Epstein–Barr virus antibody level and gastric cancer risk in Korea: a nested case–control study

Background: Few cohort studies have investigated Epstein–Barr virus (EBV) infection before the occurrence of gastric cancer. Methods: Among 14 440 cohort participants, 100 incident gastric cancer cases were individually matched to two controls. Epstein–Barr virus antibodies IgG and IgA against viral capsid antigen (VCA), EBV nuclear antigen (EBNA) antibody IgG, and early antigen (EA) antibody IgG were measured using enzyme immunoassays (EIAs). Results: The highest titres of VCA IgG (odds ratio (OR): 1.37, 95% confidence interval (CI): 0.62–3.06) or EBNA IgG (OR: 0.87, 95% CI: 0.51–1.46) were not associated with gastric cancer risk. Conclusion: Higher levels of VCA IgG or EBNA IgG were not associated with increased risk of gastric adenocarcinoma in Koreans.

Epstein -Barr virus infection is believed to be related to 5 -16% of gastric adenocarcinomas, with variations across populations (Burgess et al, 2002;Macsween and Crawford, 2003;Correa et al, 2004;Herrera-Goepfert et al, 2005). Most studies to date have been tumour tissue-based, providing information about the role of EBV in gastric carcinogenesis, but providing limited information on the association between EBV infection and gastric cancer, particularly adenocarcinoma -the most prominent histology (Tokunaga and Land, 1998;Chang and Kim, 2005;Akiba et al, 2008). The few epidemiological studies of EBV infection and gastric cancer, have mainly been limited by small size and non-prospective design (Levine et al, 1995;Shinkura et al, 2000). A recent case -control study with 185 gastric cancers did not take account of the possible contribution of Helicobacter pylori to the question (Koshiol et al, 2007).
We carried out a nested case -control study within the Korean Multi-Center Cohort (KMCC) study of the possible relationship between EBV-specific antibody titres and gastric cancer, adjusting for H. pylori infection status.

MATERIALS AND METHODS
Individuals in this nested case -control study were participants in the KMCC (Yoo et al, 2002;Gwack et al, 2006). Men and women aged X30 years were recruited from 1993 to 2002. Informed consent was obtained from all participants who were enrolled in this study, and the Institutional Review Board of Seoul National University Hospital approved the study protocol . Blood samples were collected after an overnight fast, processed within 8 h, stored at À201C for 2 -3 days, and transferred to a central site where they were preserved at À801C.
A total of 14 440 men and women aged X30 years were recruited from 1993 to 2002. Incident cancer cases were identified by a computerised record linkage to the Korea Central Cancer Registry database and the National Health Insurance database using resident registration numbers. The last date of follow-up was 31 December 2002, and 100 incident gastric cancer cases were identified among the 14 440 cohort members by the end of the follow-up period. Among participants who were alive and had no period of follow-up loss at the date when each case was diagnosed, controls were selected and matched 1 : 2 by incidence density sampling for the year of enrolment, area of residence, age within 5 years, and gender. The follow-up period was defined as the date of recruitment to the end of follow-up or to death. For the case group, the end of the follow-up date was defined as the date of diagnosis from the Korea Central Cancer Registry database.
The laboratory was blinded to the case -control status. For quality control of EBV VCA IgG and EBNA IgG, one negative and one positive control was included in each plate of 96-wells. Sensitivity for EBV VCA IgG was reported to be 98.3% and specificity 98.1% compared with an immunofluorescence assay. The sensitivity for EBNA IgG was reported to be 94.4% (90.0 -97.3%) and specificity 98.9% (93.8 -100%) compared with another ELISA system currently in commercial distribution. Four kinds of control samples were used for quality control of EBV VCA IgA and EA IgG, according to the manufacturer's guidelines: one negative control, one cutoff control, one weakly positive control, and one positive control in each plate. The relative sensitivity and specificity for EBV VCA IgA and EA IgG was reported to be 495% compared with another ELISA method. These ELISA methods from various companies have been shown to correlate closely with immunofluorescence serology (Bruu et al, 2000).
Seropositivity for H. pylori antibody was identified by testing the expression of H. pylori-specific antibody IgG using immunoblot kits (Helico Blot 2.1; MP Biomedicals Asia Pacific, Singapore, Singapore).
Antibody titres were categorised into binary variables as negative or positive, including indeterminate as positive for statistical analysis. The titres of VCA IgG and EBNA IgG were re-categorised into three tertile groups based on titres of control groups.
Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using a multivariate conditional logistic regression model. The SAS software version 9.1 (SAS Institute, Cary, NC, USA) was used for all statistical analyses. Table 1 shows the basic characteristics between cases and controls. Adenocarcinoma comprised 87% of the gastric cancer cases, 2% were other tumours, including single cases of malignant carcinoid tumour and squamous cell carcinoma in situ; 11% were unspecified.

