Customize your JAMA Network experience by selecting one or more topics from the list below.
Objective To describe pathogens identified through routine clinical practice and factors associated with identifying Shiga toxin–producing Escherichia coli (STEC) infection in patients with postdiarrheal hemolytic uremic syndrome (D+HUS).
Design Population-based active surveillance.
Setting Hospitals in the FoodNet surveillance areas from 2000 through 2010.
Participants Children younger than 18 years with D+HUS.
Main Exposures Testing for STEC and demographic and clinical characteristics.
Main Outcome Measures Percentage of patients with evidence of infection with likely HUS-causing agents and associations between exposures and evidence of STEC infection.
Results Of 617 patients, 436 (70.7%) had evidence of infection with likely HUS-causing agents: STEC O157 (401 patients), non-O157 STEC (21 patients), O157 and non-O157 STEC (1 patient), Streptococcus pneumoniae (11 patients), and other pathogens (2 patients). Among patients without microbiological evidence of STEC, 76.9% of those tested had serologic evidence of STEC infection. Children more likely to have evidence of STEC infections included those patients tested for STEC less than 4 days after diarrhea onset, 12 months or older (71.6% vs 27.8% if <12 months of age), with infections as part of an outbreak (94.3% vs 67.3%), with bloody diarrhea (77.2% vs 40.4%), with onset during June through September (76.9% vs 60.1%), with a leukocyte count greater than 18 000/μL (to convert to ×109/L, multiply by 0.001) (75.7% vs 65.3%), or with only moderate anemia (hemoglobin >7.0 g/dL [to convert to grams per liter, multiply by 10] or hematocrit greater than 20% [to convert to a proportion of 1, multiply by 0.01]) (75.1% vs 66.3%). However, many of these associations were weaker among children with thorough STEC testing.
Conclusions Early stool collection for E coli O157 culture and Shiga toxin testing of all children with possible bacterial enteric infection will increase detection of STEC strains causing HUS. In the absence of microbiological evidence of STEC, serologic testing should be performed.
Shiga toxin–producing Escherichia coli (STEC) cause illnesses ranging from mild diarrhea to postdiarrheal hemolytic uremic syndrome (D+HUS). Children have the highest incidence of D+HUS, a potentially fatal thrombotic microangiopathy.1,2 Although STEC O157 is the most frequently reported serogroup,3 more than 50 other serogroups are estimated to cause two-thirds of STEC illnesses.4 Detection of non-O157 STEC infections is increasing as more laboratories use assays to detect Shiga toxins.5- 7
Population-based descriptions of STEC strains causing D+HUS in the United States are outdated and limited to single states.8- 14 The 1 US study of national scope found that STEC O157 accounted for more than 80% of D+HUS.15 This study, like many, used a microbiological testing protocol that was more comprehensive than that used in typical clinical practice. Although studies with rigorous microbiological protocols have detected STEC in nearly 90% of cases,16 routine clinical practice yields a lower percentage, limiting population-based descriptions of the types of STEC causing D+HUS.
We analyzed 11 years of population-based D+HUS surveillance data collected by FoodNet to describe infectious agents identified through routine testing and the factors associated with establishing antecedent STEC infection, the aim being identification of practices that increase detection of STEC-related D+HUS.
A confirmed case was defined as an illness diagnosed as D+HUS in a child younger than 18 years with (1) an HUS diagnosis within 21 days after the onset of self-reported diarrhea (or any diarrhea during the 3 weeks before HUS diagnosis if there was evidence of STEC infection), (2) anemia (hemoglobin or hematocrit below age- and sex-specific thresholds),17 (3) thrombocytopenia (platelet count <150 × 103/μL [to convert to ×109/L, multiply by 1]), (4) azotemia (serum creatinine level ≥1.0 mg/dL [to convert to micromoles per liter, multiply by 88.4] if <13 years old and ≥1.5 mg/dL if ≥13 years old), and (5) red blood cell fragmentation. A probable case met all criteria except the last.
FoodNet sites (Connecticut, Georgia, Maryland, Minnesota, New Mexico, Oregon, Tennessee, and selected counties in California, Colorado, and New York) conducted active hospital-based surveillance for pediatric D+HUS; start year for sites varied from 1997 to 2004. In 2010, the FoodNet catchment included 47.1 million persons (15% of the US population). Each site established a network of pediatric nephrologists and hospital infection control personnel. FoodNet personnel routinely contacted network members to identify cases and completed standardized case report forms by interviewing treating physicians, abstracting medical records, or both. Data collected included demographics, medical history, microbiological and serologic findings, and selected laboratory parameters. Race, as indicated by a patient's caregiver, physician, or medical records, was recorded because it is associated with D+HUS18; categories included American Indian/Alaska Native, Asian/Pacific Islander, black, white, and unknown.
