Infectious Mononucleosis and Risk for
Multiple Sclerosis: A Meta-analysis
Evan L. Thacker, SM,1Fariba Mirzaei, MD, MPH1,2and Alberto Ascherio, MD, DrPH1,2
Objective: To characterize the association between infectious mononucleosis (IM), a frequent clinical manifestation of
primary Epstein–Barr virus infection after childhood, and the risk for multiple sclerosis (MS). Methods: We conducted
a systematic review and meta-analysis of case–control and cohort studies of IM and MS. Results: The combined relative
risk of MS after IM from 14 studies was 2.3 (95% confidence interval, 1.7–3.0; p < 10?8). Potential sources of heter-
ogeneity (ie, study design, MS definition, and latitude) barely influenced our results. Interpretation: We conclude that
Epstein–Barr virus infection manifesting as IM in adolescents and young adults is a risk factor for MS.
Ann Neurol 2006;59:499–503
Viral infections have long been suspected of playing a
role in multiple sclerosis (MS).1Increasing evidence
implicates infection with Epstein–Barr virus (EBV) as
an important MS risk factor.2The similarities in the
epidemiology of infectious mononucleosis (IM) and
MS led investigators to consider EBV in the cause of
MS some 25 years ago.1Both IM and MS occur in
young adults, both follow a latitude gradient, and both
are rare in populations where infections occur at an
early age, suggesting that late infection with EBV, ev-
idenced by occurrence of IM, is an important causal
factor in MS.1,3
However, studies that have evaluated the relation of
IM and MS risk have not produced consistent results.
A number of case–control studies and a well-designed
cohort study found an increased risk for MS in indi-
viduals with a history of IM,3,5–9whereas several case–
control studies and two cohort studies did not find any
difference.4,10–16In this review, we systematically iden-
tify and combine the relevant studies of the association
between IM history and MS to produce a summary
measure of the association.
Subjects and Methods
We searched Medline without language restrictions from
1965 through March 2005 using the following terms: Ep-
stein–Barr virus or EBV or human herpesvirus 4 or HHV-4
or infectious mononucleosis or glandular fever, and multiple
sclerosis or MS or disseminated sclerosis. We also reviewed
bibliographies, searched the Science Citation Index Ex-
panded database, contacted authors and experts, and
searched Medline for all case–control studies of MS to iden-
tify reports including IM only as a minor topic.
Publications reporting results from analytic epidemiological
studies of IM and MS were eligible for extraction. Two of us
independently extracted relative risks (RRs), confidence in-
tervals (CIs), and other information from each study, resolv-
ing discrepancies by joint review. We assigned latitude for
each study as the latitude of the nearest major city to where
the study was conducted or where the majority of study sub-
We based our meta-analysis on DerSimonian and Laird’s17
random-effects model implemented in Stata version 9.0
(Stata Corporation, Stanford, CA).18According to a priori
criteria jointly evaluated by two of us, studies had to have
crude or adjusted RR and CI reported in the article, estima-
ble from data provided in the article or from additional in-
formation obtained by personal contact with authors to be
included in the meta-analysis. After completing the search,
however, we noticed that two studies reported a lack of sig-
nificant association between IM and MS without providing a
RR estimate or CI. Because exclusion of these studies could
bias the summary RR, we included these studies assuming a
RR of 1.0 and estimating CI based on the number of re-
ported MS cases and the likely population prevalence of IM.
For studies reporting adjusted results, the adjusted results
were entered into the meta-analysis to minimize the effects of
confounding. For studies with only crude results, the crude
From the Departments of1Nutrition and2Epidemiology, Harvard
School of Public Health, Boston, MA.
Received Nov 10, 2005, and in revised form Jan 6, 2006. Accepted
for publication Jan 10, 2006.
(www.interscience.wiley.com). DOI: 10.1002/ana.20820
onlineFeb 23,2006inWiley InterScience
E.L.T. and F.M. contributed equally to this study.
Address correspondence to Mr. Thacker, Harvard School of Public
Health, Department of Nutrition, 655 Huntington Avenue, Boston,
MA 02115. E-mail: email@example.com
© 2006 American Neurological Association
Published by Wiley-Liss, Inc., through Wiley Subscription Services
results were entered into the meta-analysis. Studies were
weighted by the precision of the RR, based on the 95% CI.
To check the robustness of our meta-analysis, we repeated
the analysis under various assumptions of precision in studies
for which we estimated CIs without complete data, and we
assessed the influence of individual studies on the combined
RR by dropping each study one at a time. Furthermore, we
used stratified analyses and meta-analysis regression to assess
the effects of several variables on heterogeneity among the
individual study’s RRs, including study design (case–control
vs cohort), MS case definition (ie, definite vs definite ?
probable; incident vs prevalent), reported association (crude
vs adjusted), and latitude of study locations. A funnel plot
and a regression asymmetry plot were used to check for pub-
lication bias.19,20p less than 0.05 was considered significant.
