American Journal of Gastroenterology ISSN 0002-9270


C 2006 by Am. Coll. of Gastroenterology doi: 10.1111/j.1572-0241.2006.00741.x
Published by Blackwell Publishing

Rotavirus Infection Frequency and Risk of Celiac Disease

Autoimmunity in Early Childhood: A Longitudinal Study
∗ ∗
Lars C. Stene, Ph.D.,1,2, Margo C. Honeyman, Ph.D.,3, Edward J. Hoffenberg, M.D.,4 Joel E. Haas, M.D.,5
Ronald J. Sokol, M.D.,4 Lisa Emery, M.S.P.H.,6 Iman Taki, M.S.P.H.,6 Jill M. Norris, Ph.D.,6
Henry A. Erlich, Ph.D.,7 George S. Eisenbarth, M.D., Ph.D.,1 and Marian Rewers, M.D., Ph.D.1,6
1
Barbara Davis Center for Childhood Diabetes, University of Colorado School of Medicine, Aurora, Colorado;
2
Division of Epidemiology, Norwegian Institute of Public Health, Oslo, Norway; 3 Autoimmunity and
Transplantation Division, Walter and Eliza Hall Institute of Medical Research, Parkville, Melbourne, Victoria,
Australia; 4 Section of Pediatric Gastroenterology, Hepatology and Nutrition, Department of Pediatrics, The
Children’s Hospital, University of Colorado at Denver and Health Sciences Center, Denver, Colorado;
5
Department of Pathology, The Children’s Hospital, University of Colorado at Denver and Health Sciences
Center, Denver, Colorado; 6 Department of Preventive Medicine and Biometrics, University of Colorado at
Denver and Health Sciences Center, Denver, Colorado; and 7 Department of Human Genetics, Roche Molecular
Systems, Inc, Alameda, California

OBJECTIVE: Few studies have assessed the role of specific gastrointestinal infections in celiac disease. We
investigated whether increased frequency of rotavirus infection, a common cause of gastrointestinal
infection and inflammation, predicts increased risk of celiac disease autoimmunity.
METHODS: A cohort of 1,931 children from the Denver metropolitan area who carried celiac disease human
leukocyte antigen (HLA) risk alleles were followed from infancy for development of celiac disease
autoimmunity, defined as positivity at two or more subsequent clinic visits for tissue
transglutaminase (tTG) autoantibodies measured using a radioimmunoassay with human
recombinant tTG. Blood samples were obtained at ages 9, 15, and 24 months, and annually
thereafter. Rotavirus antibodies were assayed using an indirect enzyme immunoassay in serial
serum samples from each case and two matched controls. Frequency of infections were estimated
by the number of increases (>2 assay coefficient of variation) in rotavirus antibody between clinic
visits.
RESULTS: Fifty-four cases developed celiac disease autoimmunity at a median age of 4.4 yr. Thirty-six had an
intestinal biopsy, of which 27 (75%) were positive for celiac disease. Frequent rotavirus infections
predicted a higher risk of celiac disease autoimmunity (compared with zero infections, rate ratio
1.94, 95% confidence interval [CI] 0.39–9.56, for one infection and rate ratio 3.76, 95% CI
0.76–18.7, for ≥2 infections, rate ratio for trend per increase in number of infections = 1.94, 95%
CI 1.04–3.61, p = 0.037). The result was similar after adjustment for gender, ethnic group,
maternal education, breast-feeding, day-care attendance, number of siblings, season of birth, and
number of HLA DR3-DQ2 haplotypes.
CONCLUSIONS: This prospective study provides the first indication that a high frequency of rotavirus infections may
increase the risk of celiac disease autoimmunity in childhood in genetically predisposed individuals.
(Am J Gastroenterol 2006;101:2333–2340)

INTRODUCTION DQA1∗05 and DQB1∗02 alleles (DQ2). The remaining pa-

Celiac disease is characterized by small intestinal villous at-
rophy that resolves with removal of dietary wheat, rye, and
barley, and is associated with increased morbidity and mor-
tality (1). There is a clear genetic predisposition to celiac
disease. Most patients carry human leukocyte antigen (HLA)
∗
These two authors contributed equally to this work.
tients mostly carry HLA-DQA1∗0301 and −DQB1∗0302 al-
leles (DQ8) (2). While high-risk HLA alleles and exposure
to gluten are highly prevalent, the majority of people with
genetic susceptibility who are exposed to dietary gluten does
not develop celiac disease. Only part of the familial aggrega-
tion observed for celiac disease seems to be explained by the
HLA genes. Therefore, additional risk factors probably play
a role. To date, few other genetic factors have been identified

