ASSOCIATED WITH A RARE ESCHERICHIA COLI SEROTYPE.
Lee W. Riley, MD, Robert S Remis MD, MPH, Steven D Helgerson, MD, MPH,
Harry B McGee, MPH, Joy G Wells, MS, Betty R Davis, MS, Richard J Herbert
MD, Ellen S Olcott, RN, Linda M Johnson, RN, MS, Nancy T Hargrtett, PhD,
Paul A Blake, MD, MPH and Mithchell L Cohen MD.
We investigated two outbreaks of an unusual gastrointestinal illness that
affected at least 47 people in Oregon and Michigan in February through March
and May through June 1982. The illness was characterized by severe crampy
abdominal pain, initially watery diarrhoea followed by grossly bloody diarrhoea,
and little or no fever. It was associated with eating at restaurants
belonging to the same fast food restaurant chain in Oregon (P less than
0.005) and Michigan (P=0.0005) (ie. McDonald's) and with eating
any of three sandwiches containing three ingredients in common (beef patty,
rehydrated onions and pickles). Stool cultures did not yield previously
recognized pathogens. However a rare Escherichia coli serotype, 0157:H7
that was not invasive or toxigenic by standard tests was isolated from 9
of 12 stools collected within four days of onset of illness in both outbreaks
combined, and from a beef patty from a suspected lot of meat in Michigan.
The only known previous isolation of this serotype was from a sporadic case
of haemorrhagic colitis in 1975. This report describes a clinically distinctive
gastrointestinal illness associated with E. coli 0157:h7, apparently transmitted
by undercooked meat (N Engl J Med 1983; 308:681-5).
In the first half of 1982 two outbreaks of an unusual gastrointestinal
ilness characterized by sudden onset of severe abdominal cramps and grossly
bloody diarrhoea, with no fever or low grade fever, occured in Oregan and
Michigan. Isolated cases of a similar illness had recently been reported
from Japan and the United States, but the etiologic agent had not been identified.
The outbreaks in Oregan and Michigan led to intensive epidemiologic and
laboratory investigations. In this report we describe the illness and the
evidence that it is associated with a rare serotype of Escherichia coli
that is neither invasive nor enterotoxigenic according to standard tests
and is not a recognised enteropathogenic E. coli.
We defined as a case an illness characterized by severe abdominal cramps,
grossly bloody diarrhoea, and stool examinations that did not yield shigella,
salmonella, campylobacter, ova or parasites. To find cases, we contacted
local physicians, reviews records of chief complaints in emergency rooms
and discharge records from December 1981 to February 1982 in parts of Oregon,
and from May to June 1982 in parts of Michigan, and began active surveillance
in all hospitals in the affected areas and nearby countries. In both states
we conducted case-control studies with either one or two age-matched and
neighborhood-matched controls for each case, using a questionnaire developed
after intensive interviews of reported cases. We examined specific food
exposures at restaurants implicated by the Michigan neighborhood case-control
study; by comparing foods eaten by cases and by controls selected from persons
who had visited the restaurants with the cases and had remained well. Food-handling
procedures, food delivery and turnover, and employee records were reviewed
at the implicated restaurants. In Michigan, grill temperature was measured
on one implicated restaurant by means of a rapid readout surface pyrometer
(Pyrcon, type 4000A, Alnor instuments). Logistic regression analysis, the
binomial test, and the Pike-Morrow extensions of the Mantel-Haenszel test
were used for statistical analysis.
Stool specimans from cases were examined at the local at the local hospital
and state health department laboratories for salmonella, shigella, campylobacter,
ova and parasites in both states. Stool specimans from some cases and controls
were frozen (-70C) until examination at the Centers for Disease Control
(CDC). In Oregon, stool specimans were also collected from 45 persons who
visited emergency rooms because of non bloody diarrhoea. Environmental samples,
including food, were collected at implicated food establishments. At the
CDC, stool specimans from the outbreaks were examined for salmonella, shigella,
pathogenic vibrios, T. entrocelitica, campylobacter species, bacilus species,
enterotoxigenic and enteroinvasive E coli and anaaerobes (including Closttridium
difficile and toxin). Five E. coli isolates from each stool were serotyped.
