 patients had confirmed nasopharyngeal carcinoma. In Nicholls' study [15], seventy-eight NPC patients (15.1%) were plasma EBV-DNA negative who had similar 5-year overall survival and cancer-specific survival to those EBV-DNA positive counterparts by stage. If only plasma EBV-DNA was used as the population screening tool, 60.0%, 23.0%, 14.5% and 5.0% of stage I, II, III and IVA NPC may be missed. The golden standard for cancer prognosis is pathological examination following the complicated and painful procedures of biopsy, which may not be feasible by some patients [69]. In addition, endoscopic, computed tomography (CT) and Magnetic resonance imaging (MRI) has been the imaging modality of choice for cancer diagnosis and staging [70,71]. In practice, due to the current limitations of a single serum index, multiple assays (nasal endoscopy, CT and MRI) and biopsies [70,72], it is still necessary develop methods to increase NPC early diagnostic rate. Several strengths of present meta-analysis should be highlighted. This study compered five EBV-related diagnostic markers for NPC with comprehensive calculations of their diagnostic performance, shedding light on the value of these tests in clinical settings. In the context of current availability of studies on EBV-DNA and immunoglobin antibody tests in the literature, this meta-analysis covers a large sample size pooled from rigorously included studies, and the results were stable. However, this meta-analysis should be interpreted with caution due to certain limitations. First, there was large heterogeneity among the included studies with differences in characteristics of the study and participants. The inability to obtain raw data on patient age and gender may have led to the heterogeneity and hindered a more detailed analysis. Second, most of the included populations were Chinese, which could lead to population selection bias and should not allow for generalization to other ethnicity groups, rendering further research needed. Third, technical methods for testing of EBV-related markers vary across different studies, including the inconsistent cut-off values and different antigen sets used despite of a same generic name. Use of enzymelinked immunosorbent assay or immunofluorescence assay may have contributed to the different results [73]. Finally, most of the studies included NPC cases and controls in a single institution or from a same geographic region, which could have influenced the results of the study. Therefore, a larger, prospective,

Conclusions
Notwithstanding heterogeneity of currently available data, the studies included in our analysis are of a large sample size and high-quality, thus providing a considerable power. EBV-DNA, VCA-IgA, EBNA1-IgA and Rta-IgG detection have high accuracy in early diagnosis NPC and can improve the effectiveness of screening. In addition, EBV-DNA detection has the higher diagnosis accuracy in NPC. On the other hand, EA-IgA is suitable