Article Text

This article has a correction. Please see:


The Scotland and Newcastle epidemiological study of Hodgkin’s disease: impact of histopathological review and EBV status on incidence estimates
  1. R F Jarrett1,
  2. A S Krajewski2,
  3. B Angus3,
  4. J Freeland1,
  5. P R Taylor4,
  6. G M Taylor5,
  7. F E Alexander6
  1. 1LRF Virus Centre, Institute of Comparative Medicine, University of Glasgow, Glasgow G61 1QH, UK
  2. 2Department of Pathology, Northampton General NHS Trust, Northampton NN1 5BD, UK
  3. 3Department of Pathology, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 4LP, UK
  4. 4Department of Haematology, University of Newcastle upon Tyne
  5. 5Immunogenetics Laboratory, University of Manchester, St Mary’s Hospital, Manchester M13 0JH, UK
  6. 6Department of Community Health, Public Health Sciences, University of Edinburgh, Medical School, Teviot Place, Edinburgh EH8 9AG, UK
  1. Correspondence to:
 Professor R F Jarrett
 LRF Virus Centre, Institute of Comparative Medicine, Faculty of Veterinary Medicine, University of Glasgow, Glasgow, G61 1QH, UK;


Aims: The epidemiological and pathological features of Hodgkin lymphoma (HL) are complex. The Epstein-Barr virus (EBV) is consistently associated with a proportion of cases, and these cases are thought to represent a distinct aetiological subgroup of HL. The aim of the present analysis was to determine the age and sex specific incidence of EBV associated and non-associated HL, analysed separately, using data derived from a population based study–the Scotland and Newcastle epidemiological study of Hodgkin’s disease (SNEHD). This study also provided a unique opportunity to evaluate accuracy in the current diagnosis and classification of HL.

Methods: SNEHD analysed consecutive cases of HL diagnosed in the study area between 1993 and 1997. Diagnostic biopsy material was retrieved, EBV status of tumours was determined, and histological review was performed.

Results: In total, 622 cases were eligible for the study, and EBV studies and histopathological review were performed on biopsy material from 537 and 549 cases, respectively. Accuracy in the overall diagnosis of HL and classification of nodular sclerosis HL was good, but diagnosis of HL in the elderly and classification of other subtypes was less reliable. One third of classic HL cases were EBV associated, and age specific incidence curves for EBV associated and non-associated cases were distinct.

Conclusions: Comparison of age specific incidence curves for EBV associated and non-associated HL supports the hypothesis that these are two distinct aetiological entities. Accuracy in the diagnosis of HL is generally good, but certain subgroups of cases continue to present diagnostic difficulties.

  • Hodgkin lymphoma
  • human herpesvirus 4
  • epidemiology
  • histopathological review
  • CHL, classic Hodgkin lymphoma
  • EBER, Epstein-Barr virus encoded RNA
  • EBV, Epstein-Barr virus
  • HL, Hodgkin lymphoma
  • HRS, Hodgkin and Reed-Sternberg
  • IHC, immunohistochemistry
  • IM, infectious mononucleosis
  • LMP1, latent membrane protein 1
  • LDHL, lymphocyte depleted Hodgkin lymphoma
  • LRCHL, lymphocyte rich classic Hodgkin lymphoma
  • MCHL, mixed cellularity Hodgkin lymphoma
  • NHL, non-Hodgkin lymphoma
  • NLPHL, lymphocyte predominant Hodgkin lymphoma
  • NOS, not otherwise specified
  • NSHL, nodular sclerosing Hodgkin lymphoma
  • SNEHD, Scotland and Newcastle epidemiological study of Hodgkin’s disease
  • SNLG, Scotland and Newcastle lymphoma group