Our search found 295 publications. We excluded 280
studies; common exclusions were reports on biochem-
istry, virology, and immunology and studies of risk fac-
tors for MS other than IM. Of the 15 studies retained
for full extraction, we excluded 1 study21because it
contained insufficient data to calculate the RR and we
received no reply from our attempt to retrieve the data
from the authors. The remaining 14 studies3–16are
summarized in Table 1. Two of these12,13reported
null results without specifying the RR and CI. Al-
though we contacted the authors of these articles, nei-
ther could provide us with the necessary data, so we
assumed RR ? 1 and calculated the CI as described
The 14 studies in the meta-analysis included 11 ca-
se–control studies3,5–8,11–16and three cohort stud-
ies.4,9,10RRs, reported as odds ratios or risk ratios,
ranged from 0.8 to 17, with a median of 2.5. Meta-
analysis (Fig 1) produced a combined RR of 2.3 (95%
CI, 1.7–3.0; p ? 10?8). There was no significant het-
erogeneity among the 14 individual study results
(Qdf?13? 16.074; p ? 0.245).
Stratified meta-analyses showed a slightly higher
combined RR for cohort studies versus case–control
studies, studies of incident MS cases versus prevalent
cases, and studies limited to definite MS cases versus
definite and probable cases, but none of these differ-
ences was statistically significant. Latitude of study lo-
cations was not sufficiently varied to detect heteroge-
neity due to latitude. Six studies reporting only crude
results had significantly lower RRs than studies report-
Table 1. Studies of the Association between Infectious Mononucleosis and Multiple Sclerosis
Criteria Case Source
Control or Comparison
AscertainmentRR (95% CI)
145:1454.4:1Definite/probable MS MS cohort (prevalent) Friends of casesInterview or
153:1531.7:1 Definite MS
1. Definite MS
2. Optic neuritis
3. Isolated demyelina-
Hospital (incident)Hospital1 (0.19–5.37)
16:61 4.3:1Community (preva-
Community Interview1 (0.09–5.55)a
2. Blood donors
104:1502.1:1 Community (preva-
2. Hospital staff
Interview 1 (0.31–3.06)a
155:1551.63:1 Interview 3.03 (0.24–160.55)
301:1416All female Definite/probable MS
Poser criteria and
Cohort study (inci-
dent and prevalent)
MS center (prevalent)
Cohort study2.2 (1.6–3.0)
140:1311.8:1 Blood donors 0.8 (0.3–2.2)
53:5312.3:1MS Society (Incident) Friends of casesQuestion-
1. MS Society
2. Neurologists (prev-
3 cases Not reportedDefinite MS
Poser criteria or
Allison criteria and
Hospital (incident)Regional populationMedical
MS registry (incident)National populationHeterophile
6 cases Not reported Hospital records (inci-
Hospital records2.17 (0.79–4.77)
aConfidence intervals for these studies could not be calculated from data given in the articles or furnished by authors, and were calculated based
on the number of reported MS cases and the likely population prevalence of infectious mononucleosis.
RR ? relative risk; CI ? confidence interval; MS ? multiple sclerosis.
Annals of Neurology Vol 59 No 3March 2006
ing adjusted results (p ? 0.01). However, most of
these crude studies were imprecise, contributing little
weight to the meta-analysis. Studies reporting adjusted
RRs, taken as a subgroup, produced a combined RR of
2.5 (95% CI, 2.0–3.2; p ? 10?12), with no significant
heterogeneity (Qdf?7? 7.389; p ? 0.390), which is
similar to the main meta-analysis. Results from the
stratified analyses and meta-analysis regression are
given in Table 2.
We experimented with various precision levels for
the two studies reporting null results, for which we cal-
culated CIs.12,13Even assuming tight precision (RR ?
1.0; 95% CI, 0.9–1.1), the combined RR remained
significantly above the null, the combined CI covering
the point estimate of 2.3 from the primary meta-
analysis. This strongly suggests that the true association
is indeed nonnull, likely close to 2.3. We identified
one unpublished study of IM and MS (by Dr B.M.
Zaadstra), reporting a RR of 2.1 and 95% CI of 1.7 to
2.8. Because we were unable to obtain details about
this study, we did not include it in our main analysis.
Adding these data barely changed our result: combined
RR ? 2.2 (95% CI, 1.8–2.7; p ? 10?14), with no
significant heterogeneity (Qdf?14 ? 16.249; p ?
Excluding studies from the main analysis one at a
time demonstrated that no single study was overly in-
fluential. The average combined RR excluding single
studies was virtually identical to the combined RR
from the main meta-analysis. Tests and plots for pub-
lication bias gave null results.
