Ten-Year Trends and Risk Factors for Non-O157 Shiga Toxin Producing Escherichia coli Found Through Shiga Toxin Testing, Connecticut,

MAJOR ARTICLE Ten-Year Trends and Risk Factors for Non-O157 Shiga Toxin Producing Escherichia coli Found Through Shiga Toxin Testing, Connecticut, 2000 2009 James L. Hadler, 1 Paula Clogher, 1 Sharon Hurd, 1 Quyen Phan, 2 Mona Mandour, 3 Kelley Bemis, 1 and Ruthanne Marcus 1 1 Emerging Infections Program, Division of Epidemiology of Microbial Diseases, Yale University School of Public Health, New Haven, Connecticut; 2 Emerging Infections Program, Connecticut Department of Public Health, Hartford, Connecticut; and 3 State Laboratory, Connecticut Department of Public Health, Hartford, Connecticut Background. The epidemiology over time of non-o157 Shiga toxin producing Escherichia coli (STEC) is unknown. Since 1999, increasing numbers of laboratories in Connecticut have been testing for ST rather than culturing for O157, enabling identification of non-o157 STEC. Methods. Beginning in 2000, Connecticut laboratories were required to submit ST-positive broths to the State Laboratory for isolation and typing of STEC. The ratio of non-o157:o157 from laboratories conducting ST testing was used to determine state-level estimates for non-o157 STEC. Patients with STEC were interviewed for exposure factors in the 7 days preceding illness. Incidence trends, clinical features, and epidemiology of non-o157 and O157 STEC infections were compared. Results. From 1 January 2000 through 31 December 2009, ST testing detected 392 (59%) of 663 reported STEC infections; 229 (58%) of the isolates were non-o157. The estimated incidence of STEC infection decreased by 34%. O157 and the top 4 non-o157 serogroups, O111, O103, O26, and O45, were a stable percentage of all STEC isolates over the 10-year period. Bloody diarrhea, hospitalization, and hemolytic uremic syndrome were more common in patients with O157 STEC than in patients with non-o157 STEC infection. Exposure risks of patients with non-o157 STEC infection differed from those of patients with O157 STEC infection primarily in international travel (15.3% vs 2.5%; P,.01). Non-O157 types differed from each other with respect to several epidemiologic and exposure features. Conclusions. Both O157 and non-o157 STEC infection incidence decreased from 2000 through 2009. Although infection due to O157 is the most common and clinically severe STEC infection, it accounts for a minority of all clinically significant STEC infections. STEC appear to be a diverse group of organisms that have some differences as well as many epidemiologic and exposure features in common. Infections due to Shiga toxin producing Escherichia coli (STEC) are an important public health problem. E. coli O157:H7 (O157) is the most common STEC in the United States and was first recognized as a human Received 26 January 2011; accepted 5 May 2011. Correspondence: James L. Hadler, MD, MPH, Yale University School of Public Health, Emerging Infections Program, One Church Street, 7th fl, New Haven, CT 06510 (hadler-epi@att.net). Clinical Infectious Diseases 2011;53(3):269 276 Ó The Author 2011. Published by Oxford University Press on behalf of the Infectious Diseases Society of America. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com. 1058-4838/2011/533-0008$14.00 DOI: 10.1093/cid/cir377 pathogen in 1982 [1]. Since then, systematic surveillance has found that O157 is a common cause of both sporadic and outbreak-associated disease [2]. Non-O157 STEC strains are recognized to be an important cause of diarrheal illness globally, with.100 serotypes associated with outbreaks and sporadic cases [3]. Non-O157 STEC have increasingly been recognized as a cause of illness in the United States [4 10], although much less has been learned about their epidemiology, clinical spectrum of illness and modes of transmission, compared with O157 STEC. Historically, testing methods developed for STEC in the United States have focused on culture methods for Trends in Non-O157 STEC in Connecticut d CID 2011:53 (1 August) d 269

