8/13/2019 11156_ftp

1/5

AN EPIDEMIOLOGIC STUDY OF INDEX AND FAMILY INFECTIOUS

MONONUCLEOSIS AND ADULT HODGKINS DISEASE (HD): EVIDENCE FOR A

SPECIFIC ASSOCIATION WITH EBVVE HD IN YOUNG ADULTS

Freda E. ALEXANDER1

*, Davia J. LAWRENCE2

, June FREELAND3

, Andrew S. KRAJEWSKI4

, Brian ANGUS5

, G. Malcolm TAYLOR6

andRuth F. JARRETT3

1Department of Community Health, Public Health Sciences, University of Edinburgh, Medical School, Teviot Place, Edinburgh, UK2Clinical Trials and Statistics Unit, Institute of Cancer Research, Sutton, Surrey, UK3Leukaemia Research Fund Virus Centre, Department of Veterinary Pathology, University of Glasgow, Glasgow, UK4Department of Pathology, Northampton General Hospital NHS Trust, Northampton, UK5Department of Pathology, University of Newcastle on Tyne, UK6Immunogenetics Laboratory, University of Manchester, St Marys Hospital, Manchester, UK

Infectious mononucleosis (IM) is an established risk factorfor Hodgkins disease (HD). A substantial minority (33%) ofcases of HD have Epstein-Barr virus (EBV) DNA within themalignant cells (are EBVve). It is unclear whether risk afterIM applies specifically to EBVve HD. We report the resultsof a population-based case-control study of HD in adults (n

408 cases of classical HD, 513 controls) aged 1674 years; thecase series included 113 EBVve and 243 EBV-ve HD. Analysescompared total HD, EBVve HD and EBV-ve HD with thecontrols and EBVve HD with EBV-ve HD cases using, mainly,logistic regression. Regression analyses were adjusted forgender, age-group and socioeconomic status, and were per-formed for the whole age range and separately for young ( 35 years); formal tests of effectmodification by age were included. For the young adults,reported IM in index or relative was strongly and significantlyassociated with EBVve HD when compared to controls(odds ratio [OR] 2.94, 95% confidence interval [CI]: 1.087.98 and OR 5.22, 95% CI: 2.1512.68, respectively). Theseresults may be interpreted as indications that late first expo-sure to EBV increases risk of HD, especially in young adults;this applies primarily to EBVve HD. 2003 Wiley-Liss, Inc.

Key words: Epstein-Barr virus; Hodgkins disease; epidemiology;risk factor

Abundant epidemiologic evidence supports the hypothesis thatHodgkins disease (HD) aetiology differs for cases diagnosed inthe young adult peak and diagnosed in older people (aged 50).1

Under the late host-response model,2 HD in the young adult peakarises as a sequel to relatively late first infection with a commoninfectious agenta similar pattern to that seen for infectiousmononucleosis (IM) in relation to the Epstein-Barr virus (EBV).Large cohort studies3,4 have demonstrated that HD risk is in-creased around 3-fold in the followup of subjects with IM com-pared to the general population. This could indicate that EBV iscausally implicated in HD or, alternatively, that IM is a marker ofa type of lifestyle that predisposes to late first exposure to manyinfectious agents. A direct link seemed probable when EBV viralDNA was identified within the HD tumour cells.5,6 It is nowestablished that EBV is associated with a substantial minority(referred to here as EBVve) of cases of HD, and IARC hasrecently classified EBV as a probable human carcinogen in relationto HD.7

Subsequent research has indicated the complexity of the asso-ciation between EBV and HD. For example, the proportion ofcases that are EBVve is markedly lower in the young adult peak(1634 years) than for HD in children or older people.8 Becausethe late host-response model applies to the young adult peak, thissuggests that EBV is not the elusive agent involved in that model,if indeed there is just 1 agent.

One key question is whether IM is associated specifically withEBVve HD or, alternatively, associated as strongly with the cases

that lack EBV (the EBV-ve cases) that constitute the majority ofcases in the young adult peak. Very few studies have looked atepidemiologic risk factors for HD with cases classified as EBVve

or EBV-ve. Only 2 studies have collected personal data to addressthis question: a US case-series that compared EBVve and EBV-ve

HD and included cases aged 16 55 years at diagnosis,9 and a UKcase-control study10,11 that was restricted to subjects aged 1624years and analysed as both a case-series and case-control study.These 2 studies generated results that are potentially inconsistent,and suggest different answers to the key question. Sleckman et al.9

found no evidence that a history of IM conveyed specific risk forEBVve compared to EBV-ve HD. In contrast, Alexanderet al.10,11

reported evidence that increased risk after IM was specific toEBVve HD, although there was also a small excess risk ofEBV-ve HD after IM. These 2 results are consistent only if the ageat diagnosis of HD becomes critical to the association of IM withsubsequent HD.