2333
2334 Stene et al.
(1), and the few studies evaluating environmental factors have
mostly used retrospective designs and focused on clinically
identified symptomatic cases of celiac disease (3–6). Spe-
cific and sensitive screening tests for serum tissue transglu-
taminase (tTG) or endomysial antibodies have been used to
identify people with “silent” forms of the disease, demon-
strating that celiac disease is much more common than pre-
viously thought (7–12). Furthermore, it has been shown that
screening-identified cases of celiac disease have significant
morbidity (13). Prospective studies that include not only clini-
cally identified patients, but also screening-identified patients
should facilitate the search for environmental risk factors
likely to trigger celiac disease autoimmunity.
Intestinal infection and inflammation may initiate the se-
cretion of intracellular tTG and lead to increased intesti-
nal permeability. Increased intestinal permeability, frequently
observed in celiac disease, permits absorption of intact
gliadin molecules and may initiate the immune process lead-
ing to celiac disease (2). Few studies have investigated the
role of specific infectious agents in the development of celiac
disease. Rotavirus infection, one of the most common causes
of acute gastroenteritis in children worldwide (14), repre-
sents a plausible risk factor. Rotavirus infections have pre-
viously been implicated in the autoimmunity leading to type
1 diabetes (15). They have been shown to induce inflamma-
tory cytokine expression in the intestine, increase intestinal
permeability, and lead to transient structural changes in the
intestine (14). Most children in the community have had an
infection with rotavirus by 3 yr of age. Repeated infections are
common, detectable by an increase in the level of rotavirus-
specific antibodies (16, 17). It seems that HLA-DQ2, present
in as many as 20–30% of healthy subjects in Caucasoid pop-
ulations (18), is necessary but not sufficient for development
of celiac disease (2). Because rotavirus infections are ubiq-
uitous, it can be inferred by simple reasoning that if a single
rotavirus infection was sufficient to cause celiac disease, one
would have expected the prevalence of celiac disease to ap-
proach that of HLA-DQ2 in the population. However, the ob-
served disease prevalence is less than 1% in most populations
(11, 12). Thus, a single infection with rotavirus is unlikely
to be a sufficient cause of celiac disease. We therefore hy-
pothesized that an increased frequency of infections might
contribute to the development of celiac disease. The objec-
tive of this study was to investigate whether an increased fre-
quency of serologically defined rotavirus infections predicts
an increased risk of subsequent celiac disease autoimmunity
in children.
METHODS
Subjects
Subjects were participants in a prospective study of the natu-
ral history and environmental triggers of diabetes and celiac
disease autoimmunity in genetically predisposed children re-
cruited from two sources (8, 19). The study was approved
by the Colorado Multiple Institutional Review Board, and
informed consent was obtained from parents of all children
in the study. From December 1993 to January 2003, 86%
of families of children born at St. Joseph’s hospital in Den-
ver who were approached gave informed consent to genetic
screening. Over 27,800 cord blood samples from newborns
have been screened for diabetes and celiac disease-associated
HLA genotypes. Children born at St. Joseph’s hospital have
a distribution of ethnic groups similar to that of the general
population of the Denver metropolitan area and include chil-
dren classified by their mothers as non-Hispanic white (56%),
Hispanic (30%), African American (7%), Asian American
(2%), or biracial/other (5%) (10). Families in which par-
ents had difficulties understanding English or whose infant
had a severe congenital malformation or disease were ex-
cluded. Newborns were included for follow-up if they carried
HLA-DRB1∗03 (DR3-DQ2 haplotype) or DRB1∗04 together
with DQB1∗0302 (DR4-DQ8 haplotype) or both. DRB1∗03
is found almost exclusively together with DQA1∗0501 and
DQB1∗0201 (DR3-DQ2 haplotype).
The second recruitment source involved siblings or off-
spring of persons with type 1 diabetes attending clinics in the
Denver metropolitan area, primarily the Barbara Davis Cen-
ter for Childhood Diabetes. First-degree relatives of persons
with type 1 diabetes have been previously shown to have an
increased risk of celiac disease (20), and they have increased
prevalence of HLA DR3-DQ2 and −DR4-DQ8 haplotypes.
First-degree relatives of persons with type 1 diabetes were
included in the celiac disease autoimmunity study regardless
of their HLA-genotype, but all cases and controls included in
the current study had at least one DR3-DQ2 or DR4-DQ8 hap-
lotype, except seven controls. For each of the seven matched
sets of a case and two controls of which one control did not
carry DR3-DQ2 or DR4-DQ8, the other control did so, thus
ensuring that each matched set included a control child with
at least one DR3-DQ2 or DR4-DQ8 haplotype (see also para-
graph on selection of controls below).
Samples of whole blood in EDTA were sent to the HLA typ-
ing laboratory at Roche Molecular Systems, Inc., Alameda,
CA for PCR-based class II genotyping, as previously de-
scribed (19). Most participants among the newborns from
the general population were recruited within a day of birth,
and scheduled for clinic visits at ages 9, 15, and 24 months,
and annually thereafter. Blood samples were taken when chil-
dren were generally healthy, without symptoms of disease.
Some of the first-degree relatives of persons with type 1 di-
abetes were recruited at a later age. Venous blood samples
were collected at each clinic visit, and children who tested
positive for autoantibodies were scheduled for more frequent
visits (3–6 months). Additional information was obtained at
structured interviews with the families when the child was
3 and 6 months old, and at each clinic visit. Up to February
2003, a total of 1,931 children had been tested for tTG au-
toantibodies at least once, and 1,404 had been tested at least
twice.

