Analysis of 1560 inpatient and outpatient Escherichia coli isolates from across Canada – Results from the CANWARD 2007 study

1Department of Medical Microbiology and Infectious Diseases, Faculty of Medicine, University of Manitoba; 2Clinical Microbiology, Health Sciences Centre; 3Clinical Microbiology, St Boniface General Hospital, Winnipeg, Manitoba Correspondence: Dr Philippe RS Lagacé-Wiens, Diagnostic Services Manitoba, St Boniface General Hospital, L4025-409 Taché Avenue, Winnipeg, Manitoba R2H 2A6. Telephone 204-237-2483, fax 204-237-6065, e-mail plagacewiens@sbgh.mb.ca CANWARD 2007

sulfonamides should be used cautiously and in consideration of local resistance patterns for infections caused by E coli, due to lower susceptibility rates.Independent factors associated with antimicrobial resistance were age, inpatient status and isolation from a sterile site.These factors should be considered when empirically treating infections likely caused by E coli.Local antimicrobial prescribing practices, in particular the liberal use of fluoroquinolones, and inadequate infection control practices may be reducing susceptibility rates.

MethODS
E coli isolates were obtained as part of the Canadian Ward Surveillance Study (CANWARD 2007), which collected isolates submitted to 12 clinical microbiology laboratories from tertiary care hospitals in seven provinces across Canada.Submitting sites and collection strategy are described elsewhere in the present supplement (2).Isolates had to be deemed clinically significant by the referring laboratory's current specimen work-up protocol.Demographic information collected with each isolate included patient age, sex, site of infection and the location of patient contact (surgical or medical ward, emergency room, intensive care unit [ICU] or hospital clinic).A minimum number of isolates from each hospital location and anatomical site was requested to provide more power to the study.The implication of this collection strategy is that the anatomical distribution of pathogen isolation and inpatient versus outpatient distribution does not reflect the true distribution in the population studied.Isolates were collected within both primary and secondary study objectives and only isolates collected within the primary objective were considered in this analysis.For statistical analysis, age was divided into four categories: 20 years and younger, 21 to 60 years, 61 to 80 years, and 81 years and older, and location of patient contact was divided into either inpatient (wards and ICUs) or outpatient (emergency room and clinics).Information on previous antimicrobial exposure, hospitalization duration and underlying medical conditions was not available.Antimicrobial susceptibility to amoxicillin-clavulanate, cefazolin, cefepime, ceftriaxone, ciprofloxacin, gentamicin, nitrofurantoin, levofloxacin, meropenem, ertapenem, piperacillin-tazobactam, tigecycline and trimethoprim-sulfamethoxazole was determined using broth dilution as described elsewhere in the present supplement (2).
Screening for ESBL production was achieved using a 1 µg/mL or greater ceftriaxone breakpoint and confirmation was with the Clinical and Laboratory Standards Institute-recommended disk diffusion method (7).Univariate analysis using the c 2 (or Fisher's exact test where required) was undertaken to identify relationships between susceptibility to each of the antimicrobials and ESBL production; and the following variables: sex, age group, inpatient/outpatient status and isolation from a sterile site (blood, cerebrospinal fluid, synovial fluid).Relationships where the P<0.20 in the univariate analysis were included in a multivariate nominal logistic regression model to determine independent explanatory variables.Initially, a full factorial multiple logistic regression analysis was performed using the potential explanatory variables identified in the univariate analysis for each antimicrobial, and then a backward selection so that all factors remaining in the model were statistically significant at a 5% level (P<0.05).Statistical analysis was undertaken using JMP software version 7.0 (SAS Institute Inc, USA).

