Longitudinal study of antimicrobial-resistant commensal Escherichia coli in the faeces of horses in an equine hospital

Preventive Veterinary Medicine 100 (2011) 134–145

Contents lists available at ScienceDirect

Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed

Longitudinal study of antimicrobial-resistant commensal Escherichia coli in the faeces of horses in an equine hospital Thomas W. Maddox a,∗ , Nicola J. Williams a , Peter D. Clegg b , Andrew J. O’Donnell c , Susan Dawson a , Gina L. Pinchbeck d a

National Centre for Zoonosis Research, School of Veterinary Sciences, Leahurst Campus, University of Liverpool, CH64 7TE, UK Department of Musculoskeletal Biology, Institute of Ageing and Chronic Disease, School of Veterinary Sciences, Leahurst Campus, University of Liverpool, CH64 7TE, UK c Treforest Veterinary Clinic, 16 River Street, Treforest, CF37 1TD, UK d Department of Epidemiology and Population Health, Institute of Infection and Global Health, Leahurst Campus, University of Liverpool, CH64 7TE, UK b

a r t i c l e

i n f o

Keywords: Antimicrobial resistance Escherichia coli Risk factors Horses Extended spectrum ␤-lactamases

a b s t r a c t The increasing prevalence of antimicrobial resistance in bacteria represents a considerable problem for human and veterinary medicine, causing complications in the treatment of infections. Resistance in Escherichia coli from horses has been documented in commensal and pathogenic strains, but little information exists regarding the prevalence of such bacteria in hospitalised horses or associated risk factors. A longitudinal cohort study was conducted of 103 horses admitted to a referral equine hospital for more than 48 h, with faecal samples collected on hospital admission and subsequently every two days until discharge. Horses undergoing radioactive gamma scintigraphic examination, unweaned foals and mares with un-weaned foals were excluded. Data were collected from enrolled animals, including antimicrobial treatment history and hospitalisation details. Samples were cultured for resistant E. coli; isolates had their antimicrobial resistance profile determined. High sample prevalence for resistant E. coli was identified for all antimicrobials examined except co-amoxiclav. The prevalence of resistance was consistently lower at admission, rising to a peak 4 days post-admission. Risk factors were analysed using multilevel, multivariable modelling, which identified significant clustering of resistance outcomes within horses. For all outcomes except trimethoprim resistance, the day the sample was obtained was significant, with increased risk of resistance for samples taken on day 2 or later. Antimicrobial treatment in the previous seven days and increased total daily dosages of cotrimoxazole prescribed in the hospital in the previous 24–48 h were associated with increased risk. Location within the hospital and admission reason were significant risk factors for some resistance outcomes. High levels of multidrug-resistant E. coli (47.7% of samples) and extended spectrum ␤-lactamase-producing E. coli (27.3% of samples) were recovered; such bacteria could significantly complicate treatment if they were the cause of infection and may represent a risk to personnel in close contact with hospitalised horses. © 2011 Elsevier B.V. All rights reserved.

1. Introduction

∗ Corresponding author. Tel.: +44 151 795 6100; fax: +44 151 794 6005. E-mail address: [email protected] (T.W. Maddox). 0167-5877/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.prevetmed.2011.02.006

Escherichia coli is a Gram-negative bacteria of the Enterobacteriaceae family and part of the normal commensal gastrointestinal flora of most animals and humans, including horses (Holland et al., 1996; van Duijkeren et al., 2000).

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

As a commensal, E. coli is frequently exposed to antimicrobial agents used in the treatment of infections caused by other organisms, potentially allowing it to acquire resistance determinants and act as a reservoir for resistance genes (Smith, 1969; Saenz et al., 2004; Hart et al., 2006; Karami et al., 2007). Some E. coli strains can be pathogenic, causing gastrointestinal disease and extra-intestinal infections of non-gastrointestinal sites. The expression of a variety of enteropathogenicity determinants can render gastrointestinal E. coli capable of causing diarrhoea and enteritis in a number of species (Meng et al., 1998). Extra-intestinal disease in animals can involve the uterus, mammary glands and urinary tract (Nataro and Kaper, 1998; Albihn et al., 2003; Lanz et al., 2003) and septicaemia in younger animals (Hirsh et al., 1993). Antimicrobial resistance among bacteria is recognised as an important and increasing problem in human and veterinary medicine, with significant economic implications, as well as increased patient morbidity and mortality (Paladino et al., 2002; Ogeer-Gyles et al., 2006). Resistance in E. coli isolated from animals was identified as an emerging problem over forty years ago (McKay et al., 1965), and following this, multidrug antimicrobial resistance (resistance to three or more antimicrobial classes) was identified in equine E. coli isolates (Harihara and Barnum, 1973). Subsequently resistance has been reported with increasing frequency to a widening range of antimicrobial classes, such as the penicillins, cephalosporins, aminoglycosides, tetracyclines and potentiated sulphonamides (Anzai et al., 1987; Lavoie et al., 1991; Singh et al., 1992; Bucknell et al., 1997; Dunowska et al., 2006; Vo et al., 2007). Multidrugresistant (MDR) bacteria, including E. coli, have been the cause of disease outbreaks in equine hospitals on multiple occasions (Seguin et al., 1999; Ward et al., 2005; Russell et al., 2008; van Duijkeren et al., 2009). Bacterial resistance to antimicrobials can be intrinsic or acquired, and may be achieved by production of antimicrobial inactivating enzymes, modification of the target molecule or protection of the antimicrobial target (Livermore, 2003). Intrinsic resistance is the consequence of a structural or functional trait allowing tolerance of an antimicrobial drug by all members of a bacterial group; it is usually expressed by chromosomal genes and vertically inherited. Acquired resistance is a trait associated with only some strains of a bacterial species or genus. It can be achieved via chromosomal mutation, or more usually by horizontal acquisition of foreign genetic material including resistance genes on conjugative plasmids or other mobile genetic elements (Roupas and Pitton, 1974; Hall and Collis, 1995). Extended spectrum ␤-lactamase (ESBL) enzymes produced by E. coli represent a resistance mechanism of particular significance, providing resistance to ␤lactam antimicrobials, including the third-generation cephalosporins (e.g., cefotaxime, ceftriaxone, ceftazidime). The majority of ESBL enzymes are mutations derived from the classical TEM or SHV ␤-lactamases (Payne et al., 1989) and, being plasmid-borne, have become widespread within ˜ et al., 2003; Vo et al., 2007). many bacterial species (Brinas However, cefotaximase enzymes (CTX-M) represent a distinct class of ESBL that have become prevalent in E. coli from