The study of Alexander and colleagues also found that familyhistory of IM was a significant risk factor for EBVve HD in theage range 1624 years.12 Most affected relatives were siblings,and this was interpreted as further evidence that relatively late andrecent exposure to EBV increased risk of EBVve HD in youngadults; alternative interpretations involving confounding socialfactors and inherited susceptibility to EBV are also available.

We now report the results of a population based case-controlstudy (SNEHD: Scotland and Newcastle Epidemiological Study ofHodgkins Disease) for which all newly diagnosed adult cases ofHD in a defined geographic area of Scotland and Northern Englandwere eligible. We aimed, first, to test the hypothesis that indexhistory of IM, family history of IM or the 2 combined conveysboth a small general risk of HD and a larger, specific risk ofEBVve HD. Second, we aimed to test the modified hypothesisunder which the specific risk of EBVve HD was confined toyoung adults. It is only this second hypothesis that is compatible

Grant sponsor: Kay Kendall Leukaemia Fund; Grant sponsor:LRF Spe-cialist Programme Grant.

*Correspondence to: Department of Community Health, Public HealthSciences, University of Edinburgh Medical School, Teviot Place, Edin-burgh EH89AC, UK. Fax: 44-131-556-6793.E-mail: freda.alexander@ed.ac.uk

Received 14 May 2002; Revised 21 October 2002, 9 December 2002;Accepted 11 December 2002

DOI 10.1002/ijc.11156

Int. J. Cancer:107, 298302 (2003)

2003 Wiley-Liss, Inc.

Publication of the International Union Against Cancer

8/13/2019 11156_ftp

2/5

with the published work of both Sleckman et al.9 and Alexanderetal.10,11

We restricted our analysis to classical HD because lymphocytepredominance HD is now considered a separate disease entity witha distinct aetiology.13

MATERIAL AND METHODSSubjects

Eligible cases were all people aged 1674 years diagnosed withHD while resident in Scotland (excluding the Western Isles andDumfries and Galloway) and the Northern Region of Englandfrom January 1, 1993 to July 31, 1997. Controls were randomlyselected from computerised regional primary care lists (regionswere Family Health Service Authorities (FHSAs) in England andHealth Boards in Scotland). Control numbers were selected so thatcases and controls would have similar frequencies by age group(1624, 2534, . . . 6574), sex and region. An additional eligi-bility criterion imposed on cases and controls was that they musthave been born in the UK.

Cases were identified in the first instance from the ongoinglymphoma registration system already in place for the study area:the Scotland and Newcastle Lymphoma Group.14 Annual crosschecks with the relevant cancer registries (Scottish Cancer registry,Northern region cancer registry) provided a second source ofcases. Provided that their treating consultants gave consent, caseswere then approached by letter. The approach letter provided fullinformation about our study and enabled subjects to give informedconsent. When a patient died before being approached or beforeinterview, the general practitioner (GP) was asked to give the nameof a close relative and consent to approach the relative. Relativeswho gave informed consent then provided proxy interviews.

Controls were approached by letter in a similar way after receiptof consent to approach from their GP. When we failed to recruit aselected control we replaced this control with a second (or later)choice randomly selected from patients of the same age and sex onthe list of the GP practice of the first control. This process wasintended to retain to some extent the socioeconomic status of thearea of residence of our first choice controls.

Each case was given as a reference date the date of histologicdiagnosis. Each control was allocated a reference date that wasdesigned to have a similar relationship with date of interview asdate of diagnosis for cases: random numbers were used to generatea year (19931997) with the month and day being that of thesubjects birthday.

