Plasma Epstein-Barr virus DNA predicts outcome in advanced Hodgkin lymphoma: correlative analysis from a large North American cooperative group trial

Epstein-Barr virus (EBV) is a gammaherpesvirus present in Hodgkin-Reed-Sternberg cells of some cases of Hodgkin lymphoma (HL). An increased incidence of EBV(+) HL is seen in males, young children, older adults, Hispanics, mixed cellularity or lymphocyte depleted histologic subtypes, and persons from economically developing countries.^1-3  Viral nucleic acid (EBER) in situ hybridization (ISH) on tissue sections is a standard technique for determining EBV status in HL.^4,5 

Plasma or serum EBV-DNA has been shown to be detectable by polymerase chain reaction (PCR) in patients with EBV(+) HL and to have good concordance with EBV status as determined by EBER-ISH on tumor tissue.^6-8  By contrast, measurement of EBV-DNA in peripheral blood mononuclear cells has not been shown to correlate well with EBER-ISH or disease burden in HL.^7-9 

A small study showed plasma EBV-DNA to be detectable in EBV(+) HL patients with active disease, but not in EBV(+) HL patients in remission or those with EBV(-) tumors.⁸  Others have reported an association between detectable plasma EBV-DNA and higher stage disease.¹⁰  A recent study of patients with newly diagnosed or relapsed HL found pretreatment plasma EBV-DNA to be associated with EBV status by EBER-ISH and noted that plasma EBV-DNA levels declined with treatment.¹¹ 

Herein, we examine the relationship in HL between plasma EBV-DNA measurement and EBV tumor status and explore the utility of plasma EBV-DNA as a prognostic marker at diagnosis and 6-month follow-up.

All patients had previously untreated, locally extensive (mediastinal mass greater than one-third of the intrathoracic diameter on posteroanterior chest radiograph) or advanced (stage III or IV) histologically proven classical HL. Patients were prospectively enrolled from April 1999 to June 2006 in a multicenter, phase III randomized controlled clinical trial comparing 2 treatment regimens (doxorubicin, bleomycin, vinblastine, dacarbazine vs Stanford V) as first-line therapy for HL (Eastern Cooperative Oncology Group 2496).¹²  Beginning in 2003, plasma specimens were collected at baseline, during treatment, and at follow-up. The research protocol was approved by the institutional review boards or research ethics committees at participating sites. This study was conducted in accordance with the Declaration of Helsinki.

A tissue microarray was constructed from available formalin-fixed, paraffin-embedded tissue blocks. The array included duplicate 1.5-mm diameter cores of tumor specimens. ISH for EBER was performed using the INFORM EBER probe (Ventana, Tucson, AZ). Slides were stained on an automated stainer (Ventana Benchmark XT) using the Ventana ISH/iView Blue detection kit. A known positive control was used. Specimens with Hodgkin-Reed-Sternberg cells with nuclear staining were considered positive. Hodgkin-Reed-Sternberg cells were identified by morphology and immunohistochemistry for CD30 (clone BerH2, Dako North America; dilution 1:30) performed on an automated stainer (Ventana Benchmark XT) using a multimer detection kit (UltraView Universal DAB).

Blood was collected in heparinized tubes and shipped overnight at ambient temperature. Specimens collected at baseline (pretreatment) and at month 6 were assayed and analyzed. Plasma was separated by centrifugation and DNA was isolated from 250 µL of plasma using the QIAamp DNA blood mini kit (Qiagen Inc, Valencia, CA) according to manufacturer instructions. A primer pair and probe corresponding to the BamH-W region of the EBV genome (5′-CCCAACACTCCACCACACC-3′, 5′- TCTTAGGAGCTGTCCGAGGG-3′, 5′-(6-FAM)CACACACTACACACACCCACCCGTCTC (BHQ-1)-3′) were used. Namalwa DNA (Namalwa cell line genomic DNA, ATCC #CRL-1432) was used for calibration.

A receiver operating characteristic curve was used to determine the cutoff for plasma EBV-DNA with optimal sensitivity, specificity, and concordance with tumor EBV status by EBER-ISH.

Fisher’s exact test was used to compare categorical variables between patient groups. To evaluate age as a continuous variable, the Wilcoxon rank sum test was used for comparisons between 2 groups and the Kruskal-Wallis test was used for comparisons among 3 or more groups.

