Heterogeneity of Risk Factors and Antibody Profiles in Epstein‐Barr Virus Genome–Positive and –Negative Hodgkin Lymphoma

Background. Hodgkin lymphoma (HL) tumors that contain the Epstein-Barr virus (EBV) genome may differ etiologically from EBV-negative HL tumors. Methods. A case-case study examining heterogeneity of risk factors between disease subgroups compared personal characteristics and EBV antibodies between 95 EBV-positive and 303 EBV-negative patients with HL. Results. We confirmed previous associations of EBV-positive HL with older age, male sex, and mixed-cellularity (MC) histological subtypes. EBV-positive patients were less educated and more likely to have smoked cigarettes and had more prevalent and higher EBV antibody titers, compared with EBV-negative patients. After adjustment for all independent risk factors, those most strongly associated with EBV-positive HL were histological subtypes (odds ratio [OR] for MC vs. nodular sclerosis histology, 3.2; 95% confidence interval [CI], 1.4–7.2), elevated antiviral capsid antigen level (OR, 3.1; 95% CI, 1.6–6.0), and less education (OR, 0.7; 95% CI, 0.5–1.0). Cigarette smoking and a low anti-Epstein-arr nuclear protein (EBNA) 1 :anti-EBNA-2 ratio were also marginally associated with EBV-positive HL. Conclusions. EBV-positive HL is more common among individuals who have markers of diminished cellular immunity and an abnormal EBV antibody response. EBV appears to participate in the etiology of EBV-positive HL but may not be involved in EBV-negative HL. Hodgkin lymphoma (HL) has long been postulated to have an infectious etiology. The evidence that supports this theory includes clinical symptoms, such as persistent fever, night sweats, weight loss, and lymphadenopathy [1]; the abundance of reactive lymphocytes surrounding the sparse malignant Hodgkin and Reed-Sternberg (HRS) cells of HL tumors [2, 3]; the unusual bimodal age-incidence curve [4]; and previous epidemiological findings that factors related to delayed childhood infection are predictive of HL in young and middle-aged adults [5, 6]. Monoclonal Epstein-Barr virus (EBV) DNA and its gene products can be detected in 25%-50% of HL biopsy samples, in which the virus is localized to the HRS cells [2, 3]. EBV is a ubiquitous human herpesvirus that infects >90% of the worldwide population, usually during childhood, and typically persists latently for life [7]. When primary infection is postponed until adolescence, EBV can cause infectious mononucleosis (IM) in up to half of patients [8, 9]. IM, in turn, is associated with a 3-fold greater risk of subsequent HL [6, 10–16]. Additional evidence that links EBV to HL includes altered antibody titers to EBV antigens, which suggests viral reactivation and enhanced replication, preceding a diagnosis of HL [17–19]. These findings, and the fact that EBV has well-characterized oncogenic properties in vivo, implicate EBV as the major suspect in HL etiology [20]. Although EBV carriers produce antibodies to a variety of lytic and latent EBV antigens, these antibodies appear to offer little or no control over the virus in latently infected cells; rather, immune control is mediated primarily by cytotoxic lymphocytes (CTLs) [21, 22]. However, the level and prevalence of antibodies to lytic and latent viral proteins appear to reflect the prevalence of these proteins expressed in the carrier and, as such, can indirectly gauge the host response to EBV infection. In normal hosts, primary EBV infection triggers the production of antibodies against the viral capsid antigen (VCA) and early antigen (EA), which signifies viral replication, as well as the latent Epstein-Barr nuclear protein (EBNA)-2. Elevated IgG antibody titers against VCA may suggest a relatively severe primary infection or reactivated EBV [23], whereas elevated antibodies against EA are believed to reflect active viral replication and a less efficient host immune response [22, 23]. The resolution of early infection is reflected in the appearance of anti-EBNA-1 and the decline of anti-EBNA-2, which results in an anti-EBNA-1 :anti-EBNA-2 ratio >1.0 in normal carriers [22]. A persistent anti-EBNA-1: anti-EBNA-2 ratio ⩽1.0 is empirically interpreted as a marker of defective host immune control of latent EBV infection [24, 25] and has been detected in patients with HL (N.E.M., E. T.L., K. Dupnik, and B. Birmann unpublished data) and in patients with severe IM, chronic EBV infection, AIDS, and congenital immune deficiency [24]. Because evidence of the EBV genome exists in only a proportion of HL tumors, the question arises of whether EBV-positive and -negative patients share the same etiology. Previous studies have found that EBV-positive HL appears to be more common in males, children and