Interview data

Trained interviewers conducted face-to-face interviews, nor-mally in subjects homes, to complete our study questionnaire.This questionnaire elicited personal medical history of a number ofconditions relevant to our study hypotheses, including IM (glan-dular fever). Events were only included in the analyses if theyoccurred before the reference date. A list of first-degree familymembers was also obtained, together with forenames and dates of

birth/death. Respondents were asked whether they were aware ofa previous diagnosis of a similar list of medical conditions for eachof these relatives. This second list also contained IM. No particularemphasis was placed on IM but it is possible that responding caseswere aware of a potential link between IM and HD. In response tothe question, Have you previously had glandular fever?, respon-dents could answer yes, definitely, yes, suspected or no.Those answering yes were then asked if the diagnosis wasconfirmed by analysis of a blood sample. In the analyses reportedhere in detail, we have taken as prior IM an episode of illness thatbegan before the diagnosis of HD and was described as yes,definitely or yes, suspected, but analyses were replicated withalternative definitions. Because the prodromal symptoms of HDcan resemble IM, we also performed analyses in which illnessoccurring within the year before the reference date was excluded.

Family history of IM (in a first-degree relative) was, mainly,analysed as a dichotomy: yes or no. However, we also attemptedto introduce an improved measure that incorporated adjustment fornumber of relatives at risk. Specifically, we calculated a weightedsum over available relatives based on their ages at reference date(weights of 0.06 for 04 years, 0.35 for 514 years and 1 for allothers). The choice of weights was taken from an independent

source.15

Family risk was then the number of affected first-degreerelatives divided by this weighted sum of relatives.

Where we reported the affected relative (parent, sibling, child)and more than 1 relative was affected, we took the relationshipfrom the relative whose IM occurred closest in time to the refer-ence date. Similarly, the time gap from IM in a relative to thereference date was calculated as the shortest of the available times.

We did not have detailed information on whether these familymembers were living together at the times when the relatives hadIM, nor whether they met frequently. Nevertheless, we included anadditional exposure variable index and/or family IM as a yes/nodichotomy. This is meaningful only if index and family IM areinterpreted as indicators of a similar underlying biologic process:most probably, infection by EBV, but also, possibly, inheritedsusceptibility to EBV and, if the affected relative is a sibling,lifestyle conducive to late first exposure to EBV.

Socioeconomic status classification for eligible cases and con-trols was taken from the Carstairs index or depcat16 of the smallcensus area in which they lived.

Laboratory methods

Paraffin-embedded biopsy material and, if possible, immunohis-tochemically-stained sections were retrieved along with the rele-vant reports. Where satisfactory staining was not available, immu-nohistochemic staining was performed using the following panelof monoclonal antibodies: CD45, CD20, CD3, CD15, CD30 andEMA using a standard ABC peroxidase method (Vector Labora-tories, Peterborough, UK) and a Dako Techmate automated stainer(Dako Ltd., Cambridgeshire, UK). Histopathologic review wascarried out by B.A. and A.S.K., and if the review diagnosisdiffered from the original diagnosis or the diagnosis was consid-ered problematic, the case was referred for discussion before

reaching a diagnosis. Cases were classified according to the Re-vised European-American Lymphoma (REAL) classification,13

and those diagnosed as lymphocyte predominant were excludedfrom the statistical analysis. EBV status was determined by per-forming EBV EBER in situ hybridisation on sections of paraffin-embedded material using a commercially available probe (VectorLaboratories) and hybridization kit (Dako Ltd). In cases in whichsignificant numbers of bystander cells appeared to be staining on theEBER analysis, LMP-1 immunohistochemistry was also performedusing the CS1-4 monoclonal antibodies (Dako Ltd.) and an ABC kit(Vectastain, Vector Laboratories). Cases were designated EBVve ifReed-Sternberg cells and variants stained positively in either of theseassays.

Statistical methods

Except where nonparametric methods are specifically indicated,

all analyses were based on unconditional logistic regression.17

Four comparisons were systematically conducted:

1. HD compared to controls

2. EBVve HD compared to controls

3. EBV-ve HD compared to controls

4. EBVve HD compared to EBV-ve HD (the case-series anal-ysis)

These comparisons yielded estimates of risk, expressed as oddsratios (OR) accompanied by 95% confidence intervals (CI). AnOR 1 indicates that the exposure is associated with the first ofthe pair (e.g., with HD). All analyses were adjusted for gender anddepcat in the regression modelling; analyses were further adjusted

299INFECTIOUS MONONUCLEOSIS AND ADULT HD

8/13/2019 11156_ftp

3/5

by age (1634, 3549, 50) except when data were restricted toyoung adults. Formal homogeneity tests of specificity of associa-tions to the young adults tested the effect of adding a term forexposure only in young adults to a model that included the mainaffect of the exposure, age group, gender and depcat.