Statistics from

Hodgkin lymphoma (HL) is a heterogeneous condition that most probably comprises more than one aetiological entity.1 Recent data suggest that the Epstein-Barr virus (EBV) is associated with approximately one third of cases in developed countries, and this association is believed to be causal.2–4 Data describing the association between EBV and HL have mostly been derived from studies examining archival samples and reporting the proportion of positive samples within subgroups of cases.3–5 It is clear from these studies that EBV is more often associated with mixed cellularity HL (MCHL) than nodular sclerosis HL (NSHL), and that HL in children and older adults is more likely to be EBV associated than is HL in young adults.5–8 The only population based studies that have classified cases according to EBV status have been small or restricted by age or sex.9–12 The epidemiology of HL is complex, and further understanding of the aetiology of this heterogeneous condition will only be gained by performing epidemiological studies with separate analyses of EBV associated and non-associated cases, because these are thought to represent two distinct aetiological entities. Therefore, we set up the Scotland and Newcastle epidemiological study of Hodgkin’s disease (SNEHD), a population based, case–control study of adult HL (then referred to as Hodgkin’s disease). Our main aim was to identify risk factor profiles for the development of HL, with stratification of cases by EBV status of tumours. Because SNEHD is population based and relatively large, it has enabled us, for the first time, to generate separate age and sex specific incidence curves for EBV associated and non-associated HL.

“The epidemiology of Hodgkin lymphoma is complex, and further understanding of the aetiology of this heterogeneous condition will only be gained by performing epidemiological studies with separate analyses of Epstein-Barr virus associated and non-associated cases, because these are thought to represent two distinct aetiological entities”

Unlike non-Hodgkin lymphomas (NHL), the subclassification of HL has remained relatively stable since the Rye classification was introduced in 1966.13–17 Because it is now recognised that Hodgkin and Reed-Sternberg (HRS) cells, the malignant cells in HL, are derived from lymphocytes the term Hodgkin lymphoma is preferred to Hodgkin’s disease.15,18 The main change in subclassification is that nodular lymphocyte predominant HL (NLPHL) with a nodular growth pattern is now considered a separate disease entity, with a distinct immunophenotype and cellular origin.14–16,18 It is therefore classified separately from other subtypes, which are collectively referred to as classic HL (CHL).17 CHL includes the subtypes NSHL, MCHL, lymphocyte depleted HL (LDHL), and the additional subtype lymphocyte rich classic (LRCHL). Many cases previously diagnosed as LDHL are now included within NSHL, or reclassified as NHLs.17 Diagnostic accuracy in the diagnosis and subclassification of HL has been assessed previously19–32; however, the only population based study to deal with this issue in recent years was restricted to female patients.33 The inclusion of careful histopathological review in SNEHD presented an opportunity to evaluate accuracy in the diagnosis of HL in current practice.



SNEHD is a case–control study, but only index cases are included in the present report. Eligibility criteria for inclusion in SNEHD were: diagnosis of HL between 1 January 1993 and 31 July 1997; age at diagnosis of 16–74 years; residence in Scotland (excluding Dumfries and Galloway and the Western Isles) or the Northern Region of England; and, where known, birth in the UK. Cases were identified from the Scotland and Newcastle Lymphoma Group (SNLG) Registry, the ongoing system for registration of lymphomas already in place in this geographical area. Annual crosschecks were performed with relevant cancer registries (Scottish Cancer Registry and Northern Region Cancer Registry), and this provided a second source of cases. Ethical permission was obtained from all local research ethics committees in the study area, and the pathology working party of the SNLG approved the histological review.