Measurement of tTG Autoantibodies and Case Definition
A radioimmunoassay with in vitro transcribed and translated
human tTG cDNA was used to detect IgA antibodies to tTG
in serum samples as previously described (8, 21, 22). The re-
sult is given as an index ([unknown sample cpm − negative
control cpm]/[positive control cpm − negative control cpm]).
The a priori defined cutoff for positivity was three times the
100th percentile in 184 healthy control subjects, correspond-
ing to an index of 0.05 (21). Before 1998 endomysial an-
tibody (EMA) was used to screen the children in the study
cohort. When the tTG autoantibody radioimmunoassay was
established, all sera previously tested positive on EMA were
tested and confirmed positive for tTG autoantibodies. After
positive anti-tTG tests on two or more consecutive examina-
tions 3–12 months apart, clinical evaluation and small-bowel
biopsy was recommended. In some cases, families requested
a biopsy after a single positive anti-tTG test. The case defini-
tion of celiac disease autoimmunity in this study was positiv-
ity for tTG autoantibodies on two or more consecutive clinic
visits, or positivity on a single visit plus a diagnosis of celiac
disease by intestinal biopsy. We have previously shown in
this study population that positivity for tTG autoantibodies
on two or more consecutive clinic visits corresponds to a 70–
83% positive predictive value for biopsy-confirmed celiac
disease without long-term follow-up (8).
Intestinal Biopsies
Small intestinal biopsy specimens were obtained by Carey
capsule (Wilson-Cook Medical, Inc., Winston/Salem, NC,
N = 4) or by upper gastrointestinal endoscopy (N = 32)
with two to four specimens from the descending duodenum.
A single pathologist (JEH), blinded to clinical information,
assessed the biopsy specimens according to the scoring sys-
tem described by Marsh (23). A score of two (enlarged crypts
and increased number of intraepithelial lymphocytes) or three
(any degree of villous atrophy) was considered confirmatory
for celiac disease (8).
Selection of Controls
Two matched controls were selected for each case, using a
nested case-control design with risk set sampling (24). Con-
trols were matched for the presence of a first-degree relative
with type 1 diabetes (21 of the 54 cases had a first-degree
relative with type 1 diabetes), age at first clinic visit, number
of clinic visits, and follow-up time (at least as long follow-up
time as the corresponding case (24)). The risk set sampling
(also called density sampling) assures that risk ratios (as op-
posed to odds ratios) can be estimated using conditional logis-
tic regression (24). Additional potential confounding factors
were adjusted for in the data analysis (see “Results” section).
Measurement of Serum Rotavirus Antibodies and
Definition of Exposure
IgA and IgG antibodies to rotavirus were tested blindly as
previously described, using an indirect enzyme immunoas-
say with simian rotavirus SA11 as the antigen (15–17). All
samples from each case and its matched controls were always
Rotavirus and Celiac Disease Autoimmunity 2335
assayed simultaneously on the same microtiter tray, in order to
minimize assay variation. Thirty randomly selected samples
were split and the blinded duplicates tested together with the
other samples. The mean within-assay coefficient of variation
(CV) was 9.1% for IgA− and 8.9% for IgG absorbance.
An operational definition of the number of rotavirus in-
fections (an “infection index,” adapted from an earlier study
(15)) was developed prior to inspecting the data. The cut-
off values for serum rotavirus antibody positivity at the first
clinic visit were defined as the mean + 2SD among 40 con-
trol children with median age 9 months (range 7–11 months)
without previous serological evidence of rotavirus infection.
The first sign of infection was defined for a serum rotavirus
IgA or IgG above the cutoff at the first clinic visit. The in-
fection index increased by one for every >2 within-assay
CV increase in either IgA or IgG between two consecutive
clinic visits (i.e., >2 CV increase since the previous clinic
visit), given that the last value was above the cutoff value.
Only rotavirus infections occurring before and up to the
clinic visit when a case child first tested positive for anti-tTG,
and the corresponding ages for the matched controls, were
counted.
Statistical Analysis
We decided a priori to use the rotavirus infection index as
the main exposure variable. Prestudy power analysis, based
on the distribution of the rotavirus infection index in a small
pilot study, showed that 54 cases and 108 controls would pro-
vide 80% power to detect a rate ratio of 3.0 comparing an
infection index of ≥2 with ≤1. Data were analyzed using
SPSS and STATA. Conditional logistic regression analysis
was used to estimate the rate ratios with 95% confidence in-
tervals (CIs), accounting for the matched design. The number
of infections was entered as a categorical variable in the re-
gression model, coded 0, 1, and 2 for zero, one, and ≥2 infec-
tions, respectively. Tests for trend were obtained by treating
the number of infections as a continuous variable in the lo-
gistic regression model. Potential confounding factors such
as duration of breast-feeding and season of birth were ad-
justed for by including these variables as covariates in the
regression models. p-values less than 0.05 or 95% CIs for the
rate ratio not overlapping 1.00 were regarded as statistically
significant.
RESULTS
During the current study period, 54 children met the crite-
ria for celiac disease autoimmunity. Of 36 cases with one
DRB1∗03 allele, 14 also carried DRB1∗ 04-DQB1∗ 0302, and
among the 53 control children who carried one DRB1∗ 03 al-
lele, 20 also carried DRB1∗ 04-DQB1∗ 0302. Among the eight
cases with no DRB1∗ 03 allele (but with at least one DRB1∗ 04-
DQB1∗ 0302), 4 carried DRB1∗ 04-DQB1∗ 0302/DRB1∗ 04-
DQB1∗ 0302, and among the 50 controls with no DRB1∗ 03
allele (but with at least one DRB1∗ 04-DQB1∗ 0302), 14 car-
ried DRB1∗ 04-DQB1∗ 0302/DRB1∗ 04-DQB1∗ 0302.
2336 Stene et al.