The stool specimans were also examined for viruses by electron microscopy,
by immunoelectron microscopy with acute phase and convalescent phase serum,
and by culture in rhesus monkey and human fibroblast tissue cells. The 45
diarrheal stool specimans obtained at emergency rooms were examined for
E. coli 0157 and kiebsiella. The foods were examined for E. coli and Bacillus
E. coli 0157:H7 E. coli 0157:H7, klebridle oxytoca an B _____ isolates were
tested for invasiveness by the Sereny test, for heat stable toxin production
by the suckling mouse assay and for heat labile toxin production by the
Y-1 adrenal cell test. E coli 0157:H7 was tested in an infant rabbit assay
(Potter ME: personal communication). The E. coli serotyping records of the
US Department of Agriculture Animal Laboratories at Ames, Iowa, the Pennsylvania
State University Veterinary Science laboratory, and the CDC Enteric Reference
Laboratory were reviewed for previous identifications of E coli 0157:H7.
The following case report is typical of the cases seen in both outbreaks.
A 56 year old man was awakened by severe abdominal cramps in the right lower
quadrant. Later the same morning, watery diarrhoea developed, with bowel
movements every 15 to 30 minutes. The patient initially noted small amounts
of blood, but later the same day the diarrhoea became grossly bloody, with
bright red blood, described as "all blood and no stool." He had
slight nausea but no vomiting. He was hospitalized on the following day
with continuous crampy abdominal pain had frequent bloody diarrhoea. He
was afebrile and on abdominal examination had no guarding, rebound tenderness,
or distension. The white cell count was 17,900 with a slight shift to the
left. A barium enema revealed edema of the ascending and traverse colon,
with areas of spasm. Examinations of three stool specimans collected within
three days after the onset of illness did not detect salmonella, shigella,
campylobacter, versinia, ova or parasites. The patient was treated with
intravenous fluids and doxycycline and the bloody discharge subsided by
the fifth hospital day. He was discharged the next morning.
Clinical and Epidemiologic investigations
In Oregon, 25 persons from seven municipalities in one country and one person
from an adjacent county became ill between February 5 and March 15. The
median age was 28 years, with a range of 8 to 76 years; there were 16 males
and 10 females. All consulted a physician, and 19 (73 per cent) were hospitalized.
The duration of illness ranged from two to nine days, with a median of four
days. In Michigan, 21 persons had onset of illness between May28 and June
27, The median age was 17 years, with a range of 4 to 58 years. All consulted
a physician and 14 (67 per cent ) were hospitalised. The illness lasted
from three to more than seven days. Three patients in both outbreaks had
temeratures of 38.5C or more. The white cell count ranged from 7,600 to
19,600 (mean 14,000) in Oregon and from 7,600 to 17,400 (mean 13,000) in
Michigan, with a alight to moderate shift to the left. The erythrocyte sedimentation
rates, serum electrolyte concentrations, liver function tests, prothrombin
times, and urinalyses were normal in all patients in whom these tests were
done. Sigmoidoscopy performed in 10 patients revealed moderately hyperemic
mucosa in 3. In six of seven patients barium enemas demonstrated marked
submucosal edema with spasm and a "thumb printing" pattern in
the ascending and traverse colon of 23 patients whose treatment histories
were available received tetracycline compounds (eight patients) or erythromycin
(three patients). The mean duration of illness of the group treated with
antimicrobials was not significantly diiderent from that of the untreated
group. There were mo deaths, complications or sequelae in any of the cases.
In Oregon, 25 of 26 cases and 47 neighborhood controls were interviewed.
During the two weeks before onset of illness, 21 of 25 cases (84 per cent)
but only 13 of 47 controls (28 per cent) had eaten at restaurant 1, one
of a chain of fast food restaurants (Chain A) (P less than 0.005 by logistic
regression analysis). Three of the four who did not recall having eaten
at Resataurant 1 had eaten at another Chain A restaurant within a week before
the onset of illness.
The patients who ate at one of the three restaurants of Chain A in the county
were more likely (21 of 24) to have eaten one of the chain's speciality
hamburgers than were neighborhood controls (11 of 20) who had eaten at the
same restaurants (P less than 0.05, logistic regression analysis). The three
patients who did not eat a speciality hamburger ate a regular hamburger
(one patient) or a cheeseburger (two patients). The three types of sandwiches
shared three ingredients, which were always served together - ie, reconstituted
dehydrated onions, standard size hamburger meat patties, and pickles. Each
of these three ingredients was eaten by a higher proportion of cases than
neighborhood controls (P less than 0.05, Pike - Morrow extension of the
Mantel - Haenszel test), but no single ingredient could be independently
associated with disease, because they had always been served together.