older adults, nonwhite persons, economically less-developed regions, and persons with the mixed-cellularity (MC) histological subtype of HL [15, 26-33]. Few previous studies have interviewed patients to identify additional risk factors for EBV genome positivity among patients with HL [15, 32, 34]. Two small studies [35, 36] evaluated EBV serology and the presence of its genome in HL tumors and found no correlation. We conducted a case-case comparison that evaluated personal, histological, and serological differences between patients with EBV-positive and EBV-negative HL. Study population. In 1997 we initiated a population-based case-control study that included the combined populations of the greater Boston, Massachusetts, metropolitan area and the state of Connecticut, to clarify the role of EBV in the etiology of HL. The patients included in the present study were diagnosed at age 15–79 years, lived in the defined area, and had no reported evidence of HIV infection. Within the study area, 743 eligible patients with HL were diagnosed between 1 August 1997 and 31 December 2001. Permission for contacting patients was sought from the treating physician for each identified patient and was granted for 685 (92%) of 743 patients. Each patient was then sent an initial explanatory letter inviting their participation in the study. Of the eligible patients contacted for interview by study staff, 335 (84%) of 398 patients in Massachusetts and 240 (86%) of 280 patients in Connecticut participated. Ten patients were excluded by study pathologists for not having HL, leaving a total of 565 patients with HL for the study. This research protocol was approved by the institutional review boards and ethics committees of the Harvard School of Public Health, the Yale University School of Medicine, and the Johns Hopkins Medical School. We assume full responsibility for analyses and interpretation of these data. All study participants or, if they were minors, their guardians gave written informed consent at the time of enrollment in the study. Data collection. The explanatory letter was followed by a telephone request for scheduling a structured telephone interview. Next-of-kin interviews were performed for 2 patients, and 1 patient who did not wish to perform a telephone interview completed an abbreviated, mailed questionnaire. The interviews gathered information on several exposures, including personal demographic characteristics, occupation, childhood social environment at age ∼8 years, family structure, parental education and occupation, tobacco use, and personal and family medical history. Data on median household income in census tract and percentage of census tract below poverty level, according to the United States Census, were obtained on the basis of each patient's residential street address [37]. Histopathologic testing. The study pathologists (M.B., R.B.M., and E.G.W.) reviewed all available pathologic material to verify the diagnosis of HL. Archived tissue blocks were retrieved and examined for 411 (71%) of 575 patients. For those patients without adequate, available, or consented diagnostic specimens, the original pathology reports were reviewed to determine whether a diagnosis of HL was probable, possible, unlikely, or unconfirmable. Pathology report-based diagnoses were made for 156 patients (27%), and pathological material was unavailable for 8 (1%). The histopathologic classification of tumors was based on the 1994 Revised European-American Lymphoma Classification of Lymphoid Neoplasms and the 2001 World Health Organization Classification of Hematopoietic and Lymphoid Tumors [38, 39]. These classifications distinguish among several subtypes considered broadly as “classic” HL: nodular sclerosis (NS), MC, lymphocyte depletion, and lymphocyte-rich classic HL. Nodular lymphocyte predominance (NLP) HL is considered to be a disease entity that is distinct from the classic subtypes and was excluded from the present analysis. Study pathologists also confirmed cases of classic HL that could not be further subclassified, including those for which there was an insufficient amount of diagnostic tissue for subtyping, as well as those in which NS and MC HL could not be differentiated, because of early involvement of a lymph node in an interfollicular variant (IFV) pattern. The presence of EBV in HL tissue was determined by in situ hybridization for EBV-encoded RNA transcripts and/or immunohistochemical assay for the viral latency membrane protein-1 in HRS cells [40, 41]. An HL tumor was considered to be EBV positive if results were positive for either assay; a tumor was considered to be EBV negative if both assays were negative or if only a single assay was performed and its result was negative [42]. Where