By February 28, 2004, 36 children had undergone an in-
testinal biopsy, which confirmed the presence of celiac dis-
ease in 27 cases (75%). The lowest tTG autoantibody titer
among the biopsy confirmed cases was 0.08, and the me-
dian was 0.61. Among the nonbiopsied cases, only one had
a titer lower than 0.08 (0.06), and the median was 0.38. The
distribution of titers among the nonbiopsied cases was not
statistically different from that among the biopsy-confirmed
cases (Mann-Whitney test, p = 0.46). The median age at the
first visit when the tTG autoantibody test was positive was 4.4
yr, the lowest age at first positivity being 2 yr. The median age
at the end of follow-up for the controls was 6.5 yr. There were
no significant differences in any of the variables in Table 1
between the 36 cases who agreed to have a biopsy and those
18 who did not, although there was a nonsignificant tendency
that a higher percentage of those who did not have a biopsy
were of ethnic groups other than non-Hispanic white (5 of 18
vs 3 of 36, p = 0.10).
Characteristics and factors that we considered as poten-
tial confounders in the study of rotavirus infections, includ-
ing distribution of HLA alleles, are shown in Table 1. Early
day-care attendance, increasing number of older siblings and
DR3-DQ2 haplotypes all predicted a significantly increased
risk of celiac disease autoimmunity. Only two cases and
one control had a first-degree relative (mother) with celiac
disease.
Rotavirus antibodies were tested in serum from up to seven
clinic visits per child, with a mean of three clinic visits before
development of anti-tTG in case children and the correspond-
ing age for matched controls. Table 2 shows the rate ratios
for association between number of infections and subsequent
risk of celiac disease autoimmunity. A higher frequency of
rotavirus infections predicted a higher risk of celiac disease
autoimmunity. The rate ratio for increase in rate per increase
in number of infections was 1.94 (95% CI 1.04–3.61), with
a corresponding p-value for test for trend of 0.037. After
simultaneous adjustment for all the potential confounders
(gender, ethnic group, maternal education, breast-feeding,
day care, older siblings, season of birth, and number of DR3-
DQ2 haplotypes), increasing number of rotavirus infections
still significantly predicted increased risk of celiac disease
autoimmunity.
The magnitude of the rate ratios for the effect of rotavirus
infection frequency on the risk of celiac disease autoimmu-
nity were not significantly different at different levels of the
variables listed in Table 1 (no significant interactions, all p >
0.05). For matched sets in which celiac disease was confirmed
on biopsy, the risk ratios for 1 and ≥2 rotavirus infections
(compared with zero) were 2.31 and 3.89, respectively. For
matched sets in which the case either did not have a biopsy
or the biopsy was not positive for celiac disease, the corre-
sponding risk ratios were 1.63 and 3.45.

Table 1. Characteristics of Celiac Disease Autoimmunity Cases and Controls

Cases∗ (N = 54) Controls (N = 108) Rate Ratio (95% CI)†
Female gender 31 (57.4%) 50 (46.3%) 1.54 (0.81–2.93)
Ethnicity other than non-Hispanic white‡ 8 (14.8%) 25 (23.1%) 0.57 (0.23–1.39)
Maternal education >12 yr§ 47 (90.4%) 85 (79.4%) 2.46 (0.88–6.86)
Median months of breast-feeding 6 (IQR 2, 13) 5 (IQR 1, 11) 1.026 (0.988–1.066)
Day care before 2 yr of age 36 (66.7%) 53 (49.1%) 2.00 (1.02–3.92)
No older siblings# 18 (33.3%) 47 (43.5%) 1.00 (reference)
One older sibling 20 (37.0%) 45 (41.7%) 1.17 (0.53–2.59)
≥2 older siblings 16 (29.6%) 16 (14.8%) 2.65 (1.06–6.61)
Test for trend for older siblings p = 0.048
Born January–March 11 (20.4%) 28 (25.9%) 1.00 (reference)
Born April–June 15 (27.8%) 22 (20.4%) 1.76 (0.66–4.67)
Born July–September 14 (25.9%) 30 (27.8%) 1.21 (0.47–3.11)
Born October–December 14 (25.9%) 28 (25.9%) 1.29 (0.50–3.32)
p-Value for season of birth p = 0.71
0 HLA-DRB1∗ 03 alleles∗∗ 8 (14.8%) 50 (46.3%) 1.0 (reference)
1 HLA-DRB1∗ 03 alleles 36 (66.7%) 53 (49.1%) 6.19 (2.04–18.8)
2 HLA-DRB1∗ 03 alleles 10 (18.5%) 5 (4.6%) 16.4 (3.34–80.8)
Test for trend HLA-DRB1∗ 03 alleles p < 0.001
Age (yr) at end of follow-up (median, range)†† 4.4 (2, 10) 6.5 (2, 15) n.a.
Because of the design of the study, 21 of the 54 cases had a first-degree relative with type 1 diabetes. Controls were matched for the presence of a first-degree relative with type 1
diabetes.
∗
Case status (celiac disease autoimmunity) was defined by positivity for serum tissue transglutaminase antibodies at two consecutive clinic visits (approximately 3–6 months
apart), or seropositive for tissue transglutaminase autoantibodies at one clinic visit plus positive biopsy.
†
Estimated using conditional logistic regression. CI = confidence interval.
‡
Ethnic group was self-reported. Groups other than non-Hispanic white were mainly Hispanic.
§
One case and two control children had missing data for maternal education.

