Background

Nasopharyngeal carcinoma (NPC) is an epithelial malignancy that occurs in the nasopharynx. According to the World Health Organization (WHO), it was estimated to be 120 434 new NPC cases and 73 482 deaths in the world in 2022, mainly distributed in East Asia, Southeast Asia and North Africa [1]. Approximately 70% of NPC patients are already in advanced stages when they are diagnosed. Through comprehensive treatment, such as surgery, chemotherapy and radiotherapy, 5-year survival rate for advanced stages ranges from 51.4 to 76.0%, compared to 82.6–86.6% for early stages, so early detection is needed to reduce the disease burden [2,3,4,5].

The pathogenesis of NPC is closely related to Epstein-Barr virus (EBV) infection, and detection of EBV-related biomarkers is crucial for early detection of NPC. EBV serum antibody (antibodies are labeled as ‘anti-EBV’, or more precisely, they are antibodies against EBV-encoded antigens) detection has the characteristics of rapidity, simplicity and low cost. Enzyme-linked immunosorbent assay (ELISA) has standardized operation procedure and the interpretation of results is not subject to subjective influence of researchers.

So far, there have been a number of studies on the detection of serum EBV-related IgA antibodies or antibody panels associated with IgA by ELISA, but the results were not consistent. Although several meta-analyses have investigated the diagnostic value of relevant antibody tests for NPC, they involved only antibodies alone or in combination, and did not analyze and compare all possible IgA antibody panels simultaneously [6,7,8,9].

Regarding the above issues and preliminary search results, we used a diagnostic test accuracy meta-analysis to evaluate the serological diagnostic value of nine antibody panels of EBV-related IgA antibodies by ELISA, in order to find suitable detection biomarkers/panels and provide more evidence for NPC screening and early detection.

Methods

This meta-analysis was conducted according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement and registered on the International Prospective Register of Systematic Reviews (PROSPERO) with registration number CRD42023426984 [10, 11].

Search strategy

1) Databases: PubMed, Embase, China National Knowledge Infrastructure (CNKI) and Chinese BioMedical Literature Database (CBM). The first two were used for searching English literature, whereas the latter two were specifically for Chinese literature, given that a large number of related studies have been published in Chinese journals. 2) Search period: from January 1st, 2000 to September 30th, 2023. 3) Key words: nasopharyngeal carcinoma, IgA, screening, early detection, early diagnosis, sensitivity and specificity. 4) Languages: Chinese and English. 5) References for other system reviews or meta-analyses.

Inclusion criteria

1) The case group of the study were comprised of newly diagnosed or untreated NPC patients confirmed by histopathology, and the control group were healthy individuals or general population. At least 20 cases and 20 controls were included in each study. 2) The study applied ELISA to detect serum EBV-IgA antibodies alone or in parallel, within nine antibody panels including EBNA1-IgA, VCA-IgA, EA-IgA, Zta-IgA, EBNA1-IgA + VCA-IgA, VCA-IgA + EA-IgA, VCA-IgA + Rta-IgG, EBNA1-IgA + VCA-IgA + Zta-IgA, VCA-IgA + EA-IgA + Rta-IgG. 3) If similar studies were found originated from the same population, only the most complete one was selected. If the same antibody panel in one study was tested multiple times with ELISA kits from different manufacturers, the mean value was calculated and included. 4) Only studies that provide sufficient data were included.

Exclusion criteria

1) Studies in which NPC patients in the case group also had other tumors or the control group were patients from hospital with other head and neck diseases. 2) Studies for which full texts were not available.

Study selection, data extraction and quality assessment

Study selection, data extraction, and quality assessment were performed independently by two researchers, and a third researcher was consulted in the case of disagreement. According to the search strategy and inclusion/exclusion criteria, literature search, duplicate removal, title abstract screening, and full text screening were carried out via EndNote 20. The extracted data included first author, year of publication, journal name, funding sources, study area, number of cases, number of controls, antibody panels, and the numbers for true positive, false positive, false negative and true negative. The Quality Assessment of Diagnostic Accuracy Studies tool (QUADAS-2) was used to evaluate the risk of bias and applicability of the included literature. The tool is consisted of four key domains: patient selection, index test, reference standard and flow and timing. Each was assessed in terms of risk of bias based on signalling questions and the first three in terms of concerns regarding applicability. The process was conducted in RevMan5.4 [12].