Samples and histopathological review

Pathology reports and diagnostic biopsy samples, including sections stained using immunohistochemistry (IHC), were retrieved where possible. Immunohistochemical staining was carried out in 232 cases using a panel of monoclonal antibodies reactive with: CD45; CD20; CD3; CD30; CD15; and epithelial membrane antigen (Dako, Ely, Cambridgeshire, UK). Immunostaining was performed using a Dako Techmate automated staining system (Dako) and reactivity was detected using an ABC technique, according to the manufacturer’s instructions (Vector Laboratories, Peterborough, UK). The procedure for histopathological review was designed so that the reviewing pathologists, BA and ASK, would not review cases in which they had made the initial diagnosis and classification. Review diagnoses were returned to the LRF virus centre and, where the review and referral diagnosis differed, or where the diagnosis was considered problematic, cases were referred for discussion and review (BA, ASK, and RFJ) using a multiheaded microscope. Reviewers were blind to the EBV status of the lesions. Cases were initially categorised according to the revised European-American lymphoma classification system,14 but the more recent World Health Organisation nomenclature is reported here.34

EBV status

The EBV status of tumours was determined using EBV encoded RNA (EBER) in situ hybridisation. The EBER assay was carried out using a commercially available oligonucleotide probe (Vector Laboratories) and hybridisation detection kit (Dako). In the few cases in which results were difficult to interpret, or in which there appeared to be large numbers of EBV positive cells in the reactive component of lesions, latent membrane protein 1 (LMP1) IHC was performed. After antigen retrieval, the CS1–4 cocktail of monoclonal antibodies (Dako) was applied to paraffin wax embedded sections and reactivity was detected using an ABC kit (Vector Laboratories) incorporating diaminobenzidine (Dako) as the chromogenic substrate. Cases were considered EBV associated or EBV+ HL if HRS cells were positive in either assay.

Statistical methods

The κ coefficient was used to measure the level of agreement, corrected for chance, between referral and review diagnoses in six diagnostic categories: NLPHL, MCHL, LDHL, NSHL, LRCHL, and CHL not otherwise specified (NOS). The age and sex specific incidence rates were computed for age groups 16–24, 25–34, 35–49, 50–59, 60–69, and older people using as denominators population counts from the 1991 census. Comparisons of proportions have been tested using Pearson’s χ2 test. All statistical analyses were implemented using SPSS.



A total of 622 individuals were identified who fulfilled eligibility criteria and had a confirmed diagnosis of HL. Biopsy material was retrieved from 561 patients and in 549 was considered suitable for review. Appropriate material for histopathological review was not available in 73 cases; these cases have been retained in some analyses.

Histopathological review

Haematoxylin and eosin stained sections from 549 cases were subjected to histopathological review. IHC staining was performed in 232 cases and in an additional 113 cases the original IHC was available; therefore, IHC was reviewed in 345 cases. In a further 117 cases, IHC results were obtained from the original pathology report. In the remaining 87 cases, IHC could not be assessed; however, in 72 of these IHC was considered unnecessary for diagnostic purposes. Histological subtyping of all tumours was not possible, but attempts were made to distinguish NLPHL from CHL, and the diagnostic category CHL NOS was therefore introduced.

HL was confirmed in 511 of the 549 cases (93%) subjected to expert review, including all cases in which the IHC could not be assessed (table 1). In 26 cases (4.7%), the diagnosis of HL was revised; in 23 instances (4.1%), the diagnosis was changed to NHL, and in three cases only reactive changes were observed in the material submitted for review. Among cases reclassified as NHL, diffuse large B cell lymphoma (n = 9) and T cell rich B cell lymphoma (n = 9) were the most common review diagnoses. Cases with a referral diagnosis of NLPHL (n = 5), MCHL (n = 7), and HL NOS (n = 7) were over-represented among cases with a revised diagnosis, as were patients in the older age group (see below). For the 33 cases with an initial diagnosis of HL NOS or CHL NOS, HL was confirmed in only 21 cases (63.6%). In 12 of the 549 cases (2%), the diagnosis of HL was considered uncertain following pathological review, either because the biopsy material was inadequate (n = 10) or because a firm diagnosis could not be reached without further investigation (n = 2). These cases have been classified as “not HL” in this and all further analyses of SNEHD.