In Michigan, 18 of 21 cases and their age matched neighborhood controls
were interviewed (Matched controls could not be found for two cases.) Seventeen
of 18 cases and 4 of 16 controls had eaten at either Restaurant 2 or Restaurant
3 of Chain A within 10 days before the onset of illness (P = 0.0005), binomial
test.) No other exposures were significantly associated with illness. Again,
cases were more likely than their restaurant controls to have eaten the
same three food items implicated in Oregon (17 of 17 vs. 12 of 19; P less
than 0.05, Pike - Morrow extension of the Mantel - Haenszel test). The mean
period between single exposures to the implicated foods and onset of crampy
abominal pain was 3.9 days in Oregon and 3.8 days in Michigan. The attack
rate for persons eating sandwiches that included the three ingredients was
estimated to be about 1 case per 1000 sandwiches in Oregon, and 1.8 cases
per 1000 for speciality hamburgers in Michigan, and 0.6 case per 1000 regular
hamburgers and cheeseburgers in Michigan. The speciality hamburger had twice
the quantity of meat (two patties vs. one) and onions as the regular hamburger
Only one case of bloody diarrhoea occured among employees at the three restaurants.
None of the family members of the cases in Oregon and Michigan had bloody
diarrhoea. In Michigan, 4 of 13 persons who had accompanied the cases to
the implicated restaurants and had eaten one of the implicated foods had
cramps and diarrhoea without blood in one to seven days, whereas none of
the 12 who had not eaten these foods were ill (P = 0.06, Fisher's exact
test, one tailed).
There were no obvious defects in equipment of food handling practices in
the Oregon or Michigan restaurants. In Michigan, inadequate stock rotation
of some foods was observed in Restaurant 2, and during busy periods certain
parts of the grill were cooler than the temperature standard established
by Chain A.
E. coli 0157:H7, B pumilus, and K oxytoca were the only bacteris isolated
from three or more cases in either outbreak. E. coli 0157:H7 was recovered
from stool of three of six Oregon cases and none of 10 neighborhood controls
(P = 0.03). Fisher's exact test, one tailed). An additional patient with
E. coli 0157:H7 was identified by screening E. coli isolates sent to the
CDC from Oregon. K oxytoca was isolated from two cases and no controls (it
was isolated in Oregon from one of the controls); B pumilus was isolated
from two cases and one control. Y enterocolitica was recovered from one
case only on cold enrichment, suggesting small numbers of organisms. Examination
of the 45 control emergency room diarrhoeal stool specimens yielded K oxytoca
from two samples and E. coli 0157:H7 from none. In the virologic studies
we did not detect viral particles on electron microscopy or immunoelectron
microscopy or in tissue cultures. E. coli 0157:H7 was recovered from 6 of
14 specimens from cases and none of 4 specimens from controls in Michigan.
K oxytoca and B pumilus were each isolated from three cases. In both states
combined, E. coli 0157:H7 was isolated from 9 of 12 stools collected within
four days after the onset of illness, but from none of 7 stools collected
seven or more days after onset (P = 0.002, Fisher's exact test, one tailed).
In both outbreaks 0157:H7 was the predominant E. coli isolated (median of
four of five isolates serotyped). The serotypes of K oxytoca from patients'
stools were different. The E. coli and K oxytoca were sensitive to all anti
microbial agents tested, and B pumilus had two different antimicrobial-susceptibility
E. coli 0157:H7 was not islolated from food samples collected in Oregon
but was isolated from a frozen raw, standard size hamburger patty from a
suspected lot used at the Michigan restaurants during the outbreak period.
This patty had been stored at a processing plant in another state, as part
of a quality control program, and had never been in either restaurant. E.
coli 0157:H7 did not produce either heat labile or heat stable enterotoxin,
nor were isolates invasive on Sereny testing or tissue culture assays. The
organism, however, did produce a nonbloody diarrhoea in infant rabbits.
The laboratories of the US Department of Agriculture and Pennsylvania State
University reported no E. coli 0157:H7 from animals in the United States.
The CDC laboratory detected only one strain of E. coli 0157:H7 among over
3000 E. coli organisms serotyped since 1973; it was isolated from a 50 year
old California woman in 1975 during an acute, self limited afebrile illness
with severe abdominal cramps followed by grossly bloody diarrhoea.