The statistical packages SPSS and EGRET were used for thestatistical analyses.

RESULTS

Altogether, 78% of eligible cases were recruited, but only 46%of our first choice controls were recruited. Recruitment of firstchoice controls was lowest for the youngest (1624 years atreference date) and oldest (6574 at reference date) subjects (38%and 40%, respectively). In the youngest age group, male recruit-ment was much lower than female (21% compared to 28%), but itwas higher in the oldest age group (43% and 36%, respectively).Control recruitment was strongly associated with deprivation ofarea of residence, with 61% of those living in areas with depcat 1 (most affluent) but just 36% of those with depcat 7 (mostdeprived) recruited. Case recruitment decreased with age over 55years (around 80% for all age groups under 55, 68% for 5564 and60% for 6574). In general, case recruitment was slightly higher in

females (80%) than males (75%).The numbers of interviewed subjects (cases of classical HD and

controls) available for analyses are shown in Table I by age groupand status. This excludes 54 interviewed cases of HD whosedisease was classified as lymphocyte predominance HD.

Analyses of case subgroups compared to controls by reportedprior IM (Table II) showed several statistically significant results,including those for total HD, which were of similar magnitude atall ages and in young adults and older people separately. EBVve

HD was strongly and significantly associated with index IM inyoung adults; while this association was also evident when thewhole age range was analysed, it was of smaller magnitude and notstatistically significant in the older people alone. In contrast, theassociations for EBV-ve HD were of similar magnitude across theage range, although only statistically significant when all ageswere combined. In the case-series analysis (EBVve and EBV-ve

cases compared), the OR for young adults is modestly elevated,but this failed to approach statistical significance; the OR for olderpeople is close to unity.

When similar analyses were conducted for family IM, statisti-cally significant associations were confined to the young adults,and were most marked in EBVve cases compared to the controls(Table II). Although the association remained statistically signif-icant when total HD was compared to the controls, the case-seriesanalysis indicated a specific association with EBVve cases. Thehomogeneity test demonstrated the statistical significance of thedifferences between young adults and older people.

Similar results were found when reports of IM within the yearbefore reference date were excluded. We repeated these analyseswith alternative definitions of index IM (taking those who reportedmerely suspicion of IM as no IM, and also taking suspicious butunconfirmed IM as missing data). Both of these alternative defi-nitions gave very similar results to those we report here. Exclusion

of cases for whom interviews had been conducted with surrogates(8 EBVve and 5 EBV-ve, of whom 1 was positive for index IMand 3 for family IM) made only marginal differences to the ORs.

Times from index IM to reference date were compared using theMann-Whitney test; those for EBVve cases were shorter thanthose for controls and EBV-ve cases (p 0.16 and 0.075, respec-tively).

The relationship between index case and family member and thetime between familial IM and reference date were examined forsubjects reporting family IM. For the young adults, the majority(75%) of affected relatives were siblings, compared to 35% forolder people, for whom 51% of affected relatives were children(p 0.001, Pearsons chi-square). The time gap between illness ina relative and reference date was shorter for the young ( 35) thanolder adults ( 35) (p 0.006, Mann-Whitney test). There was noevidence that these intervals were shorter for EBVve compared toEBV-ve cases, but when these cases were combined, the time gapswere shorter than for the controls (p 0.046, Mann-Whitney test).Finally, we restricted analyses of the time gaps to sibling IM;intervals were shortest for EBVve cases and longest for controls(p 0.09, Kruskal-Wallis test), and the differences for the EBVve

and EBV-ve cases taken together and compared to the controlsachieved statistical significance (p 0.03, Mann-Whitney test).

An alternative analysis of family IM took numbers of familymembers at risk into account. When risk in family members (nowtaken as a continuous variable) was analysed by the Mann-Whit-ney test, significantly higher risk was found for EBVve HDcompared to controls (p 0.001), total HD compared to controls(p 0.007) and EBVve HD compared to EBV-ve HD (p 0.016), but the excess for EBV-ve HD compared to controls was ofborderline statistical significance (p 0.09).

A small number of subjects reported positively for both indexand family IM (n 8 EBVve HD,n 3 EBV-ve HD andn 5controls). Despite the small numbers, we conducted a multivariateanalysis for the young adults with an extension of the previousmodels that included (i) terms for both index and family IMsimultaneously and (ii) a further term for their interaction. Therewas no evidence of statistical interaction (all p-values 0.3); theadjusted analyses reduced, in general, the ORs reported in Table II

but confirmed the importance of prior IM (both in index andrelative) in relation to HD, especially when EBVve. For example,adjusted ORs for EBVve HD compared to controls were 2.50(95% CI: 0.877.22, p 0.09) for index IM and 4.89 (95% CI:1.9812.07, p 0.001) for family IM.