Rate ratio per month increase in duration of breast-feeding. Two control children had missing data for duration of breast-feeding. IQR = interquartile range, 25th and 75th
percentiles.
#
Number of older siblings, including half- and step-siblings.
∗∗
Cases and controls were selected from a cohort defined by increased frequency of celiac disease-associated HLA alleles DRB1∗ 03 (DR3-DQ2 haplotype) or DRB1∗ 04-DQB1∗ 0302
(DR4-DQ8 haplotype).
††
Age at clinic visit when child was first positive for tissue transglutaminase autoantibodies (cases), or last clinic visit at which child was negative for tissue transglutaminase
autoantibodies (controls).
 Rotavirus and Celiac Disease Autoimmunity 2337

Table 2. Serological Evidence of Rotavirus Infections in Cases of Celiac Disease Autoimmunity and Matched Controls from a Prospective,
Nested Case-Control Study of Children at Genetically Increased Risk of Celiac Disease
Rate Ratio (95% CI)†
Cases∗ Controls
Rotavirus infections‡ (N = 54) (N = 108) Unadjusted Adjusted§
0 3 12 1.00 (reference) 1.00 (reference)
1 18 46 1.94 (0.39–9.56) 0.70 (0.09–5.53)
≥2 33 50 3.76 (0.76–18.7) 3.24 (0.39–27.3)
Test for trend p = 0.037 p = 0.035
∗
Case status was defined by positivity for serum tissue transglutaminase antibodies at two consecutive clinic visits (approximately 3–6 months apart), or seropositive for tissue
transglutaminase autoantibodies at one clinic visit plus positive biopsy, and include biopsy-confirmed cases.
†
Estimated using conditional logistic regression.
‡
Estimated number of infections during the follow-up period. The infection index counted one if serum rotavirus IgA or IgG absorbance was above cutoff on the first clinic visit
and increased one for every increase >2 intraassay CV in either serum rotavirus IgA or IgG between two consecutive clinic visits. Only infections up to the time when a case was
first positive for tissue transglutaminase autoantibodies (and corresponding time for controls) were counted.
§
Adjusted for gender, ethnic group, maternal education, duration of breast-feeding, day-care attendance before 2 yr of age, number of older siblings, season of birth, and number
of HLA DRB1∗ 03 alleles (DR3-DQ2 haplotypes). Two control children had missing data for duration of breast-feeding, and two cases and one control child had missing data for
maternal education.