Failure-free survival (FFS) was defined as the time from treatment randomization to progression, relapse, or death. The Kaplan-Meier method¹³  was used to estimate FFS and the log-rank test was used to compare the outcomes between groups. To evaluate plasma EBV-DNA as a continuous variable, values were log transformed (a value of 1 × 10⁻⁸ was added to all values before log transformation to handle zeros) and standardized to have a standard deviation of 1. After univariate assessment to identify factors with statistical and biological relatedness to outcome, a multivariate Cox proportional hazards model was constructed to evaluate the associations between covariates of interest, including pretreatment plasma EBV-DNA status, International Prognostic Score (IPS), histology, and treatment arm, and the outcome of interest, FFS. Pairwise multivariate Cox proportional hazards regression models were also constructed to test the independent prognostic utility of pretreatment plasma EBV-DNA status in combination with each component of the IPS.

Statistical analyses were performed by H.L. and F.H. using the SAS program, version 9.2 (SAS Institute, Cary, NC), and 2-sided P values less than .05 were considered significant. Prism, version 5, was used to generate Figures 1 and 2.

The clinical trial enrolled 794 eligible patients between 1999 and 2006. Tumor specimens from 315 patients were available for the tissue microarray. Pretreatment plasma specimens were available from 274 patients. There were 116 patients whose diagnostic specimens were included in the tissue microarray and who also had pretreatment plasma specimens. Among these were 24 patients whose tumors were EBER-ISH(+) and 92 patients whose tumors were EBER-ISH(-). For EBER-ISH(+) patients, the median pretreatment plasma EBV-DNA value was 3161 viral copies/100 µL, whereas for EBER-ISH(-) patients, the median pretreatment plasma EBV-DNA value was 0 viral copies/100 µL (Figure 1). Receiver operating characteristic analysis indicated that a plasma EBV-DNA cutoff of >60 viral copies/100 µL optimized the concordance with EBER-ISH. Using this value, plasma EBV-DNA quantification yielded a 96% concordance, 92% sensitivity, and 97% specificity for EBV status by EBER-ISH (Figure 2). This cutoff was applied to both pretreatment (n = 274) and month 6 (n = 132) plasma specimens to designate each as plasma EBV(+) or plasma EBV(-).

Demographic, histologic, and prognostic characteristics for all eligible patients in the clinical trial (n = 794), patients with tumor specimens for EBER-ISH (n = 315), and patients with pretreatment plasma specimens (n = 274), with comparisons between pretreatment plasma EBV(+) and plasma EBV(-) patients, are shown in Table 1. Patients ranged in age from 16 to 83 years, with no difference between plasma EBV(+) and plasma EBV(-) groups. There were differences in the histologic subtypes based on pretreatment plasma EBV status, with plasma EBV(+) patients having more mixed cellularity subtype (24% vs 9%), more classical HL not further classifiable subtype (24% vs 11%), and fewer nodular sclerosis subtype (43% vs 76%). The proportion of patients with each poor prognostic factor of the IPS are also shown in Table 1, with comparisons between plasma EBV(+) and plasma EBV(-) groups.

FFS estimates did not differ between EBER-ISH(+) and EBER-ISH(-) groups, log-rank P = .43 (Figure 3A). By contrast, pretreatment plasma EBV(+) patients, stratified by the cutoff of 60 copies/100 μL plasma, had inferior FFS compared with pretreatment EBV(-) patients, log-rank P = .009 (Figure 3B). We chose the plasma cutoff of 60 copies/100 µL for the FFS outcome analysis because this threshold had biological justification with optimized concordance with EBV tumor status by EBER-ISH. Given the discrepancy in FFS outcomes by these 2 EBV determination techniques, we performed additional univariate Cox analyses of FFS outcomes using arbitrary cutoffs of 0, 50, 100, and 200 viral copies/100 µL plasma as well as plasma EBV-DNA as a continuous variable. Pretreatment plasma EBV-DNA positivity was at least marginally significantly associated with inferior FFS using any of these cutoffs or as a continuous variable (supplemental Table 1).

Given the unexpected finding that pretreatment plasma EBV-DNA emerged as a prognostic variable for FFS outcomes, whereas EBER-ISH status did not, further investigations into baseline prognostic characteristics of these patient groups were undertaken. Patients with plasma specimens only (n = 158), EBER-ISH tissue specimens only (n = 199), both tissue and plasma specimens (n = 116), and neither tissue nor plasma specimens (n = 321) did not differ by age, in the proportion with any of the 7 components of the IPS, or in 3-year FFS estimates (supplemental Table 2). For the 116 patients with both plasma and tissue specimens, the hazard ratio for FFS was 1.9 (95% CI 0.9-4.1, P = .09) by plasma EBV-DNA status (positive vs negative, cutoff of 60 copies/100 μL plasma) and was 1.4 (95% CI 0.6-3.2, P = .37) by EBER-ISH status (positive vs negative).