Data analysis

Bivariate models were used for statistical analysis. Firstly, the Bivariate I2 value was calculated to detect heterogeneity among the included studies; if I2 > 0.5, heterogeneity was considered present. Secondly, the threshold effect was assessed using the Spearman correlation coefficient between the logarithm of sensitivity and the logarithm of (1-specificity) for each study; if P < 0.05, the threshold effect was present, only a summary receiver operating characteristic (SROC) curve was plotted; otherwise, pooled sensitivity, specificity, diagnostic odds ratio (DOR) and the corresponding 95% confidence intervals (95%CIs) were also obtained. Subsequently, the included studies were classified by study area (whether high-risk) or number of cases (whether ≤ 100). Based on the number of studies in the two categories, subgroup analysis and sensitivity analysis were performed. Additionally, meta-regression analysis was also conducted on these two factors. The difference was statistically significant when P < 0.05. All these analyses were carried out using Meta-DiSc and Meta-DiSc 2.0 [13, 14]. Finally, Deek’s asymmetry test was conducted using the midas package in Stata 12.0 to evaluate publication bias in each antibody panel; a P-value of < 0.10 indicated the presence of publication bias.

Results

Study selection

Based on the search strategy, a total of 762 articles were found. Then, 217 duplicates were removed, leaving 545 articles. According to the inclusion and exclusion criteria, 380 articles were removed due to title and abstract screening, and then 95 articles were removed by full text screening. Finally, 70 articles were included in this meta-analysis [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84]. The literature screening process was shown in Fig. 1.

Fig. 1
figure 1

Study selection flowchart

Full size image

Study characteristics

Table 1 Main characteristics of the 70 included articles
Full size table

A total of 70 articles were included, published from 2002 to 2023, involving 11 863 NPC cases and 34 995 controls. All studies were conducted in China, and 50 of them were from high-risk areas (Guangdong, Guangxi, Hainan, and Hong Kong). The main characteristics of the included literature and antibody panels were shown in Tables 1 and 2.

Quality appraisal

QUADAS-2 was used to evaluate the quality of the included articles, and the results indicated that they were of high quality. The risk of bias was primarily attributed to flow and timing issues, followed by patient selection concerns. This was because the intervals between the index test and the reference standard were unclear in some studies, and it remained unclear whether subject selection in some studies involved continuous inclusion and random sampling. The quality assessment results were shown in Fig. S10.1 and Fig. S10.2 in Supplementary Materials.

Heterogeneity, threshold effect and publication bias

The heterogeneity, threshold effect and publication bias of the included studies for nine antibody panels were calculated separately. The results indicated that heterogeneity was present among all panels, whereas no significant threshold effect was observed, suggesting that the heterogeneity was primarily attributed to non-threshold factors. With the exception of VCA-IgA (P = 0.005) and EBNA1-IgA + VCA-IgA (P = 0.042), no publication bias was detected in the other panels (see Table 2).

Table 2 Main characteristics, heterogeneity, threshold effect, publication bias and meta-analysis results of the nine antibody panels
Full size table

Meta-analysis, subgroup analysis and sensitivity analysis

The pooled sensitivity, specificity and DOR of the nine antibody panels were presented in Table 2. Specifically, the sensitivity (95%CI) of EBNA1-IgA + VCA-IgA [0.928 (0.898, 0.950)], VCA-IgA + Rta-IgG [0.925 (0.890, 0.949)], EBNA1-IgA + VCA-IgA + Zta-IgA [0.962 (0.909, 0.985)] and VCA-IgA + EA-IgA + Rta-IgG [0.945 (0.918, 0.964)] was higher. Additionally, EBNA1-IgA + VCA-IgA [107.647 (61.173, 189.430)], VCA-IgA + Rta-IgG [105.988 (60.118, 186.857)], EBNA1-IgA + VCA-IgA + Zta-IgA [344.450 (136.351, 870.153)] had higher DOR (95%CI). The SROC curves for each panel were depicted in Fig. 2. Observing the positions of each summary point and corresponding confidence ellipses, the results for EBNA1-IgA + VCA-IgA and VCA-IgA + Rta-IgG were better. The results of the subgroup analysis classified by study area or number of cases were shown in Fig. S11.1 and Fig. S11.2 in the Supplementary Materials. Moreover, the findings from each meta-regression analysis were not statistically significant. As demonstrated in the sensitivity analysis results presented in Table S10 of the Supplementary Materials, there was no significant change in the DOR values. Therefore, parallel detection of serum VCA-IgA and EBNA1-IgA/Rta-IgG antibodies using ELISA may constitute suitable serological detection strategies for NPC screening and early detection. However, EBNA1-IgA + VCA-IgA + Zta-IgA may potentially represent an even better panel, simply because the small number of studies or individuals included in the analysis limited its evaluation and interpretation. And the characteristics and forest plots for each panel were provided in the Supplementary Materials.