detection of O157, and these methods do not detect non-o157 serotypes. Recently, methods for the rapid detection of Shiga toxin (ST) in stool specimens have become commercially available and are being increasingly used by clinical laboratories. In Connecticut, the number of laboratories using non culturebased methods increased from 4 of 24 doing any kind of STEC detection in 1999 to 16 of 23 in 2009. E. coli O157 infections have been laboratory reportable in Connecticut since 1992. In January 2000, in response to the growing clinical use of ST testing, the Connecticut Department of Public Health (DPH) added ST-related disease to the list of Laboratory Reportable Significant Findings [11]. In addition, clinical laboratories are required to submit ST-positive broth cultures to the DPH Laboratory for confirmation and identification of the causative organism. The advent of non culture-based ST tests provided an opportunity to examine the relative occurrence of O157 and non-o157 STEC infection, as well as the epidemiology and risk factors for non-o157 STEC infection. We report results from a decade of surveillance for STEC in Connecticut. METHODS Laboratory Analysis During the period 1 January 2000 through 31 December 2009, all E. coli O157 isolates and stool-inoculated broths with test results positive for ST identified by clinical laboratories throughout the state were required to be submitted to the DPH Laboratory for confirmation and further testing. At the DPH Laboratory, ST-positive broths were plated on Sorbitol MacConkey (SMAC) agar and SMAC agar with cefixime and tellurite added (CT-SMAC). Beginning in 2006, inoculation of MacConkey Broth (MB) was added to this step. All plates and the MB were incubated for 18 24 hours. If any sorbitol-negative colonies grew, one of each colony type as well as a sweep of each plate and the MB were screened for the O157 antigen using the Remel RIM E. coli O157:H7 Latex Agglutination Kit. Colonies positive for O157 were set up for flagellar H7 testing. If the sorbitol-negative colonies were O157 negative, all sorbitol-positive colony types and sweeps as well as the MB and the original broth were tested for ST using the Meridian Premier EHEC EIA. Any enzyme immunoassay positive colony pick was then subcultured to a SMAC for purity and serological testing. A triple sugar iron agar slant, motility, and indole were also set up to confirm the isolate as E. coli. After incubation, non-o157 isolates identified as E. coli were then screened, with use of slide agglutination antisera, for the following E. coli O serotypes: O26, O45, O103, O111, O121, and O145. All non- O157 STEC isolates were forwarded to the Centers for Disease Control and Prevention for further typing and polymerase chain reaction. Epidemiologic Analysis Demographic information and selected clinical features were collected from case report forms and supplemented as needed by calling the physician or hospital of record for all culture-confirmed STEC infections during the period 2000 2009. In addition, efforts were made to interview all culture-confirmed case patients identified during the period 1 February 2000 31 January 2003 and 1 April 2004 31 December 2009 by telephone using a standardized questionnaire. Household contacts (eg, spouse or parent) were interviewed on behalf of case-patients who died or pediatric cases,13 years of age. Case-patients 13 17 years of age were interviewed once parental or guardian consent was granted. Clinical information collected on the questionnaire included symptoms (eg, bloody diarrhea) and complications (eg, hospitalization, hemolytic-uremic syndrome [HUS]). Exposure information was collected for the 7 days before illness onset on a number of factors, including international travel and 3 factors that were found to be associated with illness in previous USbased studies of sporadically occurring E. coli O157: consumption of pink hamburger or exposure to raw ground beef; drinking untreated surface water from a pond, lake or river; and contact with farm animals, particularly cattle [12, 13]. For the latter, we included a positive answer to any of 3 questions: did you visit a petting zoo with cattle, did you have direct contact with cattle, and did you live on a farm (cattle not specified)? Statistical Analysis The v 2 test for trend was used for analysis of trends in the incidence of all O157 infections, O157 infections identified from direct culture, O157 infections