135

humans and other animals (Hopkins et al., 2006; Liebana et al., 2006). Several factors have been found to be associated with antimicrobial-resistant E. coli in animals. Exposure to antimicrobial drugs is generally accepted to increase the prevalence of resistance; some environmental factors and increased age (for very young animals) have also been associated (Berge et al., 2005). The use of some antimicrobial drugs, and hospitalisation, were found to be associated with antimicrobial resistance in 216 horses in a North American study (Dunowska et al., 2006). Currently only a limited number of studies have examined the prevalence of resistance in groups of hospitalised horses (Dunowska et al., 2006), and repeated sampling of horses to attempt to document longitudinal changes in resistance has not been reported. The aim of this study was to determine the prevalence of antimicrobial-resistant E. coli in horses on entry to an equine hospital and during hospitalisation, and to identify risk factors associated with its presence, as preliminary data suggested prevalence increased after entry to the hospital irrespective of individual antimicrobial treatment. The study hypothesis was that exposure to antimicrobial treatment in the hospital would increase the odds of a horse having antimicrobialresistant E. coli in its subsequent faecal samples and that prevalence of antimicrobial-resistant E. coli carriage would be increased during hospitalisation compared to admission. 2. Materials and methods 2.1. Study overview Repeated faecal samples were collected from horses in an equine hospital over an 18 month period (from December 2007 through May 2009). Antimicrobial-resistant E. coli were isolated from samples and investigated to determine their susceptibility to a panel of seven antimicrobial agents. The presence of resistance in samples was analysed using multilevel multivariable modelling. 2.2. Study population The study population consisted of horses admitted to the University of Liverpool’s Philip Leverhulme Equine Hospital (PLEH). Horses admitted for less than 48 h were excluded, as were patients undergoing radioactive gamma scintigraphic examination, un-weaned foals and mares with un-weaned foals. One or two recruitment days per normal working week were selected on a convenience basis (determined by laboratory capability), and on these days all eligible horses were enrolled. Informed ethical consent was obtained from all owners. Sample size estimates for a longitudinal study investigating horses exposed and non-exposed to antimicrobials were conducted using the EpiInfo software package (EpiInfo version 6 EpiInfo 6, CDC & WHO, Geneva). Assuming a ratio of 1:2 for untreated:treated with antimicrobials, and an expected prevalence of antimicrobial resistance of 50% in the unexposed group (as pilot studies indicated prevalence would be high), a total of 144 horses would

136

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

Table 1 Dichotomous variables considered for inclusion in the final multivariable models, with the number and percentage of samples in each category. Variable

Number of samples (%) No

Yes

Any antimicrobial on sampling day Penicillin on sampling day Gentamicin on sampling day Trimethoprim on sampling day Ceftiofur on sampling day Enrofloxacin on sampling day

332 (72.6) 376 (82.3) 449 (98.2) 416 (91.0) 453 (99.1) 449 (98.2)

125 (27.4) 81 (17.7) 8 (1.8) 41 (9.0) 4 (0.9) 8 (1.8)

Any antimicrobial in previous 2 days Penicillin in previous 2 days Gentamicin in previous 2 days Trimethoprim in previous 2 days Ceftiofur in previous 2 days Enrofloxacin in previous 2 days

313 (68.5) 351 (76.8) 448 (98.0) 411 (89.9) 453 (99.1) 450 (98.5)

144 (31.5) 106 (23.2) 9 (2.0) 46 (10.1) 4 (0.9) 7 (1.5)

Any antimicrobial in previous 7 days Penicillin in previous 7 days Gentamicin in previous 7 days Trimethoprim in previous 7 days Ceftiofur in previous 7 days Enrofloxacin in previous 7 days

284 (62.1) 308 (67.4) 443 (96.9) 401 (87.7) 450 (98.5) 451 (98.7)

173 (37.9) 149 (32.6) 14 (3.1) 56 (12.3) 7 (1.5) 6 (1.3)

Surgical procedure in 24 h prior to sample Surgical procedure in 48 h prior to sample Prior hospitalisation in previous 3 months

435 (95.2)

22 (4.8)

401 (87.7)

56 (12.3)

405 (88.6)

52 (11.4)

be required to detect odds ratios of three or more, with 95% confidence and 80% power. This calculation was based on one sample per horse; obtaining multiple samples from each horse would increase the power. 2.3. Sample collection A fresh faecal sample was obtained from each horse within 12 h of admission and then subsequently every two days until discharge from the hospital. Signalment data, admission reason, previous veterinary history (including antimicrobial therapy prior to admission and prior hospitalisation), details of the horse’s home yard and management were collected by a questionnaire administered to the owner (see Tables 1–3 for further details of information collected). 2.4. Information from hospital database records Whilst hospitalised, details of drug administration (antimicrobial or otherwise), veterinary procedures and location within the hospital were recorded for each horse. The PLEH clinical records system was also interrogated for the following information pertaining to the study period: all horses admitted and discharged on each day, and the total amounts of all systemic antimicrobials prescribed on each day. From these data, the total number of horses hospitalised and the total number of defined daily doses (DDD) prescribed for all antimicrobials were determined for each day of the study period. DDD per day was calculated as the daily total amount of an antimicrobial prescribed in the hospital divided by the total daily dose required for an

Table 2 Categorical variables considered for inclusion in the final multivariable model, with the number and percentage of samples in each category. Variable

Categories

Number of samples (%)

Day of hospitalisation

Day 0 Day 2 Day 4 Day 6 Day 7 or later

103 (22.5) 103 (22.5) 99 (21.7) 71 (15.5) 81 (17.7)

Sex

Female entire Male entire Male neutered

150 (32.8) 25 (5.5) 282 (61.7)

Admission condition

Musculoskeletal Medical Gastrointestinal (medical) Gastrointestinal (surgical) Soft tissue surgery

140 (30.6) 120 (26.3) 162 (35.4) 16 (3.5) 19 (4.2)

Hospital yard on sampling day

Medical yard

163 (35.7)

Orthopaedic yard Overflow yard Colic/intensive care Pony yard Mixed yard Temporary yard

