The first major step toward defining the pathogenic potential of E. coli O157:H7 was made in 1983, when isolates of this serotype were found to produce a toxin [27]. This toxin has been called both "Shiga-like toxin," because of its close relation to Shiga toxin, and "Vero-toxin," because of its toxicity to Vero (green monkey kidney) cells in tissue culture. Subsequent studies [28-34] have shown that E. coli O157:H7 produces at least two distinct Shiga-like toxins and that O157:H7 is only one serotype of Shiga-like toxin-producing E. coli. To distinguish their clinical and pathologic features from those of enterotoxigenic, enteroinvasive, and enteropathogenic E. coli, these Shiga-like toxin-producing strains of E. coli, which cause hemorrhagic colitis, have also been called enterohemorrhagic E. coli [35]. Serotypes other than O157:H7 can cause illness similar to that caused by E. coli O157:H7, because many strains of E. coli can produce Shiga-like toxin. Moreover, many serotypes of Shiga-like toxin-producing E. coli other than O157:H7 have been isolated from patients with hemorrhagic colitis or the hemolytic- uremic syndrome. Although few studies have been done to determine the incidence of diarrheal illness caused by Shiga-like toxin-producing E. coli other than O157:H7, none of these organisms is a common cause of diarrhea. Thus, a 2-year survey in Canada reported that 0.7% of stools (36 of 5415) showed Shiga-like toxin-producing E. coli other than O157:H7; this rate was higher than that for shigellae, and the serotypes most often isolated were O26:H11 and O103:H2 [6]. Serotype O157:H7 was also identified in 7 of these 36 stools, bringing into question the precise role of the non-O157:H7 serotype in causing the diarrhea. Shiga-like toxin-producing E. coli other than O157:H7, such as O2:H5, have also been isolated from patients with ulcerative colitis, but, again, their precise role in causing this disease or its exacerbations is unclear [36]. What is clear is that many more studies are needed to define the roles of Shiga-like toxin-producing coliforms other than O157:H7 in causing human disease.

Many species of E. coli produce Shiga-like toxin, but we focus on E. coli O157:H7 because it is the most common. The term "Shiga-like toxin" is preferred to "Verotoxin" because it emphasizes the common mechanism of action of the toxins produced by certain species of E. coli and the Shiga toxin elaborated by shigellae, as well as the wide target cell of injury, that is, cells with the surface receptor globotriaosyl ceramide.

Methods

Top Methods Conclusion Author & Article Info References

Articles on E. coli O157:H7 were identified through MEDLINE and through review of bibliographies of relevant articles. All articles and case reports describing E. coli O157:H7 and its infection were selected. Data were abstracted without judgments about study design. Data quality and validity were assessed by independent author reviews.

Clinical Manifestations

Infection with E. coli O157:H7 presents with a wide spectrum of clinical manifestations, including severe abdominal cramps with little or no fever and watery diarrhea that often progresses to grossly bloody diarrhea [37]. Infection can be asymptomatic [13, 15, 19] or can present with only nonbloody diarrhea [4, 5, 8, 10, 11, 13, 16, 20, 26]. Extraintestinal involvement, including cardiac and neurologic manifestations, has been reported [38], and infection can be associated with the hemolytic-uremic syndrome [14, 16, 18, 19, 24] and thrombotic thrombocytopenic purpura [16, 39-41]. The disease can be fatal [15, 16, 18, 19, 37].

Asymptomatic Infection and Nonbloody Diarrhea

Cases of asymptomatic E. coli O157:H7 infection have occasionally been detected in outbreaks [13, 15, 19], but the incidence rates are difficult to estimate because stool samples from asymptomatic persons are rarely obtained for culturing. In an outbreak in Canada that involved kindergarten children, 31% of those exposed to the implicated source (19 of 62) were asymptomatic [17]. Fifty-three percent of the asymptomatic children (10 of 19) had laboratory evidence of E. coli O157:H7.

Nonbloody diarrhea without progression to hemorrhagic colitis has also been reported [4, 5, 8, 10, 11, 13, 16, 20, 26]. It was noted in 18% of culture-confirmed cases (3 of 17) in a nursing home outbreak of E. coli O157:H7 infection [13] and in as many as two of four children with culture-positive stools in a day care center [20]. Although cases of nonbloody diarrhea occur in outbreaks, the routine screening of all stool specimens for E. coli O157:H7 during one outbreak period showed that the number of these cases is probably small [16]. Patients with nonbloody diarrhea have less severe disease, are less likely to develop the hemolytic-uremic syndrome, and are less likely to die than are patients with bloody stools [11]. However, nonbloody diarrhea progressing to the hemolytic-uremic syndrome has been reported. In one 2-year prospective study, two of nine confirmed cases of E. coli O157:H7-associated hemolytic-uremic syndrome were preceded by nonbloody diarrhea [26]. Patients with bloody stools have a longer duration of diarrhea and report abdominal cramps and vomiting more often than do those with nonbloody diarrhea [20].

Hemorrhagic Colitis

Hemorrhagic colitis caused by E. coli O157:H7 is a clinical syndrome that consists of abdominal cramps; diarrhea that progresses to become bloody; radiologic or endoscopic evidence of colonic mucosal edema, erosion, or hemorrhage; and the absence of conventional enteric organisms in the stool [42]. This syndrome was first reported in 1971 [43], when five young adults developed reversible segmental colitis that rapidly resolved within 2 weeks without specific therapy. The term "evanescent colitis" was used to describe this entity as a clinical syndrome distinct from ulcerative, granulomatous, and ischemic colitis [43]. After the outbreaks of hemorrhagic colitis in Michigan and Oregon [1] and subsequent similar outbreaks, E. coli O157:H7 came to be recognized as an important etiologic agent for hemorrhagic colitis.

Hemorrhagic colitis may be the only manifestation of E. coli O157:H7 infection, or it may herald the development of the hemolytic-uremic syndrome. Thirty-eight percent to 61% of such infections result in hemorrhagic colitis [44], and surveillance studies [8, 9] have found E. coli O157:H7 in 27% to 36% of sporadic cases of hemorrhagic colitis. Infection with E. coli O157:H7 usually begins with the sudden onset of severe abdominal cramps, which are followed within hours by watery diarrhea that progresses to grossly bloody stools [37]. Upper gastrointestinal symptoms, such as nausea and vomiting, occur early and may be prominent. The incubation period ranges from 1 to 9 days (mean, 3.1 to 3.9 days) during community outbreaks [1, 14, 16] and from 1 to 14 days (mean, 4 to 8 days) in institutional settings [11, 13, 17, 19]. Medical attention is usually sought 2 to 3 days after the onset of diarrhea or abdominal pain, primarily because of bloody diarrhea [5], which is the most common symptom of E. coli O157:H7 infection [37]. The median duration of diarrhea is 3.0 to 7.5 days (range, 1 to more than 31 days) [10, 11, 37], and patients report a median of 10 to 11 bowel movements (range, 3 to more than 30 bowel movements) on the worst day of diarrhea [19, 37]. Diarrhea lasts longer in children (mean, 9.1 ± 2.0 days) than in adults (mean, 6.6 ± 1.1 days) [5], and it lasts longer in persons with bloody stools (mean, 12.2 days) than in those with nonbloody stools (mean, 6.8 days) [20]. Bloody stools develop a median of 0 to 1 days (range, 0 to 8 days) after the onset of diarrhea [11, 37] and last a median of 2 to 5 days (range, 1 to 22 days) [5, 10, 11, 37]. The amount of blood in each stool ranges from streaks to more than 4 cups on the worst day (median, 4.5 tablespoons) [37]. Most of the infected children (91%) in an outbreak in a day care center produced less than 1 tablespoon of blood per stool [20], whereas patients in other outbreaks have reported that bowel movements were essentially all blood with little fecal material [1, 5]. Other symptoms related to E. coli O157:H7 infection include severe abdominal cramps, right lower quadrant pain, nausea, vomiting, fever, and chills [1, 4, 10, 26].