identified through ST screening, and non-o157 infections. Estimated STEC incidence for each year was determined as follows. The ratio of non-o157 STEC isolates to O157 isolates identified through ST screening was calculated for each year. This ratio was then applied to the number of O157 isolates identified by direct culture to determine the estimated number of non-o157 isolates that would have been found that year had ST screening been used instead of culture. The total estimated number of STEC isolates was the sum of the number of STEC isolates identified by ST testing plus the number of O157 isolates identified by culture plus the estimated number of non-o157 isolates. Annual Connecticut estimated population denominators were used. Differences between groups of STEC with respect to demographic and clinical features were determined using v 2 analysis. To determine whether non-o157 STEC isolates should be compared with all O157 isolates or only those detected by ST testing, we first compared O157 isolates detected by laboratories doing culture only with O157 isolates detected by laboratories doing ST testing. We expected that there might be differences, because.25% of STEC from ST testing came from a large 270 d CID 2011:53 (1 August) d Hadler et al

outpatient referral laboratory, which was the only laboratory exclusively testing outpatients. Because we did find differences (see Results), we limited non-o157 comparisons to O157 isolates identified in laboratories doing ST testing. In addition, because some clinical features might be age-related, we also examined clinical features stratified by age-group. For the v 2 analyses, first, the percentage of patients with the given characteristic who had an O157 isolate detected after initial ST screening was compared with the percentage of patients who had an O157 isolate detected from direct culture. The percentage of persons with the given characteristic who had any non-o157 STEC isolate was then compared with the percentage of persons with an O157 isolate detected from initial ST testing. Next, the distribution of the percentage of cases with the given characteristic in persons with each of the 4 leading non-o157 serotypes and with other serotypes was examined using v 2 2 3 5 analysis. Finally, the percentage of cases of infection due to each leading non-o157 serotype with the given characteristic was compared with that for cases of infection due to all other non- O157 serotypes. A similar analysis was done for exposure factors, except that the percentage of individuals with each exposure in persons with non-o157 isolates was compared with the percentage in persons with O157 isolates regardless of how the O157 isolate was initially identified. RESULTS A total of 663 cases of STEC infection were reported over a 10- year period. Of the 392 STEC infections detected by ST testing, 58% were non-o157, and 42% were O157 (Figure 1; Table 1). Among the mostly hospital-based laboratories using ST testing, there was 1 outpatient-only laboratory, which identified 172 STEC isolates (61 O157 and 111 non-o157 STEC isolates). Not included in these results are 78 broths reported as ST positive that were sent from laboratories for which ST positivity could not be confirmed at the state laboratory and 29 confirmed STpositive broths from which no STEC isolates were identified. The reported and estimated incidence of STEC overall, by detection method and by O157 status, are shown in Figure 2. Figure 1. Shiga toxin producing Escherichia coli (STEC) isolates by method of initial detection and STEC serotype, Connecticut, 2000 2009. Overall reported incidence of O157 has been stable since 2001. However, there is an increasing trend in the yearly percentage of all O157 STEC isolates that are detected through ST testing (23% in 2000 to 48% in 2009; P,.001). The percentage of ST-detected isolates that were non-o157 did not change significantly from year to year, and there was no trend from 2000 to 2009 (range by year, 41% 76%; median percentage, 61%). The measured incidence of non-o157 isolates increased slightly, although insignificantly (P 5.19), whereas the estimated incidence plus the measured incidence of non-o157 isolates clearly decreased (1.69 isolates per 100,000 population in 2000 