possible, tumor immunophenotypes were classified as typical, cell-like, or cell-like on the basis of immuno-peroxidase reactivity patterns with cell-surface markers. Typical HRS cells were defined as those lacking both and cell markers; cell-like cells were positive for CD20 and/or CD79a, and cell-like cells were positive for CD3 and/or CD45RO. The interpretation of immunohistochemical assays was done by consensus of all 3 study pathologists. Serology. Postdiagnosis blood samples were collected by clinical or study staff, from 469 (82%) of 575 initial patients. Serum samples were promptly processed and stored at 20° before blinded testing at Virolab. The specific antibodies measured were IgG to EBV VCA, EA-restricted (EA-R), EA-diffuse (EA-D), EBNA complex, EBNA-1, and EBNA-2 and IgA to VCA and EA-D. If a patient had detectable anti-EA-R and anti-EA-D IgG, the higher value was taken to be the titer against EA, as in other studies [7]. Nonspecific IgG was measured against the EBV-negative lymphoblastoid cell line BJAB; the 7 patients who had a ratio of titers to EBNA complex versus BJAB of <4.0 were excluded as having nonspecific antibodies. Methods for these assays have been described elsewhere [25]. Titers were expressed as the reciprocal of the highest dilution to register a positive reaction. Detectable titers were defined as either 1:20 anti-VCA IgA or IgG and anti-BJAB IgG and 1:5 for the other antibodies. In accordance with the results of our previous studies [18, 43], EBV antibody titers were dichotomized at the upper 15% (or nearest cutoff) of the distribution among all patients combined. On the basis of these criteria, elevated titers were defined as 5s 1:2560 anti-VCA IgG, 5s 1:20 anti-VCA IgA, ⩾1:160 anti-EA IgG, ⩾1:5 anti-EA-D IgA, ⩾1:1280 anti-EBNA complex, ⩾1:1280 anti-EBNA-1, and ⩾1:160 anti-EBNA-2. An abnormal proportion of antibody titers against anti-EBNA-1 and anti-EBNA-2 was defined as a ratio ⩽1.0. Statistical methods. The case-series design was used to perform a case-case comparison of the relative risk for developing EBV-positive HL with the relative risk for developing EBV-negative HL [44, 45]. Odds ratios (ORs), 95% confidence intervals (CIs), and corresponding values for 2 tests of association were computed using multiple logistic regression. All exposures with >2 levels were analyzed first as categorical variables and then as ordinal variables. Likelihood-ratio tests were performed to determine whether these exposures should be entered into statistical models as single ordinal terms. Where monotonic trends were observed for association with the log odds of EBV-positive versus EBV-negative HL, exposures were coded as ordinal variables in statistical models. After evaluating its linearity of association with EBV-positive HL, age was modeled as a continuous variable, were considered on the basis of and tests that compared models with and without additional variables. the of associations with EBV-positive HL, the were by additional factors to differences in the patient The presence of was by of tests for the of in models. were compared by that and of antibody were compared by of all with Statistical analyses were performed with were available for histopathologic from patients with HL. with 8 additional patients who tumor but pathology reports information on previous tests for EBV and patients with HL, there were patients for on EBV were Of 95 positive for the presence of the EBV genome in tumor 303 negative for the EBV and 3 had The distribution of demographic characteristics among the 95 EBV-positive and 303 EBV-negative patients is in 2 the associations of risk factors with EBV genome for age and only or for age, sex, education level, and smoking In these EBV-negative patients as the EBV-positive patients were more as likely as EBV-negative patients to be older 95% CI, patients were compared with those EBV-positive patients were also more likely EBV-negative patients to be of MC (OR, 95% CI, or (OR, 3.2; 95% CI, histological and there was a higher proportion of male patients (OR, 95% CI, in risk of EBV-positive HL by was there was a low proportion of nonwhite patients in our study. EBV genome did not appear to be associated with tumor of demographic characteristics among Epstein-Barr virus (EBV) and patients with lymphoma and odds ratios and 95% confidence intervals for association with risk of Epstein-Barr virus (EBV) vs. Hodgkin lymphoma Previous case-control studies have found an association between HL and markers of during In our case-case we found that these factors did not differ between EBV-positive and EBV-negative patients, the of EBV-positive patients had levels of education those of EBV-negative patients On EBV-positive patients were also less educated