The incidence of fever ranges from 0% to 32% [1, 5, 10, 11, 16, 19, 20] and was as high as 64% among persons with bloody diarrhea in one day care center outbreak [20]. When present, fever is usually mild, unlike that seen with the bloody diarrhea of shigellosis, amebiasis, campylobacteriosis, or enteroinvasive E. coli infection [1]. Abdominal distension and tenderness may be present, but the results of physical examination are usually normal [8, 19].

Laboratory studies usually show leukocytosis with moderate left shift and a mean leukocyte count of 13.0 to 14.0 x 10⁹/L (range, 6.2 to 20.0 x 10⁹/L) [1, 15, 19]. Hematocrit is generally not significantly decreased despite the bloody nature of the stools. The results of other studies, including erythrocyte sedimentation rate, serum electrolyte concentrations, liver test results, prothrombin times, and urinalysis results, are typically normal [1]. Mucus and leukocytes may be present in the stool but are usually seen in scant to moderate amounts [18, 19]. In an outbreak that involved 19 patients, 13 patients (68%) had 0 or fewer than 1 polymorphonuclear leukocyte per high-power field in their stools; 2 patients had 1 to 5; 2 patients had 5 to 30; and 2 patients had more than 30 [37].

The results of roentgenologic and colonoscopic examinations often suggest a diagnosis of inflammatory bowel disease or ischemic colitis. Abdominal plain films show an ileus in 60% to 100% of patients; distension of the small bowel, cecum, and ascending colon with little gas in the left colon is common; and bowel obstruction is sometimes considered [19, 37]. In barium enema studies, a pattern of submucosal edema resembling thumbprinting in the ascending and transverse colon was noted in six of seven patients in the index study [1] and in all three patients with bloody diarrhea in another study [19]. The colonic mucosa, especially in the ascending colon, may have a shaggy appearance with thickened folds [18]. Technetium—red blood cell scanning of the abdomen usually shows radiotracer within the right colon and extending into the transverse colon if bleeding is active [19]. The results of sigmoidoscopic and colonoscopic examinations show an increasing frequency and severity of mucosal abnormalities from the rectum to the cecum [37, 45]. Edema, erythema, and superficial ulceration, usually in a patchy distribution, are the most important features; other, less common findings include dusky-appearing mucosa and frank hemorrhage [37, 45]. The most severe abnormalities occur in the cecum and the ascending colon; in one study [45], one or both of these sites were affected in all patients in whom the sites were examined. The rectosigmoid mucosa appears normal to moderately hyperemic and sometimes has superficial erosions [1, 5, 15, 19].

Mean durations of 4 to 8 days (range, 2 to 14 days) have been reported [1, 4, 8]. Hospitalization rates [4, 14, 16, 18, 19] have ranged from 13% in a junior high school outbreak [10] and a nursing home outbreak [15] to 73% in a community outbreak in Oregon [1]; the mean hospital stay is 6 to 14 days [4, 18, 19, 37]. Most hospitalized patients had hemorrhagic colitis manifested by crampy abdominal pain, grossly bloody diarrhea, abdominal distention, and low-grade fever [10, 19]. Most hospitalized patients with hemorrhagic colitis recovered completely within 1 week without specific therapy or complications [5]. However, complications from hemorrhagic colitis associated with E. coli O157:H7 have been reported. In a nursing home outbreak, 3 of 34 residents had upper gastrointestinal bleeding documented during nasogastric suction to relieve abdominal distention; 2 of 34 residents had anemia with hematocrits less than 0.30; and 1 of 34 residents became hypotensive and subsequently had cerebrovascular ischemia [19].

In one outbreak, antibiotic treatment during the exposure period and before symptom onset was reported to be a risk factor for person-to-person transmission of E. coli O157:H7 [11]. An association between antibiotic use after symptom onset and an increased case-fatality rate was also documented. Other factors that may increase susceptibility to E. coli O157:H7 include young and old age and gastrectomy; this finding suggests that gastric acidity may play a protective role in the pathogenesis of this infection [11].

The Hemolytic-Uremic Syndrome

The hemolytic-uremic syndrome is a distinct entity characterized by microangiopathic hemolytic anemia, thrombocytopenia, and renal failure. Many factors have been implicated in the development of this syndrome, including genetics, pregnancy, drugs (such as oral contraceptives and estrogen therapy), toxins, chemicals, viruses, and bacteria [46-48]. On the basis of epidemiologic, clinical, and laboratory data, two major subgroups of the syndrome have been identified: typical (or epidemic) and atypical (or sporadic) [47-49]. The syndrome occurs predominantly in infants and young children and is the most common cause of acute renal failure in children [46]; the mortality rate is 5% to 10%, and persistent disability occurs in one third of survivors [50].

Most patients with the hemolytic-uremic syndrome have gastrointestinal prodromes [48, 51], and nonbloody or bloody diarrhea is the most common presenting symptom. Other presenting symptoms include abdominal cramps, vomiting, fever, lethargy, seizure, pallor and respiratory distress; some patients have no distinct prodromes [52]. The initial gastrointestinal manifestations have been confused with ulcerative, ischemic, and pseudo-membranous colitis; cecal polyps; intussusception; and appendicitis [51, 53, 54]. Radiographic abnormalities include bowel edema, transient segmental narrowing, and bowel stenosis [55], and sigmoidoscopy has shown friable mucosa, rectal ulceration, pseudomembranes, and diffuse colitis [51]. The cecum and ascending colon are most commonly involved and often show hemorrhage and necrosis. Gastrointestinal complications of the hemolytic-uremic syndrome are common and include nonspecific inflammation [53], chronic ischemic colitis with colonic stricture [56], and full-thickness necrosis resulting in colonic perforation [51, 57]. Rectal prolapse, toxic megacolon, and ascites have also been reported. Although central nervous system complications are not part of the classic triad of conditions associated with the hemolytic-uremic syndrome, they may occur in 30% to 50% of patients [22, 58-60]. Common neurologic manifestations include irritability and lethargy, but serious complications, such as seizure and coma, can occur [59, 61]. Decerebrate posturing, hemiparesis, and focal seizures have also been reported. Patients may sustain long-term neurologic sequelae, including generalized seizures, psychomotor or mental retardation, and cortical blindness [59, 61].

Escherichia coli O157:H7 has been implicated as an important etiologic agent in the typical form of the hemolytic-uremic syndrome [11, 17, 20, 48, 62, 63], and it is the most common pathogen isolated from patients with this condition [26, 63-67]. It was isolated from 46% of patients (13 of 26) in a 10-year, retrospective, population-based study of the hemolytic-uremic syndrome [52]; from 58% of patients with the hemolytic-uremic syndrome (7 of 12) in a prospective study in the Pacific Northwest [63]; and from 64% of children (9 of 14) with the hemolytic-uremic syndrome in British Columbia, Canada [26]. Infection with E. coli O157:H7 has been associated with both sporadic cases [68] and outbreaks [69] of the hemolytic-uremic syndrome. In a family outbreak of the hemolytic-uremic syndrome that involved five siblings, Shiga-like toxin was isolated from each child's stool, and E. coli O157:H7 was isolated from the stools of two of the children [69]. Similarly, evidence of Shiga-like toxin-producing E. coli infection was detected in 75% of children (30 of 40) with "idiopathic" hemolytic-uremic syndrome [62]; 3 of the 12 isolates were serotype O157:H7. The occurrence of the hemolytic-uremic syndrome during outbreaks of E. coli O157:H7 infection is also well documented; as many as 30% of patients with hemorrhagic colitis associated with E. coli O157:H7 developed the hemolytic-uremic syndrome [70]. Rates of the hemolytic-uremic syndrome in outbreaks of E. coli O157:H7 infection ranged from 3% to 53% among residents of two institutions for mentally retarded persons [11, 14, 16, 18-20]. Indeed, the hemolytic-uremic syndrome is sometimes the clue to the recognition of E. coli O157:H7 outbreaks [17]. When the results from four studies of sporadic cases of E. coli O157:H7 infection or hemorrhagic colitis were combined, it was found that 9% of patients had developed the hemolytic-uremic syndrome [3-58, 71]. Other studies of children with E. coli O157:H7 infection showed a progression rate of 6% to 8% [21-23]. Considering the different study populations and the fact that some persons may not have bloody diarrhea, the true rate of progression to the hemolytic-uremic syndrome from E. coli O157:H7 infection is estimated to be 2% to 7% [71].