to 1.43 isolates per 100,000 population in 2009; P,.001), as did the estimated overall STEC incidence (4.16 isolates per 100,000 population in 2000 to 2.73 isolates per 100,000 population in 2009; P,.001). Among the 229 isolates of non-o157 STEC, 44 (19.2%) were O111, 43 (18.8%) were O103, 39 (17.0%) were O26, and 26 (11.4%) were O45. The remaining 77 isolates included 59 with 26 additional serotypes, 10 classified as rough, and 8 with undetermined serotype. Although there was substantial variation in the percentage of isolates that were 1 of the 4 leading serotypes in any given year (eg, the percentage of isolates that were O111 ranged from 4% to 30%, with a median value of 20%), none of the variations were statistically significant nor were there consistent trends over time. Demographic Information Significant differences by sex and by age were found for O157 isolates detected by culture, compared with O157 isolates detected by ST testing (Table 1). Thus, we limited comparisons of non-o157 serotypes to O157 isolates detected by ST testing. When examined by sex, the majority of patients with O157 and non-o157 STEC infection were female. There were notable differences between several non-o157 serotypes and O157 as well as differences between non-o157 serotypes (Table 1). Patients with infection due to O26 and other non-o157 serotypes were more likely to be female than were patients with infection due to O157 (P,.05). Among patients with infection due to each of the 5 serogroups of non-o157 STEC, the percentage of female patients ranged from 40.9% to 69.2% (P,.05). Comparing each non-o157 subtype to the others combined, those patients that had O111 isolates were more likely than were other patients to be male (P,.01). When examined by age, the majority of patients in both the O157 and non-o157 groups were in the 0 17-year-old age groups (Table 1). However, when comparing all patients with non-o157 isolates to patients with O157 isolates, the patients with non-o157 isolates included relatively fewer individuals who were,17 years of age (55% vs 69%; P,.01) and a higher percentage of 18 64-year-olds. Among the 5 groups of patients with different non-o157 isolates, there were wide age Trends in Non-O157 STEC in Connecticut d CID 2011:53 (1 August) d 271

Table 1. Selected Demographic Information for Individuals With Shiga Toxin Producing Escherichia coli (STEC) Infection by STEC O Serotype and Method of Detection of O157, Connecticut, 2000 2009 Age Female sex 0 9 Years 10 17 Years 18 64 Years $65 Years STEC serotype and O157 detection method Total n % n % n % n % n % All O157 434 250 57.6 148 34.1 111 25.6 132 30.4 43 9.9 O157 culture 271 167 61.6 87 32.1 59 21.8 93 34.2 32 11.9 O157 ST broth 163 83 50.9 *,a 61 37.4 52 31.9 *,a 39 23.8 *,a 11 6.7 All non-o157 229 134 58.5 *,c 70 30.6 **,c 56 24.5 80 34.9 *,b 23 10.0 *,c 0111 44 18 40.9 **,d 19 43.2 *,d 14 31.8 9 20.5 2 4.5 0103 43 26 60.5 9 20.9 15 34.9 19 44.2 **,b 0 0 *,d 026 39 27 69.2 *,b 16 41.0 7 17.9 10 25.6 6 15.4 045 26 11 42.3 2 7.7 *,b,**,d 8 30.8 11 42.3 *,b 5 19.2 *,b Other non-o157 77 52 67.5 *,b,d 24 31.2 12 15.6 **,b,*,d 31 40.3 **,b 10 13.0 NOTE. * P,.05, ** P,.01. a Compared with O157 culture. b Compared with O157 ST broth. c Non-O157 overall chi-square. d Compared to all other non-o157 serotypes. distribution differences, particularly with respect to the percentage of patients in the 0 9-year-old age group (range, 39% 75%; P 5.01). Comparing each non-o157 serotype group with the others combined, those that had O111 isolates were more likely to be 0 9 years of age (P 5.04), whereas those that had 045 isolates were less likely to be in that age group (P,.01). At the other end of the age spectrum, those that had O103 isolates were less likely to be $65 years old (P 5.02) Clinical Information In comparing patients with O157 isolates detected by culture with those who had O157 isolates detected by ST testing, we found that there was a statistically lower rate of hospitalization Figure 2. Reported and estimated incidence of Shiga toxin producing Escherichia coli (STEC) by STEC detection method, serotype and year, Connecticut, 2000 2009. a Includes estimate of non-o157 isolates that would have been found had ST testing been done on all specimens for which