When index and/or family IM were taken together as a singleexposure variable (Table III, all ORs for comparisons of casesubgroups with the controls showed significant elevations for thetotal age range and for the young adults. However, the magnitudeswere larger for the young adults and for EBVve cases; thecase-series analysis showed a significant positive association withEBVve cases in the young adults. All ORs for older people wereclose to, or less than, unity. The formal test of homogeneityconfirmed the age group differences.

DISCUSSION

This is the first large case-control study of adult HD to haveincluded EBV status of subjects, and the largest case-series com-parison of epidemiologic variables for HD by EBV status. Ourpresent study is also, we believe, the first to report results restrictedto classical HD and to have conducted histologic review thatincluded immunohistochemistry. Smaller previous studies are acase-series comparison of 100 cases aged 1855 years,9 a case-control study (n 103 cases) of 16- to 24-year-olds only10 and acase-series comparison based on the same subjects.11 These studiesyielded potentially inconsistent results (see Introduction).

SNEHD has potential methodologic problems: recall bias, se-lection bias, misclassification of prodromal HD as IM and lack oflaboratory confirmation of IM diagnoses. We have noted that IM

TABLE I NUMBERS OF SUBJECTS AVAILABLE FOR ANALYSIS

StatusAge range

35 3549 50 & over All adults

EBVveclassical HD

44 21 48 113

EBVveclassical HD

145 49 49 243

Classical HD,EBV status notavailable

17 15 20 52

Controls 235 117 161 513

300 ALEXANDER ET AL.

8/13/2019 11156_ftp

4/5

was not highlighted in the questionnaire in any way; however,cases may have become aware outside our study that a relationship

between IM and HD is an important area of scientific research.Selection bias is certainly present, in that recruitment of our firstchoice controls was strongly associated with the socioeconomicstatus of their area of residence. We tried to prevent influence onour case-control comparisons in 2 ways: first, by replacing controlswith second and later choices from the same GP (hence, broadly,the same area of residence) and second, by adjusting all logisticregression analyses by depcat. Neither of these succeeded in re-moving the effects of socioeconomic status imbalance in ourcontrols.18 It is possible, therefore, that our case-control compar-isons for total HD may have been influenced by recall bias andselection bias. However, cases had no knowledge of the EBVstatus of their tumour. The same set of controls were used in allcase-control comparisons; neither recall nor selection bias couldhave influenced differences between EBVve and EBV-ve HD

when compared to controls or when compared to each other in thecase-series analysis.

Lack of laboratory confirmation of IM should have led tonondifferential misclassification bias and, hence, biased our re-ported ORs towards the null. Differential misclassification canonly arise if prodromal HD was misclassified as IM in cases; thismay have occurred for total IM but there is no reason to suggestthat it applies preferentially to EBVve or EBV-ve cases. Weattempted to overcome this problem by replicating our analyseswith exclusion of subjects with IM diagnosed within 1 year of thereference date. We believe that differential misclassification mayhave influenced our results for total HD but not for differencesbetween EBVve and EBV-ve HD when compared to controls orwhen compared to each other in the case-series analysis.

Despite the large size of the overall study, our numbers arerelatively small within age-group categories. Nevertheless, we

TABLE II CLASSICAL HD, ASSOCIATIONS WITH REPORTED INDEX IM AND REPORTED FAMILY IM

All adult ages1 Young adults (1634 yrs)2 Older persons (35 and older)3

Index prior IM4

HD vs. ctrl OR (CI) 2.48 (1.474.18) 2.61 (1.305.23) 2.28 (1.025.10)p 0.001 0.005 0.042

n1(%)n2(%) 43 (10.5%)/24 (4.7%) 27 (13.1%)/13 (5.6%) 16 (7.9%)/11 (4.0%)EBV vs. ctrl OR (CI) 2.59 (1.245.43) 2.94 (1.087.98) 2.17 (0.716.66)

p 0.016 0.045 0.19n1(%)n2(%) 12 (10.6%)/24 (4.7%) 7 (15.9%)/13 (5.5%) 5 (7.2%)/11 (4.0%)