Two outbreaks of a clinically distinctive diarrhoeal illness occured three
months apart in two widely separated areas of the country among persons
who had eaten at restaurants of a single fast food chain. A rare E. coli
serotype 0157:H7 was isolated from ill patients in both outbreaks and from
a retained speciman of hamburger patty from a suspected lot in the Michigan
outbreak. We hypothesize that on two occasions, E. coli 0157:h7 contaminated
the meat before it was made into hamburger patties, survived the cooking
procedures at the restaurants, and caused illness among some people who
ate the meat. Apparently failure to use meat by lots permitted cases to
occur during a prolonged period in both outbreaks. The low attack rate in
both outbreaks suggested a low level of contamination of a large volume
of raw meat, reduction of the innoculum by cooking, or unknown host susceptibility
factors. The pathogenesis of the illness and the source contamination of
the raw meat by this E. coli serotype are not known.
The evidence that E. coli 0157:H7 was the etiologic agent in these two outbreaks
may be summarised as follows. In two outbreaks this serotype was isolated
from ill persons but not from healthy persons, ill persons with other forms
of diarrhoea, or patients who had completely recovered from bloody diarrhoea.
The only other isolate of E. coli 0157:H7 identified at the CDC was from
a patient who had an identical disease in 1975. Preliminary studies indicate
that this E. coli produces non bloody diarrhoea in infant rabbits, whereas
E. coli isolates from two human control stool specimans do not (Potter ME:
personal communication). The clinical presentation of this illness may be
distinguished from that of the bloody diarrhoea or dysentry described in
Shigellosis, amebiasis, campylobacteriosis, or invasive E. coli gastrointestinal
bleeding. However, like other causes of bloody diarrhoea, E. coli 0157;H7
may produce a spectrum of illness. In michigan there was evidence of no
bloody diarrhoea among persons who ate at the same restaurants of Chain
A. our case definition, which required the prescence of bloody diarrhoea,
would have excluded milder cases of this illness.
The case of bloody diarrhoea associated with E. coli 0157:H7 in 1975 suggests
that this disease has occured sporadically in the past. Descriptions of
sporadic cases of a similar illness, frequently associated with antimicrobial
use and distinct from pseudomembranous colitis, have been reported from
several countries, including Japan and the United States. Sakurai et al
reported that colonscopy recealed areas of diffuse mucosal haemorrhage or
erosion, mostly in the right colon: biopsy specimans of the mucosa showed
little or no inflammatory change. In the Oregon and Michigan outbreaks,
only one patient in Oregon had been receiving and antibiotic (penicillin)
before the onset of illness.
Other reports have described cases of haemorrhagic colitis not associated
with antimicrobial agents or cases of so called ischemic colitis in young
adults. Some of these may represent sporadic cases of the same illness that
we have described in these outbreaks.
Acute haemorrhagic enterocolitis has been reported in patients with K oxytoca
islolated from their stools, but no control stool cultures were obtained.
In the two outbreaks that we investigated, strains of K oxytoca were of
several serotypes and were isolated from cases and controls. Although B
pumilus can induce enterocolitis in guinea pigs given clindamyein, the B
pumilus isolates obtained in the two outbreaks were isolated from both cases
and controls appeared to include more than one strain.
E. coli can cause diarrhoea by direct invasion of the intestinal mucosa
and by eleboration of heat stable enterotoxins or of heat labile resembling
cholera toxin. Our laboratory studies have shown that strain 0157:h7 does
not cause disease by these mechanisms. Strains of E. coli 0157:H7 that cause
diarrhoea by poorly defined mechanisms have been studied, but they do not
typically cause bloody diarrhoea. E. coli 0157:H7 may cause diarrhoea by
an as yet unknown mechanism, perhaps by the production of previously unrecognized
Isolation of E. coli 0157:H7, a rare serotype, from cases in two outbreaks
of bloody diarrhoea, from the suspected vehicle for the outbreaks, and from
a sporadic case of bloody diarrhoea in 1975 strongly suggests, but does
not prove, that it caused the illness; proof may require studies using animal
models and perhaps human volunteers. Similarly, the epidemiology, clinical
spectrum, and pathogenesis of this unusual illness and the reservoir of
the putative etiologic agent are still poorly understood or unknown and
require continued clinical, epidemiologic, and laboratory studies.