DISCUSSION in biopsy-confirmed celiac disease cases with adequate sta-

In this first prospective study of a specific infectious agent
and development of tTG autoantibodies and celiac disease,
we have found that an increased frequency of rotavirus infec-
tions predicts increased risk of celiac disease autoimmunity
in children.
Our outcome was celiac disease autoimmunity, defined as
positivity for tTG autoantibodies at least twice, which has
previously been shown to correspond to a high positive pre-
dictive value for celiac disease (8). A major strength of the
present study is that serological evidence of prior rotavirus
infection was measured repeatedly in samples obtained be-
fore celiac disease autoimmunity developed. The methods
used in the present study have been validated and are likely
to detect the majority of infections within the sampling inter-
vals (16, 17). Although our definition may be less stringent
than that used in clinical practice, it is important to note that
it was defined a priori, based on a previous study (15), and
adapted for prospective follow-up with blood samples from
generally asymptomatic children. As fecal rotavirus shed-
ding can be transient, recurrent infections may be detected
more accurately by serology with sampling intervals as in
our study. While future studies could also benefit from sub-
ject inclusion at an earlier age, more frequent sampling in-
tervals, and stool studies for acute rotavirus infections, this
is unlikely to be feasible in large-scale studies such as ours.
Because the intervals between blood sampling after 1 yr of
age was 12 months (all tTG seroconversion occurred after
age 1 yr), a positive rotavirus test in the same sample interval
as the seroconversion for tTG autoantibodies could in theory
mean that the infection occurred along with the autoanti-
body seroconversion. However, 79% of all rotavirus infec-
tions measured in cases occurred before this sample interval
and 83% of infections in matched controls occurred before
the corresponding interval. Excluding infections in this inter-
val would have excluded a relatively large proportion of the
follow-up time, including the time just prior to seroconver-
sion for autoantibodies. Future studies should also attempt to
include a sufficient number of subjects to allow subanalyses
tistical power. Another strength is that we were able to adjust
for a number of potential confounding factors such as sea-
son of birth, duration of breast-feeding, day-care attendance,
number of older siblings, and maternal education (marker for
socioeconomic status). Although we cannot strictly exclude
the possibility that those who developed celiac disease au-
toimmunity had inherently different immunological response
to rotavirus, our prospective design and adjustment for HLA
genotype makes this less likely. Our study population was
selected based on the children carrying HLA risk alleles for
celiac disease and type 1 diabetes. The majority of our study
population consists of non-Hispanic whites of Northern Eu-
ropean decent, and the majority of celiac disease patients in
our population is likely to carry the susceptibility haplotypes
included in our study cohort. Although individuals carrying
the HLA-DRB1∗ 07-DQA1∗ 0201-DQB1∗ 0202/DRB1∗ 11/12-
DQA1∗ 0505-DQB1∗ 0301 (DR7-DQ2/DR5-DQ7, that is
DQ2 molecule encoded in trans) may be underrepresented
in our study cohort, this is unlikely to have any major influ-
ence on the relation between rotavirus infections and celiac
disease autoimmunity because the functional DQ2 molecule
encoded by individuals carrying this genotype is likely to be
very similar to that encoded by the DR3-DQ2 haplotype (2).
Gastrointestinal infection has long been suggested as a pos-
sible contributing cause of celiac disease, but to our knowl-
edge, no prospective study using biomarkers of infection
has previously been reported. Two recent studies indicated
that general infectious symptoms are associated with higher
risk of clinically identified, early onset celiac disease (5, 25).
Another recent study from Sweden indicated increased fre-
quency of rod-shaped bacteria on the intestinal epithelium of
celiac disease patients (6). The authors suggested a potential
etiological role of these bacteria (6), but the possibility that
bacterial colonization was a consequence of disease cannot
be excluded. The present study also suggested increased risk
of celiac disease autoimmunity associated with early day-care
attendance and increased number of older siblings. These are
both indicators of increased exposure to other children and
thus to infections. Adenovirus infection was suggested as a
2338 Stene et al.
potential trigger of celiac disease 20 yr ago, as there is se-
quence similarity between peptides of alpha-gliadin and the
54-kDa E1b protein of adenovirus 12 (26). However, there
is little direct evidence for molecular mimicry as an impor-
tant mechanism of autoimmunity in humans (27), and several
studies have been inconsistent in their conclusions about the
frequency of adenovirus infections in celiac disease cases
and controls (3, 28, 29). Other viruses may be suggested