Escherichia coli O157:H7 accounts for most or a major part of cases of the hemolytic-uremic syndrome in North America, but, in studies of children in Argentina, where the incidence of the syndrome is among the highest in the world and where meat consumption is almost universal, investigators isolated E. coli O157:H7 in only 2% of children with the hemolytic-uremic syndrome (1 of 51), and 48% of these children (15 of 31) had evidence of free fecal toxin [72]. This suggests that Shiga-like toxin-producing E. coli other than O157:H7 are an important cause of the hemolytic-uremic syndrome in Argentina. Combining data on seroconversion, free fecal toxin, and DNA-probe positivity, the frequency of Shiga-like toxin-associated diarrhea in Argentine children is 32% (14 of 44) [72]. The high incidence of Shiga-like toxin-associated diarrhea in young Argentine children probably explains the high frequency and occurrence of the hemolytic-uremic syndrome seen in this age group. Although the syndrome is considered to be predominantly a disorder of childhood, its development after hemorrhagic colitis caused by E. coli O157:H7 has been reported in two young adults (25 and 36 years of age, respectively) [8, 73], in elderly patients [19], and in 22% of affected residents in one nursing home outbreak [11].

Studies of the hemolytic-uremic syndrome have been limited in their ability to ascertain temporal trends in the syndrome, but its incidence has clearly been increasing. A population-based study from King County in Washington State [74] documented an increase in the incidence of the hemolytic-uremic syndrome in that area during the 15-year study period. A 10-year, retrospective, population-based study in Minnesota [52] also reported a significant increase in the incidence of the syndrome; most of the increase was attributed to disease in young children. The increasing incidence of the syndrome coincided with the emergence of E. coli O157:H7 as an important pathogen, and the seasonal pattern of cases of the hemolytic-uremic syndrome was similar to that of E. coli O157:H7 infections [52], further strengthening the evidence for the association of the hemolytic-uremic syndrome with E. coli O157:H7.

Although most patients who have E. coli O157:H7-associated hemolytic-uremic syndrome had bloody diarrhea before their illness, some had nonbloody diarrhea [26], and others had no prodromal illness [16]. The hemolytic-uremic syndrome developed abruptly 5 to 10 days after the onset of diarrhea during outbreaks [18, 37], but the time between the onset of diarrhea and the diagnosis of the hemolytic-uremic syndrome varied (mean, 6.5 to 7.7 days) [63, 75]. Leukocytosis with a left shift on presentation, high temperature [18], and extreme young or old age [18, 37, 52] appear to be indicators for progression to the hemolytic-uremic syndrome, and the initial leukocyte count correlates positively with adverse outcome [76]. Conditions on day 3 of illness were also shown to be strongly predictive of progression to the hemolyticuremic syndrome; persons who subsequently developed the syndrome were significantly more likely to have a documented fever or leukocytosis (leukocyte count, more than 12.0 x 10⁹/L) on day 3 of illness [18]. The prolonged use of antimotility or antidiarrheal agents has also been proposed as a risk factor [77, 78]. The selective development of the hemolytic-uremic syndrome among patients infected with E. coli O157:H7 may be related to host susceptibility factors, such as preexisting immunity, inoculum size, virulence of the strain, or other unknown factors [10, 79, 80]. The disproportionate number of cases of the hemolytic-uremic syndrome that occur in children younger than 5 years of age further suggests that host factors may be important [52]. Strain characteristics may also play a role. One study [81] found that the risk for the hemolytic-uremic syndrome is increased when isolates contain only Shiga-like toxin II genes; this suggests that Shiga-like toxin II may be more virulent than Shiga-like toxin I. Although the case-fatality rate in childhood cases of the hemolytic-uremic syndrome is typically about 5% to 10% [46, 62], the case-fatality rate associated with the hemolytic-uremic syndrome during an outbreak of hemorrhagic colitis was 88% among elderly patients [11].

Thrombotic Thrombocytopenic Purpura

Thrombotic thrombocytopenic purpura consists of a pentad of findings: thrombocytopenia, microangiopathic hemolytic anemia, fever, renal failure, and neurologic symptoms. Thrombotic thrombocytopenic purpura is thought to represent a more extensive form of the clinical spectrum of vascular diseases that produces the hemolytic-uremic syndrome [47], and the criteria for diagnosis of thrombotic thrombocytopenic purpura are the same as those for the hemolytic-uremic syndrome, with the addition of fever and new-onset neurologic deficit. Postmortem examinations usually show widespread vascular lesions with platelet-fibrin thrombi [82]. Many agents or conditions, including drugs, toxins, infectious agents, pregnancy, and underlying immunologic diseases, have been implicated as causes of thrombotic thrombocytopenic purpura [83].

Infection with E. coli O157:H7 has been implicated in some cases of thrombotic thrombocytopenic purpura [39-41], and this condition has served as a clue in the recognition of outbreaks of E. coli O157:H7 infection [16]. Although such progression is rare, as many as 8% of patients (3 of 37) with hemorrhagic colitis associated with E. coli O157:H7 infection progressed to thrombotic thrombocytopenic purpura in one outbreak [16]. All documented cases of thrombotic thrombocytopenic purpura associated with E. coli O157:H7 have occurred in adults, and progression to thrombotic thrombocytopenic purpura in children with E. coli O157:H7 infection has not been reported.

Death

Case-fatality rates for E. coli O157:H7 infection range from 3% to 36% among elderly residents of nursing homes [11, 15, 19] and residents of institutions for mental retardation [18]. The risk for death is strongly related to age: Patients at extremes of age are at increased risk for E. coli O157:H7-associated diarrhea as well as for the hemolytic-uremic syndrome, thrombotic thrombocytopenic purpura, and death [37]. In elderly persons, especially those with serious underlying diseases, infection can be particularly severe and often resembles ischemic colitis [19]. Deaths in elderly persons appear to be caused by various events. In one outbreak, two elderly persons died [37]. One was a 78-year-old woman with thrombotic thrombocytopenic purpura who had a grand mal seizure followed by severe hypotension and coma; she died on the 20th day of illness. The other was a 70-year-old woman with thrombotic thrombocytopenic purpura who developed bilateral pneumonitis and died on the 29th day of illness [37]. In one nursing home outbreak in which four persons died, one patient died secondary to congestive heart failure from fluid overload; one had fever and Clostridium perfringens bacteremia; and two died after developing high fever (> 38.9 °C) with no other identifiable source of infection [19]. In another large nursing home outbreak, 17 of 19 deaths were attributed directly or indirectly to E. coli infection, and the causes of death included colitis, pulmonary edema, pneumonia, myocardial infarction, and the hemolytic-uremic syndrome [11].

Epidemiology

Escherichia coli O157:H7 was first recognized as a cause of human illness in two separate outbreaks of hemorrhagic colitis in Michigan and Oregon in 1982 [1]. The organisms were transmitted by the same source of undercooked beef, and Shiga-like toxin-producing strains of E. coli O157:H7 were isolated from the stools of the affected persons and from a sample of the implicated beef burgers but from no healthy controls. Increasing numbers of diseases related to E. coli O157:H7 have been reported since 1982; most have been sporadic [2-9], but many institutional and community-wide outbreaks [10-1924, 84-88] have occurred in nursing homes [11, 15, 19], schools [10, 17], and day care centers [20] or have been related to eating at fast food restaurants [1, 16, 84, 87], drinking untreated municipal water or fresh-pressed apple cider [24, 88], or swimming in lake water [85] (Table 1).