ReSuLtS
Of 7881 total organisms, 1702 E coli (21.6%) were collected from the CANWARD 2007 study, making it the most common organism isolated from patients in Canadian hospitals overall.Of these, 1560 fell within the primary objective and the remaining 142 were submitted as putative ESBL producers for separate analysis and excluded from the present analysis.The mean age of patients infected with E coli was 56.9 years; 12.3% of E coli isolates were from patients younger than 21 years, 34.7% were 21 to 60 years of age, 33.9% were 61 to 80 years of age and 19.1% were older than 80 years of age.There were more samples from women (59.3%); with both sexes combined, 50.5% were invasive isolates (all bloodstream), and 40.7% were from urine, 6.4% from respiratory sources and 2.4% from wounds.Note that the sampling strategy was biased to include a surplus of bloodstream isolates to have greater numbers of these for analysis and this does not represent the true source distribution of E coli infections.The distribution among provinces was British Columbia, 9.7%; Alberta, 7.6%; Saskatchewan, 9.1%; Manitoba, 9.2%; Ontario, 28.3%; Quebec, 29.2% and Nova Scotia, 6.9%.Isolates were not obtained from Newfoundland, Nunavut, the Northwest Territories, Yukon, New Brunswick or Prince Edward Island.
Univariate statistical analysis revealed no relationship (P≥0.20) between any of the demographic variables and susceptibility to meropenem, ertapenem, tigecycline and cefoxitin.The multivariate model is illustrated in Table 2. Piperacillin-tazobactam is excluded from the table because no variables were associated with susceptibility to this antimicrobial in the multivariate model.Age was the strongest predictor of reduced susceptibility, being independently associated with reduced susceptibility to fluoroquinolones, ceftriaxone, cefepime, gentamicin and nitrofurantoin, ESBL production and MDR.Separate analysis of isolates that were MDR but not ESBL producers showed that the association between age and MDR isolated was largely driven by ESBL producers and no association existed between age and MDR isolates that were not ESBL producers (data not shown).Both stepwise regression analyses applied to age categories and linear regression applied to actual age (data not shown) confirmed that increasing age was associated with decreasing susceptibility to all agents and increased likelihood of ESBL production.Isolation from a sterile site was only independently associated with reduced susceptibility to trimethoprim-sulfamethoxazole.Inpatient status was associated with reduced susceptibility to amoxicillin-clavulanate, cefazolin, fluoroquinolones and trimethoprim-sulfamethoxazole.Sex was not independently associated with reduced susceptibility to any antimicrobials.Differences in susceptibility between sexes were driven by higher inpatient prevalence and higher mean and median age in men.