EBV vs. ctrl OR (CI) 2.11 (1.143.90) 1.88 (0.854.14) 2.47 (0.936.55)p 0.019 0.12 0.08

n1(%)n2(%)

22 (9.1%)/24 (4.7%) 14 (9.7%)/13 (5.6%) 8 (8.2%)/11 (4.0%)

EBV vs. EBV OR (CI) 1.37 (0.632.98) 1.76 (0.634.88) 0.98 (0.293.27)p 0.43 0.29 0.97

n1(%)n2(%) 12 (10.6%)/22 (9.1%) 7 (15.9%)/14 (9.7%) 5 (7.2%)/8 (8.2%)Family IM5

HD vs. ctrl OR (CI) 1.23 (0.791.90) 2.43 (1.234.78) 0.67 (0.361.26)p 0.36 0.018 0.21

n1(%)n2(%) 44 (11.0%)/47 (9.3%) 27 (13.3%)/14 (6.0%) 17 (8.6%)/33 (12.2%)EBV vs. ctrl OR (CI) 1.81 (0.973.38) 5.22 (2.1512.68) 0.67 (0.241.82)

p 0.07 0.001 0.41n1(%)n2(%) 16 (14.5%)/48 (9.5%) 11 (25.0%)/15 (6.4%) 5 (7.6%)/33 (12.2%)

EBV vs. ctrl OR (CI) 1.24 (0.742.08) 1.84 (0.853.96) 0.83 (0.391.76)p 0.43 0.12 0.62

n1(%)n2(%) 26 (10.9%)/47 (9.3%) 15 (10.6%)/14 (6.0%) 11 (11.5%)/33 (12.2%)EBV vs. EBV OR (CI) 1.65 (0.813.35) 3.35 (1.338.44) 0.63 (0.202.00)

p 0.18 0.012 0.42n1(%)n2(%) 16 (14.5%)/26 (10.9%) 11 (25.0%)/15 (10.6%) 5 (7.6%)/11 (11.5%)

1Adjusted for gender, deprivation of area of residence and age group (1634, 3549, older).2Adjusted for gender and deprivation of area ofresidence.3Adjusted for gender, deprivation of area of residence and age group (3549, older). 4Homogeneity test by age at diagnosis: all pvalues exceeded 0.4.5Homogeneity best by age at diagnosis:p 0.005 (HDvs. ctrl), 0.002 (EBVve HDvs. crtrl), 0.1 (EBVve HDvs.ctrl),0.04 (EBVve HD vs. EBVve HD).

TABLE III CLASSICAL HD, ASSOCIATION WITH REPORTED IM IN INDEX AND/OR FAMILY MEMBER (COMBINED EXPOSURE VARIABLE)

All adult ages1 Young adults (1634yrs)2 Older persons (35 and older)3

HD vs. ctrl OR (CI) 1.68 (1.172.41) 2.47 (1.474.15) 1.11 (0.661.88)p 0.005 0.001 0.69

n1(%)/n2(%)4 79 (19.8%)/66 (13.1%) 49 (24.1%)/27 (11.6%) 30 (15.2%)/39 (14.4%)

EBV vs. ctrl OR (CI) 2.03 (1.193.46) 3.85 (1.818.19) 1.08 (0.482.39)

p 0.012 0.001 0.86n1(%)/n2 (%) 24 (21.8%)/67 (13.3%) 15 (34.1%/28 (12.0%) 9 (13.6%)/39 (14.4%)

EBV vs. ctrl OR (CI) 1.60 (1.042.44) 1.90 (1.053.41) 1.25 (0.652.39)p 0.03 0.031 0.50

n1(%)/n2(%) 45 (18.9%)/66 (13.1%) 28 (19.7%)/27 (11.6%) 17 (17.7%)/39 (14.4%)EBV vs. EBV OR (CI) 1.43 (0.792.60) 2.46 (1.115.49) 0.73 (0.291.85)

p 0.24 0.03 0.51n1(%)/n2(%) 24 (21.8%)/45 (18.9%) 15 (34.1%)/28 (19.7%) 9 (13.6%)/17 (17.7%)

1Adjusted for gender, deprivation of area of residence and age group (1634, 3549, older); homogeneity test for age at diagnosis; p 0.03(HDvs. ctrl), 0.02 (EBVve HDvs. ctrl), 0.3 (EBVve HD vs. ctrl), 0.07 (EBVve HDvs. EBVve HD).2Adjusted for gender and deprivationof area of residence.3Adjusted for gender, deprivation of area of residence and age group (35 49, older).4Numbers and percentages ve in 1stgroup/in 2nd group.

