Is Hodgkin Lymphoma Associated with Hepatitis B and C Viruses? A Systematic Review and Meta-analysis

Hodgkin lymphoma (HL), a B-cell lymphoid malignancy affecting the lymphatic system, accounted for an estimated 83,087 malignancies diagnosed in 2020, representing 0.43% of all incident cancers worldwide (1). A hallmark of HL epidemiology is the variation in the age distribution at diagnosis; high-income regions display a bimodal distribution with distinct peaks in incidence observed between ages 15 and 34 and in those older than age 50 (2). The Epstein–Barr virus (EBV), a member of the Herpesviridae family, was classified as a cause of HL in 1997 (3, 4). Apart from the EBV, the etiology of HL remains largely unknown.

Hepatitis B virus (HBV) and hepatitis C virus (HCV) have been identified in lymphoid tissues and are known to exert direct effects on lymphoid cells (5, 6). In 2009, the International Agency for Research on Cancer concluded that chronic HCV infection can cause non-Hodgkin lymphoma (7). A positive association between chronic HBV infection and non-Hodgkin lymphoma has also been observed (8), suggesting a pathophysiologic link between hepatitis virus infections and lymphatic neoplasms. Because hematologic malignancies represent a heterogeneous group of cancers that share biological and morphologic features, chronic HBV/HCV infections may also influence the development of other lymphomas (8, 9). Although mechanistic pathways have not been elucidated, evidence suggests that neoplastic transformation may result from indirect oncogenic activity of HBV and HCV on B cells, resulting in chronic antigen stimulation (10).

HBV chronically infects roughly 5% of the world's population (11). Approximately 95% of those initially infected with HBV during infancy progress to chronic infection compared with just 5% of individuals initially infected in adulthood (12). Chronic HCV infection has an estimated prevalence of 1% globally (13). Of those positive for HCV, 75%–80% will develop chronic HCV infection (13, 14). Evidence of HBV and HCV detection in cancer tissues similar to HL, coupled with elevated HL incidence among those positive for these viruses (15, 16), provides a basis to assess the potential role of HBV and HCV infections in the development of HL (8).

A meta-analysis of 10 studies conducted by Dalia and colleagues (2015) reported a pooled odds ratio (OR) of 1.54 [95% confidence interval (CI), 1.05–2.27] for the association between HBV and HL (17). Several studies, including cohort studies (15, 18, 19), were subsequently published after Dalia and colleagues' search date. For the association between HCV and HL, Dal Maso and Franceschi (2006) reported a pooled relative risk (RR) of 1.46 (95% CI, 1.00–2.13) based on four studies (20), three of which reported nonstatistically significant elevated risks (21–23), and one a nonsignificant reduced risk (20). All four studies were conducted in Italy or Spain. Thus, an up-to-date systematic review including studies conducted outside of Southern Europe is warranted. For these reasons, we performed a systematic review and meta-analysis to summarize the current evidence and evaluate whether each HBV and HCV are associated with the development of HL.

The protocol for this review was registered with the PROSPERO international prospective register of systematic reviews (registration number: CRD42020188483) and is accessible at https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=188483. We adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (24).

With the support of a Health Sciences Librarian with expertise in conducting systematic literature searches, we developed a search strategy using a combination of Medical Subject Headings (MeSH) and keywords for HBV, HCV, HL, and their hypothesized associations. We searched the Embase, MEDLINE, and Web of Science databases from each database's inception to March 9, 2021, without date or language restrictions (the full search strategies appear in Supplementary Table S1). Non-English texts were translated with a free web-based software (DeepL).

We included observational studies (cohort, case–control, and nested case–control) that reported either a measure of association for HBV or HCV and HL or sufficient data to allow us to calculate ORs, RRs, or hazard ratios (HR). We also included in the systematic review, but not in the meta-analysis, studies that reported a standardized incidence ratio (SIR). Conference abstracts, proceedings, and posters were excluded. Records had to have at least 10 primary HL cases that had not started cancer treatment, and comparator groups had to be free of hematologic and HBV/HCV-associated malignancies (i.e., liver and bile duct cancers). Studies relying on self-reported hepatitis status were excluded. Additionally, studies utilizing HBV detection methods other than hepatitis B surface antigen (HBsAg) or medical records were excluded. For the main analyses, studies with a Newcastle-Ottawa Scale (NOS) score of three or less, indicating low methodological quality, were excluded, but were retained for a sensitivity analysis. When study populations overlapped, we included the study with the highest number of HL cases.