Table 1

Comparison between referral and review diagnoses

The histological subtype at initial diagnosis was known for 534 cases; the referral and review diagnoses were identical in 386 instances, giving 72.3% agreement overall and 75.6% for cases confirmed as HL. There were 23 changes (4.3%) between NLPHL and CHL, now considered distinct disease entities. Additional subtype changes occurred in 86 instances, with MCHL to NSHL being the most frequent diagnostic change (n = 34). Confirmation rates (the percentage of referral diagnoses confirmed on review) differed for the different histological subtypes, being good for NSHL (88%), but less good for NLPHL (64%) and MCHL (59%). For female patients the confirmation rate for NSHL was slightly higher (92%), but dropped to 46% and 43% for NLPHL and MCHL, respectively. There were 12 cases with a referral diagnosis of LDHL, but only one of these was confirmed on review, giving a confirmation rate of 8.3%; only five cases in the total series and two of the reviewed cases were classified as LDHL, confirming that this subtype is almost disappearing with the use of modern diagnostic criteria.17 Identification rates (the percentage of review diagnoses that were the same as initial diagnosis) showed less variation among the major subtypes (HLNS, 80%; NLPHL, 80%; MCHL, 77%); this value was much lower for LRCHL (31%), but this reflects the introduction of this subtype as a provisional entity during the study period.14 The κ coefficient for chance corrected agreement between referral and review diagnosis using six categories (NLPHL, MCHL, LDHL, NSHL, LRCHL, and CHL NOS) is 0.62, confirming a substantial level of agreement.

Diagnostic changes were analysed with respect to patient age in three age groups: 16–34 years, 35–49 years, and > 50 years. Changes in diagnosis and subtype were more common for the older adult age group, and differences were statistically significant for “all changes” (p < 0.001), and changes from HL to “diagnosis uncertain”. There was also a higher proportion of cases in the older adult group in which histological subtyping could not be reliably performed (p = 0.01).

Data from the 511 confirmed cases of HL, plus the 73 cases not subjected to histopathological review, were used to derive the incidence of HL by age, sex, and histological subtype (fig 1). A bimodal curve with an incidence peak in young adulthood (15–34 years) and a second peak in older adults (⩾ 55 years) is evident for “all HL” and CHL. Overall, 61.3% of cases were classified as NSHL and 21.4% as MCHL. NSHL largely accounts for the young adult age incidence peak, whereas the incidence of MCHL shows less variation by age (fig 1B). Age incidence patterns for the sexes are different; there is a male excess up to 65 years, with the curve for women showing more evidence of bimodality (fig 1C).

Figure 1

Age specific incidence rates of Hodgkin lymphoma (HL) for each 100 000 person years. (A) All HL, classic HL, and nodular lymphocyte predominant HL (NLPHL). (B) Classic HL, nodular sclerosis HL (NSHL), and mixed cellularity HL (MCHL). (C) Classic HL by sex.

EBV status

EBV status was determined for 537 cases, including 461 cases with a review diagnosis of CHL; in 154 (33%) of these last cases all of the HRS cells were EBER positive and therefore designated EBV associated or EBV+ HL (fig 2; table 2). LMP1 staining identified no EBV+ HL cases that were not picked up using the EBER assay. Consistent with previous studies, there were significant differences by histological subtype, with 63 of 105 (60%) MCHL and 76 of 320 (23.8%) NSHL cases being EBV associated, respectively (p < 0.001). Differences by age group were also highly significant, with patients who have CHL in the older adult group (71 of 137; 51.8%) being more likely to have EBV+ HL than young adult patients (54 of 236; 22.8%) (p < 0.001). Table 2 shows the number of patients with EBV+ HL in each five year age group. There was a male excess within EBV+ HL (male to female ratio, 102:52) but not EBV− HL (male to female ratio, 152:155) (p < 0.001).