DISCuSSION
Low susceptibility of ICU E coli isolates to fluoroquinolones and trimethoprim-sulfamethoxazole was not unexpected given the wide use of these antimicrobials in both inpatients and outpatients.In particular, the dramatic increase in fluoroquinolone resistance has been observed in many settings (8)(9)(10).Our observations suggest that first-generation cephalosporins and amoxicillin-clavulanate are still useful agents for infections caused by E coli in that susceptibility rates remain near 90% overall.This is particularly true of outpatient isolates where susceptibility is greater than 90% for both these agents.On the contrary, low susceptibility to fluoroquinolones even in the outpatient setting (84%) begins to bring into question the use of these agents as first line for infections commonly caused by E coli, such as urinary tract infections.Trimethoprimsulfamethoxazole susceptibility rates are below 80% in both inpatient and outpatient settings and should only be used for infections empirically in the context of supportive data from local antibiograms or definitive susceptibility data.
In our multivariate model, increasing age was independently associated with reduced susceptibility to fluoroquinolones, nitrofurantoin, ceftriaxone, cefepime, gentamicin and ESBL production.The association between age and fluoroquinolone susceptibility has been demonstrated previously and is likely due to increasing exposure to fluoroquinolones over time and avoidance of fluoroquinolone use in children (11)(12)(13)(14).This suggests that fluoroquinolones as empirical therapy may be appropriate in younger patients if indicated.However, in routine pediatric practice, these drugs are best avoided because of concerns for toxicity.Age has also been described as a risk factor for infection with an ESBL-producing organism (15,16), and may be associated with increasing exposure to health care settings and antibiotics.Interestingly, exposure to fluoroquinolones has been described as a risk factor for acquisition of an ESBL producer, which may partially explain that age is associated with both resistance to fluoroquinolones and ESBL-producing E coli (16).The association between age and cefepime and ceftriaxone resistance is likely driven by the same factors responsible for the increase in infections caused by ESBL producers, because ESBL production correlates well with resistance to both cefepime and ceftriaxone.Because resistance to cefazolin was not associated with increasing age, it would appear that increasing age does not correlate with the acquisition of cephalosporinases capable of hydrolyzing on first-generation cephalosporins, but does correlate with acquisition of cephalosporinases with activity against advanced cephalosporins (ie, ESBLs).It is interesting that increasing age predicted resistance to gentamicin and nitrofurantoin.Age has been reported as a risk factor for resistance to several antimicrobials before and is likely related to an increased cumulative antimicrobial pressure (17).ESBL production has also been shown to be strongly associated with gentamicin resistance (18), as is the case in our data (gentamicin susceptibility in ESBL producers is 41.5% versus 93.2% in non-ESBL producers, P<0.001), which suggests that the increasing prevalence of ESBL producers with increasing age is playing a role in the increasing resistance to gentamicin.Isolation from a sterile site was significantly independently associated with resistance to trimethoprim-sulfamethoxazole.This observation was made in a previous study of ICU E coli isolates and was thought to be potentially associated with the unmeasured variable of length of stay (12).However, in the present study, this observation was independent of inpatient status, which suggests that another mechanism may be at play.Possible explanations include co-presence of virulence determinants and trimethoprim-sulfamethoxazole resistance genes on mobile elements (integrons) or data biased by widespread use of trimethoprim-sulfamethoxazole empirically in urinary tract infections caused by trimethoprim-sulfamethoxazoleresistant organisms that subsequently progress to bacteremia, thereby inflating the number of trimethoprim-sulfamethoxazoleresistant invasive isolates.
Predictably, inpatient isolates had lower susceptibility to several antibiotics, including amoxicillin-clavulanate, fluoroquinolones, cefazolin and trimethoprim-sulfamethoxazole.Interestingly, susceptibility to antimicrobials commonly used in the inpatient setting (ceftriaxone, cefepime, gentamicin, carbapenems and piperacillin-tazobactam) did not appear to be significantly affected by inpatient status.This is reassuring in that these antimicrobials maintain good activity overall in the hospital setting.The reason that antimicrobials commonly used in the community are most affected by inpatient status is not known, but may be due to general practitioners using these antimicrobials to treat outpatients and selection bias occurring because poor response due to antimicrobial resistance requires admission for parenteral antimicrobials.
Interestingly, sex was not a predictor of susceptibility to any of the antimicrobials tested after adjusting for other factors in the multivariate model.Although large differences were seen between susceptibility to fluoroquinolones, both inpatient status and age appeared to be confounding factors in the effect of sex on fluoroquinolone resistance.The absence of a sex effect contradicts the findings of others (16,17,19), but few of these studies adjusted both for inpatient status and age and other studies have demonstrated no association with sex (12,20).
Meropenem, ertapenem, piperacillin-tazobactam, tigecycline and cefoxitin were not significantly associated with any demographic variable in the multivariate model.Low overall resistance rates accounts for these observations.
Our study had some limitations.We could not collect patient information such as length of stay, previous antimicrobial exposure and underlying disease.Although of great interest for the prediction of antimicrobial resistance, the effect of these variables cannot be determined with our data.Also, our isolates reflect only information from the 12 centres studied and our data may not reflect the antimicrobial susceptibility patterns of all hospitals in Canada.However, this study does provide valuable information about the factors predicting antimicrobial susceptibility of E coli in one of the largest of inpatient and outpatient populations in Canada studied to date.
ACKNOWLeDGeMeNtS: Funding for the CANWARD 2007 study was provided in part by the University of Manitoba, Health Sciences Center in Winnipeg, National Microbiology Laboratory-Health Canada, Abbott, Affinium Inc, Astellas, Bayer, Janssen Ortho Inc, Merck, Oryx, Pfizer Canada, TaiGen, Targanta and Wyeth Inc. Special thanks to Nancy Laing, Barb Weshnoweski, Ravi Vashisht, Franil Tailor, Lisa Bittner and Haley Butcher for technological assistance.The authors wish to thank M Tarka for expert secretarial assistance.

APPeNDIx 1
The authors would like to thank the investigators and laboratory site staff at each medical centre that participated in the CANWARD 2007 study: Vancouver Hospital, Vancouver, British Columbia -Dr D Roscoe; University of Alberta Hospitals, Edmonton, Alberta -Dr R Rennie; Royal University Hospital, Saskatoon, Saskatchewan -Dr J Blondeau; Health Sciences Centre, Winnipeg, Manitoba -Drs D Hoban and

TABLE 2 Percentage of Escherichia coli isolates susceptible to selected antimicrobials by explanatory variables significant in the multivariate nominal logistic regression model
CLA Amoxicillin-clavulanate; CFZ Cefazolin; CIP Ciprofloxacin; CPM Cefepime; CRO Ceftriaxone; ESBL Extended-spectrum beta-lactamase producers; GM Gentamicin; LVX Levofloxacin; MDR Resistant to three or more classes of antimicrobials tested; NF Nitrofurantoin; NS No statistically significant association between susceptibility to the antimicrobial and demographic variable at the 0.05 level in the multivariate model; SXT Trimethoprim-sulfamethoxazole