It is estimated that 0.6% to 2.4% of all cases of diarrhea [3, 6, 25, 26] and 15% to 36% of all cases of bloody diarrhea or hemorrhagic colitis [5, 7-9, 89] are associated with E. coli O157:H7 (Table 2). In a 1-year (1985-1986), population-based study in the Puget Sound area of Washington State [3], E. coli O157:H7 was the third most common cause of bacterial diarrhea. Among the 4539 patients who submitted stool specimens, E. coli O157:H7 was isolated in 25 cases (0.6%) and followed Campylobacter (165 cases, 3.6%) and Salmonella organisms (70 cases, 1.5%) in frequency. Shigella organisms were isolated slightly less frequently than E. coli O157:H7 (23 cases, 0.5%). The population-based incidence rates in the same study [3] were 8 cases per 100 000 person-years for E. coli O157:H7; 50 cases per 100 000 person years for Campylobacter organisms; 21 cases per 100 000 person-years for Salmonella organisms; and 7 per 100 000 person-years for Shigella organisms. Other prospective studies [6, 26] have found E. coli O157:H7 to be second to Salmonella organisms in areas of Canada (2.4%) and to Campylobacter organisms in Great Britain (1.9%) as the most common cause of bacterial diarrhea. In both studies [6, 26], E. coli O157:H7 was isolated more often than Shigella, Yersinia, or Aeromonas organisms. In a prospective study limited to persons with grossly bloody diarrhea in Calgary, Canada [5], E. coli O157:H7 was isolated from 15% of patients (19 of 125). A 21-month surveillance study in the United States, established after initial outbreaks, identified 103 cases of hemorrhagic colitis [8], 28 of which (27%) were associated with E. coli O157:H7. Escherichia coli O157:H7 was also found in 36% of sporadic cases (30 of 83) of hemorrhagic colitis in a British surveillance study [9]. Thus, the frequency of E. coli O157:H7 in infectious diarrhea rivals that of other major bacterial organisms, and E. coli O157:H7 is an important cause of bloody diarrhea and hemorrhagic colitis.

The estimated incidences cited in the above studies, however, are problematic and probably underestimate the true incidence of E. coli O157:H7 infection. Certainly, the best way to examine the incidence of an organism is to do prospective, laboratory-based studies within defined populations, and such studies have been the major sources of data on the reported incidences of E. coli O157:H7. However, case reporting through a surveillance system is affected by many factors: the variety and severity of clinical manifestations; the number of infected persons seeking medical attention; whether a stool culture is ordered and its timing in relation to the onset of illness and possible use of antibiotics; whether the laboratory tests correctly identify the organism; and whether the results are reported to public health officials. Clinical laboratories are becoming increasingly familiar with the varied spectrum of illness produced by E. coli O157:H7, and the ability to screen for this organism is becoming more widespread.

The minimum estimated attack rates of E. coli O157:H7 among persons who consumed the suspected food product were 3.5% in a community outbreak [14] and 8% in a junior high school outbreak [10]. These estimated attack rates include only cases in which patients had bloody diarrhea or a positive stool culture; thus, "possible" cases or those with milder symptoms were excluded. In a nursing home outbreak, the estimated attack rates from both food-borne and person-to-person transmission were 33% among the nursing home residents and 13% among the staff [11]. The attack rate was reported to be as high as 67% (42 of 63 persons) in a kindergarten outbreak involving unpasteurized milk [17].

Most reported cases of E. coli O157:H7 infection have occurred in the United States, Canada, and Great Britain, but cases have also been documented in Japan [38], Australia [90], South Africa [91], Europe [92, 93], Argentina [72], and Chile [94]. Escherichia coli O157:H7 has been detected in most areas of the United States; the largest numbers of isolations have been found in Washington State, Oregon, Minnesota, and Massachusetts [95]. The geographical position of these states and of the two countries other than the United States (Canada and Great Britain) in which most reported cases have occurred suggests a predominance of infections in northern latitudes [95]. However, the high number of reported cases in these regions may also reflect increased awareness among physicians in those areas and the fact that case reporting is required in some states.

Escherichia coli O157:H7 infections occur in all age groups, and the young are most often affected. In one study [4], the age-specific annual incidence rate was highest for children younger than 5 years of age (6.1 cases per 100 000 persons compared with an overall incidence rate of 2.1 cases per 100 000 persons). The lowest rate was for adults 50 to 59 years of age, who had an annual incidence rate of 0.5 per 100 000 persons [4]. The trends in age-specific incidence of the hemolytic-uremic syndrome in the pediatric population parallel those in the incidence of E. coli O157:H7 infection. A 10-year retrospective, population-based study of the hemolytic-uremic syndrome in Minnesota reported a substantial increase in the incidence of the hemolytic-uremic syndrome during the study period, and a disproportionate number of cases occurred in children younger than 5 years of age [52]. Although E. coli O157:H7 infections occur most often in young children and elderly persons, the elderly—especially those in institutional settings—have the highest morbidity and mortality rates [11, 19, 37]. Escherichia coli O157:H7 generally affects both sexes equally, and no data are available on the ethnicity-specific incidence rate of infection; most outbreaks seem to have affected patients with an ethnic distribution similar to that of the general population.

The rate of E. coli O157:H7 infection follows a seasonal pattern, with a peak incidence from June through September [4, 6, 70, 95]. Sixty percent of E. coli O157:H7 infections and 73% of cases of the hemolytic-uremic syndrome and thrombotic thrombocytopenic purpura presented with bloody diarrhea between June and September, and patients affected during the summer months were younger than those seen during the rest of the year [4]. In contrast to the pattern seen with Salmonella infection, the number of cases of E. coli infection does not increase after the December holiday period [70].

Food-borne transmission of E. coli O157:H7 is the most important means of infection. Transmission has primarily been linked to undercooked meat, and, during the 1982 outbreaks, the organism was cultured from a suspected lot of hamburger patties [1]. Sources other than undercooked hamburger meat [15-19] that have been implicated in transmission of E. coli O157:H7 include heat-processed meat patties, which should be pathogen-free [10]; roast beef [14, 96]; ham, turkey, and cheese sandwiches [11]; and potatoes [9]. Unpasteurized milk has been implicated as the vehicle for two cases of the hemolytic-uremic syndrome [97] and for a kindergarten outbreak of E. coli O157:H7 infection [17]; E. coli O157:H7 was isolated from the feces of healthy cows who had supplied raw milk consumed by the patients affected in the outbreak. Even fresh-pressed, unpreserved apple cider, a seemingly unlikely vehicle, was implicated in one outbreak [88]. The transmission probably occurred through the pressing of apples contaminated on the ground or during the production process. The isolation of E. coli O157:H7 from milk and from the feces of healthy cattle [17, 97, 98] and the fact that hamburger is a major vehicle associated with food-borne outbreaks of E. coli O157:H7 infection [1, 15, 19] suggest that cattle are an important reservoir for the pathogen. Escherichia coli O157:H7 has been isolated more often from dairy than from beef cattle [97], but both beef and dairy cattle are thought to be principal domestic reservoirs for the organism. In one study [99], a particularly high rate of isolation of the organism from beef (31%) was found to correlate with the increasing incidence of human infection in the region studied. Because cattle are an important reservoir for E. coli O157:H7, the apparent increase in E. coli O157:H7 infections during the past several years suggests that an epizootic infection may be occurring in the animal reservoir [52]. Escherichia coli O157:H7 has also been isolated from 1.5% to 3.7% of retail samples of beef, pork, poultry, and lamb from grocery stores in Canada and the United States [99]. Because of the wide array of contaminated food products, the precise sources of organisms are often difficult to trace and thus remain unknown in most cases.