The results of this study combined with those of sero-
logical investigations22–24suggest that the risk for MS,
which is extremely low in individuals who are not in-
fected with EBV, increases after infection. Replication
of the positive association between IM and MS in
many studies suggests that the association is not a
product of chance. It is possible that the positive asso-
ciation in case–control studies arose spuriously through
recall bias, as many of these studies ascertained the oc-
currence of IM retrospectively through interview or
questionnaire, but the consistency of results between
case–control and cohort studies suggests that the influ-
ence of this source of error was at most modest. A spu-
rious positive association between IM and MS could
also have occurred if individuals with history of IM
were more likely to be diagnosed with MS than indi-
viduals without such history. This appears to be an un-
likely proposition, because in the majority of studies in
the meta-analysis,3,8–15the diagnosis of MS was based
Fig 1. Random-effects meta-analysis of 14 studies of the associ-
ation between infectious mononucleosis and multiple sclerosis.
Confidence intervals are truncated at 20. RR ? relative risk.
Table 2. Stratified Meta-analysis and Meta-analysis Regression for Assessing Potential Sources of Heterogeneity among Studies of the
Association between Infectious Mononucleosis and Multiple Sclerosis
VariableStudies, n RR (95% CI)
p for Meta-analysis
Incident ? prevalent3,4,14
MS diagnosis certaintya
Definite ? probable3,5–7,9,12,14
aThree studies4,15,16did not report the MS diagnosis certainty.
RR ? relative risk; CI ? confidence interval; MS ? multiple sclerosis; NA ? not applicable.
Thacker et al: Infectious Mono and Risk for MS
on standard criteria, minimizing the possibility that
knowledge of exposure would influence diagnosis.
Using the results of this meta-analysis and results
from other investigations, we can conceptualize a
model of the relation between EBV infection and MS
that accounts for early-life infection versus infection in
adolescence or adulthood, which often manifests as
IM24(Fig 2). According to this model, the risk for MS
is close to zero among individuals who are not EBV
infected, intermediate among individuals infected with
EBV in early childhood, and highest among individu-
als first infected with EBV in adolescence or later in
life. This model predicts that many cases of MS could
be prevented by a suitable vaccine that protects against
EBV infection. A smaller but still substantial number
of cases could be prevented by exposing children to
EBV infection early in life. A critical test of the pro-
posed model is difficult in the absence of a safe and
effective vaccine, but corroborating evidence could be
accrued by conducting large longitudinal studies of
EBV infection and MS risk, by investigating the ge-
netic and environmental factors (including other infec-
tions) that interact with EBV, and by elucidating the
molecular mechanism that may link EBV or the im-
mune response to EBV to MS. These studies could be
extended to other autoimmune diseases, particularly to
systemic lupus erythematosus, which also is strongly as-
sociated with EBV infection.28An early event in sys-
temic lupus erythematosus appears to be the appear-
ance of antibodies to EBV that cross-react with a lupus
autoantigen.29Molecular mimicry mechanisms have
also been proposed for MS,30,31but their causative rel-
evance remains to be proved. Whereas there is no
doubt that EBV is at best only one component in the
complex pathway of events that cause MS, it could
prove to be one of the components that is more vul-
nerable to intervention.
1. Ascherio A, Munger K. Multiple sclerosis. In Nelson LM, ed.
Neuroepidemiology: from principles to practice. Oxford: Ox-
ford University Press, 2004.
2. Ascherio A, Munch M. Epstein-Barr virus and multiple sclero-
sis. Epidemiology 2000;11:220–224.
3. Hernan MA, Zhang SM, Lipworth L, et al. Multiple sclerosis
and age at infection with common viruses. Epidemiology 2001;
Fig 2. Schematic representation of multiple sclerosis (MS) incidence according to Epstein–Barr virus (EBV) infection. The shape of
the incidence curve labeled “Early EBV infection without IM” is based on the typical age-specific incidence of MS in most popula-
tions; incidence begins to increase in adolescence, peaks around age 25 to 30 years, and declines to nearly zero by age 60.25The
age-specific incidence for the group with no EBV infection has been drawn at one-tenth the incidence among EBV-positive individ-
uals, based on the results of a previous review,26and that of individuals with late EBV infection and infectious mononucleosis (IM)
has been estimated to be 2.3 times higher than that of EBV-positive individuals without history of IM (this meta-analysis). More
accurate curves could be drawn by taking into account the proportion of individuals in the population who are infected with EBV
early in childhood and the age-specific prevalence of history of IM. We have ignored these adjustments for simplicity, and because
these proportions vary across developed countries.27
Annals of Neurology Vol 59No 3 March 2006
4. Goldacre MJ, Wotton CJ, Seagroatt V, Yeates D. Multiple scle-
rosis after infectious mononucleosis: record linkage study. J Epi-
demiol Community Health 2004;58:1032–1035.