Two reviewers (CM and KV) independently screened record titles and/or abstracts for relevance using Rayyan (25). Full texts were retrieved for records deemed potentially eligible by either reviewer. After piloting on 10 studies, a standardized data extraction form was used to record information including year of publication, study design and location, inclusion and exclusion criteria for enrolling participants, number of participants, demographic characteristics, hepatitis diagnostic method and positivity definition, HL diagnostic method, and the reported crude and adjusted measures of association, including adjustment variables. We also extracted additional information for case–control (ascertainment period, source of cases and controls, and matching variables) and cohort (source of the cohort and comparators, and person-time measures of follow-up) studies. CM extracted the data that were verified by ZG and KV; discrepancies were resolved through discussion. If exposure or outcome information was missing, the corresponding author was contacted.

Two authors (CM and ZG) independently assessed methodological study quality using the NOS. The NOS is a “star”-based rating system that evaluates participant selection, comparability of cases and controls, and outcome assessment in nonrandomized studies (26). NOS scores of 7–9 (nine stars is the maximum that can be assigned) indicate high quality, scores of 4–6 fair quality, and scores of 0–3 low quality (26, 27).

We calculated pooled RRs with 95% CIs for the association between HBV infection and HL, and between HCV infection and HL. Because HL is a rare malignancy, the OR can approximate the RR, thereby allowing estimates from case–control and cohort studies to be pooled (28). If at least three studies reported HRs for a given exposure, these were pooled separately, otherwise, HRs were pooled with RRs. We assessed heterogeneity across studies using the Cochran Q test (29) with its corresponding P value and the index of consistency (I²) statistic (30). We used a fixed effects model if the I² indicated low heterogeneity (≤25%), otherwise random effect pooled estimates were used for the main analysis. The results of both fixed and random effects models are presented for comparison. For studies not reporting a measure of association, the open-source software, OpenEpi, was used to calculate ORs or RRs (31). If studies had zero HBV or HCV positive cases and/or controls, 0.5 persons or 0.5% prevalence, whichever produced lower hepatitis prevalence, was imputed before calculating the OR or RR.

To explore sources of heterogeneity, we conducted subgroup analyses based on the following covariates defined a priori: study design (case–control or cohort), region (African, Americas, Eastern Mediterranean, European, Western Pacific) as classified by the World Health Organization (32), hepatitis prevalence (HBV: high ≥8.0%, high intermediate 5.0%–7.9%, low intermediate 2.0%–4.9%, low <2%, varies; HCV: high 2.9%–6.7%, high moderate 1.3%–2.9%, low moderate 0.8%–1.3%, low <0.8%, varies) as defined by the Centers for Disease Control and Prevention (33, 34), hepatitis detection method [HBsAg, record-based, HCV antibody test (anti-HCV), anti-HCV with nucleic acid testing (NAT)], and the source of controls within case–control studies (blood donors, healthy individuals, hospital-based, population-based, and registry-based), and whether possible confounding variables were considered or not (i.e., in the design by matching and/or in the analysis by adjusting).

In addition to the subgroup analyses, we performed sensitivity analyses to assess whether the pooled estimates were influenced by the addition of studies conducted among immunocompromised populations (i.e., people living with HIV and organ transplant recipients), and then studies with low NOS scores. Leave-one-out analyses were used to assess the effect of individual studies on the overall pooled estimates. To explore publication bias, we generated funnel plots of the study effect sizes (Hedges' g) against the standard error of study estimates and used Egger's test to determine if study estimates were related to the size of included studies (35, 36). We performed the meta-analyses with R statistical software (version 3.6.1; ref. 37) using the meta (38) and metafor (39) packages.

Our search identified 2,730 unique records, of which 204 full texts were reviewed (Fig. 1). No record meeting the inclusion criteria was published in a non-English language. Among the 30 records that met the inclusion criteria for the systematic review component, seven provided data on HBV alone, 18 reported on HCV alone, and 5 reported on both HBV and HCV (Table 1). The majority were case–control studies (66.7% of HBV and 78.2% of HCV studies). Most (66.7%) HBV studies used HBsAg detection and among the HCV studies, 48.3% used anti-HCV with NAT detection. A summary of studies included in the systematic review is provided in Table 1 and the individual details of cohort and case–control studies are reported in Supplementary Tables S2 and S3, respectively.