Non-food-borne vehicles have also been implicated in the spread of E. coli O157:H7. Water-borne transmission has been implicated in two outbreaks [12, 24], and transmission by person-to-person contact or by fomites has been suggested in sporadic cases [5, 100] and outbreaks [10, 11, 13, 19, 20]. Secondary person-to-person contact can be an important method of spread in institutional settings, especially day care centers [20, 45] and nursing homes [11, 19], but it is less common in community-wide outbreaks. Nosocomial E. coli O157:H7 infections have also been reported [101, 102].

Pathology

Escherichia coli O157:H7 infection produces its most severe abnormalities in the ascending and transverse colon [103, 104]; this is consistent with endoscopic and radiologic findings showing right-sided predominance [1, 8, 37]. Colonic tissues show a spectrum of appearances ranging from normal to gross dilation with hyperemia of the involved segments [19]. In one study [104], all specimens showed patchy, shallow mucosal ulcers with partial loss of normal mucosal folds, and many ulcers were covered by a thin layer of yellow or green exudate. Extreme submucosal edema, hemorrhage, and thickening of the bowel wall were present and, in one case, were so severe that the lumen of the ascending colon was almost obliterated.

Microscopically, no single histologic feature is diagnostic of E. coli O157:H7 infection, but the colonic pathology in colitis associated with E. coli O157:H7 often resembles a combination of ischemic colitis and infectious injury similar to that seen in toxin-mediated Clostridium difficile-associated colitis [103]. Submucosal hemorrhage, edema, and fibrin exudation are the most prominent features; ulceration, hemorrhage, capillary thrombi, and mild neutrophil infiltration in the mucosa are less common [103, 104]. Immunocytochemical examination showed that the submucosal plasma cells were primarily IgG, IgA, and IgM cells [104]. In one study of 11 patients [103], all 20 colonic specimens showed variable hemorrhage and edema in the lamina propria. Nine patients had colonic pathology similar to the pattern of injury associated with acute ischemic colitis [103]: focal coagulative necrosis, hemorrhage, and acute inflammation in the superficial mucosa and preservation of the deep colonic crypts. Five patients showed both neutrophilic infiltration of the lamina propria and crypts and formation of crypt abscesses, resembling the pattern of injury seen in infectious colitis [103]. Pseudomembranous lesions similar to those in C. difficile-associated pseudomembranous colitis may also be present [40, 103-105]. The ischemic and infectious patterns of injury can be seen alone or in combination [103]; occasionally, normal specimens have also been described [5, 19, 40, 105, 106]. In one case of nonbloody diarrhea, the ascending colon showed only patchy eosinophilic infiltrates [19]. No single histologic feature is diagnostic of colitis associated with E. coli O157:H7, but the combination of infectious and ischemic patterns of injury, especially in association with capillary microthrombi and a compatible clinical picture, should suggest the diagnosis [103]. Obtaining more than one biopsy specimen from any patient increases the likelihood of identifying an abnormality, because abnormalities are often patchy [103].

Light microscopy showed no evidence of bacterial adherence or invasion in either diseased areas or normal mucosa [19, 104]. To date, immunocytochemical [104] or immunofluorescent [19] studies for E. coli O157 and H7 antigens have also failed to detect the organism in tissues. In a recent pilot study [107], immunohistochemical staining with peroxidase-labeled antibody to E. coli O157:H7 successfully detected the organism in biopsy or surgical specimens from four patients known to have colitis associated with E. coli O157:H7 and from six patients with ischemic colitis [107].

Microbiology

With one exception [16], all E. coli O157:H7 isolates have been reported to produce Shiga-like toxins [5, 10, 11, 20, 27, 52, 108]. These distinct toxins were first discovered in 1977 in certain E. coli strains associated with diarrheal disease [109], and they were termed "Shiga-like" because of their structural similarities to Shiga toxin. Shiga-like toxins have also been called Verotoxins because of their cytotoxic effect on Vero cells in tissue culture, but the toxin activity (and presumably the globotriaosyl ceramide receptor) extends to many other cell lines, including HeLa cells, intestinal villus cells, endothelial cells, and Burkitt lymphoma cells [110-114]. Shiga-like toxins were isolated in England from stools of children with bloody diarrhea [115], but they were not linked to E. coli O157:H7 infection until 1982, when Shiga-like toxin was first isolated from an outbreak of hemorrhagic colitis in a Canadian institution for elderly patients [27].

Escherichia coli O157:H7 strains have been shown to produce at least two distinct Shiga-like toxins, Shiga-like toxin I and Shiga-like toxin II, that have different immunologic and physicochemical properties [27-32, 34]. Shiga-like toxin I has many of the same biological properties as Shiga toxin, from which it is almost indistinguishable except at the nucleotide and protein level [29, 30, 116]. Shiga-like toxin I has the same isoelectric point and relative heat stabilities as Shiga toxin, and it can be neutralized by antiserum to purified Shiga toxin [28-30117, 118]. Shiga-like toxins have the same subunit structure as Shiga toxin; this consists of one active "A" subunit and five "B" binding subunits [30, 119]. Nucleotide sequencing and deduced amino acid composition show that Shiga-like toxin I and Shiga toxin share a greater than 99% gene homology and that their structures differ in only one amino acid on subunit A [116, 120, 121]. In contrast, Shiga-like toxin II [31-34] is genetically related to but antigenically distinct from Shiga-like toxin I. Shiga-like toxin II shares less than 60% of its DNA homology with Shiga toxin or Shiga-like toxin I [122, 123], and it lacks cross-neutralization with anti-Shiga-like toxin I or anti-Shiga toxin antibodies [31, 32, 34]. Shiga-like toxin II has more sequence and antigenic variability than Shiga-like toxin I, and a growing number of closely related Shiga-like toxins have been identified that belong to the Shiga-like toxin II family [124].

Although different in their molecular sequences and immunologic properties, Shiga-like toxins I and II share the same cell receptor and the same intracellular mechanism of action in vitro. Both bind to the same surface membrane receptor, globotriaosyl ceramide, which is the major Shiga-like toxin-binding glycolipid in Vero cells [110, 111]. Globotriaosyl ceramide is highly expressed in the cortex of human kidney [112] and is found in primary human endothelial cell cultures [111]. Although it has not yet been identified, it is thought to be functional in human enterocytes [111]. On binding to globotriaosyl ceramide receptors, Shiga-like toxins I and II inhibit protein synthesis by N-glycosidase cleavage at a specific site of an adenine residue on the 28s ribosomal subunit [125-128]. In addition to sharing the same intracellular mechanism of action, Shiga-like toxins I and II are reported to have similar biological activities [34]. Both have been shown to be cytotoxic to Vero and HeLa cells, enterotoxic to rabbit ileal loops, and paralytic-lethal for laboratory mice [30, 34, 129]. It is unclear which toxin is more virulent, but given an equal amount of protein in cell lysates, Shiga-like toxin II is more lethal for mice and less cytotoxic than Shiga-like toxin I [34].

Shiga-like toxins I and II are both bacteriophage encoded, and toxin production appears to be a consequence of lysogenization with one or more toxin-converting phages [130-133]. Some strains of E. coli O157:H7 produce only one Shiga-like toxin, and some produce both. Analysis of E. coli O157:H7 isolates obtained during outbreaks and from patients with the hemolytic-uremic syndrome has shown that most isolates produce either Shiga-like toxin II alone or Shiga-like toxins I and II (the numbers range from 93% of isolates producing only Shiga-like toxin II [16] to 100% of isolates producing Shiga-like toxins I and II in another outbreak [10]). In one study [108], Shiga-like toxin I alone was found in only 1 of 26 strains of E. coli O157:H7, and only 1 of 14 isolates in a community outbreak [16] had neither Shiga-like toxin I nor Shiga-like toxin II [16]. However, the absence of Shiga-like toxin production may be due to the instability of Shiga-like toxin genes, which may result in the loss of toxin genes during repeated laboratory culturing [134].

