MAJOR ARTICLE

Diarrhea Etiology in a Pediatric Emergency

Department: A Case Control Study
Donna M. Denno,1,2,5 Nurmohammad Shaikh,3 Jenny R. Stapp,5 Xuan Qin,5,6 Carolyn M. Hutter,7 Valerie Hoffman,5
Jody C. Mooney,5 Kelly M. Wood,5 Harold J. Stevens,3 Robert Jones,8 Phillip I. Tarr,3,4,a and Eileen J. Klein1,5,a
1
Department of Pediatrics, 2Global Health, University of Washington, Seattle; 3Division of Gastroenterology and Nutrition, Department of Pediatrics,
and 4Department of Molecular Microbiology, Washington University School of Medicine, St Louis, Missouri; 5Seattle Childrens Hospital, 6Department
of Laboratory Medicine, University of Washington, 7Department of Public Health Sciences, Division of Public Health Sciences, Fred Hutchinson Cancer
Center, and 8Craic Computing LLC, Seattle, Washington

Background. The etiology of childhood diarrhea is frequently unknown.

Methods. We sought Aeromonas, Campylobacter, Escherichia coli O157:H7, Pleisiomonas shigelloides, Salmo-
nella, Shigella, Vibrio, and Yersinia (by culture), adenoviruses, astroviruses, noroviruses, rotavirus, and Shiga

Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016

toxin-producing E. coli (STEC; by enzyme immunoassay), Clostridium difcile (by cytotoxicity), parasites (by mi-
croscopy), and enteroaggregative E. coli (EAEC; by polymerase chain reaction [PCR] analysis) in the stools of 254
children with diarrhea presenting to a pediatric emergency facility. Age- and geographic-matched community
controls without diarrhea (n = 452) were similarly studied, except bacterial cultures of the stool were limited only
to cases.
Results. Twenty-nine (11.4%) case stools contained 13 Salmonella, 10 STEC (6 O157:H7 and 4 non-O157:H7
serotypes), 5 Campylobacter, and 2 Shigella. PCR-dened EAEC were present more often in case (3.2%) specimens
than in control (0.9%) specimens (adjusted odds ratio [OR], 3.9; 95% condence interval [CI], 1.113.7), and
their adherence phenotypes were variable. Rotavirus, astrovirus, and adenovirus were more common among cases
than controls, but both groups contained noroviruses and C. difcile at similar rates. PCR evidence of hyperviru-
lent C. difcile was found in case and control stools; parasites were much more common in control specimens.
Conclusions. EAEC are associated with childhood diarrhea in Seattle, but the optimal way to identify these
agents warrants determination. Children without diarrhea harbor diarrheagenic pathogens, including hyperviru-
lent C. difcile. Our data support the importance of taking into account host susceptibility, microbial density, and
organism virulence traits in future case-control studies, not merely categorizing candidate pathogens as being
present or absent.

The study of acute childhood diarrhea presents 2 major
challenges: this disorder often eludes etiologic determi-
nation [15], and many putatively diarrheagenic agents
detected in stools are of undetermined causative signi-
cance. For example, in a recent Seattle Childrens Hos-
pital (SCH) emergency department (SCHED) study, 6%
Received 13 February 2012; accepted 1 June 2012; electronically published 14
June 2012.
a
P. I. T. and E. J. K. contributed equally to this work.
Correspondence: E. J. Klein, MD, Division of Emergency Medicine, Seattle
Childrens Hospital, 4800 Sandpoint Way NE, B-5506, Seattle, WA 98105 (eileen.
klein@seattlechildrens.org).
Clinical Infectious Diseases 2012;55(7):897904
The Author 2012. Published by Oxford University Press on behalf of the Infectious
Diseases Society of America. All rights reserved. For Permissions, please e-mail:
journals.permissions@oup.com.
DOI: 10.1093/cid/cis553
of diarrheal stools contained Clostridium difcile cyto-
toxin [1], but the role of C. difcile in childhood diarrhea
is a complex issue because the stools of many healthy
infants and children contain this organism [6, 7]. Here
we report a comprehensive case-control analysis of di-
arrhea etiology at the same venue to determine if candi-
date pathogens such as C. difcile and enteroaggregative
Escherichia coli (EAEC) are, in fact, associated with
diarrhea.
METHODS
Study Design and Enrollment
This prospective case-control study was approved by
the SCH and Washington University Institutional