Among the 21 studies that were meta-analyzed, 9 reported on HBV and 15 reported on HCV. Only four reported the age and sex of HL cases, and no study reported a separate measure of association by age group or sex. Only one study directly assessed EBV tumor status and reported that all HL cases were EBV positive (40). No studies investigated HBV-HCV coinfection as a risk factor for HL. Two studies focused on populations with underlying immunodeficiencies [people living with HIV infection and HBV and/or HCV (ref. 18), and organ transplant recipients with HCV (ref. 41)]. We had to calculate the measure of association for 13 studies (Supplementary Tables S2 and S3 indicate which studies), 5 (40, 42–45) of which had no hepatitis-positive HL cases and were imputed.

Two cohort studies comparing HL incidence to the general population reported SIRs of 2.76 (95% CI, 1.37–4.95; ref. 16) for HBV, and 21.7 (95% CI, 6.70–67.90; ref. 15) for HCV; however, this HCV study examined the effect of HIV positivity compared with the general population. One cohort study conducted among individuals registered for opioid substitution therapy reported an SIR of 1.44 (95% CI, 0.39–3.69; ref. 46) for HCV. Two HBV cohort studies (19, 47) and one HCV cohort study (48) reported HRs and were thus pooled with studies that reported RRs.

Six of the eight (75.0%) included cohort studies and 14 of the 23 (61.0%) case–control studies were rated high methodological quality (Supplementary Tables S4 and S5). The most common reason for low or fair study quality was not reporting response/participation rates in the exposure assessment.

Overall, the pooled RR was elevated but not statistically significant from nine studies that reported on the association between HBV infection and HL (Fig. 2A). No statistically significant association was found between HBV infection and HL development among subgroups that had at least two studies (Table 2). Pooling estimates from 15 studies that assessed HCV infection showed no evidence of an association (Fig. 2B). Forest plots for corresponding subgroupings are reported in Supplementary Figs. S1–S12. We observed a statistically significant association in the three HCV studies conducted in the Western Pacific region (Table 3; Supplementary Fig. S4) and for the six HCV studies that used hospital-based controls (Table 3; Supplementary Fig. S10). Among HBV studies, only one study did not account for possible confounding by matching in the design and/or adjusting in the analysis (Supplementary Fig. S11), whereas among the HCV studies, pooling the eight studies that did not account for possible confounding showed increased risk (Supplementary Fig. S12).

Leave-one-out analyses showed that omitting each Tian and colleagues (2020), Becker and colleagues (2012), and Engels and colleagues (2010) from the main HBV analysis produced statistically significant associations for HL (RR: 1.43; 95% CI, 1.01–2.01, RR: 1.51; 95% CI, 1.05–2.17, and RR: 1.52; 95% CI, 1.05–2.21, respectively; Supplementary Fig. S13A); however, for HCV, omitting individual studies did not alter the conclusions (Supplementary Fig. S13B). We did not observe asymmetry in the funnel plots for HBV and HCV, as confirmed by Egger's test (HBV: P = 0.96, HCV: P = 0.19; Supplementary Fig. S14).

Adding one cohort study comprised of people living with HIV (18) to the HBV meta-analysis yielded a statistically significant elevated risk of HL (Table 2; Supplementary Fig. S15). Adding this study (18), and one other among organ transplant recipients (41), to the HCV meta-analysis did not alter the pooled estimate (Table 3; Supplementary Fig. S16). Adding one low NOS score study to the HBV analysis did not alter the pooled estimate and its associated CIs (Table 2; Supplementary Fig. S17); however, for HCV, adding four low NOS score studies produced a statistically significant elevated RR of 2.20 (95% CI, 1.13–4.27; Table 3; Supplementary Fig. S18).

Compared with previous meta-analyses that identified positive associations between each of HBV and HCV seropositivity with HL (17, 20), the results of our meta-analysis depict an inconclusive association with HBV and HL based on nine studies, and HCV and HL based on 15 studies. These differences are likely explained by the inclusion of studies published after the previous meta-analyses search dates. Specifically, our analyses included two HBV (19, 49) and 12 HCV (42, 43, 45, 48, 50–57) studies not included in previous systematic reviews and meta-analyses.