Fig. 2
figure 2

SROC curves for each panel. AUC (95%CI): EBNA1-IgA=0.94 (0.91, 0.96); VCA-IgA=0.93 (0.90, 0.95); EA-IgA=0.92 (0.89, 0.94); Zta-IgA=0.93 (0.90, 0.95); EBNA1-IgA+VCA-IgA=0.97 (0.95, 0.98); VCA-IgA+EA-IgA=0.94 (0.91, 0.95); VCA-IgA+Rta-IgG=0.96 (0.94, 0.98); EBNA1-IgA+VCA-IgA+Zta-IgA=0.99 (0.97, 0.99); VCA-IgA+EA-IgA+Rta-IgG=0.96 (0.94, 0.98)

Full size image

Discussion

In China, NPC is mainly distributed in Guangdong, Guangxi, Hainan and Hong Kong [85]. Given that all 70 articles included in our study were from China, we designated these regions as high-risk areas for subsequent analysis. The results of this study revealed heterogeneity among the included studies for nine antibody panels. All panels did not have threshold effects, allowing us to calculate their pooled sensitivity, specificity, and DOR. Publication bias was detected for VCA-IgA and EBNA1-IgA + VCA-IgA. In conclusion, parallel detection of serum EBNA1-IgA, VCA-IgA and Zta-IgA antibodies by ELISA showed better pooled sensitivity and DOR among these nine panels. Under the condition that fewer indicators need to be detected, serum VCA-IgA and EBNA1-IgA/Rta-IgG antibody panel has comparable performance.

In addition to EBV-related antibodies serving as diagnostic markers for NPC, EBV-DNA load is also one of the common diagnostic methods. Other biomarkers, including heat shock protein 70 (HSP70), sialic acid (SA) and microRNA (miRNA), are closely associated with the development of NPC. A prospective study has shown that screening with two positive plasma EBV-DNA tests approximately four weeks apart could help in the early detection of NPC in asymptomatic individuals [86]. Another study has found and validated that P85 Ab, a novel serological biomarker for NPC screening, had good screening performance [87]. Although the results in this article are generally more accurate than those of several previous meta-analyses, the gap is not particularly significant [6,7,8].

This study is the first meta-analysis to determine the usefulness of EBNA1-IgA + VCA-IgA + Zta-IgA and comprehensively evaluates the performance of nine serum EBV-related IgA antibody panels detected by ELISA. The overall quality of the included studies was relatively high, and the inclusion and exclusion criteria of the article were considered comprehensively, aiming to align with the screening scenarios of a large population.

However, it also has some limitations. Since all research sources were from China, the conclusions may not be universally applicable to other regions or populations. Most of the included diagnostic trials were single-center studies, and the sample sizes of many studies were not particularly large. Due to the different designs of the included studies and the differences in sex, age, and tumor stage among the study population, the source of heterogeneity cannot be clearly explained. Although the design of this study is as close to the population screening situation as possible, the conditions still need to be verified in large-scale populations. In addition, this article focused solely on analyzing serum IgA antibodies or antibody panels detected by ELISA, and there are numerous other biomarkers and panels with promising diagnostic potential that could be explored in future research based on the current findings.

Conclusion

In summary, parallel detection of serum EBNA1-IgA, VCA-IgA and Zta-IgA antibodies using ELISA demonstrates better pooled sensitivity and DOR among these nine panels generally. When using a reduced number of indicators, serum VCA-IgA and EBNA1-IgA/Rta-IgG antibody panel exhibits comparable diagnostic performance.