Etiology of Childhood Diarrhea CID 2012:55 (1 October) 897

Review Boards. Case subjects were children presenting with
diarrhea from November 2003 to November 2005 to the
SCHED (a facility with >60 000 patient encounters during that
interval), whose parents agreed to participate. Families provid-
ed demographic data, illness characteristics, medication, and
travel history on standardized questionnaires, on which notice
was provided that responding on the questionnaire signied
participation consent.
Potential controls were recruited before and during case en-
rollment into a reserve cohort using 72 family and pediatric
practices in King County (which includes Seattle and its
suburbs) and southern Snohomish County, Washington,
where circa 90% of SCHED patients reside. Multilingual re-
cruitment materials describing the study [8] were displayed;
brochures contained return mail enrollment forms for inter-
ested participants. For each case subject, a computer algorithm
selected 2 community controls from the control reserve
cohort, matched sequentially to the closest zip code of resi-
dence of the patient to reduce confounding by geography and
demography, age (closest 50%, up to 19 years old), and
nally sex.
One of 2 research assistants called families of each computer-
selected control to explain the study, determine eligibility,
obtain consent, and administer the same travel and medica-
tion questions used for cases. Controls could participate only
once and were ineligible if they had diarrhea in the preceding
30 days or had a sibling chosen as a control for the same case.
Qualifying controls were then sent a sterile specimen cup,
collection hat, an insulated box, aqueous freezer bricks, and
instructions to freeze the bricks. Control stools were placed
into the cup and returned in the supplied box with the frozen
bricks via courier to the SCH clinical microbiology laboratory.
A small monetary acknowledgement was provided in appreci-
ation for participation. US census data [9] and inter-zip code
driving distances (Google Maps, http://maps.google.com/
maps?hl=en&tab=wl on 12 April 2012) were used to approxi-
mate median household incomes and median distances
between cases and corresponding controls, respectively.
Laboratory Analysis
Fresh case and control stools were evaluated for Shiga toxin
(Stx)-producing E. coli (STEC) using an Stx enzyme immuno-
assay (EIA) on overnight broth cultures, parasites (by micros-
copy) and C. difcile (by cytotoxicity assay), and plated on
MacConkey agar; case stools were cultured for bacteria gener-
ally considered to be diarrheagenic (Aeromonas species, Cam-
pylobacter species, STEC O157:H7, Pleisiomonas shigelloides,
Salmonella, Shigella, Vibrio, and Yersinia species) [10, 11].
Stools were then frozen at 80C for subsequent batch-testing
for adenoviruses, astroviruses, noroviruses 1 and 2, and rotavi-
rus (IDEIA EIA kits K6021, K6042, K6043, K6020,
898 CID 2012:55 (1 October) Denno et al
respectively, Dako, Carpinteria, California). Non-O157:H7
STEC were recovered by isolating, and subsequently testing up
to 20 lactose fermenting colonies by polymerase chain reaction
(PCR) [10].
We identied EAEC by PCR testing 3 colonies per subject
for aatA and aggR (conditions and primers in Supplementary
Table 1), subject to colony availability (Supplementary
Table 2). The stx1 and stx2 PCR was performed on pure cul-
tures of single colonies of all O157:H7 and non-O157:H7
STEC, and 8 colony pools of EAEC.
We extracted DNA (QIAmp, Qiagen, Valencia, California)
from frozen stools that tested positive in the C. difcile cyto-
toxicity assay and from their matched cytotoxin-negative con-
trols. We then employed PCR with primers specic for the
universal bacterial 16S ribosomal RNA (rRNA) gene (UB16S)
and the C. difcile 16S rRNA gene (CD16S), and, if these
were present, C. difcile toxins A (tcdA) and B (tcdB),
binary toxin (cdtA and cdtB), and intact or deleted tcdC (tcdC,
Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016
when absent (deleted) from C. difcile is associated with
enhanced expression of toxins A and B). We also tested
22 cytotoxin-negative stools from controls that were matched
to cytotoxin-positive cases for CD16S using the same
methodology.
We examined individual EAEC to determine their adher-
ence phenotypes. HeLa cells were grown to nearconuence in
Dulbecco modied eagle medium (DMEM) with 10% heat-
inactivated fetal calf serum (FCS) in glass chamber slides
(5% CO2, 37C). Bacteria were grown (37C, overnight) in
Luria-Bertani (LB) broth with shaking under aerobic condi-
tions, and 0.3 mL were added to 2.7 mL of fresh LB broth and
grown under the same conditions for 6 hours, washed in
DMEM, and suspended in adherence media (DMEM contain-
ing 5% FCS and 0.5% D-mannose), which was then added to
the chamber slides after removing the media under which the
HeLa cells were growing. After incubation (1 hour, 5% CO2,
37C), we aspirated the adherence media and washed the
monolayers 3 times with DMEM. After adding fresh adher-
ence media and incubating (3 more hours) in the same condi-
tions, we washed the cells 3 times ( phosphate-buffered saline
[PBS]), and xed (100% cold methanol, 5 minutes) and
stained (0.5% crystal violet) them before applying a coverslip.
An observer unaware of the identity of the bacteria in each
chamber scored the assays as nonadherent or adherent and
noted the characteristics of the positives. Positive and negative
controls were E. coli strains EAEC O42 and ORN172,
respectively.
Only subjects whose specimens were tested completely for
all organisms of interest (STEC and viruses by EIA, C. difcile
by cytotoxicity, parasites by microscopy, and EAEC by PCR
for cases and controls, classic bacterial pathogens by culture
for cases) were included in this analysis.
Statistical Analysis
We modeled power with different potential sample and effect
sizes; with sample sizes of 250 case and 450 control subjects,
the project had 74%, 89%, and 95% chances of nding differ-
ences of 2%, 3%, and 4% frequency in the case group, com-
pared with the control group. We used STATA version 11
(Statacorp, College Station, Texas) to measure associations
between organism and case status, and conditional logistic re-
gression analyses to calculate univariate matched odds ratios
(ORs) and 95% condence intervals (CIs). To address con-
founding by demographic and socioeconomic factors (includ-
ing potential residual confounding by matching variable), we
calculated adjusted ORs (aOR) and 95% CIs using multivari-
ate conditional logistic models that included variables found
to be associated with disease status in our data, or, if cells con-
tained no or 1 subject, we calculated P values using the exact
version of McNemar test ( pexact).
(interquartile range, 40; range, 8121). Cases were matched to
controls living in the same (31%), contiguous (36%), or adja-
cent to a contiguous (14%), zip code; the median distance
between zip codes of residence of cases and matched controls
was 11.9 km. Controls zip codes had higher mean incomes
than those of the cases, and there was a higher proportion of
whites among controls and Hispanics among cases, but no
other signicant differences between groups. For these
reasons, we adjusted for mean income by zip code, race and
ethnicity in the multivariate regression analyses.
Twenty-nine (11.4%) case stools contained 30 bacterial
pathogens: 13 Salmonella (5 Salmonella serovar Typhimurium,
2 Salmonella serovar Heidelberg, and 1 each Salmonella sub-
genus I Group B (4,512:i,-) Salmonella subspecies I serotype
4,12,:I,-, Salmonella serovar Newport, Salmonella serovar Min-
nesota, Salmonella serovar Enteritidis, and Salmonella serovar
Brandenburg), 10 STEC , 5 Campylobacter (including 1 non-
jejuni), and 2 Shigella sonnei.

Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016

The single specimen that was negative on the STEC EIA
RESULTS but which contained E. coli O157:H7 on the corresponding
sorbitol MacConkey agar plate possessed stx1 and stx2 genes
We enrolled 254 cases and 452 matched controls (Table 1; (Supplementary Table 4); a pure broth culture of this isolate
mean, 1.8 matched controls per case; standard deviation, 0.5; was positive in the EIA. The stools containing STEC O26:H11
range, 14). A median of 39 days elapsed between receipt and O111:nonmotile also contained rotavirus and S. serovar
of case stools and their corresponding controls stools Brandenburg, respectively, whereas those that contained STEC
O121:nonmotile, O177:nonmotile, and O157:H7 possessed no
other organism of interest (Tables 2, 3 and Supplementary
Table 1. Demographic and Specimen Characteristics of Cases Table 4). The statistical signicance of the differences in posi-
and Controls
tivity between STEC among cases and controls varied from
Case Control
pexact = .13 for non-O157:H7 overall, to pexact = .50 for non-
Characteristic (n = 254) (n = 452) P O157:H7 where no other organism of interest was identied,
Female sex 121 (47.6) 216 (48.1) a
.95 to pexact = .06 for the 5 cases of STEC O157:H7 identied by
Age, median months 20 (1046) 24(1149) .49 Stx EIA. We necessarily excluded the sixth case identied only
Raceb by SMAC agar screening from this particular analysis as con-
White 142 (58.0) 345 (77.0) <.001 trols were not similarly screened.
Asian/Pacific Islander 28 (11.4) 56 (12.5) An average of 2.92 colonies per subject (median 3, range
African American 25 (9.8) 39 (8.6) 03) were tested for the 2 EAEC loci. EAEC from 4 cases and
Native American 10 (4.8) 8 (1.8) 2 controls contained both loci, whereas EAEC from 4 cases
Other 44 (17.3) 42 (9.3) and 2 controls contained one or the other locus (Supplementary
Dont know 26 (10.2) 2 (0.4)
Table 2). Adherence phenotypes of EAEC varied and included
No response 9 (3.5) 4 (0.9)
nonadherent, sparsely and densely adherent, and adherent in
Hispanic ethnicity 69 (27.2) 70 (15.5) <.0001
a stippled pattern over cells (Supplemental Figure 1). Adher-
Dont know 13 (5.2) 2 (0.4)
No response 0 (0) 3 (0.7)
ence phenotypes correlated with neither case nor control
Mean household $54 099 $55 849 .02 subject status, or EAEC genotype, and varied even between
income of zip codec EAEC from the same individuals. All EAEC were stxnegative.
Data are expressed as No. (%) or median (interquartile range), unless C. difcile cytotoxin positivity was similar in cases and con-
otherwise indicated. trols (aOR: 0.9, CI: .42). Point estimates did not meaningfully
a
Sex not provided for 3 controls. change when further adjusted for a history of antibiotic use
b
Percentages exceed 100 because of multiple responses by some subjects.
P value is for comparison of white to nonwhite.
(data not shown), but did increase in magnitude when restrict-
c
Based on US census data for 1999 on median household income of ing analysis to subject sets older than one and three years of
households in participants zip code of residence. age (aORs: 1.7 [CI: .55.4] and 4.2 [CI: .725.4], respectively).