only O157 culture was done. Assumes the ration by year of non- O157 to O157 isolates from ST testing applies equally to O157 isolates for that year from culture. in the latter group; therefore, we again limited comparisons of non-o157 isolates to O157 isolates detected through ST testing (Table 2). Overall, individuals with non-o157 isolates tended to have less severe disease, compared with individuals with O157 isolates. Individuals with non-o157 isolates were less likely to have bloody diarrhea, less likely to be hospitalized, and less likely to have HUS (P,.01 for all). Among those with non-o157 serotypes, there were differences between serotype groups with respect to the percentage of patients with bloody diarrhea (range, 44% 79%; P 5.03), but differences in the percentage of patients who were hospitalized were not statistically significant (range, 7% 23%). When examined by age group, there was a statistically significant trend of increasing percentages of patients hospitalized with increasing age for both the O157 and non-o157 groups but not for the percentages of patients with bloody diarrhea or with HUS (Table 3). When age group specific percentages of patients with each clinical feature were compared, the percentage of patients with non-o157 serotypes who had each feature was always lower than that for patients with O157 serotypes, with varying degrees of statistical significance. Numbers of patients by age group were too small to make meaningful comparisons between those with different non-o157 serotypes. Exposure Information Overall, exposure questionnaire data were available for 545 of the 663 patients with STEC, including 375 with O157 isolates and 170 with non-o157 isolates. The only exposure with significant differences between all patients with non-o157 isolates and those with O157 isolates was international travel in the week before symptom onset (15.3% vs 2.5%; P,.001) (Table 4). 272 d CID 2011:53 (1 August) d Hadler et al

Table 2. Clinical Features and Complications of Shiga Toxin Producing Escherichia coli (STEC) Infection by STEC O Serotype and Method of O157 Detection, Connecticut, 2000 2009 Bloody diarrhea Hospitalization Hemolytic-uremic syndrome STEC serotype and O157 detection method Total n/n a % n % n % All O157 434 316/362 87.3 163 37.6 45 10.4 O157 culture 271 200/223 89.7 122 45.0 31 11.4 O157 ST broth 163 111/131 84.7 41 25.2 **,b 14 8.6 All non-o157 229 93/154 60.4 **,c,*,d 32 14.0 **,c 1,0.5 *,c O111 44 21/31 67.7 *,c 3 6.8 **,c 0 0 *,c O103 43 25/35 71.4 5 11.6 0 0 *,c O26 39 15/29 51.7 **,c 7 17.9 0 0 *,c O45 26 15/19 78.9 6 23.1 0 0 Other 77 17/40 42.5 **,c,*,e 11 14.3 1 1.3 *,e NOTE. a Determination of bloody diarrhea is dependent on patient interview, which was not able to be done on all patients. * P,.05, ** P,.01. b Compared with O157 culture. c Compared with O157 ST. d Non-O157 overall chi-square. e Compared with all other non-o157 STEC serotypes. Among those who traveled, South and Central America, including the Caribbean, were the most common destinations (in 16 of 18 patients with non-o157 isolates and 5 of 8 patients with O157 isolates). The prevalence of international travel among those with non-o157 serotypes showed more variation than was expected by chance (range, 0% 30.3%; P 5.05). Persons with O111 serotypes were more likely than were those with other non-o157 serotypes to have traveled (P,.01), and those with O45 serotypes were less likely to have traveled (P 5.04). There was a suggestion of a difference between patients with O157 and those with non-o157 STEC with respect to consumption of pink hamburger or handling raw ground beef (25.5% vs 17.8%; P 5.06) and for untreated surface water consumption (7.5% vs 13.9%; P 5.07). Among those with non- O157 STEC, those with infection due to serotype O103 were more likely than were those with all other non-o157 STEC to have undercooked hamburger exposure (31.4% vs 13.7%; P 5.02). When limiting analysis to domestically acquired cases, the relative differences presented above were larger for those with O157 serotypes, compared with those with non-o157 Table 3. Clinical Features and Complications by Age and ST Serotype, Shiga Toxin Producing Escherichia coli (STEC) Detected by ST Testing, Connecticut, 2000 2009 Bloody