Studies included in our analysis used different approaches to assess the presence of hepatitis—where each testing method has its own limitations. First, record-based hepatitis assessment methods do not encompass systematic HBV or HCV screening and, therefore, could underestimate the number of exposed individuals (58). Second, even the gold-standard method for detecting HBV infection, HBsAg, does not account for occult HBV infection (i.e., low-level HBV-DNA in serum; ref. 59). Therefore, individuals with an occult infection testing negative for HBV markers in sera could be positive for HBV-DNA, which may result in underestimating the number exposed. No HBV study included in the main analysis incorporated HBV-DNA testing in their hepatitis positivity definition. Finally, individuals testing anti-HCV positive can include those without chronic HCV infection at the time of testing, such as those who spontaneously clear their infection or those cured by antiviral treatment. Five included studies used anti-HCV testing alone. Although their pooled RR was higher than the pooled estimate for studies based on anti-HCV with NAT, the estimates did not differ statistically.

Although our findings do not support an association between each HBV and HCV infection with HL in the general population, the available data did not allow us to meta-analytically assess this association among immunocompromised populations. We identified one study of people living with HIV. This study reported a statistically significantly elevated risk for HL among those with HBV-HIV coinfection; however, a similar association was not observed in those with HCV-HIV coinfection (18). Increased HL incidence in people living with HIV compared with the general population has been reported in the United States (SIR: 5.6; 95% CI, 3.90–7.80; ref. 60) and among HCV-HIV coinfected individuals in Spain (SIR: 21.7; 95% CI, 6.7–67.9; ref. 15). Future research focused on the possible role of HBV and HCV in HL development among immunocompromised populations could provide insight into the etiology of HL beyond the role of the EBV.

The quality of included studies varied substantially. When the four low NOS score studies were included in HCV analysis, the pooled RR indicated a statistically significant doubling of risk. Low-quality studies were all performed in the Eastern Mediterranean regions, three of the four used blood donors as controls, and all inadequately described the comparability of cases and controls. Blood donors tend to be healthier compared with the general population, and, as part of the blood donation process, they may be screened for the hepatitis viruses. This would result in the control group having lower HBV and/or HCV prevalence, and thus bias the measure of association away from the null. Although we excluded low-quality studies, the main analysis included three (55, 57, 61) studies that used blood donors as controls, and another eight (21, 23, 34, 49–52, 62) studies used hospital-based controls. Among HCV studies using hospital-based controls, we observed an elevated and statistically significant pooled RR, though it is possible that selection bias influenced this result.

As a systematic review and meta-analysis, this study is limited by the availability and quality of the evidence. Although 15 studies reported on HCV, only nine studies included HBV—thus limiting inferences that can be made overall and especially among subgroups. Importantly, most studies assessed HL as a subgroup among other lymphomas or other cancers. Because HL represents a smaller proportion of lymphomas (compared with non-Hodgkin lymphomas) and an even smaller proportion among many different types of cancers, demographic characteristics reported for the relevant study population of patients with HL could not be deduced. For example, HL cases comprised as little as 0.02% (47) of the larger study population for which age and sex characteristics were reported. This prevented us from performing subgroup analyses by age and sex—variables known to be important to describing the epidemiology of HL (2). Only six (21, 22, 41, 44, 45, 63) out of the 30 studies reported sex characteristics specific to HL, and no studies assessed this relationship in younger populations. We were also unable to assess the potential influence of hepatitis treatment on the development of HL as only two of the 15 HCV studies and one of the nine HBV studies were published after the introduction of effective treatment. Of the two HCV studies, only one (52) reported cancer diagnosis dates and, therefore, subgroup analysis was not possible. Additionally, the one study that assessed EBV tumor status found that all HL cases were EBV positive (40) and therefore no assessment of coinfection could be made. However, our study has several strengths, including adherence to best practices for conducting systematic reviews (i.e., systematic search of three databases, independent review, and quality assessment), restriction of the main analysis to fair- and high-quality studies, and subgroup and sensitivity analyses to provide insight into factors that may affect a potential association.

In conclusion, our findings demonstrate an inconclusive relationship between HBV and HCV in the development of HL. More research and detailed reporting will be important in assessing if age, sex, HIV status, and EBV tumor status influence the potential role of HBV and HCV infection in the development of HL.

No disclosures were reported.

The authors would like to thank Genevieve Gore, Liaison Librarian at McGill University, for her guidance in developing the search strategy. This work was partially funded by the Cancer Research Society via an endowment to the Division of Cancer Epidemiology at McGill University. It was also supported by funding from the Canadian Institutes of Health Research (grant FDN-143347 to E.L. Franco).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.