Pathogenesis

Several animal models have been developed for use in the study of the pathogenesis of E. coli O157:H7 infection [135-141]. Two studies [137, 139] have consistently induced nonbloody diarrhea in infant rabbits (5 to 10 days old in one study and 3 to 11 days old in the other) that were nasogastrically inoculated with E. coli O157:H7. The organism failed to produce diarrhea in older rabbits, 2-week-old guinea pigs, 3-week-old mice, and young rhesus monkeys [137]. In another animal study [138], orally administered E. coli O157:H7 produced watery diarrhea in gnotobiotic pigs. Epithelial distortion and detachment, effacement or fusion of the intestinal microvilli, and projections or invaginations of the plasma membrane occurred at bacterial attachment sites [138]. However, in none of these animal models were investigators able to reproduce grossly bloody diarrhea with a predominance of colonic disease as seen in humans. Such a pattern was seen in only one study [139], in which partially purified Shiga-like toxin from culture filtrates of E. coli O157:H7 was intragastrically administered to infant rabbits; one rabbit developed bloody diarrhea and extensive colonic pathology. In that study [139], rabbits given Shiga-like toxins alone and those inoculated with a high Shiga-like toxin-producing strain of E. coli O157:H7 produced the same histopathology, providing evidence to show that a toxin plays a role in pathogenesis. These histologic lesions included a predominance in the mid- and distal colon, characterized by apoptosis (defined as "individual cell death") in the surface epithelium, increased mitotic activity in the crypts, mucin depletion, and a mild to moderate infiltration of neutrophils in the lamina propria and epithelium [139]. Other intestinal morphologic changes include the premature expulsion of mature columnar absorptive epithelia of the rabbit intestinal villus in vivo as a result of the direct and selective action of Shiga-like toxin [114]. The goblet mucus cells remain attached and the crypt epithelia proliferate to maintain epithelial integrity.

On the basis of animal models and evidence from clinical, microbiological, and epidemiologic studies, toxinemia has been implicated as the primary pathogenic event in the production of the spectrum of illnesses associated with E. coli O157:H7 infection [40, 142]. Karmali and colleagues [142] proposed that Shiga-like toxin is the direct etiologic agent in the pathogenesis of both hemorrhagic colitis and the hemolytic-uremic syndrome. Observations that persons with diarrhea, hemorrhagic colitis, or the hemolytic-uremic syndrome produced high levels of Shiga-like toxins but that controls did not suggest that toxin production is important in pathogenesis [143]. Evidence that Shiga-like toxins I and II are enterotoxic to rabbit ileal loops [34], that Shiga-like toxin I acts directly on the epithelium of the intestinal villus in rabbits [114], and that Shiga-like toxin may produce a specific cytotoxic effect on the colons of mice, leading to colonic hemorrhage [31], supports the hypothesis that Shiga-like toxin is causally involved in the pathogenesis of E. coli O157:H7 infection.

Escherichia coli O157:H7 has been closely linked with both hemorrhagic colitis and the hemolytic-uremic syndrome. The clinical and radiologic features of hemorrhagic colitis resemble the gastrointestinal prodrome of the hemolytic-uremic syndrome and suggest that these conditions have a common vasculitic process [142]. The histopathologic lesions of platelet-fibrin thrombi in the microvasculature of various organs in the hemolytic-uremic syndrome are also consistent with systemic toxinemia. In vitro studies have shown that E. coli Shiga-like toxin I shows a direct cytotoxic response in vascular endothelial cells [144], and, when injected into rabbits, Shiga-like toxin I produces thrombotic microangiopathic lesions similar to those seen in humans with the hemolytic-uremic syndrome [105, 145]. It has been proposed that microvascular thrombi form through a direct cytotoxic effect on vascular endothelium or a direct effect on platelet aggregation after infection with E. coli O157:H7 [40]. It has been shown that E. coli O157:H7 Shiga-like toxins decrease prostacyclin synthesis [146] and that Shiga-like toxins incubated with platelet-poor plasma can lead to platelet aggregation [147]. It is therefore reasonable to propose that, after infection with the organisms and release of toxins, damage to vascular endothelium is accompanied by a decreased synthesis of prostacyclin, an increased agglutination of platelets, and the exposure of subendothelium beneath the disrupted surface endothelium. A cascade of coagulative events is thus triggered, leading to intravascular thrombi formation. Ischemic changes precipitated by platelet-fibrin thrombi in the colonic microvasculature result in hemorrhagic colitis. Patients who develop the hemolytic-uremic syndrome or thrombotic thrombocytopenic purpura represent a clinical spectrum arising from the same underlying disease process, and they differ mainly in the distribution of their thrombotic lesions [40]. Platelet-fibrin thrombi are predominantly located in the kidneys in patients with the hemolytic-uremic syndrome, but they appear to be more disseminated in persons with thrombotic thrombocytopenic purpura, with involvement in the pancreas, adrenal glands, heart, brain, and kidneys [82]. The hypothesis of vascular ischemia secondary to thrombi formation in the pathogenesis of E. coli O157:H7 infection is supported by the report of a patient with hemorrhagic colitis and thrombotic thrombocytopenic purpura [40]. Barium enema studies showed marked thickening of the mucosa and thumbprinting along the transverse and descending colon suggestive of ischemic colitis with submucosal edema or hemorrhage. Colonoscopy showed acute severe colitis, and biopsy showed focal ulceration of the mucosa and capillary-platelet thrombi in the submucosa [40]. Furthermore, injecting Shiga-like toxin I into gnotobiotic pigs produces vascular damage and ischemic necrosis in the intestines and brain, resembling the lesions seen in humans with hemorrhagic colitis and thrombotic thrombocytopenic purpura [148]. In short, current thinking on the pathogenesis of E. coli O157:H7 infection is that the organism releases its toxins in the bowel and that they are absorbed into the circulation, producing vascular endothelial damage with subsequent local intravascular coagulation and fibrin deposition and ultimately resulting in various clinical features of E. coli O157:H7 infection.

Patients with hemorrhagic colitis and the hemolytic-uremic syndrome have shown an increase in Shiga-like toxin-neutralizing antibody titers [62, 68, 72]. This serologic finding further supports the idea that Shiga-like toxin is important in E. coli O157:H7 infection and suggests that antibodies to Shiga-like toxin may play a protective or pathogenic role. It has even been suggested that the more severe clinical courses seen among patients at extremes of age is caused by an absence of specific neutralizing antibodies [142].

Escherichia coli O157:H7 may possess a complex of virulence determinants other than Shiga-like toxins. An animal study of gnotobiotic pigs inoculated with two strains of E. coli O157:H7—one that produced high levels of Shiga-like toxin and one that produced only moderate levels of a different Shiga-like toxin—showed similar clinical manifestations and histopathology [141]. This observation undermines the idea that Shiga-like toxins play a pathogenic role and suggests that at least one other virulent factor exists. Although E. coli O157:H7 does not invade epithelial cells, investigators have postulated that it colonizes the bowel through fimbrial adherence to the cells [149]. Animal models have shown that it adheres to the luminal surface of the colon, cecum, gut-associated lymphoid tissues, and, to a lesser extent, the small intestinal epithelium of infected rabbits [139]. A similar study with gnotobiotic pigs showed diffuse bacterial attachment to the epithelial surface of the cecum and colon and focal adherence to the ileum and rectum [138]. Microscopically, these organisms produce lesions with a characteristic attaching and effacing pattern [150-153]. It was subsequently found that, in tissue culture, fimbriated E. coli O157:H7 isolates adhered to Henle 407 cells, a human-derived intestinal cell line, and that a 60-Md plasmid present in most isolates [154] encoded expression of the fimbrial antigen of E. coli O157:H7 [149]. Observations that plasmid-cured and thus nonfimbriated strains of E. coli O157:H7 lost their ability to adhere to intestinal cells [149] suggest that the biological activity of cell adherence is mediated by the E. coli O157:H7 fimbria.