diarrhea Hospitalization Hemolytic-uremic syndrome STEC serotype, age group n/n a % n/n % n/n % O157 0 9 Years 39/48 81.3 9/61 14.8 8/60 13.3 10 17 Years 40/46 87.0 11/52 21.2 01/52 1.9 18 64 Years 25/29 86.2 15/39 38.5 4/39 10.3 $65 Years 7/8 87.5 6/11 54.6 1/10 10.0 P for trend NS,.001 NS All non-o157 0 9 Years 20/45 44.4 **,b 4/70 5.7 1/70 1.4 **,b 10 17 Years 31/44 70.5 5/56 8.9 0/56 0 18 64 Years 39/59 66.1 *,b 17/80 21.3 *,b 0/80 0 **,b $65 Years 8/14 57.1 6/23 26.1 0/23 0 P for trend NS.001 NS NOTE. * P,.05, ** P,.01. a Determination of bloody diarrhea is dependent on patient interview, which was not able to be done on all patients. b compared to O157, same age group. Trends in Non-O157 STEC in Connecticut d CID 2011:53 (1 August) d 273

Table 4. Prevalence of Selected Exposure Factors for Individuals With Shiga Toxin Producing Escherichia coli (STEC) by STEC O serotype, Connecticut, 2000 2009 Ate pink hamburger or handled raw ground beef Petting zoo with cattle/farm Drank from untreated surface water a International travel STEC serotype n/n % n/n % n/n % n/n % All O157 89/349 25.5 22/349 6.3 15/201 7.5 9/353 2.5 All non-o157 27/152 17.8 10/154 6.5 15/108 13.9 *,c 24/157 15.3 **,b,*,c O111 3/33 9.1 2/33 6.1 4/26 15.4 10/33 30.3 **,b,**,d O103 11/35 31.4 *,d 2/35 5.7 5/21 23.8 5/35 14.3 **,b O26 4/29 13.8 3/29 10.3 2/22 9.1 4/30 13.3 *,b O45 2/18 11.1 1/18 5.6 2/12 16.7 0/18 0 *,d Other 7/37 19.9 2/39 5.1 2/27 7.4 5/41 12.2 **,b NOTE. a No data on this variable prior to 2004. * P,.05, ** P,.01. b Compared with O157. c Non-O157 overall chi-square. d Compared to all other non-o157 STEC serotypes. serotypes, and larger for those with O103 serotypes, compared with those with other non-o157 serotypes, for pink hamburger consumption or handling of raw ground beef (differences of 7.4% and 19.0%, respectively) and were smaller for those with non-o157 serotypes, compared with those with O157 serotypes, for domestic water consumption (difference, 3.4%). DISCUSSION Non-O157 E. coli have been increasingly recognized as being important pathogens in the United States [3 10]. However, there have been few population-based studies that have compared the relative incidence of infection due to O157 STEC with that of infection due to non-o157 STEC [6 8, 10, 14, 15], the relative severity of illness [14, 15], and comparative exposures [14], and the studies that have investigated these issues have not had sufficient data to examine trends. The Connecticut 10-year trend data add substantially to this experience. There are several important observations from this data: (1) as a group, non- O157 STEC strains are consistently more common than are O157 STEC strains, although O157 is consistently the most common individual serotype; (2) when accounting for differences in the use of ST testing over time, the incidence of STEC infection in Connecticut appears to have progressively decreased, by approximately 34%, from 2000 through 2009; (3) there is stability in the relative occurrence of the leading STEC serotypes; (4) O157 is associated with the most-severe disease; (5) non-o157 serotypes, as a group, and O157 have many epidemiologic characteristics and risk factors in common; and (6) among non-o157 serotypes, there is substantial diversity in epidemiology and risk factors. In 1999, national estimates of infection due to O157 and non-o157 STEC assumed that the two groups produced equal numbers of clinically important infections, hospitalizations, and deaths [16]. However, data collected through surveillance in several states since then has provided evidence that the incidence of non-o157 STEC is higher [7, 9, 10, 14]. New national estimates of food-borne illness incidence reflect this; non-o157 strains of STEC are estimated to cause 1.75 times thenumberofillnessesaso157stec[2].thisreportfurthers these observations by demonstrating that, over a 10-year period, non-o157 isolates have consistently been more common than O157 isolates when testing for both is performed. It is likely that, as additional laboratories and states adopt the recommendations to routinely test all stool specimens submitted for enteric pathogen testing for ST [5], a clearer picture of the relative frequency and importance of non-o157 serotypes will emerge. Nationally, the