301INFECTIOUS MONONUCLEOSIS AND ADULT HD

8/13/2019 11156_ftp

5/5

consider it essential to analyse young adults and older peopleseparately because there is overwhelming evidence that the aeti-ologies are distinct (see Introduction).

Our data confirm our hypothesis that index IM conveys, or is amarker for, increased risk of HD that applies outside the youngadult peak and outside the EBVve subgroup of cases. The mag-nitude of the risk we report is between 2 and 3 and is consistent

with the cohort studies estimate of a 3-fold increase. The highestOR we report is for EBVve cases in young adults compared tocontrols, and there is a suggestion in our data that there is a specificadditional risk of EBVve HD in this age-group after IM. Thisconfirms results reported from our previous case-control study.10

When IM in first-degree blood relatives was analysed, we foundevidence of increased risk of, specifically, EBVve HD in youngadults. Here, IM in first-degree blood relatives is capable of 2 (notmutually exclusive) interpretations: it may indicate an inherited hostresponse to EBV infection, and it may indicate likelihood of firstinfection by EBV beyond early childhoodeither by direct infectionfrom the affected relative or, in the case of siblings, a childhoodlifestyle conducive to subsequent delayed infection by EBV. Al-though we collected extensive data on dates of birth of relatives anddates of IM in relatives. we do not know how frequently and when

relatives actually met, and so our data cannot be unambiguouslyinterpreted in terms of the second alternative. Nevertheless, the resultswe reportspecifically: (i) the difference in relationship of relative byage at reference date (i.e., siblings predominating as affected relativesin our young adult subjects but not subsequently); (ii) age-groupspecificity; and (iii) the shorter time intervals from relative IM toreference date for casestaken together, point to family IM in youngadults being a marker for late exposure of index subjects to EBV.

In the belief that this is true, we considered the compositeexposure variable (own and/or family IM) as an indicator ofpotential late first exposure to EBV. Although this interpretation isopen to question, the results provide consistency that increasestheir plausibility: first, there is a general increase in risk of totalHD across all ages and involving both EBVve and EBV-ve HDand second, there is a specific risk of EBVve HD in young adults.

This is consistent with the previously published epidemiologicdata for IM and HD by EBV status.911

The most important existing models for HD aetiology are themultiple aetiology model1 and the polio model.3 Under the multi-ple aetiology model, HD in children, young adults and olderpeople has distinct aetiologies. Under the polio model, the sameinfectious agent contributes to the aetiology of HD in children andyoung adults but infection with the agent during childhood con-veys much reduced risk of subsequent HD compared to infection

as a young adult. This higher risk is embodied in the late host-response model,2 which also seeks to explain the young adult peakunder the multiple aetiology model. The identification of EBVve

and EBV-ve HD as 2 large subsets of total HD has led to debate asto whether either model has adequate complexity.19

It is clear that the polio model, with EBV as infectious agent, doesnot explain the epidemiology of all HD because most cases are notEBVve and some cases have never been infected by EBV (i.e., areseronegative [Gallagheret al., unpublished], lack EBV in PCR anal-yses of peripheral blood and have no cytotoxic T-cell responses toEBV [Khan et al., unpublished]). Our main conclusion after thepresent analyses is that the polio model is an attractive one forpaediatric and young adult EBVve HD. We have clearly demon-strated that young adults who have, themselves, been exposed to EBVlate and succumbed to IM, together with those whose relatives havedone so, are at substantially increased risk of HD that (at least when

relatives are included) is specifically EBVve

. The specificity of thisrisk attenuates with age and time. Separation of individual effects ofage and time is almost impossible. Our data add nothing to knowledgeof paediatric EBVve HD but there is a growing body of epidemio-logic data indicating increased frequency in places where exposure toEBV is likely to occur early.6 In the UK, both residence in an area oflow SES and South Asian ethnicity contribute independently to EBVpositivity in pediatric HD.20

For EBV-ve HD, we provided evidence that IM conveys in-creased risk of subsequent HD. This is strongest in the young adultpeak but persists beyond it. Although this evidence is less strong,we prefer to interpret it, provisionally and in accordance with theliterature, as indicating a genuine increased risk consequent uponbehaviour and lifestyle that predispose to late first infection byEBV and also other common infectious agents. Our data provideno evidence that EBV, itself, is causally associated with EBV-ve

HD but offer limited support for the late host-response model forEBV-ve HD in young adults.