To augment case finding, all FoodNet sites, except New Mexico, reviewed hospital discharge data. Records of hospitalizations assigned any of the following International Classification of Disease, Clinical Modification, Ninth Revision codes were reviewed: (1) 283.11 (HUS), (2) 584.X and 283.X and 287.X (acute renal failure, hemolytic anemia, and thrombocytopenia), and (3) 446.6 and 008.X or 009.X (thrombotic thrombocytopenic purpura with diarrhea caused by E coli or unknown pathogen). Because of delayed availability of records, at the time of analysis, 5 of 9 sites had completed review for patients hospitalized through 2010; 3 had completed reviews through 2009 and 1 through 2008. This surveillance was deemed nonresearch by the Centers for Disease Control and Prevention (CDC).
All testing was conducted at the discretion of attending physicians. Our surveillance protocol did not specify any required testing. Results collected included stool culture for E coli O157 (eg, culture on sorbitol-MacConkey or other selective and differential agar), stool assays for Shiga toxins, and positive studies for other pathogens, including non-O157 STEC, in stool and other specimens. Tests may have been performed in a clinical laboratory, a state public health laboratory, or at the CDC. Methods differed among laboratories. For example, to detect evidence of Shiga toxins, most clinical laboratories used enzyme immunoassay, but the CDC and some state laboratories used polymerase chain reaction assays.5 The CDC tested serum samples received for IgG or IgM to the lipopolysaccharide of E coli O157, O111, or both by enzyme immunoassay.19 We obtained Shiga toxin 1 (Stx1) and 2 (Stx2) profiles for STEC isolates reported from 2007 through 2010. We obtained additional toxin profiles and polymerase chain reaction results for eae (intimin) and ehxA (enterohemolysin) for non-O157 STEC isolates from 2000 through 2010 from the CDC National E coli Reference Laboratory.
Cases were categorized into 3 groups based on evidence of STEC: (1) microbiological evidence through culture (regardless of serologic evidence), (2) serologic evidence only, and (3) no evidence of STEC (either no testing performed or no positive STEC findings). Microbiological evidence was further categorized as confirmed (isolation of an E coli that produced Shiga toxin or, for E coli O157, presence of the H7 antigen) or probable (isolation of E coli O157 without evidence of the H7 antigen or Shiga toxin).
Pathogens we defined as likely HUS-causing agents were STEC, other Shiga toxin–producing Enterobacteriaceae, Streptococcus pneumoniae, human immunodeficiency virus, Epstein-Barr virus, varicella-zoster virus, and influenza A.
The number of cases ascertained from 2000 through 2010 determined the sample size. A 2-tailed significance level of .05 was used for all tests. Mean annual incidence rates of D+HUS by year of age were calculated using US Census Bureau data. The Cochran-Armitage test was used to assess trends by year. Univariate associations between evidence of STEC infection and clinical factors were assessed using the Fisher exact test; patients infected with other HUS-causing agents and patients with only evidence of Shiga toxin were excluded. Potential confounding by geography was assessed by change in summary Mantel-Haenszel odds ratios following stratification by the FoodNet site.
We performed post hoc analyses to assess the extent to which factors associated with STEC evidence varied by completeness of STEC testing. We defined thorough STEC testing in 2 ways: complete stool testing (ie, performance of E coli O157 culture and Shiga toxin testing) and complete stool testing or serologic testing. We constructed 4 × 2 tables, with each row representing 1 exposure combination of 2 dichotomous factors: thorough STEC testing and a given clinical factor. These tables contain all data needed for assessment of interaction on additive and multiplicative scales.20 We calculated the risk of having documented STEC evidence for patients in each exposure category. We calculated risk differences for each clinical factor separately for patients with and without thorough STEC testing.
We identified 617 (523 confirmed and 94 probable) patients with D+HUS. The median age of patients was 3.7 years (range, 2.9 months to 17.8 years). The mean annual incidence of D+HUS was 0.54 case per 100 000 children. Incidence trended downward with age from a peak in children aged 12 to 36 months. A total of 55.3% of the patients were girls. Of the 532 patients (86.2%) with race reported, more were white (90.4%) and fewer were black (4.3%) than would be expected from the population under surveillance (71.3% white and 19.3% black). Diarrhea began in June through September for 59.3% of patients.