Table 2

Epstein-Barr virus (EBV) association in classic Hodgkin lymphoma by age group

Figure 2

Representative Epstein-Barr virus encoded RNA in situ hybridisations showing the characteristic staining pattern in the nuclei of Hodgkin and Reed-Sternberg cells. (A) Low power and (B) high power view.

Age and sex specific incidence curves for EBV associated and non-associated HL in adults were determined and, as predicted, these have different shapes (fig 3). The non-EBV associated cases show the young adult peak, and the incidence then declines with age. In contrast, the curve for EBV+ HL has a peak in the older adult age group; a smaller peak is also observed at the start of the young adult age range, consistent with our previous prediction.11 Bimodality of the incidence curve for EBV+ HL persists when male and female patients are analysed separately (fig 3B).

Figure 3

Age specific incidence rates of Hodgkin lymphoma for each 100 000 person years. (A) Classic Hodgkin lymphoma (CHL), EBV+ CHL and EBV− CHL. (B) EBV+ CHL by sex.


SNEHD provided a unique opportunity to investigate the age incidence of HL, stratified by EBV status, and the impact of expert histological review on HL diagnosis and classification. It is the largest population based study of HL to examine the EBV status of lesions; the only similar survey is a population based analysis of 395 women aged 19–79 in northern California.33 Although the effects of histological review on the diagnosis and classification of HL have been well documented,19–25,27–32 few analyses have been performed on population based data sets,22,24 and most were performed before the introduction of modern diagnostic criteria and immunohistochemical staining techniques.19–21,23,25,27–30

“Only two of the reviewed cases were considered to be lymphocyte depleted Hodgkin lymphoma (LDHL), consistent with the recognition that many cases previously classified as LDHL are either nodular sclerosing Hodgkin lymphoma or non-Hodgkin lymphoma”

In SNEHD, histological review confirmed the diagnosis of HL in most cases (93.1%); only 4.7% of cases were considered to have diagnoses other than HL, a similar figure to the 3% reported by Glaser et al.33 A further 2.2% of SNEHD diagnoses were considered uncertain following histological review, and assessment of further biopsy material was deemed necessary before a correct diagnosis could be reached. Overall, this represents a considerable improvement over many previous comparisons.19,20,22,27,29,35 There was more disagreement concerning the histological subtyping of lesions, with 24% of confirmed HL cases having a different referral and review diagnosis. Consistent with many previous studies, the confirmation rate for a referral diagnosis of NSHL was good, but for MCHL and NLPHL reliability was much lower.21,23–33 Only two of the reviewed cases were considered to be LDHL, consistent with the recognition that many cases previously classified as LDHL are either NSHL or NHL.17 Diagnostic reliability was lower for older patients (⩾ 50 years), with more misdiagnoses, subtype changes, and uncertainties in this age group. This may, in part, reflect the increased incidence of NHL in older patients; however, it was noted that the quality of biopsy material from older patients was often poor. Therefore, despite improvements in the diagnosis of HL over time, studies evaluating diagnostic accuracy since the implementation of the Rye classification consistently report underdiagnosis of NSHL (although this subtype has good confirmation rates), and less reliable diagnosis of non-NSHL subtypes and HL in older persons.19–33,36,37

Take home messages

  • Comparison of the age specific incidence curves for Epstein-Barr virus associated and non-associated Hodgkin lymphoma (HL) supports the hypothesis that these form two distinct aetiological entities

  • Although accuracy in the diagnosis of HL is generally good, certain subgroups of cases and the elderly continue to present diagnostic difficulties