Multivariate analysis (Table 2) showed that pretreatment plasma EBV positivity was associated with inferior FFS with a hazard ratio of 2.0 (95% CI 1.2-3.5, P = .01), after adjusting for IPS, treatment arm, and histology. Each component of the IPS was evaluated separately with pretreatment plasma EBV-DNA status in pairwise Cox proportional hazards models. In each model, plasma EBV-DNA positivity remained an independent predictor of inferior FFS (Table 3).

Having established the concordance between plasma EBV-DNA and EBER-ISH status, we next evaluated plasma EBV-DNA as a tumor marker on follow-up using plasma specimens obtained 6 months after the initiation of therapy. In keeping with the earlier analyses, we again used the cutoff of 60 copies/100 μL plasma. The 3-year FFS estimate for patients who were plasma EBV(+) at month 6 (n = 7) was 48% compared with 79% for patients who were plasma EBV(-) at month 6 (n = 125), log-rank P = .007 (Figure 4). The median FFS for patients who were plasma EBV(+) at month 6 was 1.3 years and has not yet been reached for patients who were plasma EBV(-). Of note, a few of the patients who were plasma EBV(+) at month 6 were pretreatment plasma EBV(-) and/or EBER-ISH(-), with failure events not restricted to those known to be EBV(+) at baseline. The clinical outcomes of the patients in the month 6 analysis, separated into 4 groups based on pretreatment EBV status (by plasma EBV or EBER-ISH) and month 6 plasma EBV-DNA status, are shown in supplemental Table 3.

This is the largest prospective analysis of plasma EBV-DNA reported in HL patients and the only such data from a randomized cooperative group trial. Pretreatment quantitative plasma EBV-DNA determinations closely approximate the results of EBER-ISH in patients with locally extensive or advanced stage classical HL. FFS analyses suggest that pretreatment and 6-month plasma EBV-DNA determinations are candidate biomarkers worthy of further investigation in patients with HL.

The close relationship between EBER-ISH and EBV-DNA in plasma is not unexpected. Studies in undifferentiated nasopharyngeal carcinoma (NPC) and lymphoma, including HL, have suggested that most of the EBV-DNA detected in plasma or serum is tumor-derived.^14,15  Others have also reported correlations between serum or plasma EBV-DNA PCR and EBER-ISH.^6,7  Our plasma assay appears to be more sensitive, perhaps reflecting the choice of primers that target a repeated region of the EBV genome.¹⁶ 

Although pretreatment plasma EBV-DNA and EBER-ISH were 96% concordant, there were differences in clinical outcomes depending on how EBV status was determined. It is worth considering possible explanations for such discrepancies. If EBV status is truly a prognostic factor in HL, it may be that plasma EBV-DNA was more accurate than EBER-ISH in a tissue microarray format in identifying patients with EBV(+) tumors. This may also explain why other studies that have relied on tissue-based techniques to determine EBV tumor status have not consistently demonstrated EBV-related differences in outcomes. Other possibilities also exist, including that plasma EBV-DNA might reflect the presence of virion DNA rather than naked viral DNA released from tumor cells. In patients with HIV, chronic active EBV, or organ transplant, as well as during primary EBV infection, plasma EBV-DNA may reflect the presence of virions rather than genomic viral DNA released from latently infected cells. Some patients with HL may fall into this category and one might imagine that such patients may have inferior outcomes, reflecting immune dysfunction, independent of EBV tumor status. This scenario also provides a possible explanation for discordant cases in which tumor was EBER-ISH(-), whereas plasma was EBV(+).

Our data show pretreatment plasma EBV-DNA to be a predictor of inferior FFS, independent of known poor prognostic factors encompassed in the IPS. Pretreatment plasma EBV-DNA positivity was associated with inferior FFS when evaluated as a continuous variable as well as using a range of cutoffs. It should be noted that the range of EBV(+) pretreatment viral loads was very broad (from 60 to 353 000 copies/100 µL plasma), with treatment failures distributed throughout the range. Plasma EBV-DNA has prognostic significance in NPC, and these measurements are currently used to stage NPC patients and assess treatment responses.^17,18  Additionally, pretreatment plasma EBV-DNA has been shown to be an independent predictor of NPC relapse, with each log elevation of viral DNA copy number at diagnosis corresponding to heightened risk of relapse.¹⁹  In EBV(+) extranodal natural killer/T-cell lymphoma, the use of pretreatment plasma EBV-DNA measurements to assign patients to more intensive vs less intensive therapy is currently being prospectively investigated by other researchers, based on findings that high-plasma EBV-DNA at presentation is an independent predictor of inferior FFS for these lymphomas.^20,21  If our findings in HL are confirmed in a validation sample, assessment of pretreatment plasma EBV-DNA may aid in risk-stratification.