incidence of infection due to O157 strains has been declining [17]. Although the reported number of cases of infection due to non-o157 STEC has been increasing (unpublished data), this is likely to be an artifact of increased detection. In Connecticut, the incidence of infection due to non- O157 STEC, based on reported cases, has, if anything, been increasing. However, after adjustment for non-o157 STEC that would have been detected had ST testing, instead of culture for O157 STEC, been performed initially, the estimated rates of non-o157 STEC infection show a decrease comparable to that for O157 STEC infections. In addition, the Connecticut data suggest that the relative occurrence of all of the major O serotypes is stable and that no single type is emerging in prevalence, relative to other types. Our findings with respect to the main clinical features of STEC infection are consistent with those of others [4, 14, 15]. Persons with infection due to O157 STEC had higher rates of bloody diarrhea, hospitalization, and HUS, compared with 274 d CID 2011:53 (1 August) d Hadler et al

those among patients with any other relatively common STEC serotype, and this was true for all age groups. Within the United States, there has only been 1 other reported study of the relative prevalence of risk factors for infection due to O157 STEC among persons with infection due to non-o157 serotypes [14]. As in that study, we found that international travel in the 7 days prior to symptom onset was the most distinguishing feature between those with infection due to non-o157 STEC and those with infection due to O157 STEC. This finding applied to each of the leading serotypes except O45. Two previous case-control studies of sporadic O157 cases demonstrated that consumption of pink hamburger; drinking untreated surface water from a pond, lake or river; and contact with farm animals, particularly cattle [12, 13], were associated with disease. In Connecticut, there was a suggestion that those individuals with non-o157 STEC infection were less likely than others to have eaten or handled undercooked ground beef and were more likely to have consumed untreated surface water. These findings are consistent with those in Minnesota [14]. A study in Australia [18] also found that persons with O157 were more likely than those with non-o157 to have had any hamburger prior to illness. Our findings suggest that not all non-o157 STEC strains are epidemiologically similar. Some strains (eg, O103) seem to have an epidemiologic and exposure profile that is similar to that of O157, and they likely occupy a similar ecologic niche. Others are quite different (eg, O45) and may not be able to be managed through domestic efforts to control O157 STEC. Continued population-level monitoring of the epidemiology of STEC is needed to determine longer term trends and opportunities for control. Our surveillance system has several important potential limitations. First, incidence is likely to be underestimated. Not all persons with diarrhea see a physician or undergo testing, and not all laboratories routinely test for O157 STEC [19]. Second, the true ratio of non-o157 to O157 STEC cases could be somewhat different than described. Not all laboratories conducted ST testing, and the assumption that the ratio of non- O157 to O157 isolates would be the same (if that were the case) if all had conducted ST testing could have been wrong. Third, the number of cases of infection due to each of the leading non- O157 serotypes was small, which limited the power to detect potentially important differences between them. Finally, data were missing on bloody diarrhea for 25% 33% of cases. In summary, both O157 and non-o157 STEC infections appear to have decreased in incidence over the period 2000 2009. Although infection due to O157 is the most common and clinically severe type of STEC infection, it accounts for a minority of all clinically significant STEC infections. STEC are a diverse group of organisms that have many epidemiologic and exposure features in common, but which also have many differences. Acknowledgments We thank the Connecticut Department of Public Health Laboratory and Epidemiology Program staff and Yale School of Public Health students, who have assisted with