Etiology of Childhood Diarrhea CID 2012:55 (1 October) 899

Table 2. Case-Control Results for Organisms of Interest

No. of Positive No. of Positive Unadjusted OR Adjusted ORa

Established or Candidate Pathogens Cases (%) Controls (%) (95% CI) (95% CI)
STEC EIA positiveb 9 (3.5) 0 P = .004c
O157:H7d 5 (2.0) 0 P = .06c
Non-O157:H7e 4 (1.6) 0 P = .13c
EAEC 8 (3.1) 4 (0.9) 3.4 (.911.9) 3.9 (1.113.7)
aatA 7 (2.8) 4 (0.9) 2.9 (.810.6) 3.4 (.912.6)
aggR 5 (2.0) 3 (0.7) 2.8 (.613.2) 3.7 (.817.0)
C. difficile all ages 14 (5.5) 28 (6.2) 0.9 (.51.9) 0.9 (.42)
C. difficile <36 months old 9 (5.2) 26 (8.8) 0.6 (.31.3) 0.6 (.21.4)
C. difficile > 36 months old 5 (6.3) 2 (1.3) 4.6 (.827) 4.3 (.725.4)
Parasites (any)f 1 (0.4) 16 (3.5) P = .0003c
Astrovirus 11 (4.3) 7 (1.6) 3.1 (1.28.1) 3.6 (1.310)
Adenovirus 8 (3.2) 4 (0.9) 3.8 (1.113.3) 3.9 (1.114.9)
Norovirus 1 4 (1.6) 1 (0.2) P = .38c
Norovirus 2 5 (2.0) 15 (3.3) 0.6 (.21.6) 0.9 (.32.6)
Rotavirus 110 (43.3) 1 (0.22) P < .0001c

Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016

Some of these subjects specimens contained >1 target organism. Details of dual infections, and of characteristics of the subjects and their illnesses, are
provided here and in Supplemental Tables 24. Odds ratios vary depending on the specific numbers of positive cases and controls in matched and unmatched
sets. Abbreviations: CI, confidence interval; EAEC, enteroaggregative E. coli; EIA, enzyme immunoassay; OR, odds ratio; STEC, Shiga toxin-producing E. coli.
a
Adjusted for mean income by zip code, race, and ethnicity
b
These are defined as subjects whose broth cultures tested positive for Shiga toxin (Stx) by EIA and from which a non-O157:H7 or an O157:H7 STEC was
recovered but does not include the additional case subject whose stool contained STEC O157:H7 as identified on sorbitol MacConkey agar, but whose toxin
assay was negative.
c
For organisms where no or only 1 subjects specimen tested positive, we present P values from an exact version of McNemar test (pexact), rather than ORs and
CIs.
d
These are defined as subjects whose broth cultures tested positive for Stx by EIA and from which an O157:H7 STEC was recovered but does not include the
additional case subject whose stool contained STEC O157:H7 as identified on sorbitol MacConkey agar, but whose toxin assay was negative.
e
These are defined as subjects whose broth cultures tested positive for Stx by EIA and from which a non-O157:H7 STEC was recovered.
f
Cryptosporidia (1 case and 6 control subjects), Entamoeba coli (6 control subjects), Blastocystis hominis (1 control subject), Endolimax nana (2 control subjects),
and Giardia lamblia and E. nana (1 control subject).

PCR evidence of C. difcile was lacking (ie, specimens were
negative for CD16S) in one case and one control stool that
were cytotoxicity assay positive. Each of 22 cytotoxin-negative
specimens from the controls matched to cytotoxin-positive
cases was negative for CD16S. Of the 12 cases and 27 controls
whose cytotoxin-positive stools contained PCR conrmation
of C. difcile presence, 3 (25%) and 2 (7.4%) contained a hy-
pervirulent C. difcile (the tcdC deletion).
Only 1 case subject, a 4-year-old girl who was evaluated for
4 days of diarrhea following travel to Honduras, was infected
with a parasite (Cryptosporidia), whereas 16 control subjects
produced stools with parasites. Adenoviruses were identied
in the same proportion of cases and controls as EAEC, and
astrovirus were also associated with case status but noroviruses
were not. Rotavirus was found overwhelmingly in case stools
(Table 2).
One or more organisms of interest were found in the stools
of 174 cases and 69 controls (aOR, 12.9; CI, 7.921.2).
Seventeen cases and 6 controls had >1 such microbe in their
stools (aOR, 6.0; CI, 2.415.1; Table 3). Only 1 stool (from a
control) contained 3 organisms of interest (Cryptosporidia,
adenovirus and EAEC).
DISCUSSION
The distribution of bona de enteric pathogens (Campylobac-
ter, STEC O157:H7, Salmonella, Shigella, and rotavirus) in this
series resembles the etiologic distribution in this venue
between 1998 and 2001. This constancy demonstrates small
area (greater Seattle) stability of childhood diarrhea etiology,
though both studies were conducted before widespread use of
rotavirus vaccines. Our ndings differ from recent North
American and European studies of people presenting with
acute diarrhea in the higher proportion of bacterial pathogens
we have found, with the most pronounced difference in the
higher recovery of STECs [24, 1214]. It is possible that this
high-acuity population, drawn entirely from a pediatric emer-
gency facility, is responsible for this particular difference. Ad-
ditionally, we identied a greater proportion of astrovirus in
diarrhea than these other studies, though this is an agent that