22 (4.8) 44 (9.6) 39 (8.5) 46 (10.1) 132 (28.9) 11 (2.4)

average (500 kg) horse for its main therapeutic indication; adapted from the standard World Health Organisation calculation used for human drugs (WHO, 2007; Kuster et al., 2008). The daily levels of these exposures for 0–24 h and 24–48 h prior to each sample were determined. For a sample collected on admission (day 0), exposure was classified as zero as the horse had not yet been exposed to the hospital environment. 2.5. Sample processing Samples were processed immediately or after overnight refrigeration; 2.5 g of faeces was added to a 10 ml volume of brain–heart infusion broth containing 5% glycerol and mixed using a stomacher (Stomacher 80, Seward). The surTable 3 Continuous variables considered for inclusion in the final multivariable models, with the median values and interquartile range for each variable. Variable Antimicrobial doses prescribed in hospital 0–24 h prior to sample 24–48 h prior to sample Penicillin doses prescribed in hospital 0–24 h prior to sample 24–48 h prior to sample Cotrimoxazole doses prescribed in hospital 0–24 h prior to sample 24–48 h prior to sample Gentamicin doses prescribed in hospital 0–24 h prior to sample 24–48 h prior to sample Total number of horses hospitalised 0–24 h prior to sample 24–48 h prior to sample Age of horse Number of horses on home yard

Median value (IQR) 6.6 DDD (0, 29.6) 8.7 DDD (0, 35.8) 0.8 DDD (0, 8.2) 1.5 DDD (0, 8.2) 0 DDD (0, 11) 0 DDD (0, 13) 0 DDD (0, 1.9) 0 DDD (0, 2.1) 23 horses (8, 30) 22 horses (8, 30) 6 years (9, 12) 15 horses (8, 35)

IQR = interquartile range; DDD = defined daily doses (see Section 2).

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

face of one eosin methylene blue agar (EMBA) plate and one MacConkey agar plate was inoculated for confluent bacterial growth and seven antimicrobial impregnated discs (MAST Ltd., UK) were applied to the surface of both plates, in accordance with the direct plating method of Bartoloni et al. (1998, 2006). The antimicrobials used and their potencies were: ampicillin (10 ␮g), co-amoxiclav (30 ␮g) ciprofloxacin (1 ␮g), gentamicin (10 ␮g), nalidixic acid (30 ␮g), tetracycline (30 ␮g) and trimethoprim (2.5 ␮g). The faecal suspension was streaked onto three further EMBA plates; one plate containing the cephalosporin cefotaxime (1 ␮g/ml), one containing ceftazidime (1 ␮g/ml) to select for ESBL-producing E. coli, and one with no antimicrobials incorporated. After overnight incubation at 37 ◦ C, bacterial growth morphologically consistent with E. coli was subcultured for further testing. From the MacConkey and EMBA antimicrobial disc plates, a loopful of bacterial colonies (or at least two if individual colonies were present) growing either adjacent to the antimicrobial disc or inside the general inhibition zone for each disc were selected. At least one colony was taken from each of the EMBA plates with cefotaxime and ceftazidime (if bacterial growth had occurred), and three colonies were randomly selected from the EMBA plate with no antimicrobial. Samples with no growth on the plates containing cefotaxime and ceftazidime were enriched in buffered peptone water overnight at 37 ◦ C, before being plated onto the same selective medium. 2.6. Antimicrobial sensitivity testing Isolates were prepared for antimicrobial sensitivity testing in accordance with British Society for Antimicrobial Chemotherapy (BSAC) guidelines. The surface of one IsoSensitest agar plate was inoculated for semi-confluent bacterial growth and the same seven antimicrobial discs detailed previously were applied. After overnight incubation at 37 ◦ C, the diameter in millimetres of the zones of inhibition around each of the antimicrobial discs was recorded and categorised as resistant or sensitive in accordance with BSAC recommendations. For the antimicrobials nalidixic acid, gentamicin and ciprofloxacin, if the diameter of the zone of inhibition was close to the cut-off value for determining resistance (within ±5 mm), then the results were confirmed by determining the minimum inhibitory concentration (MIC) of the antimicrobial for the isolates concerned. MIC testing was conducted as described by BSAC guidelines (BSAC, 2007). Suspected ESBL-producing isolates taken from the EMBA plates containing cefotaxime or ceftazidime were subjected to the paired disc diffusion test (MAST Ltd., UK) in accordance with the method of M’Zali et al. (2000). An Iso-Sensitest agar plate was inoculated for confluent bacterial growth and three pairs of antimicrobial discs were applied to its surface. The antimicrobial pairs used and their potencies were: cefpodoxime (30 ␮g) and cefpodoxime/clavulanic acid (30/10 ␮g), ceftazidime (30 ␮g) and ceftazidime/clavulanic acid (30/10 ␮g), cefotaxime (30 ␮g) and cefotaxime/clavulanic acid (30/10 ␮g). After overnight incubation at 37 ◦ C, the diameter in millimetres of the zones

137

of inhibition for each pair of the antimicrobial discs was recorded. An increase in zone size of 5 mm or more in the presence of clavulanic acid for any or all of the antimicrobial pairs indicates confirmation of ESBL production. E. coli ATCC 25922 was used as a reference strain for quality control in all susceptibility testing. All culture media was obtained from Lab MTM (Bury, UK). 2.7. Identification of E. coli All isolates with unique antimicrobial resistance phenotypes were confirmed as E. coli if Gram stain negative and by the following series of biochemical tests results: catalase production, non-production of oxidase enzyme, lactose fermentation, indole production, and inability to utilise citrate. Isolates with a biochemical profile consistent with E. coli were confirmed by polymerase chain reaction (PCR) assay for the E. coli specific uidA gene (McDaniels et al., 1996). 2.8. Statistical analysis All collected information and antimicrobial resistance data were entered into a spreadsheet program (Microsoft Excel 2007, Microsoft Corporation) and the dataset was reviewed and checked for coding of all variables. Independent (risk factor) variables were derived from information obtained from the owner questionnaire, clinical records and hospital records. The majority of variables were dichotomous in nature, although a small number represented categorical or continuous variables (detailed in Tables 1–3). Faecal samples were considered the level one unit of interest, the binary outcome for each sample was the presence or absence of an E. coli isolate with resistance to one of seven antimicrobials. Resistance to each of the seven antimicrobials was considered as a separate outcome. Additionally, the presence in a sample of an E. coli with multidrug resistance (to three or more antimicrobial classes) or with ESBL-mediated resistance was considered as two further outcomes. Analysis was restricted to antimicrobial agents for which resistant E. coli prevalence was ≥10% of samples. Due to repeated measures, data were clustered within horses (level two units); therefore factors affecting the occurrence of antimicrobial-resistant E. coli were examined using separate multivariable, multilevel models with a binomial distribution and logit link function. Within-horse clustering was accounted for as a random intercept in all models. In order to make allowance for the autocorrelation of repeated measures data in a longitudinal study, an additional variable was created (resistance in previous sample), defined as the presence or absence of the resistance outcome in the preceding faecal sample. For continuous variables with P-value < 0.25 (doses of each antimicrobial prescribed in the previous 0–24 and 24–48 h, and number of horses in the hospital in the previous 0–24 and 24–48 h), the functional form of the variable with respect to each outcome was assessed using generalised additive models (GAM). The GAM models were fitted using cubic spline smoothers in the S-Plus software