by a study showing that Swedish children born in summer
had a higher risk of developing celiac disease before 2 yr
of age (30). These children would have a higher risk of sum-
mer enterovirus infections neonatally, and of winter rotavirus
infections around 6 months of age, perhaps closer to wean-
ing. However, our study of North American children did not
identify a significant relationship between season of birth and
risk of celiac disease autoimmunity. Enterovirus infection, at
least maternal infection during pregnancy, is not associated
with celiac disease in children (31). Because blood samples
in our study were collected when children were symptom-
free, and because rotavirus infections are frequently asymp-
tomatic, particularly for children who are breast-fed or who
had a previous infection, our study is not well suited for relat-
ing serologically defined infections to acute diarrhea or other
clinical symptoms. The possible relation of clinically overt
diarrhea confirmed to be caused by rotavirus and later risk of
celiac disease should, however, be the topic of future studies.
Recent understanding of the pathology of celiac disease has
come from evidence that tTG, the self-antigen in celiac dis-
ease, is released to the extracellular matrix under mechanical
or inflammatory stress (2), and that contact with transglutam-
inase is necessary for the deamidation of gliadin to reveal the
nonself T-cell epitopes recognized in celiac disease (2). These
data suggest that intestinal infection could lead to upregula-
tion and release of transglutaminase with consequent deami-
dation of gliadin present in the gut. Rotavirus infections, a
common cause of gastrointestinal infection in children, can
also lead to transient increased intestinal permeability (14),
permitting antigens more free access to the mucosal immune
system of the gut. We therefore hypothesize that the com-
bined effect of transiently increased intestinal permeability
and increased presence of tTG in the gut mucosa facilitate
deamidation of cereal proteins into more immunogenic epi-
topes. The continued inflammatory response to these neoanti-
gens could then spread to the tTG itself in individuals with
HLA-DQ2 or DQ8, thus establishing autoimmunity leading
to celiac disease.
In conclusion, this is the first study to provide evidence that
frequent rotavirus infections may be a risk factor for celiac
disease autoimmunity in early childhood. Like any novel find-
ing, replication and further studies on possible mechanisms of
action will be important to establish this relationship. In addi-
tion to shedding light on possible pathophysiological mecha-
nisms and disease prevention, there are also important impli-
cations regarding development and application of rotavirus
vaccines (32).
ACKNOWLEDGMENTS
The authors would like to thank the DAISY and Barbara Davis
Center staff for their help, Dory Bugawan for assistance with
the HLA genotyping, Helen Young for assistance with the ro-
tavirus antibody testing, and Leonard Harrison and Barbara
S. Coulson for support and helpful discussions. This work
was supported by NIH grants DK 50979; DK 32493; DK
32083; Autoimmune Prevention Center AI 50864; Diabetes
Endocrinology Research Center (DERC) Clinical Investiga-
tion and Bioinformatics Core, P30 DK 57516, M01 RR00069
General Clinical Research Centers Program, National Cen-
ters for Research Resources, NIH, and by the Children’s Dia-
betes Foundation. Margo C. Honeyman was supported by the
Australian National Health and Medical Research Council.
Lars C. Stene was supported by a grant from the National
Research Council of Norway (148359/330).
STUDY HIGHLIGHTS
What Is Current Knowledge
r Dietary gluten and predisposing genes (HLA-DQ2) are
necessary for development of celiac disease, but not
sufficient.
r Although intestinal infections have long been thought
to contribute to disease development, the few studies
that directly address this hypothesis have mostly been
cross-sectional rather than prospective. Such studies
could therefore not exclude the possibility that the in-
fections have been a consequence of disease, rather
than a contributing cause.
r Furthermore, most previous studies have involved clin-
ically overt disease, and not asymptomatic (silent) dis-
ease.
What Is New Here
r By recruiting genetically susceptible children soon af-
ter birth to regular blood draws, the current study is
prospective and measures auto antibodies towards ro-
tavirus in the blood of children before the initial trig-
gering of celiac disease, as marked by appearance in the
blood of antibodies towards the enzyme tissue transg-
lutaminase (“celiac disease autoimmunity”).
r Rotavirus is one of the most common causes of gas-
trointestinal infections in young children, and our study
shows that frequent occurrences of rotavirus infections
(as marked by immune-response towards to rotavirus)
seem to predict increased risk of developing celiac dis-
ease autoimmunity.
r Although this finding needs to been confirmed, it sug-
gests that a common, potentially preventable gastroin-
testinal infection may contribute to the initiation of
celiac disease pathogenesis.