On the basis of several epidemiologic patterns, it was suggested that only a small inoculum was required to produce illness [71]. First, unlike salmonella infection, which usually results from gross mishandling during food preparation, most cases of E. coli O157:H7 infection were traced to meat contaminated by only slight undercooking and not left at a warm temperature for a period of time to allow for bacterial overgrowth. Second, an appreciable rate of E. coli O157:H7 infection has been traced to vehicles such as raw milk and municipal water, which should be associated with only a small number of organisms because of their cold temperature and dilutional nature. Last, the appreciable rate of secondary person-to-person transmission is similar to that of shigellosis, which can be transmitted by a small inoculum. This low infectious dose makes it even more important to implement public health measures, including strict regulation of food processing (see below).

Diagnosis

Most patients with E. coli O157:H7 infection that occurs in epidemics are suspected of having infectious diarrhea. More laboratories are now screening for E. coli O157:H7, but infection with this organism is often unrecognized because most clinical laboratories still do not routinely test stool samples for this organism. Other differential diagnoses that have often been considered include inflammatory bowel disease, ischemic colitis, antibiotic-associated pseudomembranous colitis, intussusception, or various causes of an acute abdomen [16, 19, 37].

The strongest evidence for E. coli O157:H7 infection is the presence of organisms in stool culture, but diagnosis can also be supported by the presence of Shiga-like toxin, an increase in serum Shiga-like toxin antibody titers, or a host of new genotypic and phenotypic assays (Table 3). Stool culture for this organism requires a special growth medium, because E. coli O157:H7 ferments lactose rapidly and thus cannot be picked out from normal fecal flora when grown on a lactose-containing medium for routine stool cultures. However, serotype O157:H7 can be distinguished from most other strains of E. coli by its slow fermentation of sorbitol. When plated on MacConkey agar (indicator medium) and sorbitol agar (selective medium), E. coli O157:H7 appears sorbitol negative at 24 hours [155, 156]. This MacConkey-sorbitol agar medium is 100% sensitive, 85% specific, and 86% accurate for detecting E. coli O157:H7 [156]. Sorbitol-negative colonies can be picked and further tested by characterizing responses to other biochemical parameters [154, 157-160], serotyping using antisera to H7 and O157 antigens, or determining the presence of Shiga-like toxins. One limitation to this approach is that the rate of isolation decreases with delay in collection of stool samples; cultures obtained more than 6 days after the onset of illness or after the administration of antibiotics often produce negative results [20, 37, 75, 154]. Escherichia coli O157:H7 was isolated from 75% to 100% of the stool samples obtained within 7 days of the onset of illness, but the recovery rates from samples collected after day 7 ranged from 0% to 33% [20, 37, 75, 154]. In one study [75], the rate of positive stool culture decreased from 100% for samples collected within 2 days of the onset of diarrhea to 92% for samples collected on days 3 through 6 and to 33% for samples collected after day 7. The duration of carriage seems to be longer in children than in adults [5]. Finally, there are sorbitol-fermenting E. coli O157 strains that have been reported to cause human disease [161], but their prevalence and significance are still unclear.

Screening specimens on sorbitol-containing MacConkey culture medium and then testing the non-sorbitol-fermenting colonies for E. coli O157:H7 by using biochemical parameters and by serotyping with O157 and H7 antisera [26, 37, 156, 162] can be laborious and time-consuming. Antisera to both H7 and O157 are now commercially available, so that after screening with sorbitol-MacConkey medium, the sorbitol-negative colonies can be rapidly confirmed with O serum and H serum in the slide agglutination test [163, 164]. Investigators have shown that the commercially available latex slide agglutination tests for O157 serum are an efficient and reliable alternative to conventional serotyping with the standard-tube agglutination test, making rapid presumptive detection of E. coli O157:H7 possible [163, 164]. However, colonies that agglutinate should be confirmed serologically, using agglutination or direct immunofluoresecent antibody tests [163-165]. An alternative screening method was reported by Farmer and Davis [155], who devised an H7 antiserum-sorbitol fermentation medium as a single-tube screening medium; strains that were presumptive positives (negative for sorbitol fermentation and positive for H7 reaction) were then tested by slide or tube agglutination with E. coli O157 serum [155].

Another sensitive method of diagnosing E. coli O157:H7 infection is to look for Shiga-like toxins. These toxins have been detected in E. coli culture broth filtrate and in stool extracts [27, 29, 68]. Demonstration of free fecal Shiga-like toxins can be made by tissue culture assays with neutralization by appropriate antisera [162, 166-168]. The disadvantage of this approach is that classic tissue culture assays using HeLa or Vero cell culture cytotoxicity [109] require appropriate facilities and are slow and cumbersome. On the other hand, testing for Shiga-like toxin allows the detection not only of E. coli O157:H7 but of Shiga-like toxin-producing serotypes other than O157:H7, which may be increasing in importance as causes of human illness. Moreover, Shiga-like toxins have been found in fecal filtrates long after E. coli cannot be cultured from stools [62, 162]: More than 4 to 9 days after an E. coli O157:H7 infection, the excretion of organisms into stools usually decreases to an undetectable amount, but free fecal Shiga-like toxin may remain measurable for as long as 4 to 6 weeks. Free fecal Shiga-like toxin assay has been reported to be more sensitive than stool culture for the organism [11, 62]. In a nursing home outbreak, the rate of isolation from stool samples was 34% and the detection rate for free fecal toxin was 50% [11]. Although the organism has never been isolated without fecal Shiga-like toxin, the latter was often present even when stool culture was negative [62].

Other methods for detecting toxins include genetic probes and immunospecific assays, which are simpler and more sensitive than culture techniques, although some may be less practical for use in clinical laboratories. Deoxyribonucleic acid hybridization assays using synthetic nucleotides or fragments of structural genes specific for the toxins can also be used to detect Shiga-like toxin-producing E. coli [67, 89, 161, 169-173]. Gene probes are sensitive and specific [169, 171]. Using colony blot hybridization, only 2 of 102 strains were toxin-probe positive when toxin was not present [170], suggesting that the use of DNA probes to detect Shiga-like toxin production is as accurate as the use of toxin-specific antibodies. These specific DNA probes were able to detect colonies of Shiga-like toxin-producing E. coli present in numbers as small as 1 in 1200 colonies [67]. In one study [169] that used synthetic oligonucleotides from selected sequences of genes encoding A-subunit of Shiga-like toxin I and B-subunit of Shiga-like toxin II at different degrees of stringency, the A-I probe had 92% sensitivity and 91% specificity for identifying Shiga-like toxin I-producing E. coli, and the B-II probe had 100% sensitivity and 97% specificity for identifying Shiga-like toxin II-producing E. coli [169]. Both probes were able to identify strains that produce variants of Shiga-like toxins. Gene probes are diagnostically useful, but the cost and concern associated with radioactive safety have limited their widespread applicability. Various enzyme-linked immunosorbent assays using polyclonal and monoclonal antibodies against Shiga-like toxins I and II to detect the presence of toxins in culture or fecal extract have also been developed [174-177]. On the basis of its specific binding to the globotriaosyl ceramide natural receptor, a modified enzyme-linked immunosorbent assay for the rapid detection of Shiga-like toxin I has been reported [178], in which toxin bound to the globotriaosyl ceramide receptor was detected by enzyme-linked immunosorbent assay with monoclonal antibodies against Shiga-like toxin I. Both techniques are highly sensitive and specific in detecting toxin production, and they promise to shorten the time to diagnosis of E. coli O157:H7 infection. Another genetic technique involves polymerase chain reaction (PCR) amplification to test for the presence of Shiga-like toxin genes [161, 179-182]. Because PCR should detect organisms in low numbers, it can detect Shiga-like toxin production when culture fails [180]. In addition, like other methods for detecting the presence of Shiga-like toxin, PCR can identify Shiga-like toxin-producing E. coli other than O157:H7. Techniques for the direct detection of Shiga-like toxin sequences in stool specimens have also been reported [180, 181], overcoming the difficulty of high-frequency loss of toxin genes with repeated cultures [134].