138

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

package (S-plus 2000, Mathsoft Inc.) with inclusion of the categorical variable of day of sample. The functional forms of the relationships were then used to inform the polynomial fits (of centred data) in the multivariable logistic regression models, which were then tested for significance. All variables that showed some association with the presence of resistant E. coli on univariable analysis (a P-value < 0.25) were considered for incorporation into a final multivariable model for that outcome. For any pair of variables with a correlation coefficient of ≥0.70, only the variable with the smallest P-value was considered for further analysis. The final models were constructed by a manual backwards stepwise procedure where variables with a Wald P-value < 0.05 were retained in the model. The resistance in previous sample variable was retained for all final models. First order interaction terms were tested for biologically plausible variables remaining in the final models. Data were analysed using the MLwiN statistical software package (MLwiN Version 2.1 Centre for Multilevel Modelling, University of Bristol). Initial univariable and multivariable calculations were performed using penalised quasi-likelihood estimates (2nd order PQL). The final models were fitted using Monte Carlo Markov Chain utilising Metropolis Hasting sampling with diffuse priors, a burn in period of 10,000 iterations and a run of 500,000 iterations. The number of iterations required was determined by examining the MCMC diagnostics, including the Brooks–Draper and Rafferty–Lewis Nhat statistics (Browne, 2009). The true prevalence (PT ) of resistance was estimated (i.e. adjusting for the clustered nature of repeated sampling) using the formula below, by incorporation of the constant parameter estimate (ˇ0 ) derived from the random intercept-only two-level models constructed for each of the outcomes considered: PT =

eˇ0 1 + eˇ0

95% confidence intervals for all adjusted prevalence estimates were constructed by examination of the standard errors of ˇ0 of the intercept-only model parameters. When both zone diameter and MIC values were available, isolates were separately categorised as sensitive or resistant in accordance with BSAC recommendations for both techniques. In order to assess the level of agreement between the two techniques, the results were entered into the SPSS software package (SPSS 16.0 for Windows, SPSS Inc., Chicago, IL) and cross-tabulated, with the corresponding kappa statistic calculated.

Table 4 Number and prevalence of samples with resistance to all antimicrobials tested, multidrug resistance and ESBL-mediated resistance. The prevalence figures have been adjusted for the clustered nature of the data using intercept-only multilevel models constructed for each outcome. Antimicrobial resistance

Ampicillin Co-amoxiclav Ciprofloxacin Gentamicin Nalidixic acid Tetracycline Trimethoprim Multidrug ESBL mediated

Samples (n = 457) No of resistant samples

% (95% CI)

232 36 121 164 147 226 297 225 131

50.1 (43, 57.2) 6.9 (4.6, 10.1) 21.7 (16.6, 27.7) 32.3 (26.1, 39.3) 28.9 (23.4, 35.1) 50.9 (43.6, 58.1) 64.6 (58.6, 70.2) 47.7 (40.7, 54.7) 27.3 (22.1, 33.2)

95% CI = 95% confidence intervals.

At least one antimicrobial-resistant E. coli isolate was recovered from 309 of the faecal samples (70.2% after adjustment for clustering, 95% confidence interval (CI) 63.0, 76.5), from a total of 94 horses (91.3%, 95% CI 85.8, 96.7). The remaining nine horses had no antimicrobial-resistant E. coli recovered from any of their samples. The total number of non-duplicate resistant E. coli isolated from all samples was 694. The overall prevalence of resistant E. coli isolates in faecal samples was over 25% for all the antimicrobials examined except co-amoxiclav (8.1%), with a high prevalence of ampicillin, tetracycline and trimethoprim resistance. Multidrug-resistant E. coli (resistant to three or more antimicrobial classes) were identified in 47.7% of samples. ESBL-producing E. coli were recovered from 131 samples (27.3%, 95% CI 22.1, 33.2), from a total of 55 horses (53.4%, 95% CI 43.8, 63.0). The sample prevalence of resistance to each antimicrobial, multidrug resistance and presence of an ESBL producing E. coli are detailed in Table 4. A total of 42 distinct antimicrobial resistance phenotypes were identified among the 694 resistant E. coli isolates. Resistance only to trimethoprim was the most frequent phenotype identified (21.3%), but several phenotypes with extensive multidrug resistance were also very common. The eleven most frequently identified resistance phenotypes are detailed in Table 5. Prevalence figures for each day of admission are shown in Fig. 1. Although the prevalence levels for resistance to each antimicrobial differed, a distinct pattern of antimicrobial resistance during hospitalisation was repeatedly shown for all drugs. On admission there was consistently low prevalence of resistance, with a sharp increase observed on days 2 and 4, and then a slight decrease for samples collected on day 6 or later.

3. Results

3.2. Agreement between disc diffusion and MIC data

3.1. Descriptive statistics

There was a high level of agreement between the disc diffusion and MIC data for the categorisation of isolates as resistant or sensitive, with kappa values for all three antimicrobials considered being above 0.81 (Landis and Koch, 1977), as detailed in Table 5. However, to verify this finding, the final models for ciprofloxacin, gentamicin and nalidixic acid resistance were repeated; firstly with any discordant

In total 110 horses were enrolled in the study, seven were subsequently excluded as they remained hospitalised for only one night in the PLEH. A total of 457 faecal samples were collected from the remaining 103 horses; the median duration of hospitalisation was 6 days (IQR 4–8 days).

