Patients were children with diarrhea attending the Royal Children's Hospital, Melbourne, between March 1 and August 31, 2003. They were considered for inclusion in the study when an obviously loose stool sample from a child <14 years of age was received at the Diagnostic Microbiology Laboratory for investigation. After their caregivers, attending physicians, and medical records had been consulted, patients were considered eligible for inclusion in the study if, during the current illness, they had passed >3 loose stools within a day or had experienced loose stools plus vomiting, abdominal pain, or rectal bleeding. Patients with chronic gastrointestinal disorders, such as inflammatory bowel or celiac disease, were excluded, as were those with cystic fibrosis, leukemia, and other immunosuppressive disorders. Repeat samples and samples from children who had received antimicrobial agents within the preceding week were also excluded.

Clinical data were obtained in accordance with a standardized pro forma and were analyzed before the results of the laboratory findings were known. Data collected included age; gender; date of onset of illness; symptoms and clinical signs, including characteristics of stools, abdominal pain, vomiting (number per day and duration), fever, abdominal tenderness, largest number of bowel movements in a 24-hour period preceding the sample collection, and extent of dehydration. Duration of diarrhea was estimated from the passage of the first loose stool to the patient's last appearance in the ward or 1 day after discharge. Patients with temperatures of >38°C, taken by tympanic thermometer, were considered febrile. Severity of illness was estimated by using the 20-point scale developed by Ruuska and Vesikari (10).

All stool specimens were the first specimen obtained from a patient on a hospital visit, and specimens were investigated within 4 hours of collection. Specimens were examined macroscopically for color and consistency and by light microscopy for leukocytes, erythrocytes, and parasitic forms (amebas, cysts, and ova) by using a saline-and-iodine wet preparation and a modified Ziehl-Neelsen stain for oocysts of Cryptosporidium spp (11). Samples were tested by enzyme immunoassay for enteric adenoviruses and rotaviruses and cultured for E. coli, Salmonella, Shigella, Yersinia, and Campylobacter spp (12).

To reduce the cost of the investigation, diarrheogenic strains of E. coli were sought only during the first 11 weeks of the study, from March 1 to May 15. Bacteria were isolated from fecal samples by direct plating on MacConkey agar (Oxoid Ltd., Basingstoke, UK). After overnight incubation at 37°C, a sterile cotton swab was used to transfer the entire growth from each plate into Luria broth containing 30% (vol/vol) glycerol, which was then frozen at –70°C until required. E. coli pathotypes were identified by polymerase chain reaction (PCR) and confirmed by Southern hybridization (9). Briefly, template DNA for use in PCR was prepared from bacteria isolated from MacConkey agar plates and grown in 2.5 mL MacConkey broth with shaking at 37°C overnight. Bacteria from this culture were washed in phosphate-buffered saline, resuspended in sterile distilled water, and heated for 10 min at 100°C. Samples were then placed on ice for 5 min and recentrifuged for 5 min at 16,000 × g. Aliquots of the supernatant were pipetted into sterile tubes, stored at –20°C for <1 week, and then diluted 1 in 10 in distilled water before being added to the PCR mix. PCR amplifications were performed in a GeneAmp PCR System 9700 thermal cycler (Applied Biosystems, Foster City, CA, USA) with the PCR primers and conditions described previously (9). Genes identified by these primers and their association with each pathotype of diarrheogenic E. coli are listed in Table 1. PCR for the lacZ gene, which is present in almost all wildtype strains of E. coli, was included as a control to ensure that negative PCR results were not due to the absence of viable bacteria in the sample or the presence of inhibitors in the reaction mixture. Samples that were PCR negative for lacZ were excluded from further analysis. At the conclusion of the PCR, 10 μL of the reaction mixture underwent electrophoresis on 2.5% 96-well format agarose gels (Electro-fast; Abgene, Epsom, UK). Gels were stained with ethidium bromide, visualized on a UV transilluminator, and photographed. A portion of the PCR product was retained for Southern blotting, which was performed by using capillary transfer of separated DNA fragments onto positively charged nylon membranes (Roche Diagnostics Ltd., Lewes, UK). Digoxigenin-labeled DNA probes were prepared from the control strains of diarrheogenic E. coli listed in Table 1 and used as described (9). PCR- and probe-positive bacteria were assigned to a pathotype according to the criteria in Table 1. Equivocal or ambiguous assays were repeated, and if still unclear, were excluded from further analysis. Atypical EPEC strains were isolated in pure culture from the original sample and then serotyped by using hyperimmune rabbit antisera to O-antigens O1 through O181 (18). These strains were also subjected to PCR to determine the intimin subtype and to investigate the presence of selected virulence-associated genes by using the PCR primers and conditions described previously (9).