Stool was cultured for STEC O157 in 569 patients (92.2%) and tested for Shiga toxin in 348 patients (56.4%). The percentage cultured for STEC O157 decreased and the percentage tested for Shiga toxin increased from 2000 through 2010 (Table 1).
Microbiological evidence of STEC infection was found in 353 patients (57.2%): 333 with STEC O157, 19 with non-O157 STEC (serogroups O111 [5 patients], O145 [5 patients], O121 [3 patients], O26 [2 patients], O103 [1 patient], O130 , unknown [2 patients]), and 1 with both STEC O157 and STEC O26 (Table 2). No significant trend was seen from 2000 through 2010 in the percentage of patients with STEC O157 isolated, but there was a gradual near-significant increase in the percentage of patients with non-O157 STEC isolated (Table 1). Shiga toxin profiles were reported for 103 STEC O157 isolates and 17 non-O157 STEC isolates. Most isolates produced Stx2 (Table 3).
Serologic testing was performed for 91 of 264 patients (34.5%) without microbiologic evidence of STEC; 70 (76.9%) had serologic evidence of infection, 68 with O157 and 2 with O111. In total, 423 patients (68.6%) had evidence of STEC infection, including 401 (94.8%) with STEC O157, 21 (5.0%) with non-O157 STEC, and 1 with both.
Evidence of other likely HUS-causing agents was found in 13 patients without evidence of STEC infection: S pneumoniae (11 patients), Shigella dysenteriae type 1 (1 patient), and pandemic H1N1 influenza A (1 patient). The 11 patients with S pneumoniae infection tended to be younger (median age, 1.8 years) than other patients, and all 11 had diarrhea onset during October through March. Blacks were more likely than whites to have S pneumoniae isolated (13.0% vs 1.1%) and more likely to become ill during October through March (52.2% vs 25.8%).
Other pathogens were found infrequently in stool (Table 2). Streptococcus pyogenes was isolated from stool in 2 of 5 patients (40.0%) with STEC O145 infections and 3 of 401 patients (0.7%) with STEC O157 infections. Children with mixed STEC and S pyogenes infections were generally older than other children with STEC infections (median age, 7.2 vs 3.6 years).
Children were more likely to have evidence of STEC infection if they were 12 months or older, were part of a recognized outbreak, or had bloody diarrhea, diarrhea onset during June through September, a white blood cell count greater than 18 000/μL(to convert to ×109/L, multiply by 0.001), only moderate anemia, or stool cultured for E coli O157 or tested for Shiga toxin less than 4 days after diarrhea onset (Table 4). There was no apparent confounding by geography.
We found evidence that the effect of 5 clinical characteristics (being part of a recognized outbreak, bloody diarrhea, diarrhea onset during June through September, white blood cell count >18 000/μL, and only moderate anemia) on the risk of having any evidence of STEC varied by completeness of STEC testing; the increase in risk was larger among patients without thorough STEC testing (Table 5). We observed similar effects of these 5 characteristics on the risk of having microbiological evidence of STEC infection; the increase in risk was larger among those without complete stool testing (data not shown).
In this first population-based description of pediatric D+HUS of national scope in the United States, most (94.8%) of the two-thirds of children with evidence of STEC infection had STEC O157 infection. The marked increase in the use of Shiga toxin testing from 2000 through 2010 was not accompanied by as significant of an increase in the percentge of D+HUS cases attributable to non-O157 STEC. Our findings reaffirm previously established associations between several clinical factors and antecedent STEC infection. In addition, we found that most of these associations were weaker among children with thorough STEC testing. Thus, our findings support the value of STEC testing of children with possible bacterial enteric infection even in the absence of factors, such as summertime illness and bloody diarrhea, that have traditionally been used to support decisions to conduct STEC testing. Furthermore, we found that 2.1% of D+HUS cases are unlikely caused by STEC infection.
Our finding that 68.6% of children with D+HUS had evidence of STEC infection is comparable to the 60% to 88% found in other studies.15,16,21- 26 However, only 34.5% of patients without microbiological evidence of STEC had serologic testing. Of those tested, 76.9% were seropositive. Applying this percentage to all children without evidence of infection by a likely HUS-causing agent would increase the estimate of cases with STEC infection to 88.5%. In contrast to routine practices in our surveillance area, studies with the highest STEC detection rates used immunomagnetic separation, a stool culture technique that increases STEC detection sensitivity.16,25- 27 Greater use of immunomagnetic separation in the United States, in addition to serologic testing, could increase the percentage of cases with serogroup-specific evidence of STEC infection further.