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

139

Fig. 1. Sample prevalence of antimicrobial resistance, multidrug resistance and ESBL-mediated resistance for each day of sampling in 457 faecal samples from 110 horses admitted to the PLEH. AMP = ampicillin; AMOX CL = co-amoxiclav; CIP = ciprofloxacin; GM = gentamicin; NAL = nalidixic acid; TET = tetracycline; TMP = trimethoprim; MDR = multidrug resistant, ESBL = ESBL mediated resistance.

Table 5 The eleven most frequent antimicrobial-resistant phenotypes identified among the 694 antimicrobial-resistant E. coli isolates. Resistance phenotype

Number of isolates

% (95% CI)

TM AMP, GM, TET, TM AMP, CIP, GM, NA, TET, TM AMP, CIP, NA, TET, TM AMP, TM AMP, GM, TM TET, TM AMP, TET, TM AMP, GM, NA, TET, TM CIP, NA, TET, TM TET

148 116 81 41 37 34 32 31 22 14 14

21.3 (18.3, 24.4) 16.7 (13.9, 19.5) 11.7 (9.3, 14.1) 5.9 (4.2, 7.7) 5.3 (3.7, 7.0) 4.9 (3.3, 6.5) 4.6 (3.1, 6.2) 4.5 (2.9, 6.0) 3.2 (1.9, 4.5) 2.0 (1.0, 3.1) 2.0 (1.0, 3.1)

95% CI = 95% confidence intervals; AMP = ampicillin; CIP = ciprofloxacin; GM = gentamicin; NAL = nalidixic acid; TET = tetracycline; TM = trimethoprim.

results omitted, and then using the resistance status as categorised by MIC values. In neither case were there any substantial changes in the models for the three outcomes. For consistency, all subsequent models presented are based on resistance outcomes as determined by disc diffusion testing (Table 6).

3.3. GAM results The generalised additive models constructed showed that total number of antimicrobial doses prescribed (DDD) in the previous 0–24 and 24–48 h, and the total number of horses hospitalised in the previous 0–24 and 24–48 h demonstrated a significant non-linear relationship (p < 0.05) with some of the outcomes examined. Appropriate polynomial terms were considered during the multivariable analysis for the relevant outcomes, however these variables did not remain in the final models.

Table 6 Cross tabulation of the results of 482 E. coli isolates classified as resistant or sensitive to the antimicrobials ciprofloxacin, gentamicin and nalidixic acid by both MIC determination and disc diffusion testing. Antimicrobial

Ciprofloxacin Result by disc diffusion

Gentamicin Result by disc diffusion

Nalidixic acid Result by disc diffusion

Result by MIC Sensitive

Resistant

Total

Sensitive Resistant

297 2

17 166

314 168

Total Kappa = 0.915, p < 0.0001

299

183

482

Sensitive Resistant

113 26

10 333

123 359

Total Kappa = 0.812, p < 0.0001

139

343

482

Sensitive Resistant

235 20

20 207

255 227

Total Kappa = 0.833, p < 0.0001

255

227

482

1.9 (1.1)

– 2.15, 20.1 4.22, 64.8 2.31, 40.7 1.36, 19.9 1.1, 6.08 1.73, 7.9 1.0, 1.03 (Ref) 6.57 16.5 9.69 5.2 2.59 3.63 1.02

2.6 (1.3)

– 0.002 <0.001 0.003 0.03 0.06 <0.001 0.04 – 1.81, 14.9 3.55, 55.9 1.99, 34.9 1.16, 15.7 0.96, 6.64 1.67, 7.48 1.0, 1.03 (Ref) 5.18 14.1 8.34 4.27 2.49 3.54 1.02

2.2 (1.0)

– 0.02 <0.001 0.005 0.02 0.05 <0.001 0.006 – 1.21, 8.21 2.90, 32.3 1.69, 22.8 1.21, 12.6 0.98, 5.86 1.77, 7.49 1.01, 1.04

2.0 (0.9)

95% CI = 95% credible intervals; P values are from the Wald chi-squared test.

(Ref) 3.15 9.66 6.21 3.90 2.40 3.64 1.02 – <0.001 <0.001 0.004 0.04 0.1 <0.001 – – 1.7, 9.45 3.22, 27.0 1.72, 17.2 1.04, 9.03 0.86, 4.44 1.97, 8.82 – (Ref) 4.01 9.33 4.45 3.06 1.96 4.17 –

95% CI OR 95% CI

P

95% CI OR OR

P

OR

95% CI

P

Gentamicin resistance Multidrug resistance Ampicillin resistance Tetracycline resistance

The final multivariable, multilevel logistic regression models are shown in Tables 7 and 8. The prevalence of co-amoxiclav resistance was too low for construction of a model. For all outcomes except trimethoprim resistance, the day the sample was obtained was significant, with increased risk of resistance for samples taken on day 2 or later. For all outcomes except ESBL-mediated resistance, having had antimicrobial treatment in the seven days prior to a sample also significantly increased the risk of resistance. The amount prescribed (DDD) of cotrimoxazole in the hospital in the 24–48 h prior to a sample was a significant risk factor for all outcomes except ciprofloxacin, tetracycline, and ESBL-mediated resistance. Admission reason or hospital yard on sampling day were significant for nalidixic acid, ESBL-mediated, trimethoprim and ciprofloxacin resistance. Patients admitted for gastrointestinal surgery were at increased risk for nalidixic acid and particularly ciprofloxacin resistance. Horses in the orthopaedic yard were at increased risk for trimethoprim and ciprofloxacin resistant E. coli, and those on the mixed yard were at increased risk for ESBL-mediated resistant E. coli. As expected, there was significant clustering of resistance outcomes within horses; however, random slope effects were not significant, suggesting that there was no major difference in covariate effects in different horses. No significant interactions (Wald P-value < 0.05) were found for the variables remaining in the final models. The MCMC diagnostics performed (Browne, 2009) indicated that the fits were smooth and regular, with adequate mixing of chains for all fixed effect variables. Sufficient iterations were performed to give certainty about the estimates for the model parameters.

Variable

3.5. Multivariable analysis

Table 7 Results of multivariable multilevel analysis for the outcomes of tetracycline, ampicillin, multidrug and gentamicin resistance in 457 faecal samples from 110 horses admitted to the PLEH.