Additional phenotypic and genotypic assays have been developed to assist in epidemiologic studies, allowing investigators to determine the extent of outbreaks, trace human outbreaks to animal sources, and differentiate and analyze linkage between strains of E. coli O157:H7. These schemes include Shiga-like toxin genotyping [170, 183-185], plasmid DNA profiling [16, 27, 154, 185, 186], bacteriophage typing [11, 170, 183, 185-187], restriction digests of plasmid [154], restriction fragment length polymorphism with a bacteriophage probe [185, 188], electrophoresis of plasmids and multilocus enzyme electrophoretic typing [189], pulsed-field gel electrophoresis of restriction fragment length polymorphism [190], and patterns of antibiotic susceptibilities [24].

Another useful diagnostic tool is serologic testing to detect antibodies to Shiga-like toxin or O157 lipopolysaccharides. Increases in serum Shiga-like toxin-neutralizing antibody titers during E. coli O157:H7 infections have been used to detect or support the diagnosis of infections [62, 68, 72]. The antibody titers ranged from 4 to 80 in acute serum specimens collected between days 4 and 18 after the onset of illness; they ranged from 32 to 1280 in convalescent serum specimens collected between days 13 and 43 [62, 68]. In one case [62], the acute and convalescent serum specimens yielded titers of 4096 and 32 000, respectively. In the same study, a fourfold or greater increase in Shiga-like toxin-neutralizing antibody titer was used to diagnose infection [62]. Fifty-nine percent of patients (16 of 27) met the requirement, and this criterion was the only evidence of infection in 15% of those tested [62]. This serologic test may be an alternative way to diagnose E. coli O157:H7 infection, especially during epidemics of this infection or when other methods fail to detect E. coli O157:H7. Similarly, serologic response to O157 lipopolysaccharides of E. coli O157:H7 has also been reported [191-195] and can be a useful adjunct for diagnosing E. coli O157:H7 infection. In one study [192], this serologic test detected evidence of E. coli O157:H7 infection in 73% of children with the hemolytic-uremic syndrome and was more sensitive than either isolation of the organism or the detection of fecal Shiga-like toxin. In studies involving patients with the hemolytic-uremic syndrome, the presence of antibodies to O157 lipopolysaccharide was able to provide evidence of E. coli O157:H7 infection when fecal bacteria or Shiga-like toxin activity could no longer be detected [192, 195]. Most IgM antibodies became undetectable 2 to 3 months after the acute phase of the hemolytic-uremic syndrome [195]. However, the interpretation of the serologic study for O157 lipopolysaccharide may be affected by possible cross-reactivity with other organisms and detection of nontoxigenic or non-H7 strains of E. coli O157.

In summary, the most common algorithm for diagnosing E. coli O157:H7 infection in current clinical practice is to culture stool specimens for the organisms using sorbitol-MacConkey agar; this can be done at local hospital laboratories. The sorbitol-negative colonies can be serotyped using commercially available anti-sera to O157 while the sample is sent to a reference laboratory. Presumptive diagnosis can also be made by biochemical testing. In either case, diagnosis is confirmed by the reference laboratory, where the O157 latex test or O157 direct fluorescent antibodies and H7 antisera are used to test for O157:H7. In addition, DNA probes are used to detect Shiga-like toxin in stools at some reference laboratories. If the initial culture is negative but clinical suspicion is still high, stool samples can be sent to a reference laboratory, where more sophisticated techniques, such as PCR for toxin genes, can be used. In practice, serologic determination of Shiga-like toxin titers is used primarily as a diagnostic aid and is not done routinely. In areas where infection with Shiga-like toxin-producing E. coli is common, Shiga-like toxin titers on one serum specimen may be difficult to interpret.

Treatment

No specific treatment currently exists for E. coli O157:H7 infection other than supportive therapy and management of complications such as anemia and renal failure. Antimicrobial agents have not been shown to modify the illness, but few conclusive data are available on individual agents. Some studies [1, 4, 8] have shown that the duration of illness in persons treated with antibiotics did not differ significantly from that in untreated persons. Although the mean duration of diarrhea was similar in patients who did and did not receive antimicrobial therapy [4, 37], one study reported a significantly longer duration of bloody diarrhea in persons treated with antibiotics than in untreated persons [4]. It has even been suggested that the use of antibiotics is a risk factor for infection and that an association exists between the use of antibiotics and increased mortality [11]. It has been postulated that antibiotics can worsen the clinical course of E. coli O157:H7 infection through two mechanisms [196]: 1) the elimination of competing bowel flora by antibiotics, leading to an overgrowth of E. coli O157:H7, and 2) lysis of or sublethal damage to the infecting organisms, with the subsequent liberation of Shiga-like toxins.

Most E. coli O157:H7 isolates are sensitive to most antimicrobial agents in vitro. Isolates of E. coli O157:H7 have been found to be uniformly susceptible to ampicillin, carbenicillin, cephalothin, chloramphenicol, gentamicin, kanamycin, nalidixic acid, norfloxacin, sulfisoxazole, tetracycline, ticarcillin, tobramycin, trimethoprim, and trimethoprim-sulfamethoxazole [5, 37, 196]. Isolates have been found to be resistant to erythromycin, metronidazole, and vancomycin [196], and some have been reported to be resistant to tetracycline [8, 24, 197]. A strain of E. coli O157:H7 from a water-borne outbreak in 1989 was resistant to streptomycin, sulfisoxazole, and tetracycline [24]. A study of antibiotic-resistant E. coli O157:H7 in Washington State showed an emergence of antibiotic resistance to streptomycin, sulfisoxazole, and tetracycline, from zero isolation (0 of 56) between 1984 and 1987 to 7.4% isolation (13 of 176) between 1989 and 1991 [197]. Studies of an E. coli O157:H7 outbreak have suggested that patients receiving ampicillin and patients receiving placebo did not differ in durations of diarrhea or bloody diarrhea, number of stools per day, or hospitalization rate. In contrast, another study [18] showed that patients treated with trimethoprim-sulfamethoxazole had longer durations of diarrhea and bloody diarrhea and were more likely to develop the hemolytic-uremic syndrome. It has been suggested that trimethoprim-sulfamethoxazole and polymyxin B increase in vitro toxin concentration released by E. coli [167, 198, 199]. This suggestion is based on a study [196] that showed a worse outcome with trimethoprim-sulfamethoxazole [18] and on the hypothesis that antibiotic therapy aggravates E. coli O157:H7 infection by sublethal damage or lysis of the infecting organism and the subsequent release of Shiga-like toxin into the gut lumen. However, without randomization, patients with more severe disease may be more likely to receive antibiotics, leading to bias in data interpretation [18]. Other studies have found no association of trimethoprim-sulfamethoxazole (or other "appropriate" antibiotics) with progression to the hemolytic-uremic syndrome [77, 200]. By decreasing Shiga-like toxin synthesis in vitro [176] and eliminating other enteric pathogens from the gut mucosa [201], it was postulated that ciprofloxacin may be useful for treating infection with this organism [36].

Use of antimotility agents has also been suggested as a risk factor for progression of E. coli O157:H7 infection to the hemolytic-uremic syndrome, because such use may allow more time for toxin absorption [77]. A positive association has been found between the use of antimotility agents and the severity of E. coli O157:H7 infection in one study of four geriatric patients [19], although contradictory findings, which show no difference in duration of diarrhea and overall illness with antidiarrheal use, have been reported [4].

Prevention

The obvious health implications of E. coli O157:H7 infection and its complications, including the hemolyticuremic syndrome and thrombotic thrombocytopenic purpura, warrant better educational and preventive measures. Several public health measures have been proposed [202], including improved case identification resulting from increased awareness of this infection; more widespread and frequent screening for the organism at health laboratory facilities; routine testi