STEC isolation and surveillance over the past 10 years, and Matthew Cartter and Robert Heimer of the Connecticut Emerging Infections Program, for reviewing this work. Financial support. This work was supported by the Centers for Disease Control and Prevention as part of the Emerging Infections Program, Foodborne Diseases Active Surveillance Network (FoodNet); Cooperative agreement U50/CCU111188 from 2000 to 2004, and Cooperative agreement U50/01CI000307 from 2005 to 2011. Potential conflicts of interest. J. L. H. has received payment for participation on the CDC Board of Scientific Counselors. All other authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed in the Acknowledgments section. References 1. Riley L, Remis R, Helgerson S, et al. Hemorrhagic colitis associated with a rare Escherichia coli serotype. N Engl J Med 1983; 308:681 5. 2. Scallan E, Hoekstra RM, Angulo FJ, et al. Foodborne illness acquired in the United States - major pathogens. Emerg Infect Dis 2011; 17:7 15. 3. Johnson KE, Thorpe CM, Sears CL. The emerging clinical importance of non-o157 Shiga toxin-producing Escherichia coli. Clin Infect Dis 2006; 43:1587 95. 4. Brooks J, Sowers E, Wells J, et al. Non-O157 Shiga toxin-producing Escherichia coli infections in the United States, 1983 2002. J Infect Dis 2005; 192:1422 9. 5. Centers for Disease Control and Prevention. Recommendations for diagnosis of Shiga toxin-producing Escherichia coli infections by clinical laboratories. Morb Mort Wkly Rep 2009; 58(RR-12):1 14. 6. Fey P, Wickert R, Rupp M, Safranek T, Hirichs S. Prevalence of non- O157:H7 shiga toxin-producing Escherichia coli in diarrheal stool samples from Nebraska. Emerg Infect Dis 2000; 6:530 3. 7. Manning S, Madera R, Schneider W, et al. Surveillance for Shiga toxinproducing Escherichia coli, Michigan, 2001 2005. Emerg Infect Dis 2007; 13:318 21. 8. Lockary VM, Hudson RF, Ball CL. Shiga toxin-producing Escherichia coli, Idaho [letter]. Emerg Infect Dis 2007; 13:1262 4. 9. Centers for Disease Control and Prevention. Laboratory-confirmed non-o157 shiga toxin-producing Escherichia coli Connecticut, 2000 2005. MMWR Morb Mortal Wkly Rep 2007; 56:29 31. 10. Connecticut Department of Public Health. Reportable diseases and laboratory findings, 2000: shiga toxin producing Escherichia coli infection. Connecticut Epidemiol 2000; 20:1 4. http://www.ct.gov/dph/ lib/dph/infectious_diseases/ctepinews/vol20no1.pdf. Accessed 23 June 2011. 11. Lathrop S, Edge K, Bareta J. Shiga toxin-producing Escherichia coli, New Mexico, USA, 2004 2007. Emerg Infect Dis 2009; 15:1289 91. 12 Kassenborg H, Hedberg C, Hoekstra M, et al. Farm visits and undercooked hamburgers as major risk factors for sporadic Escherichia coli O157:H7 infection: data from a case-control study in 5 FoodNet sites. Clin Infect Dis 2004; 38(Suppl 3):S271 S8. 13. Voetsch A, Kennedy M, Keene W, et al. Risk factors for sporadic Shiga toxin-producing Escherichia coli O157 infections in FoodNet sites, 1999 2000. Epidemiol Infect 2007; 135:993 1000. 14. Hedican E, Medus C, Besser JM, et al. Characteristics of O157 versus non-o157 Shiga toxin-producing Escherichia coli infections in Minnesota, 2000 2006. Clin Infect Dis 2009; 49:358 64. Trends in Non-O157 STEC in Connecticut d CID 2011:53 (1 August) d 275

15. Jelacic J, Damrow T, Chen G, et al. Shiga toxin-producing Escherichia coli in Montana: bacterial genotypes and clinical profiles. J Infect Dis 2003; 188:719 29. 16. Mead P, Slutsker L, Dietz V, et al. Food-related illness and death in the United States. Emerg Infect Dis 1999; 5:607 25. 17. Centers for Disease Control and Prevention. Preliminary FoodNet data on the incidence of infection with pathogens transmitted commonly through food 10 States, 2009. MMWR Morb Mortal Wkly Rep 2010; 59:418 22. 18. McPherson M, Lalor K, Combs B, Raupach J, Stafford R, Kirk MD. Serogroup-specific risk factors for Shiga toxin-producing Escherichia coli infection in Australia. Clin Infect Dis 2009; 49:249 56. 19. Angulo FJ, Voetsch AC, Vugia D, et al. Determining the burden of human illness from foodborne diseases: CDC s emerging infectious disease program foodborne diseases active surveillance network (Foodnet). Vet Clin North Am Food Anim Pract 1998; 14:165 72. 276 d CID 2011:53 (1 August) d Hadler et al