Univariable analysis found many of the variables described in Tables 1–3 to be significantly associated with the outcomes considered (data not shown). Individual horse antimicrobial treatment within the previous 24 h, 48 h and seven days were all significantly correlated with each other, both when considering individual antimicrobial types or antimicrobial treatment overall. For all outcomes considered, overall antimicrobial treatment in the previous seven days was determined as the best fit for the models. Similarly, antimicrobial defined daily dosages prescribed in the hospital in the previous 0–24 and 24–48 h were significantly correlated, as were undergoing a surgical procedure in the previous 24 and 48 h. The variable pertaining to exposure in the previous 48 h proved to be the best fit for the model for all of these variables. Variables including number of horses in the hospital in the 0–24 and 24–48 h before sampling, and undergoing surgery in the 24 and 48 h before sampling, were highly significant on univariable analysis (P < 0.0001) but failed to remain in the final models. In addition age, sex and previous hospitalisation were only found to be significantly associated with outcomes considered on univariable analysis.

P

3.4. Univariable analysis

– <0.001 <0.001 0.002 0.02 0.03 <0.001 0.03

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

Day of sample Day 0 Day 2 Day 4 Day 6 Day 7+ Resistant at previous sample Antimicrobial previous 7 days Cotrimoxazole doses prescribed (DDD) in hospital previous 48 h Variance (standard error)

140

Table 8 Results of multivariable multilevel analysis for the outcomes of nalidixic acid, ciprofloxacin, trimethoprim and ESBL-mediated resistance in 457 faecal samples from 110 horses admitted to the PLEH. Variable

Variance (standard error)

OR

95% CI

(Ref) 4.35 9.0 6.47 4.6 3.69 4.57 1.01

– 1.37, 13.8 2.8, 28.9 1.86, 22.5 1.35, 15.7 1.84, 7.41 2.28, 9.18 1.0, 1.03

(Ref) 1.83 0.69 7.06 1.72 – – – – – – –

Ciprofloxacin resistance P

Trimethoprim resistance

OR

95% CI

– 0.01 <0.001 <0.001 0.01 <0.001 <0.001 0.04

(Ref) 6.25 30.6 17.9 13.1 1.11 12.8 –

– 1.6, 24.3 7.0, 134 3.63, 88.8 2.82, 61.2 0.42, 2.92 3.84, 42.5 –

– 0.008 <0.001 <0.001 0.001 0.8 <0.001 –

– 0.82, 4.07 0.34, 1.42 1.29, 38.7 0.39, 7.58

– 0.1 0.3 0.02 0.5

(Ref) 3.69 1.18 114 10.0

– 0.94, 14.4 0.34, 4.18 4.9, 2590 0.59, 169

– – – – – – –

– – – – – – –

(Ref) 9.71 0.57 0.21 2.73 1.11 4.06

– 1.41, 66.9 0.10, 3.17 0.03, 1.74 0.58, 12.9 0.35, 3.55 0.49, 33.2

0.3 (0.4)

2.7 (1.46)

P

ESBL-mediated resistance

OR

95% CI

(Ref) 0.89 2.65 2.51 1.42 2.83 4.36 1.02

– 0.39, 2.03 1.0, 6.98 0.85, 7.4 0.52, 3.85 1.25, 6.39 2.04, 9.33 1.01, 1.04

– 0.06 0.8 0.003 0.11

– – – – –

– – – – –

– – – – –

– – – – –

– – – – –

– 0.02 0.5 0.2 0.2 0.9 0.2

(Ref) 4.94 0.75 2.45 1.96 1.22 4.45

– 1.13, 21.5 0.25, 2.21 0.67, 9.01 0.64, 5.98 0.56, 2.67 0.49, 39.8

– 0.03 0.6 0.2 0.2 0.6 0.2

(Ref) 7.71 2.59 1.88 4.73 11.6 4.23

– 0.78, 76.0 0.39, 17.0 0.26, 13.3 0.79, 28.3 2.75, 48.8 0.39, 46.5

4.2 (1.7)

P – 0.7 0.05 0.1 0.08 0.01 <0.001 0.01

OR (Ref) 27.1 97.8 94.6 36.2 1.25 – –

95% CI – 5.63, 131 17.2, 555 15.0, 595 6.42, 205 0.48, 3.25 – –

P – <0.001 <0.001 <0.001 <0.001 0.6 – –

– – – – – – 0.08 0.3 0.6 0.09 <0.001 0.2

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

Day of sample Day 0 Day 2 Day 4 Day 6 Day 7+ Resistant at previous sample Antimicrobial previous 7 days Cotrimoxazole doses prescribed (DDD) in hospital previous 48 h Admission condition Soft tissue surgery Musculoskeletal Medical Gastrointestinal (surg) Gastrointestinal (med) Hospital yard Medical yard Orthopaedic yard Overflow yard Colic/intensive care Pony yard Mixed yard Temporary yard

Nalidixic acid resistance

4.1 (1.9)

95% CI = 95% credible intervals; P values are from the Wald chi-squared test.

141

142

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

4. Discussion The overall prevalence of antimicrobial-resistant E. coli in horse faecal samples identified in this study was high (70.2%), but similar studies have shown a comparable prevalence. A study from Ireland identified multidrugresistant E. coli in 65% of hospitalised horses (Bryan et al., 2010), whilst a North American equine hospital reported resistance in 62% of E. coli isolated from horses (Dunowska et al., 2006). The prevalence of resistant E. coli on hospital admission was lower at 42.7% of samples, and within the range of other estimates of prevalence of 18–75% reported for horses in the community (Bucknell et al., 1997; Dunowska et al., 2006), although the population of horses visiting a referral equine hospital may not be representative of the general equine population. The prevalence of ESBL-producing bacteria identified in this study was higher than expected, with 53.4% of horses having at least one ESBL-positive E. coli isolated during hospitalisation. ESBL-producing E. coli have been identified in diagnostic submissions of pathogenic isolates from horses in the Netherlands; all were of the CTX-M type (Vo et al., 2007). A prevalence of 19% of horses for ESBLproducing E. coli was reported in a Czech equine hospital that examined 27 animals sampled on a single occasion; again they were positive for the CTX-M type (Dolejska et al., 2008). ESBL-producing E. coli from this study were identified as carrying ctx-m, shv and tem ESBL-associated resistance genes (data not shown). Risk factors identified for the antimicrobial-resistances were similar in many cases. This is expected, as there was correlation of resistance outcomes in many cases, demonstrated by the limited number of resistance phenotypes identified among the 694 resistant isolates. In fact, 82% of the resistant E. coli isolates demonstrated one of just 10 resistance phenotypes, even though total number possible for seven antimicrobials is 127 (27 − 1). There may be multiple reasons for this. In the case of drugs of the same antimicrobial class, for example nalidixic acid and ciprofloxacin (both quinolones), there is likely to be some cross-resistance. Even with drugs of different classes, if the involved resistances are linked in some way (in bacterial strains with multiple resistances or via genes carried on the same mobile genetic element), selection of one resistance type concurrently selects for another. The identified increase in the prevalence of antimicrobial-resistant E. coli after admission to the hospital could occur via a number of means. Possibilities include: acquisition of resistant E. coli from other hospitalised horses or the hospital environment; acquisition of genetic resistant determinants from other bacteria from similar sources; an increase in resistant E. coli already present (but previously undetectable) in the gastrointestinal flora or transfer of resistance determinants from other members of the gut flora. It is plausible that all of these possibilities may be involved in a hospital setting. The time taken for acquired resistant determinants to distribute throughout a bacterial population is unknown, but the time taken for an acquired bacterial population to become established within the equine gastrointestinal tract has been better characterised. Oral inoculation of horses with