EBV-negative patients (OR, 95% CI, The of several other risk factors for HL did not differ between EBV-positive and EBV-negative patients. Although a of and a family of are risk factors for HL those did not distinguish between EBV-positive and EBV-negative HL, as we and previous in which have also been identified as risk factors for HL were not associated with EBV genome positivity in our A of smoking was more common among patients with EBV-positive HL: the for having smoked of cigarettes during for EBV-positive versus EBV-negative patients was CI, The positive association between smoking and EBV-positive HL was of age, sex, and other demographic factors, including of previous EBV infection, as by detectable IgG antibody to VCA, was present in of patients with serological results for patients were patients, with a age of versus All patients had either EBV-negative tumors or tumors without available EBV genome were no between EBV infection and tumor histology, sex, or other were patients in our study with both tumor EBV genome and serological of had EBV-positive HL and had EBV-negative HL. The differences between EBV-positive and EBV-negative patients in prevalence of detectable antibody and in median antibody titer are in EBV-positive patients were more likely EBV-negative patients to have detectable IgA antibody against VCA (OR, 95% CI, (OR, 95% CI, and anti-EA (OR, 95% CI, IgG, we for age, sex, education level, and smoking All EBV-positive patients, but not all EBV-negative patients, had detectable IgG titers to VCA and EBNA complex for The titers of all antibodies anti-EBNA complex and anti-EBNA-1 were more among EBV-positive EBV-negative patients, whereas the median anti-EBNA-1: anti-EBNA-2 ratio was for EBV-positive patients for all of detectable Epstein-Barr virus (EBV) antibody and median titer by tumor EBV genome We examined whether the prevalence of elevated antibody titer was associated with the risk of compared with HL EBV-positive HL samples were more likely to have elevated titers against VCA and and as well as an abnormal anti-EBNA-1: anti-EBNA-2 After for antibodies to VCA, and the anti-EBNA-1: anti-EBNA-2 the associations with an anti-VCA IgG and anti-EBNA-1: anti-EBNA-2 ratio ⩽1.0 The for anti-VCA IgG was CI, and that for an anti-EBNA ratio ⩽1.0 was CI, with of EBV-positive and of EBV-negative patients having an anti-EBNA ratio ⩽1.0. odds ratios and 95% confidence intervals for association between elevated Epstein-Barr virus (EBV) antibody titer and risk of EBV vs. Hodgkin by age, sex, and histological subtype of HL that the antibodies predictive of EBV-positive elevated anti-VCA IgG and an abnormal anti-EBNA-1: anti-EBNA-2 did not differ between The only was between age and elevated anti-VCA IgG, with a higher among older patients for heterogeneity by The for elevated anti-VCA IgG was CI, among patients years, whereas was CI, patients was no heterogeneity of association by or tumor We combined our and serological for all factors associated with risk of EBV-positive HL After for age, sex, education level, smoking tumor histology, elevated anti-VCA IgG, and an anti-EBNA-1: anti-EBNA-2 ratio we found that histological elevated anti-VCA IgG, and education level had the associations with EBV-positive HL. The for the MC versus the NS histological was CI, 1.4–7.2), whereas that for elevated anti-VCA IgG was CI, 1.6–6.0), and that for patient education level was CI, 0.5–1.0). Cigarette smoking and an abnormal anti-EBNA-1: anti-EBNA-2 ratio also marginally associated with EBV-positive HL. The for smoking was CI, and that for an ratio ⩽1.0 was CI, odds ratios and 95% confidence intervals for association between independent risk factors and the risk of Epstein-Barr virus (EBV) vs. findings that patients with EBV-positive HL are more of a and with MC histological are with those of studies that were based on a variety of other populations worldwide these findings suggest that EBV genome positivity in HL tumors may be with a relative of immune immunity is to decline with age and poverty In male patients have diminished cellular immunity relative to patients could to defective and control of EBV infection, to viral higher antibody titers against and, a higher risk of cell and the of HL findings of or differences in EBV genome positivity may reflect in or could a in that is to The low proportion of EBV-positive patients in our compared with that have reported in could be to our relatively and population, as well as to previous associations with EBV genome we found that smoking was more common among EBV-positive EBV-negative patients. A case-control study reported a risk of MC HL among a that may our because the prevalence of EBV positivity was higher among MC tumors. is to both