900 CID 2012:55 (1 October) Denno et al

Table 3. Characteristics of Patients and Illnesses Associated With Pathogens or Candidate Pathogens

Stool Appearance
Established or Candidate Mean Age Abdominal
Pathogens No. (Months) Fever a Vomitinga Paina Blooda Mucusa,b Frequencya,c Durationd Travela,e WBCf
STECg 10 72 50 70 90 70 80 10 3 10 50
Campylobacter 5 85 60 60 80 40 40 16 4 0 40
Salmonella 13 57 92 69 62 62 54 10 5 23 46
Shigella 2 84 100 50 100 100 50 10 1 0 100
Clostridium difficile h 14 54 57 71 50 43 64 6 2 7 14
EAECi 8 54 63 50 63 25 63 7 2 25 25
Astrovirus 11 56 82 73 73 18 45 8 3 0 18
Adenovirus 8 38 38 88 75 0 38 6 4 0 0
Rotavirus 110 22 81 96 48 7 42 8 2 5 3
Norovirus 1 4 144 0 75 75 25 25 8 3 0 25
Norovirus 2 5 39 40 100 100 20 100 8 1 0 20
Total No. of established or 174 36 73 86 59 20 46 8 3 7 14
candidate pathogensj
All patients with multiple 17 40 65 88 47 24 65 7 2 12 18
established or candidate
k
pathogens

Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016

All patients with no 80 38 56 58 55 21 36 5 3 8 5
established or candidate
pathogens
All patients 254 37 68 77 58 20 43 7 3 7 11
Symptoms and signs are family/self-reported on questionnaire.
Abbreviations: EAEC, enteroaggregative E. coli; STEC, Shiga toxin-producing E. coli; WBC, white blood cells.
a
History and symptoms were family/self-reported on questionnaire.
b
Percentages calculated by dividing the number of subjects answering yes by total number of subjects in group (including those answering dont know or
leaving response blank) and multiplying by 100, and therefore represent minimal values.
c
Median No. of stools in 24-hour period before Seattle Childrens Hospital emergency department (SCHED) visit, among those providing a response other than
dont know.
d
Median No. of days of with diarrhea before SCHED visit, among those providing a response other than dont know.
e
Outside of the United States or Canada.
f
Defined as white blood cells identified in the stools by Gram stain or wet mount.
g
Includes patients infected with STEC O157:H7 (n = 6), O26:H11, O111:nonmotile, O121: nonmotile, and O177:nonmotile (n = 1 each). STEC-infected cases are
those subjects whose broth cultures tested positive for Shiga toxin (Stx) by enzyme immunoassay (EIA) and from which a non-O157:H7 or an O157:H7 STEC
was recovered, or whose sorbitol MacConkey agar plate culture yielded an O157:H7 STEC even though the Stx EIA was negative. The stx genotypes and
associated illness characteristics are provided in Supplementary Table 4.
h
Defined as patients whose stools were cytotoxin positive; characteristics of genotypes in individual positive specimens and corresponding subjects are
provided in Supplementary Table 3.
i
Defined as isolates positive for aatA and/or aggR loci by polymerase chain reaction; characteristics of genotypes of positive colonies and corresponding subjects
are provided in Supplementary Table 2.
j
174 organisms were detected in 157 case patients.
k
Case subjects had evidence of 2 organisms of interest in their stool: C. difficile toxin and rotavirus (n = 4), rotavirus and astrovirus (n = 2), rotavirus and EAEC
(n = 2), C. difficile toxin and adenovirus, C. difficile toxin and norovirus 1, C. difficile toxin and norovirus 2, adenovirus and astrovirus, adenovirus and rotavirus,
STEC O26:H11 and rotavirus, S. serovar Typhimurium and rotavirus, Shigella sonnei and EAEC, and STEC O111:nonmotile and S. serovar Brandenburg (n = 1
each pairing).

has been reported in hospitalized North American children in
frequencies similar to what we report [5].
The higher proportion of EAEC among cases than among
controls corroborates ndings from children in Cincinnati [3],
New Haven, and Baltimore [4] and extends these ndings by
comparing community-acquired diarrhea to community-
based age-matched controls. EAEC diarrhea is generally
nonbloody and persistent, but the characteristics of the case
subjects with these organisms in their stool in our study sug-
gests a more acute picture, occasionally including bloody diar-
rhea, although these EAEC did not contain stx genes, unlike
EAEC O104:H4, which caused a large and devastating epi-
demic in Europe in MayJune 2011. We believe that future
work should be directed to thoroughly dene the scope of