a Lactobacillus species resulted in detectable presence of the bacteria in faeces within 24 h in two studies (Weese et al., 2003, 2004). The remaining possibilities suggest that some factor results in the selection of resistant bacterial strains from the normal enteric population; either directly resulting in an increase in numbers of resistant E. coli or allowing susceptible E. coli to acquire resistance determinants from increased numbers of other resistant bacteria. Antimicrobial use would represent one obvious selection pressure, but other aspects of hospitalisation could also be responsible. For example, the stress resulting from transportation, hospitalisation or undergoing a surgical procedure could possibly increase the shedding of resistant enteric bacteria. Increased shedding of antimicrobial-resistant E. coli has been demonstrated in cold stressed pigs (Moro et al., 1998). The recruitment of a convenience sample of horses rather than a fully randomised sample may have introduced sampling bias. However, this convenience sampling was necessitated by limitations of laboratory capability and time constraints caused by a concurrent cross-sectional study being undertaken by the same investigator; therefore bias towards particular horses or days should be limited. Recruitment days (n = 69) were spread throughout the study period and the number of horses recruited represents approximately 20% of all eligible horses admitted to the hospital. Fewer horses were recruited than the sample size calculations suggested, meaning the study may have insufficient statistical power to detect some smaller effects. Additionally, the statistical power of the study was limited in its ability to detect odds ratios of less than three. However, it should be noted that sample size calculations were based on obtaining single samples, obtaining multiple samples per horse would have increased the power. The two most important risk factors, in terms of the number of resistance outcomes influenced and their corresponding odds ratios, were day of hospitalisation and treatment with antimicrobials in the seven days prior to a sample. An association with antimicrobial treatment agrees with another study that included a group of hospitalised horses (Dunowska et al., 2006), which identified aminoglycoside, cephalosporin and cotrimoxazole administration as risk factors for antimicrobial-resistance in E. coli. In the present study, although treatment with an individual antimicrobial type was often significant on univariable analysis, the overall variable of treatment with any antimicrobial within the previous seven days represented the best fit for inclusion in the multivariable models. This may be because most of the antimicrobial types were only used on a relatively small number of horses. Resistance was not restricted to samples for which there was antimicrobial treatment in the previous seven days and clearly the increased prevalence of resistance is not solely the result of antimicrobial treatment. Dunowska et al. identified hospitalisation as an independent risk factor (in addition to antimicrobial treatment) on comparison of resistant E. coli prevalence in groups of hospitalised and non-hospitalised horses (Dunowska et al., 2006). An increase in the number of resistant isolates and the extent of their resistance was identified on sampling commensal skin staphylo-

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

cocci from horses at admission and during hospitalisation (Schnellmann et al., 2006). This was increased further in horses undergoing surgery or antimicrobial treatment. Few studies in animals have documented the influence of hospital level factors such as number of hospitalised patients or total antimicrobial usage on the presence of resistance, although this is more common for human studies (Kuster et al., 2008; Pakyz et al., 2008). Although there was correlation between the number of horses hospitalised and the number of antimicrobial doses prescribed in the hospital, this was insufficient for either of these variables to be excluded from the multivariable models. The number of horses in the hospital was not a significant risk factor for any outcome. However, for most outcomes the number of doses of cotrimoxazole prescribed in the hospital was significant. Antimicrobial treatment of individual horses may result in increased shedding of resistant E. coli in faeces, translating to increased numbers of such bacteria present in the hospital environment. This increases other horses’ exposure to these bacteria (or their resistance determinants), increasing the likelihood of transmission. In this way, hospital-level antimicrobial use may impact on the individual horse, even when the horse concerned is not receiving any antimicrobial treatment. The association of cotrimoxazole use with trimethoprim resistance might be expected, as cotrimoxazole consists of trimethoprim in combination with sulphamethoxazole. The influence of cotrimoxazole on other resistance outcomes likely results from linked resistance traits, or that the defined daily doses of cotrimoxazole were correlated with total antimicrobial daily dosage. Although cotrimoxazole represented the best fit for the models, to an extent it may reflect the general level of antimicrobial use in the hospital. Decreases in the number of defined daily doses (DDDs) of some antimicrobials (third-generation cephalosporins, quinolones and vancomycin) have been associated with a decreased occurrence of vancomycin resistant enterococci in some human hospital studies (Quale et al., 1996; Kolar et al., 2006), although another study found that decreased DDDs for a number of antimicrobials did not affect the antimicrobial sensitivities recorded for several pathogens (Cook et al., 2004). The decrease in resistant E. coli prevalence noted from day six onward is interesting (reflected in the reduced odds ratios seen for samples from day six and day seven or later). However, it should be noted that confidence intervals for all of these values overlap and so this may not represent a true effect, additionally the confidence intervals for these later days widen, as fewer horses remain hospitalised. Regardless of the exact mechanism, the increased prevalence of resistant E. coli associated with hospitalisation may signify some form of disturbance of the horses’ normal gastrointestinal flora. The subsequent decrease, if genuine, could represent the re-establishment of a more normal bacterial population, particularly if the cause of the disturbance occurs early during hospitalisation or the selection pressure is not maintained. Implicit within this idea is the concept that carriage of a resistance mechanism can result in some form of fitness cost to the organism involved, resulting in a disadvantage when the selection pressure is no longer dominant (Gillespie and McHugh,