and immunity which in turn, the reactivation of EBV and to the of EBV-positive HL. In that the of smoking and EBV infection may in HL, on the basis of evidence that both smoking and EBV HRS cells the which a for and EBV to HRS cell by risk factors for HL to to EBV-positive and EBV-negative HL. of IM, family of and to products were not associated with EBV-positive HL. These findings suggest that the of EBV-positive and EBV-negative HL share at in because factors or the risk for HL of EBV genome of no association between and EBV-positive HL with the results of a study by who reported a case-case of CI, they compared a of between EBV-positive and EBV-negative patients with HL. The same study found a association between and EBV-positive patients versus control patients but not EBV-negative patients versus control patients for A was reported by who detected a elevated for of in EBV-positive patients with HL but not EBV-negative patients, compared with control patients. However, their case-case for patients with HL who were was CI, this not differ from our of CI, for the same age The association between and EBV-positive HL in our study also with the results of who found an that was elevated but not serological results that host antibody response to EBV between EBV-positive and EBV-negative patients with HL. EBV-positive patients were more likely to be EBV in of the patients had EBV-positive tumors. In EBV-positive patients were more likely EBV-negative patients to have both detectable and elevated EBV antibody The antibody titers that between EBV-positive and EBV-negative samples were against both lytic and latent virus antigens, which suggests an immune response to the virus in both of its A ratio of titers to and which has been found to be a marker of defective immunity against EBV [24, was strongly associated with EBV-positive HL. our serological results our interview which suggests abnormal immunity in patients with EBV-positive HL. The case-case design a of over the case-control study The for was in our because the participation was and enrollment was performed before the of tumor EBV genome which little of participation between EBV-positive and EBV-negative patients. was also because patients were to whether their tumors were EBV This with a case-control in which individuals may to participate in a study and/or information according to their disease Because all blood for our study were collected and the diagnosis of HL, antibody levels may have disease and may not reflect immune before the of HL. However, our findings are with those from a study of the in which an ratio ⩽1.0 was predictive of subsequent EBV-positive HL these results suggest that an abnormal response to EBV infection, and immune control of the EBV to an role in the of HL in EBV-positive patients. Although our results the that patients with EBV-positive HL have abnormal relative to patients with EBV-negative HL, a of models for HL EBV-positive and EBV-negative HL may share the same with of the viral genome from EBV-negative HRS cells viral In this of HL the viral genome could be from EBV-negative tumors by an active or the virus could or However, that a small proportion of patients with HL have no serological evidence of infection with EBV and that all patients with tumor in our study were EBV negative that EBV infection is not a for HL. to defective or EBV in HL have been These findings against the involvement of EBV in both EBV-positive and EBV-negative HL. EBV-negative HL could be by a infectious or its could be independent of an infectious The association of HL with markers of to common childhood including low low and strongly suggests that both EBV-positive and EBV-negative HL are associated with an infectious [5, However, for evidence of involvement of other in HL have negative results our results suggest that diminished cellular immunity and an abnormal EBV response are associated with a greater risk of EBV-positive HL, whereas delayed childhood infection is a risk for HL in EBV an role in EBV-positive infection or other a role in EBV-negative HL to be used in this study were obtained from the Massachusetts and the Connecticut which is of the Connecticut of Public In Massachusetts, patients were identified from the Medical Medical Medical and Boston, Medical Harvard and Medical Medical of Massachusetts Massachusetts and Massachusetts Medical Medical Medical University of Massachusetts Medical and In Connecticut, patients were identified from the Connecticut of Public Health and Medical and Medical and We a of to and School of Public for of the and its and School of Public for For at their study we are to and Hopkins Medical and University and School of and and We also and School of Public for of the we the of at all