All but 1 of the non-O157 serogroups we observed were among the 6 most common serogroups detected in ill persons in the United States.28 However, our Shiga toxin profile findings suggest that increases in detection of common non-O157 STEC serogroups as causes of diarrheal illness, occurring as a result of increased Shiga toxin testing,7 has not led to an equivalent increase in the reported incidence of D+HUS attributable to these serogroups. Among 14 D+HUS cases in our series with evidence of infection with only 1 of the 6 most common non-O157 STEC serogroups (5 serogroups represented), 12 isolates (85.7%) produced Stx2, a factor strongly associated with D+HUS.29 In contrast, the percentage of isolates of these 5 serogroups that produced Stx2 from a convenience sample of isolates from ill persons was 36.2% (217 of 600).28 The only common serogroup not detected in our D+HUS surveillance was O45, which rarely produces Stx2.28 Emergence of new strains, such as the Stx2-producing enteroaggregative E coli O104 that caused D+HUS in Europe in 2011, can occur.30 Further study is needed to understand the role of other virulence factors and the subtypes of Stx2 in STEC strains causing D+HUS in the United States.
We, similar to others, found that children with stool cultured for STEC O157 within 4 days of diarrhea onset were more likely to have STEC detected.31 Identifying an STEC infection early allows for measures to minimize transmission to others32 and to improve outcomes.33,34 Although serologic testing is less constrained by timing, microbiological evidence should always be sought because it yields an isolate that can be subtyped for outbreak detection and characterized by virulence factors.
Like others, we found that bloody diarrhea, diarrhea in summer months, and being part of an outbreak were associated with evidence of STEC infection.15,24,35 We showed that these associations are weaker in the presence of thorough STEC testing. For example, whereas overall 40.4% of patients with nonbloody diarrhea (vs 77.2% with bloody diarrhea) had evidence of STEC infection (Table 4), the percentage increased to 71.7% (vs 86.4% with bloody diarrhea) among patients with thorough testing (Table 5). Our findings argue against the practice of some laboratories to limit STEC testing to patients with bloody stools or to summer months.6
Routine laboratory parameters should not overly influence STEC testing decisions. Although, as noted by others,25 children with D+HUS and marked leukocytosis were more likely to have evidence of STEC, 46.8% of children with evidence of STEC infection had a white blood cell count less than 18 000/μL. Our finding that children whose hemoglobin level remained greater than 7.0 g/dL (or whose hematocrit was >20%) were more likely to have evidence of STEC might relate to earlier presentation for care, before the hemoglobin level could decrease lower. Alternatively, a higher hemoglobin level could be a marker of dehydration, which, in turn, could be associated with other signs and symptoms that prompt testing for STEC.36
We found that increasing use of Shiga toxin testing among patients with D+HUS from 2000 through 2010 coincided with a decrease in stool culturing for E coli O157. This finding is concerning because the fastest method to identify E coli O157 infections remains culture.5 To maximize the number of STEC infections detected, all stool specimens submitted for bacterial testing to clinical laboratories from patients with community-acquired diarrhea or suspected HUS, even in the absence of diarrhea, should be promptly cultured for E coli O157 and assayed for Shiga toxins.5,37 Because 1.4% of clinical laboratories surveyed in 2007 followed these practices,6 it is important that physicians know what tests their laboratories perform and request guideline-based testing. Isolates and Shiga toxin–positive samples should be sent to public health laboratories for characterization and isolation. To diagnose the origin of D+HUS cases without microbiological evidence of STEC in the clinical laboratory, stool samples should be sent to public health laboratories able to perform immunomagnetic separation–assisted culture and polymerase chain reaction testing for Shiga toxin, and serum samples should be sent through a public health laboratory to the CDC. The CDC has validated tests to detect anti-O157 and anti-O111 serum antibodies and, in some cases, can use unvalidated tests to detect antibodies to additional serogroups.
Analyses of data collected in nonresearch contexts are more prone to introduction of nonstatistical sources of uncertainty through measurement error, variability in testing procedures, missing data, and other factors. We did not adjust our findings to address statistical uncertainty arising from multiple comparisons because doing so may imply unjustified confidence in reported precision. Nevertheless, the factors we identified agree with other studies.