Plasma EBV-DNA measurements have proven useful in detecting relapse in NPC patients in advance of clinical symptoms or other modes of surveillance and monitoring.²²  In a small series of EBV(+) lymphomas, plasma EBV-DNA levels paralleled clinical course, with patients in remission having undetectable plasma EBV-DNA and patients with refractory disease remaining with detectable levels.²⁰  In a recent study of HL patients treated with rituximab-doxorubicin, bleomycin, vinblastine, dacarbazine, those with EBER-ISH(+) tumors had detectable pretreatment plasma EBV-DNA that declined 50-fold or completely disappeared within the first month of therapy, corresponding a to durable remission in each instance.²³  A recently published report of patients with extranodal natural killer/T-cell lymphoma, those with undetectable plasma EBV-DNA posttreatment had superior survival outcomes compared with those with detectable levels at the end of therapy.²¹  Similar findings have been shown here for HL patients, suggesting that plasma EBV-DNA may have utility as a marker of disease status and identify those with a poor treatment response. Furthermore, given that some of the patients who were plasma EBV(+) at month 6 had EBER-ISH(-) tumors, were pretreatment plasma EBV(-), or both, plasma EBV-DNA should be further explored as a prognostic marker for HL patients, regardless of EBV status at diagnosis. Plasma EBV-DNA assessment is an attractive potential marker in HL given that blood collection is very feasible and the diagnostic technology is already available across clinical and research settings. Furthermore, as an alternative or adjunct, plasma EBV-DNA monitoring may be informative in the posttreatment follow-up of patients with HL as a low-risk, high-sensitivity screen for relapsed disease. Investigation into the time course of plasma EBV-DNA clearance, appearance or reappearance of EBV-DNA in plasma, and various clinical outcomes as well as the origin of plasma EBV-DNA in discordant cases is ongoing.

Within the pretreatment plasma EBV(+) subset, nearly one-quarter were histologically unclassifiable beyond the general category of classical HL. This is in contrast to the pretreatment plasma EBV(-) subset with only 11% unclassifiable cases. This difference is intriguing, although the clinical significance is uncertain. There may be histologic differences between the tumors of plasma EBV(+) and plasma EBV(-) patients, rendering the tumors of plasma EBV(+) patients more difficult to classify within World Health Organization guidelines. Further investigation into these differences may be informative and may perhaps be related to infiltrating cell types, such as the presence or absence of tumor-associated macrophages, which have been recently described in HL to be associated with EBER-ISH positivity and/or inferior FFS.^7,24-26 

Although plasma specimens were obtained from less than one-third of patients enrolled in the larger clinical study, similarities in the baseline characteristics between patients with plasma and/or tissue specimens and those without these specimens suggest some generalizability to all clinically eligible patients in the E2496 cohort. We would caution that the results presented here are limited to patients with histologically confirmed classical HL with locally extensive or advanced stage disease. They may not be generalizable to early-stage, favorable HL; HIV-associated HL, HL developing in the posttransplant setting, or other EBV-related lymphomas. Nonetheless, in well-defined populations, pretreatment plasma EBV-DNA may be useful as a surrogate for EBER-ISH and should be considered a potential marker of disease status and treatment response in HL.

In conclusion, pretreatment plasma EBV-DNA is highly concordant with EBER-ISH in classical HL, confirming many previous reports. Pretreatment plasma EBV-DNA appears to have prognostic value and yield information beyond the IPS or its components. Plasma EBV-DNA positivity at month 6 is associated with particularly poor outcomes and may serve as an indicator of the need for further therapy. The prognostic results warrant validation.

This work was supported by National Cancer Institute, National Institutes of Health, and the Department of Health and Human Services (grants CA96888, CA95423, CA21115, CA23318, CA66636, CA16116, CA17145, and CA21076). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.

Contribution: J.A.K. compiled, analyzed, and interpreted the data and authored the manuscript; H.L. performed statistical analysis; L.L.G. and W.H. processed and analyzed laboratory data; M.V.L. provided critical input and assistance in the processing and analysis of laboratory data; F.H. assisted in statistical analysis; K.L.T. and R.D.G. constructed tissue microarrays and performed EBER-ISH; L.I.G., R.I.F., N.L.B., P.S., B.D.C., R.A., T.P.M., B.S.K., and S.J.H. codesigned the companion clinical trial and enrolled patients in the study; and R.F.A. designed the research, assisted in analysis of the data, provided critical insights, enrolled patients in the study, and coauthored the manuscript. All authors approved the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Richard F. Ambinder, Department of Hematologic Malignancies, Johns Hopkins University, 1650 Orleans St, CRB1, Room 389, Baltimore, MD 21287; e-mail: ambinder@jhu.edu.