The data suggest that diagnostic inaccuracy is unlikely to affect overall incidence data for HL although, for the reasons described above, care should be taken in the interpretation of data for subtypes other than NSHL and for persons in the older adult age group (⩾ 50 years). At the level of the individual patient, there is clearly room for improvement in HL diagnosis. Misdiagnosis of NHL and reactive nodes as HL, and misclassifications between NLPHL and CHL, occurred in ~ 9% of cases in our study, and would probably affect clinical management. It is recognised that the differentiation of NLPHL from CHL and NHL is difficult in some cases; however, although the neoplastic cells in NLPHL may show a broad spectrum of morphologies they usually show a characteristic phenotype, with expression of B cell specific markers and lack of expression of CD30 and CD15.38 In almost all cases previously classified as diffuse NLPHL, the tumour cells are CD30+ and CD15+ and show the clinical behaviour of CHL, and are properly classified as LRCHL.38,39 Furthermore, the existence of LRCHL with a nodular growth pattern is now accepted.17,38,40 Although improved diagnostic criteria should lead to the more accurate diagnosis of NLPHL, the results of our study suggest that expert review is advisable before embarking on less aggressive treatment regimens for this entity.

Incidence data, including subtype distributions, have been derived from this high quality data set. Overall, a bimodal incidence is seen, with the peak incidence in the young adult age group (fig 1). The distribution of histological subtypes in SNEHD is similar to that seen in other large, histologically reviewed data sets from Denmark and Germany,30,37 and for women resembles the distribution in northern California.33 Male patients predominate overall, but incidence curves for men and women show different features.

SNEHD is population based and relatively large; therefore, for the first time it has been possible to determine the age and sex adjusted incidence of EBV associated and non-associated HL. Incidence curves for the subgroups are different, supporting the hypothesis that these are distinct aetiological subgroups of HL. The young adult age incidence peak is largely driven by the non-EBV associated cases. A “hit and run” mechanism involving EBV has been suggested in non-EBV associated cases, but evidence for this is currently lacking;41 in particular, we have shown that young adult patients with EBV− HL are more likely to be EBV seronegative than are age matched controls.41

“For the first time it has been possible to determine the age and sex adjusted incidence of Epstein-Barr virus associated and non-associated Hodgkin lymphoma”

The age incidence curve for the EBV associated cases is flatter; there is a suggestion that this curve is bimodal, with a first peak in the 15–24 year age range and a second peak at ages > 55 years, although this is difficult to test formally. We previously suggested that there is a small peak in the incidence curve for EBV associated HL in the 15–24 year age range,11 and our present data provide support for this prediction. In SNEHD, young adult patients with EBV associated HL were more likely to report previous infectious mononucleosis (IM) and/or IM in a first degree relative than were patients in other subgroups, suggesting that delayed exposure to EBV is a specific risk factor for the development of young adult EBV+ HL.42 The number of EBV+ HL cases in the 15–24 year age incidence peak was small and precluded statistical analysis; however, 44% of these patients reported IM, or IM in a first degree family member, compared with only 17% of patients with EBV− HL and 9% of controls.

Our present study is restricted to adult cases; however, there is a body of data showing that HL in early childhood is usually EBV associated.5,7 Taken together, these data suggest that EBV associated HL is more likely to occur at three time periods in life—in early childhood, young adulthood, and older adult life. We predict that EBV associated HL in early childhood and young adulthood is associated with primary EBV infection. Overall, the data support a four disease model of HL, with three groups of EBV associated HL and a group of non-EBV associated cases.43


We would like to thank all pathologists within the SNLG who contributed material to this study. We particularly thank J White for help with tracing cases and F Jack for discussion of EBV results. SNEHD was funded by the Kay Kendall Leukaemia Fund and work at the LRF virus centre is supported by a Leukaemia Research Fund specialist programme grant.


View Abstract
  • Author Correction

    Please note that there is an error in the labelling of Figure 3. The correct figure is shown here:

    Files in this Data Supplement:

    • - Figure 3 Age specific incidence rates of Hodgkin lymphoma for each 100 000 person years. (A) Classic Hodgkin lymphoma (CHL), EBV+ CHL and EBV- CHL. (B) EBV+ CHL by sex.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles

  • Correction
    BMJ Publishing Group Ltd and Association of Clinical Pathologists