Statistical analysis of quantitative and qualitative data was performed by using InStat, Version 3.05 (GraphPad Software Inc., San Diego, CA, USA). A 2-tailed p value of <0.05 indicated statistical significance. For the analysis of clinical features associated with infection, patients whose stools yielded >1 pathogen were excluded.

After exclusion of repeat samples and samples from patients >14 years of age or with cystic fibrosis, chronic inflammatory bowel disease, leukemia, or a history of recent antimicrobial drug usage, 303 of 972 consecutive fecal samples remained for analysis. Of these, 134 were from the first period of the study, March 1–May 15, when diarrheogenic E. coli were sought together with other enteropathogens, and 169 were from the period May 16–August 31, when E. coli were not sought.

The frequency of bacterial, viral, and parasitic pathogens identified during the 2 phases of the study are shown in Table 2. During the first period, a putative etiologic agent was identified in 88 (66%) of 134 children. Diarrheogenic E. coli were found in 42 (31%) of these children, followed by enteric adenovirus (10%), Salmonella sp. (10%), Campylobacter spp. (9%), Giardia sp. (6%), rotavirus (4%), and Cryptosporidium sp. (2%). Of the 42 E. coli isolates, 30 (71%) were EPEC; 6 (14%) were Shiga toxin–producing E. coli (STEC), of which 3 were EHEC; 4 (10%) were enteroaggregative E. coli (EAEC), 1 (2%) was enterotoxigenic E. coli; and 1 (2%) was enteroinvasive E. coli. Nine children (7%) were infected with >1 pathogen, including 2 concurrently infected with EPEC and adenovirus or EPEC and rotavirus, and 1 each with EPEC and Giardia sp.; STEC and Campylobacter sp.; STEC and Giardia sp.; EAEC and Campylobacter sp., and EAEC and rotavirus.

All EPEC isolates were atypical EPEC (i.e., PCR negative for bfpA). Determination of the O:H serotype and intimin subtype of 29 of the 30 EPEC strains (1 was not viable) indicated that they were highly heterogeneous (Table 3). Although 3 strains (R41, R151, and R446) were O-nontypable:H34, intimin-α2; and 2 (R89 and R104) were O153:H7, intimin β, these isolates were neither temporally nor geographically related to each other and showed some differences in their carriage of accessory virulence-related factors (data not shown). Two other isolates (R250 and R436) were O33:H6 but had different intimin types. Ten isolates were O-serogroups that were classified as nontypable because they did not react with any of the available O-typing sera (O1–O181), and 2 isolates could not be serotyped because they were rough. Only 1 isolate (R404) belonged to an E. coli serotype, O128:H2, that is commonly associated with EPEC (4).

During the second period of the study, when E. coli was not sought, putative pathogens were identified in 99 (58.6%) of 169 children; rotavirus was the most frequent (33.7%), followed by Campylobacter (11.8%), adenovirus (7.7%), Salmonella (5.3%), Giardia (1.8%), and Cryptosporidium (0.6%) spp. Four patients were infected with >1 pathogen: 3 concurrently infected with rotavirus and adenovirus and 1 with adenovirus and Salmonella sp. Shigella and Yersinia spp. were not identified during either period of the study. The frequency of rotavirus infection during the second phase of the study was significantly greater than during the first phase (odds ratio [OR] 13.13; 95% confidence interval [CI] 5.08–33.91, p<10^–6, 2-tailed Fisher exact test), confirming the well-known association of rotavirus with winter diarrhea (12). The frequency of patients in whom no pathogens were identified was not significantly different between the 2 study periods (OR 0.74, 95% CI 0.46–1.18, p>0.2, 2-tailed Fisher exact test), despite the omission of tests for diarrheogenic E. coli during the second period. This finding suggests that E. coli did not account for a large number of cases during the second period of the study and accords with our previous observations that diarrhea due to E. coli is relatively less frequent during winter (9,19).