We show that a portion of D+HUS is likely to be caused by S pneumoniae, a known cause of nondiarrheal HUS. The suspected mechanism is bacterial neuraminidase-mediated exposure of antigens on erythrocytes, platelets, and glomerular endothelium, followed by binding of host antibodies.38 Eleven patients (5.7%) without evidence of STEC infection had S pneumoniae isolated. The illness seasonality and young age of these patients fit the epidemiology of pneumococcal HUS.39 Our findings that children with no evidence of STEC infection were more likely to have illness in nonsummer months and to be younger suggest that S pneumoniae may account for more D+HUS cases than we identified, perhaps especially in black children. Although our surveillance may have captured pneumococcal HUS cases because of coincidental presence of mild diarrhea unrelated to S pneumoniae, 4 of 9 children with S pneumoniae infection with data had bloody diarrhea; S pneumoniae infection may be an underrecognized cause of bloody diarrhea.40 One additional patient had infection with pandemic H1N1 influenza A, a potential trigger for HUS, potentially through its neuraminidase activity or its association with S pneumoniae infection.41
Some of the enteric pathogens found in children without evidence of STEC infection have been noted as possible causes of D+HUS.42- 45 However, these reports are more than 20 years old, and limited assessments for simultaneous STEC infection were conducted. Patients with STEC-related D+HUS may have simultaneous evidence of other infections.46 Although many of the non-STEC enteric infections we observed likely occurred in children in whom STEC infection was missed, D+HUS surveillance should continue collecting information on other pathogens to generate testable hypotheses. For example, it is uncertain whether the more frequent isolation of S pyogenes in the stool of slightly older children with concurrent non-O157 STEC infection is of biological importance. Streptococcus pyogenes has been isolated in stool, either alone or in the presence of STEC from children with D+HUS, and it can cause acute poststreptococcal glomerulonephritis.46- 48 Others have reported simultaneous occurrence of HUS and acute poststreptococcal glomerulonephritis.48
In summary, STEC O157 continues to account for most pediatric D+HUS cases in the United States. However, one-third of children had no evidence of STEC infection. Although 5.7% of these children had documented S pneumoniae infection, most likely had undiagnosed STEC infections. Early and complete testing for all STEC in children with diarrheal illness is needed for more complete description of the infectious causes of D+HUS.
Correspondence: Rajal K. Mody, MD, MPH, Centers for Disease Control and Prevention, Mailstop C-09, 1600 Clifton Rd NE, Atlanta, GA 30333 (firstname.lastname@example.org).
Accepted for Publication: March 9, 2012.
Published Online: August 6, 2012. doi:10.1001 /archpediatrics.2012.471
Author Contributions:Study concept and design: Mody, Jones, Hoekstra, and Griffin. Acquisition of data: Scheftel, Hurd, Smith, Lathrop, Palmer, Strockbine, and Talkington. Analysis and interpretation of data: Mody, Luna-Gierke, Comstock, Mahon, Hoekstra, and Griffin. Drafting of the manuscript: Mody and Griffin. Critical revision of the manuscript for important intellectual content: Mody, Luna-Gierke, Jones, Comstock, Scheftel, Hurd, Lathrop, Smith, Palmer, Strockbine, Talkington, Mahon, Hoekstra, and Griffin. Statistical analysis: Mody, Luna-Gierke, and Hoekstra. Administrative, technical, or material support: Griffin, Mahon, Jones, and Lathrop. Study supervision: Griffin, Mahon, Jones, and Talkington.
Financial Disclosure: None reported.
Funding/Support: This study was funded by the CDC's Emerging Infections Program.
Additional Contributions: We thank Beletschachew Shiferaw, MD, Nancy Spina, MPH, Mirasol Apostol, MPH, Effie Boothe, MSN, Tameka Hayes Webb, MPH, and additional past and present FoodNet HUS surveillance coordinators.
Mody RK, Luna-Gierke RE, Jones TF, Comstock N, Hurd S, Scheftel J, Lathrop S, Smith G, Palmer A, Strockbine N, Talkington D, Mahon BE, Hoekstra RM, Griffin PM. Infections in Pediatric Postdiarrheal Hemolytic Uremic SyndromeFactors Associated With Identifying Shiga Toxin–Producing Escherichia coli. Arch Pediatr Adolesc Med. 2012;166(10):902–909. doi:10.1001/archpediatrics.2012.471