RESULTS
Viral capsid antigen IgG, which develops early in EBV infection and persists throughout life, was positive in 94% of the cases and in 92% of the controls (P ¼ 0.4). Epstein -Barr virus nuclear antigen IgG, another antibody indicating latent infection, was positive in 79% of the cases and in 82% of the controls (P ¼ 0.4). Viral capsid antigen IgA, an indicator of recent infection, was positive in 2% of the cases and in 3% of the controls. Lytic enzyme EA IgG was positive in 12% of the cases and in 11% of the controls.
Crude and adjusted ORs for VCA IgG, VCA IgA, EBNA IgG, and EA IgG of indeterminate and positive categories compared with the negative category were not significantly associated with gastric cancer risk (Table 2). This remained the case when VCA IgG and EBNA IgG positive groups were re-categorised by the titre level of below and above median, higher titre levels. When the analysis was restricted to adenocarcinoma case pairs, the OR for the highest tertile of VCA IgG was 1.33 (95% CI: 0.59 -3.00) and that of EBNA was 0.92 (95% CI 0.53 -1.58) compared with the negative category (Table 3). For gastric adenocarcinoma pairs, including those who were diagnosed more than 2 years after enrolment, the ORs for the highest category of VCA IgG and EBNA IgG vs the negatives were 1.53 (95% CI: 0.62 -3.75) and 1.23 (95% CI: 0.64 -2.36), respectively.

DISCUSSION
In this study, more than 80% of the study participants tested positive for EBV latent infection antibodies (EBV VCA IgG and EBNA IgG), and no significant association between seropositivity for latent EBV infection and gastric cancer risk was observed. Even after restricting the analysis to adenocarcinoma pairs who were in follow-up more than 2 years after enrolment, there was no significant association between antibody levels and gastric cancer risk.
The highest category of VCA IgG and EBNA IgG titres had an increased risk of gastric adenocarcinoma in a nested case -control study of 54 cases and controls, but the difference was not significant (Levine et al, 1995). A recent Chinese nested casecontrol study suggested an opposite pattern, that a high VCA IgG level might be associated with a lower risk of cardiac gastric cancers (Koshiol et al, 2007). In our study, EBV latent infection antibodies were not associated with gastric cancer risk. These differences may be due to infection by different EBV strains, which are believed to influence EBV-infected gastric cancer worldwide in combination with different environmental factors (Herrera-Goepfert et al, 2005).
Another possible reason for these inconsistencies may be an interaction between H. pylori and EBV infection, which is poorly understood. In this study, we measured seropositivity of H. pylori  (Ouburg et al, 2005;Gwack et al, 2006). Several candidate gene approach studies suggest that EBNA genes interact with inflammatory and tumour suppressor genes, which has implications in gastric carcinogenesis (Ouburg et al, 2005).
In the analysis of EBNA, EA, and VCA antibodies, immunofluorescence assay methods are more time consuming and difficult to interpret, but are more specific, have high resolution, and stand as the 'gold standard' technique. Enzyme immunoassay (EIA) is often used in clinical laboratories because of its reliability in highthroughput analysis (Gartner et al, 2003). The validity of the EIA kits is supported by reasonably high sensitivity (95 -100%) and specificity (86 -100%), compared with the indirect immunofluorescence assay as a gold standard (Bruu et al, 2000;Gartner et al, 2003). Large-scale population-based studies have recently used EIA (Tedeschi et al, 2007).
In our study, pathological specimens were not collected so that the association between EBV antibody titre and gastric cancer was not compared among EBV-related and non-EBV-related gastric cancer cases. We cannot exclude chance effects due to our small numbers, and further studies are required. Antibodies were checked only at the time of enrolment and not re-tested, which omitted detailed information about EBV infection status during follow-up.  aOR ¼ adjusted odds ratio; CI ¼ confidence interval; EBNA ¼ EBV nuclear antigen; IgG, immunoglobulin G; VCA ¼ viral capsid antigen. a Conditional logistic regression model adjusting for age (continuous), education (none, 1 -12 years, more than 13 years), alcohol consumption (never, past, current), cigarette smoking (never, past, current), and H. pylori antibody seropositivity (yes or no).

Epstein -Barr virus antibody level and gastric cancer Y Kim et al
Our study has several strengths; it was nested in a prospective cohort study, and sera were collected before the diagnosis of gastric cancer, which may minimise any possibility of misclassification for exposure. We restricted our analysis to gastric adenocarcinoma cases. If lymphoepithelioma-like cancer was included among the 'unspecified' pathological group, the OR may have tended towards null. To exclude potentially developing cases before or near the period of enrolment, we restricted our analysis to individuals who were diagnosed after 2 years of enrolment.
In summary, our results provide little support for the hypothesis that increased levels of EBV antibodies are related to gastric adenocarcinoma.