Etiology of Childhood Diarrhea CID 2012:55 (1 October) 901

disease caused by EAEC, and consensus regarding detection
methodologies [1520] for EAEC will be very helpful in any
attempt to advance our understanding of this group of organ-
isms in childhood diarrhea. However, inconsistent phenotypes
among EAEC identied by PCR suggest that adherence assays
are an insensitive detection strategy for this set of candidate
pathogens.
SMAC agar was superior to Stx EIA for detecting STEC
O157:H7, an often noted discordance [1, 2125], and we agree
with Centers for Disease Control and Prevention recom-
mendations that Stx assays should not be used to screen for
STEC O157:H7 in lieu of SMAC agar plating [26]. The
absence of STEC in the control group is informative; few
such analyses have been performed in this country. However,
of the 4 case specimens that contained non-O157:H7 STEC, 2
also were positive for rotavirus or Salmonella. Interestingly,
stools of patients infected with non-O157:H7 STEC often
contain additional diarrheagenic agents [1, 10, 2729], whereas
STEC O157:H7 is predominantly a solitary pathogen [29, 30].
Is C. difcile a bona de cause of childhood ambulatory di-
arrhea? Our ndings fail to demonstrate more than a trend
toward association, and only in children older than 36 months
of age. Nonetheless, each of the 6 case subjects whose stools
contained C. difcile cytotoxin but no other enteric pathogens
reported at least 3 of the following symptoms or signs: visible
blood or mucus in the stool, fever, or abdominal pain, making
a plausible case that C. difcile caused their diarrhea. There
remain insufcient data on which to formulate guidance for
clinicians regarding treatment of symptomatic children whose
stools contain C. difcile. This is an important question, as
children might well be community reservoirs of this pathogen,
including hypervirulent strains, and be even more infectious if
they have diarrhea. Though we found good correlation
between cytotoxicity assay and PCR evidence of C. difcile,
the best methods for nding actionable C. difcile in speci-
mens remain elusive [31].
Despite expansion of testing to include agents not sought in
our previous project (EAEC and noroviruses), the causes of
nearly one-third of the cases of diarrhea remained undeter-
mined, although it is possible that reliance on EIA technology
for viral agents underenumerated their presence. Moreover, our
data exemplify how difcult it is to attribute microbial pathoge-
nicity in diarrhea. Most strikingly, Giardia and Cryptosporidia
were more frequently identied in our controls than in our
cases, and even STEC O157:H7 was not statistically associated
with case status. The small numbers of positives among cases
no doubt contributed to our inability to ascribe pathogenicity
to non-O157:H7 STEC and to noroviruses, and EIAs might
lack detection sensitivity. Also, some candidate pathogens, such
as Dientamoeba fragilis, were found in neither group, so our
data cannot inform debate regarding their pathogenicity [32].
902 CID 2012:55 (1 October) Denno et al
Categorical present/absent assignment to microbes might
not adequately weight factors that would incriminate such or-
ganisms as pathogens when found in cases or exculpate them
when found in controls. Factors not considered in present/
absent results include host immunopropensity toward acquir-
ing specic pathogens and then becoming ill; microbial
density in analyte as has been noted with enteropathogenic E.
coli [33]; presence and expression of genes encoding virulence
factors by the putative pathogen in the host; and ambient ora
that might alter expression of virulence. For EAEC, however,
microbial enumeration is not so straightforward, because the
dening loci are contained on plasmids of variable copy
number, which hinder attempts to determine organism
density per mass of stool. However, in this study, the mean
numbers of positive colonies per EAEC-excreting cases (2.4)
and controls (2.8) were similar.
We are also constrained by our diagnostic reliance on avail-
able reagents that target only established pathogens. Agnos-
Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016
tic and new technologic approaches to pathogen discovery and
identication are needed to more thoroughly identify causes
of diarrhea. One such example to illuminate diarrhea etiology
is mass sequencing, which has identied several candidate di-
arrheagenic viruses in childhood stool [3438]. Expanding da-
tabases of microbial sequences, developing enumeration tools
that are more sophisticated than yes/no presence of organism,
and factoring in host response to agent presence/absence will
advance our abilities to conrm or refute virulence. Until
these many case-control design and analysis issues can be re-
solved, it appears that outbreaks will remain the best way to
establish pathogenicity of candidate diarrheagenic organisms.
Finally, the presence of diarrheagenic viruses, parasites,
EAEC, and C. difcile (including potentially hypervirulent
strains) in many control stools, sometimes in greater frequen-
cy than in case specimens, raises the possibility that future