5. Operskalski EA, Visscher BR, Malmgren RM, Detels R. A case-
control study of multiple sclerosis. Neurology 1989;39:
6. Martyn CN, Cruddas M, Compston DA. Symptomatic
Epstein-Barr virus infection and multiple sclerosis. J Neurol
Neurosurg Psychiatry 1993;56:167–168.
7. Marrie RA, Wolfson C, Sturkenboom MC, et al. Multiple scle-
rosis and antecedent infections: a case-control study. Neurology
8. Ponsonby AL, van der Mei I, Dwyer T, et al. Exposure to in-
fant siblings during early life and risk of multiple sclerosis.
9. Haahr S, Koch-Henriksen N, Moller-Larsen A, et al. Increased
risk of multiple sclerosis after late Epstein-Barr virus infection: a
historical prospective study. Mult Scler 1995;1:73–77.
10. Lindberg C, Andersen O, Vahlne A, et al. Epidemiological in-
vestigation of the association between infectious mononucleosis
and multiple sclerosis. Neuroepidemiology 1991;10:62–65.
11. Souberbielle BE, Martin-Mondiere C, O’Brien ME, et al. A
case-control epidemiological study of MS in the Paris area with
particular reference to past disease history and profession. Acta
Neurol Scand 1990;82:303–310.
12. Hopkins RS, Indian RW, Pinnow E, Conomy J. Multiple scle-
rosis in Galion, Ohio: prevalence and results of a case-control
study. Neuroepidemiology 1991;10:192–199.
13. Casetta I, Granieri E, Malagu S, et al. Environmental risk fac-
tors and multiple sclerosis: a community-based, case-control
study in the province of Ferrara, Italy. Neuroepidemiology
14. Gusev E, Boiko A, Lauer K, et al. Environmental risk factors in
MS: a case-control study in Moscow. Acta Neurol Scand 1996;
15. Zorzon M, Zivadinov R, Nasuelli D, et al. Risk factors of mul-
tiple sclerosis: a case-control study. Neurol Sci 2003;24:
16. Haahr S, Plesner AM, Vestergaard BF, Hollsberg P. A role of
late Epstein-Barr virus infection in multiple sclerosis. Acta Neu-
rol Scand 2004;109:270–275.
17. DerSimonian R, Laird NM. Meta-analysis in clinical trials.
Control Clin Trials 1986;7:177–188.
18. Sharp S, Sterne J. sbe16: meta-analysis. Stata Tech Bull 1997;
19. Begg CB, Mazumdar M. Operating characteristics of a rank
correlation test for publication bias. Biometrics 1994;50:
20. Egger M, Davey Smith G, Schneider M, Minder C. Bias in
meta-analysis detected by a simple, graphical test. BMJ 1997;
21. Bachmann S, Kesselring J. Multiple sclerosis and infectious
childhood diseases. Neuroepidemiology 1998;17:154–160.
22. Ascherio A, Munger KL, Lennette ET, et al. Epstein-Barr virus
antibodies and risk of multiple sclerosis. JAMA 2001;286:
23. Alotaibi S, Kennedy J, Tellier R, et al. Epstein-Barr virus in
pediatric multiple sclerosis. JAMA 2004;291:1875–1879.
24. Levin LI, Munger KL, Rubertone MV, et al. Temporal rela-
tionship between elevation of Epstein-Barr virus antibody titers
and initial onset of neurological symptoms in multiple sclerosis.
25. Koch-Henriksen N. The Danish Multiple Sclerosis Registry: a
50-year follow-up. Mult Scler 1999;5:293–296.
26. Ascherio A, Munch M. Epstein-Barr virus and multiple sclero-
sis. Epidemiology 2000;11:220–224.
27. Niederman JC, Evans AS. Epstein-Barr virus. In: Evans AS,
Kaslow RA, eds. Viral infections of humans: epidemiology and
control. New York: Plenum Publishing, 1997.
28. James JA, Kaufman KM, Farris AD, et al. An increased preva-
lence of Epstein-Barr virus infection in young patients suggests
a possible etiology for systemic lupus erythematosus. J Clin In-
29. McClain MT, Heinlen LD, Dennis GJ, et al. Early events in
lupus humoral autoimmunity suggest initiation through molec-
ular mimicry. Nat Med 2005;11:85–89.
30. Lang HL, Jacobsen H, Ikemizu S, et al. A functional and struc-
tural basis for TCR cross-reactivity in multiple sclerosis. Nat
31. Rand KH, Houck H, Denslow ND, et al. Epstein-Barr virus
nuclear antigen-1 (EBNA-1) associated oligoclonal bands in pa-
tients with multiple sclerosis. J Neurol Sci 2000;173:32–39.
Thacker et al: Infectious Mono and Risk for MS