Reprint requests and correspondence: Marian Rewers, M.D.,
Ph.D., Barbara Davis Center for Childhood Diabetes, University
of Colorado School of Medicine, P.O. Box 6511, Mailstop A-140,
Aurora, CO 80045-6511.
Received January 25, 2006; accepted April 25, 2006.
REFERENCES
1. Farrell RJ, Kelly CP. Celiac sprue. N Engl J Med 2002;
346:180–8.
2. Koning F, Schuppan D, Cerf-Bensussan N, et al. Pathomech-
anisms in celiac disease. Best Pract Res Clin Gastroenterol
2005;19:373–87.
3. Mahon J, Blair GE, Wood GM, et al. Is persistent aden-
ovirus 12 infection involved in coeliac disease? A search
for viral DNA using the polymerase chain reaction. Gut
1991;32:1114–6.
4. Ivarsson A, Hernell O, Stenlund H, et al. Breast-feeding
protects against celiac disease. Am J Clin Nutr 2002;75:914–
21.
5. Sandberg-Bennich S, Dahlquist G, Källén B. Coeliac dis-
ease is associated with intrauterine growth and neonatal in-
fections. Acta Paediatr 2002;91:30–3.
6. Forsberg G, Fahlgren A, Hörstedt P, et al. Presence of bacte-
ria and innate immunity of intestinal epithelium in childhood
celiac disease. Am J Gastroenterol 2004;99:894–904.
7. Dieterich W, Ehnis T, Bauer M, et al. Identification of tissue
transglutaminase as the autoantigen of celiac disease. Nat
Med 1997;3:797–801.
8. Hoffenberg EJ, Bao F, Eisenbarth GS, et al. Transglutam-
inase antibodies in children with a genetic risk for celiac
disease. J Pediatr 2000;137:356–60.
9. Fasano A, Catassi C. Current approaches to diagnosis and
treatment of celiac disease: An evolving spectrum. Gas-
troenterology 2001;120:636–51.
10. Hoffenberg EJ, Mackenzie T, Barriga KJ, et al. A prospec-
tive study of the incidence of childhood celiac disease. J
Pediatr 2003;143:308–14.
11. Mäki M, Mustalahti K, Kokkonen J, et al. Prevalence of
celiac disease among children in Finland. N Engl J Med
2003;348:2517–24.
12. Rewers M. Epidemiology of celiac disease: What are the
prevalence, incidence, and progression of celiac disease?
Gastroenterology 2005;128(suppl 1):S47–51.
13. Hoffenberg EJ, Emery LM, Barriga KJ, et al. Clinical fea-
tures of children with screening-identified evidence of celiac
disease. Pediatrics 2004;113:1254–9.
14. Kapikian AZ, Hoshino Y, Chanock RM. Rotaviruses. In:
Knipe DM, Howley PM, eds. Fields virology. Philadelphia:
Lippincott Williams & Wilkins, 2001:1787–833.
15. Honeyman MC, Coulson BS, Stone NL, et al. Association
between rotavirus infection and pancreatic islet autoimmu-
nity in children at risk of developing type 1 diabetes. Dia-
betes 2000;49:1319–24.
16. Coulson BS, Grimwood K, Bishop RF, et al. Evaluation of
end-point titration, single dilution and capture enzyme im-
munoassays for measurement of antirotaviral IgA and IgM in
infantile secretions and serum. J Virol Methods 1989;26:53–
65.
17. Coulson BS, Grimwood K, Masendycz PJ, et al. Compar-
ison of rotavirus immunoglobulin A coproconversion with
other indices of rotavirus infection in a longitudinal study
in childhood. J Clin Microbiol 1990;28:1367–74.
18. Sollid LM, Markussen G, Ek J, et al. Evidence for a primary
Rotavirus and Celiac Disease Autoimmunity 2339
association of celiac disease to a particular HLA-DQ α/β
heterodimer. J Exp Med 1989;169:345–50.
19. Rewers M, Bugawan TL, Norris JM, et al. Newborn screen-
ing for HLA markers associated with IDDM: Diabetes Au-
toimmunity Study in the Young (DAISY). Diabetologia
1996;39:807–12.
20. Rewers M, Liu E, Simmons J, et al. Celiac disease associated
with type 1 diabetes mellitus. Endocrinol Metab Clin North
Am 2004;33:197–214.
21. Bao F, Yu L, Babu S, et al. One third of HLA DQ2 homozy-
gous patients with type 1 diabetes express celiac disease-
associated transglutaminase autoantibodies. J Autoimmun
1999;13:143–8.
22. Liu E, Li M, Bao F, et al. Need for quantitative assess-
ment of transglutaminase autoantibodies for celiac dis-
ease in screening-identified children. J Pediatr 2005;146:
494–9.
23. Marsh MN. Gluten, major histocompatibility complex, and
the small intestine. A molecular and immunobiologic ap-
proach to the spectrum of gluten sensitivity (‘celiac sprue’).
Gastroenterology 1992;102:330–54.
24. Rothman KJ, Greenland S. Case-control studies. In: Roth-
man KJ, Greenland S, eds. Modern epidemiology. Philadel-
phia: Lippincott-Raven, 1998:93–114.
25. Ivarsson A. On the multifactorial etiology of celiac disease:
An epidemiological approach to the Swedish epidemic. Pe-
diatrics, Department of Clinical Sciences and Epidemiology,
Department of Public Health and Clinical Medicine, Umeå
University, Umeå, Sweden, 2001.
26. Kagnoff MF, Austin RK, Hubert JJ, et al. Possible role for
a human adenovirus in the pathogenesis of celiac disease. J
Exp Med 1984;160:1544–57.
27. Davidson A, Diamond B. Autoimmune diseases. N Engl J
Med 2001;345:340–50.
28. Lähdeaho ML, Parkkonen P, Reunala T, et al. Antibodies
to E1b protein-derived peptides of enteric adenovirus type
40 are associated with celiac disease and dermatitis herpeti-
formis. Clin Immunol Immunopathol 1993;69:300–5.
29. Vesy CJ, Greenson JK, Papp AC, et al. Evaluation of celiac
disease biopsies for adenovirus 12 DNA using a multiplex
polymerase chain reaction. Mod Pathol 1993;6:61–4.
30. Ivarsson A, Hernell O, Nyström L, et al. Children born in the
summer have increased risk for coeliac disease. J Epidemiol
Community Health 2003;57:36–9.
31. Carlsson AK, Lindberg BA, Bredberg AC, et al. En-
terovirus infection during pregnancy is not a risk factor for
celiac disease in the offspring. J Pediatr Gastroenterol Nutr
2002;35:649–52.
32. Offit PA. The future of rotavirus vaccines. Semin Pediatr
Infect Dis 2002;13:190–5.
CONFLICT OF INTEREST
Guarantor: Prof. Marian Rewers
Specific author contributions: Marian Rewers is the prin-
cipal investigator, and developed the general protocol for
the DAISY study with input from Jill Norris, George S.
Eisenbarth, and Henry A. Erlich. Henry A. Erlich was re-
sponsible for the HLA genotyping and George S. Eisenbarth
was responsible for measuring the autoantibodies. Edward
J. Hoffenberg, Ronald J. Sokol and Joel E. Haas were re-
sponsible for the clinical evaluation of the antibody positive
2340 Stene et al.
children. Lisa Emery was responsible for managing and
preparing the data bases. Iman Taki was the study coordi-
nator and she retrieved and aliquoted the serum samples to
be tested for rotavirus. Margo C. Honeyman was responsible
for the rotavirus antibody testing. Lars C. Stene designed the
present nested case-control study and selected the controls.
Lars C. Stene designed the analysis strategy, did all the sta-
tistical analyses, and wrote the manuscript with input from
all authors.
Financial support: NIH grants DK 50979; DK 32493; DK
32083; Autoimmune Prevention Center AI 50864; Diabetes
Endocrinology Research Center (DERC) Clinical Investiga-
tion & Bioinformatics Core, P30 DK 57516, M01 RR00069
General Clinical Research Centers Program, National Cen-
ters for Research Resources, NIH, and by the Children’s Di-
abetes Foundation. Margo C. Honeyman was supported by
the Australian National Health and Medical Research Coun-
cil. LC Stene was supported by a grant from the National
Research Council of Norway (148359/330).
Potential competing interests: None of the authors declared
any potential competing interest.