control strategies and modeling efforts might need to account
for community reservoirs of these agents [39].
In summary, EAEC are statistically associated with diarrhea,
but denition of this class of probable pathogens needs rene-
ment, and optimal detection methodologies are lacking. Some
community-acquired acute childhood diarrhea is probably
caused by C. difcile, but we doubt that cytotoxin-producing
C. difcile always warrants treatment when found in cases of
acute diarrhea. This case-control study strengthens the case
that some non-O157:H7 STEC are pathogenic but again por-
trays the predominance of the O157:H7 serotype in a high-
acuity setting, and the superiority of SMAC agar screening for
this pathogen. Finally, asymptomatic excretion of C. difcile,
Cryptosporidia, and diarrheagenic viruses by community con-
trols highlights the need to identify host and microbial factors
and processes that differentiate asymptomatic carriage from
pathogenicity and might have implications for transmission.
The high carriage rate also poses difculties when using case-
control studies to conrm or refute pathogenicity of candidate
diarrheagenic agents.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online
(http://cid.oxfordjournals.org). Supplementary materials consist of data
provided by the author that are published to benet the reader. The
posted materials are not copyedited. The contents of all supplementary
data are the sole responsibility of the authors. Questions or messages
regarding errors should be addressed to the author.
Notes
Acknowledgments. We thank subjects and their families, Seattle
Childrens Hospital (SCH) emergency department nurses and physicians,
SCH Microbiology Laboratory staff, for their assistance with this complex
project, Tara Sweeney for technical assistance, Dako for providing the
virus detection kits at reduced cost, Ariana Jasarevic and Helen Odle for
assistance with manuscript preparation, Drs James Nataro and Paul
Orndorff for the gift of strains EAEC O42 and ORN172 , respectively, and
Dr Lars Westblade for review of the manuscript. We particularly thank
the network of community based controls and their families, and their
primary care doctors ofces and staff, for their enthusiastic assistance
with the recruitment of this unique control set.
Financial support. This study was supported by US Department of
Agriculture (NRI grant 0202238 to P. I. T.) and the Morphology and
Biobank Cores of the Washington University School of Medicine Digestive
Diseases Research Core Center (P30DK052574). Virus enzyme immunoas-
say kits were kindly provided at reduced cost by Dako, but Dako had no
role in study design or manuscript preparation.
Potential conicts of interest. All authors: No reported conicts.
All authors have submitted the ICMJE Form for Disclosure of Potential
Conicts of Interest. Conicts that the editors consider relevant to the
content of the manuscript have been disclosed.
References
1. Klein EJ, Boster DR, Stapp JR, et al. Diarrhea etiology in a Childrens
Hospital Emergency Department: a prospective cohort study. Clin
Infect Dis 2006; 43:80713.
2. Denno DM, Stapp JR, Boster DR, et al. Etiology of diarrhea in pediat-
ric outpatient settings. Ped Infect Dis J 2005; 24:1428.
3. Cohen MB, Nataro JP, Bernstein DI, Hawkins J, Roberts N, Staat MA.
Prevalence of diarrheagenic Escherichia coli in acute childhood enteri-
tis: a prospective controlled study. J Pediatr 2005; 146:5461.
4. Nataro JP, Mai V, Johnson J, et al. Diarrheagenic Escherichia coli in-
fection in Baltimore, Maryland, and New Haven, Connecticut. Clin
Infect Dis 2006; 43:4027.
5. Rodriguez-Baez N, OBrien R, Qiu SQ, Bass DM. Astrovirus, adenovi-
rus, and rotavirus in hospitalized children: prevalence and association
with gastroenteritis. J Pediatr Gastroenterol Nutr 2002; 35:648.
6. Vesikari T, Isolauri E, Maki M, Gronroos P. Clostridium difcile in
young children: association with antibiotic usage. Acta Paediatr Scand
1984; 73:8691.
7. Thompson CM Jr., Gilligan PH, Fisher MC, Long SS. Clostridium
difcile cytotoxin in a pediatric population. Am J Dis Child 1983;
137:2714.
8. Denno DM, Keene WE, Hutter CM, et al. Tri-county comprehensive
assessment of risk factors for sporadic reportable bacterial enteric in-
fection in children. J Infect Dis 2009; 199:46776.
9. US Census Bureau. Available at: http://wwwcensusgov/epcd/www/
zipstatshtml. Accessed 9 February 2012.
10. Klein EJ, Stapp JR, Clausen CR, et al. Shiga toxin-producing Escheri-
chia coli in children with diarrhea: a prospective point-of-care study.
J Pediatr 2002; 141:1727.
11. Rifkin GD, Fekety FR, Silva JJ. Antibiotic-induced colitis implication
of a toxin neutralised by Clostridium sordellii antitoxin. Lancet 1977;
2:11036.
12. Bresee JS, Marcus R, Venezia RA, et al. The etiology of severe acute
gastroenteritis among adults visiting emergency departments in the
United States. J Infect Dis 2012; 205:137481.
13. Tam CC, OBrien SJ, Tompkins DS, et al. Changes in causes of acute
gastroenteritis in the United Kingdom over 15 years: Microbiologic
ndings from 2 prospective, population-based studies of infectious in-
testinal disease. Clin Infect Dis 2012; 54:127586.