The clinical and laboratory features of patients infected with different pathogens were compared (Table 4). Patients infected with >1 pathogen were excluded from this analysis, as were those infected with Giardia or Cryptosporidium spp. or diarrheogenic E. coli other than EPEC because their numbers were too small for the results to be meaningful. For this analysis, only those patients in whom no pathogens were identified from the first study period were considered because of the possibility that some of those studied during the second period were infected with diarrheogenic E. coli.

Patients infected with EPEC were of a similar age (median 16.9 months) to those from whom no pathogens were isolated (median age 11.6 months). Of the various groups of patients defined according to the cause of diarrhea, only those infected with Campylobacter spp. (median age 34.2 months) differed significantly in age from those with EPEC (p = 0.0002, Mann-Whitney U test). Eighteen (72%) of 25 children monoinfected with EPEC were boys compared with 20 (44%) of 46 children in whom no pathogens were identified (OR = 3.34, 95% CI 1.17–9.55, p = 0.03, 2-tailed Fisher exact test), and with 49 (47%) of 104 children enrolled in phase 1 of the study who were not infected with EPEC (OR = 2.89, 95% CI 1.11–7.5, p = 0.03).

Patients infected with rotavirus or adenovirus were significantly more likely to have a history of vomiting than those with no pathogen identified or those infected with EPEC, Salmonella, or Campylobacter spp. The frequency of vomiting in patients infected with EPEC and in those with no pathogen identified was similar. Abdominal pain was reported significantly more frequently in patients infected with Campylobacter spp. than in those infected with EPEC (p = 0.007, 2-tailed Fisher exact test), adenovirus (p = 0.0005), rotavirus (p = 0.0005), or those in whom no pathogen was detected (p = 0.003).

The duration of diarrhea was significantly longer in patients with EPEC than in those infected with adenovirus (p = 0.002, 2-tailed Student t test), rotavirus (p = 0.0003), Campylobacter (p = 0.0003), Salmonella (p = 0.02), and those without an identifiable pathogen (p = 0.02). Moreover, persistent diarrhea (defined as diarrhea lasting >14 days) was significantly more common in patients infected with atypical EPEC than in those infected with adenovirus, rotavirus, Campylobacter, Salmonella, and those with no pathogen identified (Table 4). Persistent diarrhea also developed in 4 (36%) of 11 patients infected with Giardia sp. The frequency of persistent diarrhea associated with Giardia was significantly greater than that attributable to adenovirus (p = 0.03), rotavirus (p = 0.01), and Camplyobacter, but not Salmonella or atypical EPEC (p>0.1, 2-tailed Fisher exact test).

Fever was significantly more common in patients infected with rotavirus or Salmonella than in those infected with EPEC, adenovirus, or Campylobacter, and those with no pathogen identified. Dehydration of >5% occurred significantly more often in patients infected with rotavirus than in those infected with EPEC, adenovirus, Campylobacter, Salmonella, and those without an identifiable pathogen.

The disease severity score, determined according to the criteria of Ruuska and Vesikari (10), was highest in patients infected with rotavirus followed by Salmonella sp. The mean severity scores in patients infected with EPEC, adenovirus, and Campylobacter sp. and those in whom no pathogen was found were similar. Stools from patients infected with Campylobacter or Salmonella spp. were more likely to contain frank blood, although the differences between patients infected with different etiologic agents were not significant (p>0.05, 2-tailed Fisher exact test). Erythrocytes were more commonly detected on microscopic examination in patients infected with Campylobacter or Salmonella spp. than in those infected with EPEC, adenovirus, rotavirus, or no identifiable pathogen, but the differences were significant with respect to Campylobacter spp. only. Fecal leukocytes were present significantly more often in patients infected with Campylobacter or Salmonella spp. than in those infected with EPEC, adenovirus, rotavirus, or those with no identifiable pathogen.