REFERENCES

1. MacMahon M. Epidemiology of Hodgkins disease. Cancer Res 1966;26:1189.

2. Gutensohn N, Cole P. Childhood social-environment and Hodgkinsdisease. N Engl J Med 1981;304:13540.

3. Gutensohn N, Cole P. Epidemiology of Hodgkins disease. SeminOncol 1980;7:92102.

4. Hjalgrim H, Askling J, Sorensen P. Risk of Hodgkins disease andother cancers after infectious mononucleosis. J Natl Cancer Inst2000;92:1522 8.

5. Pallesen G, Hamilton-Dutoit SJ, Rowe M, Young LS. Expression ofEpstein-Barr virus latent gene products in tumour cells of Hodgkinsdisease. Lancet 1991;337:3202.

6. Jarrett RF, Armstrong AA, Alexander FE. Epidemiology of EBV andHodgkins lymphoma. Ann Oncol 1996;7:S510.

7. IARC. Proceedings of the IARC Working Group on the Evaluation ofCarcinogenic Risks to Humans. Epstein-Barr virus and Kaposis sar-coma herpesvirus/human herpesvirus 8. Lyon, France, 1724 June1997. IARC Monogr Eval Carcinog Risks Hum 1997;70:1492.

8. Armstrong AA, Alexander FE, Cartwright RA, Angus B, KrajewskiAS, Wright DH, Brown I, Lee F, Krane E, Jarrett RF. Epstein-Barrvirus and Hodgkins disease: further evidence for the three diseasehypothesis. Leukemia 1998;12:12726.

9. Sleckman BG, Mauch PM, Ambinder RF. Correlation between EBV-status and risk factor profile in Hodgkins disease. Cancer EpidemiolBiomark Prev 1998;7:111721.

10. Alexander FE, Jarrett RF, Lawrence DJ, Armstrong AA, Freeland J,Gokhale DA, Kane E, Taylor GM, Wright DH, Cartwright RA. Riskfactors for Hodgkins disease by Epstein-Barr virus (EBV) status: priorinfection by EBV and other agents. Br J Cancer 2000;82:111721.

11. Alexander FE, Jarrett RF, Cartwright RA, Armstrong AA, Gokhale D,Kane E, Gray D, Lawrence DJ, Taylor GM. Epstein-Barr virus and

HLA-DPB1-*0301 in young adult Hodgkins disease: evidence forinherited susceptibility to Epstein-Barr virus in cases that are EBVve.Cancer Epidemiol Biomark Prev 2001;10:7059.

12. Lawrence DJ. The epidemiology of Hodgkins disease with specialreference to EBV status. PhD Thesis, University of Edinburgh, Ed-inburgh, 1999.

13. Harris NL, Jaffe ES, Stein H, Banks PM, Chan JKC, Cleary ML,Delsol G, Dewolfpeeters C, Falini B, Gatter KC, Grogan TM, Isaac-son PG, et al. A revised European-American classification of lym-phoid neoplasmsa proposal from the International LymphomaStudy Group. Blood 1994;84:136192.

14. The PACE (population adjusted clinical epidemiology) strategy: anew approach to multi-centred clinical research. QJ Med 1997;90:

14751.15. Douglas AS, Brown T, Reid D. Infectious mononucleosis andHodgkins disease - A similar seasonality. Leuk Lymph, 1996; 23:323-331.

16. Jarman B, Townsend P, Carstairs V. Deprivation indices. Br J Med1991;303:523.

17. Clayton D, Hills M. Statistical methods in epidemiology. New York:Oxford University Press, 1993.

18. Law GR, Smith AG, Roman E, for the United Kingdom ChildhoodCancer Study Investigators. The importance of full participation: lessonsfrom a national case-control study. Br J Cancer 2002;86:35055.

19. Jarrett RF. Viruses and Hodgkins lymphoma. Ann Oncology 2002;13(Suppl 1):239.

20. Flavell KJ, Biddulph JP, Powell JE, Parkes SE, Redfern D, WeinrebM, Nelson P, Mann JR, Young LS, Murray PG. South Asian ethnicityand material deprivation increase the risk of Epstein-Barr virus infec-tion in childhood Hodgkins disease. Br J Cancer 2001;85:350 6.

302 ALEXANDER ET AL.