14. de Wit MA, Koopmans MP, Kortbeek LM, van Leeuwen NJ, Vinje J,
van Duynhoven YT. Etiology of gastroenteritis in sentinel general
practices in the Netherlands. Clin Infect Dis 2001; 33:2808.
15. Vidal M, Kruger E, Duran C, et al. Single multiplex PCR assay to iden-
tify simultaneously the six categories of diarrheagenic Escherichia coli
associated with enteric infections. J Clin Microbiol 2005; 43:53625.
16. Nataro JP. Enteroaggregative Escherichia coli pathogenesis. Curr Opin
Gastroenterol 2005; 21:48.
17. Steiner TS, Lima AA, Nataro JP, Guerrant RL. Enteroaggregative
Escherichia coli produce intestinal inammation and growth impair-
ment and cause interleukin-8 release from intestinal epithelial cells.
Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016
J Infect Dis 1998; 177:8896.
18. Nguyen TV, Le Van P, Le Huy C, Gia KN, Weintraub A. Detection
and characterization of diarrheagenic Escherichia coli from young chil-
dren in Hanoi, Vietnam. J Clin Microbiol 2005; 43:75560.
19. Vial PA, Mathewson JJ, DuPont HL, Guers L, Levine MM. Compari-
son of two assay methods for patterns of adherence to HEp-2 cells of
Escherichia coli from patients with diarrhea. J Clin Microbiol 1990;
28:8825.
20. Jensen C, Ethelberg S, Olesen B, et al. Attaching and effacing
Escherichia coli isolates from Danish children: clinical signicance and
microbiological characteristics. Clin Microbiol Infect 2007; 13:86372.
21. Fey PD, Wickert RS, Rupp ME, Safranek TJ, Hinrichs SH. Prevalence
of non-O157:H7 Shiga toxin-producing Escherichia coli in diarrheal
stool samples from Nebraska. Emerg Infect Dis 2000; 6:5303.
22. Carroll KC, Adamson K, Korgenski K, et al. Comparison of a com-
mercial reversed passive latex agglutination assay to an enzyme immu-
noassay for the detection of Shiga toxin-producing Escherichia coli.
Eur J Clin Microbiol Infect Dis 2003; 22:68992.
23. Manning SD, Madera RT, Schneider W, et al. Surveillance for Shiga
toxin-producing Escherichia coli, Michigan, 20012005. Emerg Infect
Dis 2007; 13:31821.
24. Teel LD, Daly JA, Jerris RC, et al. Rapid detection of Shiga toxin-pro-
ducing Escherichia coli by optical immunoassay. J Clin Microbiol
2007; 45:337780.
25. Park CH, Kim HJ, Hixon DL, Bubert A. Evaluation of the duopath
verotoxin test for detection of Shiga toxins in cultures of human
stools. J Clin Microbiol 2003; 41:26503.
26. Gould LH, Bopp C, Strockbine N, et al. Recommendations for diagno-
sis of Shiga toxinproducing Escherichia coli infections by clinical lab-
oratories. MMWR Recomm Rep 2009; 58:114.
27. Pradel N, Livrelli V, De Champs C, et al. Prevalence and characteriza-
tion of Shiga toxin-producing Escherichia coli isolated from cattle,
food, and children during a one-year prospective study in France.
J Clin Microbiol 2000; 38:102331.
28. Park CH, Gates KM, Vandel NM, Hixon DL. Isolation of Shiga-like
toxin producing Escherichia coli (O157 and non-O157) in a commu-
nity hospital. Diagn Microbiol Infect Dis 1996; 26:6972.
29. Jelacic JK, Damrow T, Chen GS, et al. Shiga toxin-producing Escheri-
chia coli in Montana: bacterial genotypes and clinical proles. J Infect
Dis 2003; 188:71929.
30. Bokete TN, Whittam TS, Wilson RA, et al. Genetic and
phenotypic analysis of Escherichia coli with enteropathogenic
Etiology of Childhood Diarrhea CID 2012:55 (1 October) 903
 characteristics isolated from Seattle children. J Infect Dis 1997;
175:13829.
31. Dubberke ER, Haslam DB, Lanzas C, et al. The ecology and pathobi-
ology of Clostridium difcile infections: an interdisciplinary challenge.
Zoonoses Public Health 2011; 58:420.
32. Barratt JL, Harkness J, Marriott D, Ellis JT, Stark D. A review of Dien-
tamoeba fragilis carriage in humans: several reasons why this organ-
ism should be considered in the diagnosis of gastrointestinal illness.
Gut Microbes 2011; 2:312.
33. Barletta F, Ochoa TJ, Mercado E, et al. Quantitative real-time poly-
merase chain reaction for enteropathogenic Escherichia coli: a tool for
investigation of asymptomatic versus symptomatic infections. Clin
Infect Dis 2011; 53:12239.
34. Finkbeiner SR, Holtz LR, Jiang Y, et al. Human stool contains a previ-
ously unrecognized diversity of novel astroviruses. Virol J 2009; 6:161.
904 CID 2012:55 (1 October) Denno et al
35. Holtz LR, Finkbeiner SR, Kirkwood CD, Wang D. Identication of a
novel picornavirus related to cosaviruses in a child with acute diar-
rhea. Virol J 2008; 5:159.
36. Finkbeiner SR, Le BM, Holtz LR, Storch GA, Wang D. Detection of
newly described astrovirus MLB1 in stool samples from children.
Emerg Infect Dis 2009; 15:4414.
37. Holtz LR, Finkbeiner SR, Zhao G, et al. Klassevirus 1, a previously
undescribed member of the family Picornaviridae, is globally wide-
spread. Virol J 2009; 6:86.
38. Finkbeiner SR, Allred AF, Tarr PI, Klein EJ, Kirkwood CD, Wang D.
Metagenomic analysis of human diarrhea: viral detection and discov-
ery. PLoS Pathog 2008; 4:e1000011.
39. Phillips G, Lopman B, Rodrigues LC, Tam CC. Asymptomatic rotavi-
rus infections in England: prevalence, characteristics, and risk factors.
Am J Epidemiol 2010; 171:102330.
Downloaded from http://cid.oxfordjournals.org/ by guest on July 11, 2016