143

1997). Conflicting evidence exists concerning any fitness cost associated with antimicrobial-resistance in E. coli. Some experimental studies have demonstrated a high fitness cost (Gualco et al., 2007), whilst others have either shown a minimal fitness cost (Enne et al., 2005) or a minimal fitness cost that can be compensated for over time (Schrag et al., 1997). Reason for admission was significant only for nalidixic acid and ciprofloxacin resistance; for both outcomes admission for a surgical gastrointestinal condition represented increased risk. As discussed previously, the outcomes for these quinolone antimicrobials were the most correlated. It is unclear whether the condition itself or some aspect of subsequent treatment is the important factor. Non-surgical gastrointestinal disease did not represent increased risk, and undergoing surgery was not a significant risk factor for any outcome on multivariable analysis. Surgery for gastrointestinal disease was not specifically evaluated as a separate risk factor. It is difficult to be certain whether hospital yard represents a direct environmental factor or is merely a reflection of the reason for admission, type of treatment or severity of disease. Although the descriptions given to the yards represent their primary intended use, they are not strictly adhered to and there is considerable overlap and movement of cases. Therefore, little weight should be attached to their designations. As such, the yard a horse is located on is perhaps more likely to represent an environmental factor, and lend further credence to the concept of acquiring resistant bacteria from the hospital environment. One yard (orthopaedic) was shown to be significantly associated with increased risk for ciprofloxacin and trimethoprim resistance, and the mixed-purpose yard (which also houses the largest number of horses) was also significantly associated with ESBL-mediated resistance. Environment sampling undertaken at the time of the study recovered large numbers of multidrug-resistant E. coli from the floors, walls and water troughs of a number of stables from all the yards investigated (A O’Donnell unpublished data). A similar situation was found in a Czech equine hospital study, where multidrug-resistant E. coli were isolated from most water troughs sampled (with ESBL-producing E. coli recovered from 11% of sampled troughs) (Dolejska et al., 2008). The emergence of multidrug-resistant bacteria in animals is a concern for public health as well as veterinary medicine (Barza, 2002; Aarestrup, 2005). Transfer of E. coli from animals to in-contact people is difficult to confirm but household members, including pets, have been shown to share E. coli strains (Murray et al., 2004; Johnson et al., 2008b; Damborg et al., 2009), and a longitudinal study documented repeated transfer of pathogenic E. coli strains between a pet and members of its owner family (Johnson et al., 2008a). Comparable work has yet to be undertaken with horses and their owners, but similar opportunities for bacterial transfer may exist. The antimicrobial-resistant E. coli recovered in this study did not appear to be causing disease and likely represent commensal strains of the organism. However, under certain circumstances such strains may be able to cause disease, either through the acquisition of virulence deter-

144

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145

minants or via inoculation of a non-gastrointestinal site (such as a post-surgical wound). Clinicians should be aware that prolonged hospitalisation and the use of antimicrobials may result in carriage of antimicrobial-resultant E. coli, with implications for the development of nosocomial or hospital acquired infection. Isolates such as the multidrug-resistant E. coli identified in this study could prove refractory to treatment by most of the antimicrobials available for use in equine medicine. Additionally, humans involved with the care of hospitalised horses are frequently in close contact with the animals and their faeces, representing considerable opportunity for the potential transfer of resistant E. coli strains to humans. 5. Conclusions This study has identified a high prevalence of antimicrobial-resistant commensal E. coli isolated from hospitalised horses. Previous antimicrobial treatment and being hospitalised for two days or more increased the odds of resistance to almost all antimicrobial types among the E. coli recovered. Additionally, overall hospital use of cotrimoxazole, location within the hospital and presenting reason for admission were associated with increased odds of resistance to some antimicrobial types. Although the E. coli involved are likely to be commensal rather than pathogenic strains, it may be possible for such strains to cause disease in horses, or to be transferred to in-contact humans, and so action to reduce their prevalence is warranted. Conflict of interest None of the authors has any financial and personal relationships with other people or organisations that could have inappropriately influenced the work presented here. Acknowledgements The authors wish to thank Gill Hutchinson, Ruth Ryvar, Amy Wedley and Thelma Roscoe for technical assistance. This work was supported by the Bransby Home of Rest for Horses (Registered Charity No: 1075601) and the UK Department for Environment, Food and Rural Affairs (Defra). References Aarestrup, F.M., 2005. Veterinary drug usage and antimicrobial resistance in bacteria of animal origin. Basic Clin. Pharmacol. Toxicol. 96, 271–281. Albihn, A., Baverud, V., Magnusson, U., 2003. Uterine microbiology and antimicrobial susceptibility in isolated bacteria from mares with fertility problems. Acta Vet. Scand. 44, 121–129. Anzai, T., Kamada, M., Ike, K., Kanemaru, T., Kumanomido, T., 1987. Drug susceptibity of Escherichia coli isolated from foals with diarrhoea and mares with metritis. Bull. Equine Res. Inst., 42–50. Bartoloni, A., Benedetti, M., Pallecchi, L., Larsson, M., Mantella, A., Strohmeyer, M., Bartalesi, F., Fernandez, C., Guzman, E., Vallejos, Y., Villagran, A.L., Guerra, H., Gotuzzo, E., Paradisi, F., Falkenberg, T., Rossolini, G.M., Kronvall, G., 2006. Evaluation of a rapid screening method for detection of antimicrobial resistance in the commensal microbiota of the gut. Trans. R. Soc. Trop. Med. Hyg. 100, 119–125.

Bartoloni, A., Cutts, F., Leoni, S., Austin, C.C., Mantella, A., Guglielmetti, P., Roselli, M., Salazar, E., Paradisi, F., 1998. Patterns of antimicrobial use and antimicrobial resistance among healthy children in Bolivia. Trop. Med. Int. Health 3, 116–123. Barza, M., 2002. Potential mechanisms of increased disease in humans from antimicrobial resistance in food animals. Clin. Infect. Dis. 34, S123–S125. Berge, A.C.B., Atwill, E.R., Sischo, W.M., 2005. Animal and farm influences on the dynamics of antibiotic resistance in faecal Escherichia coli in young dairy calves. Prev. Vet. Med. 69, 25–38. ˜ Brinas, L., Moreno, M.A., Teshager, T., Zarazaga, M., Sáenz, Y., Porrero, C., Dominguez, L., Torres, T., 2003. beta-Lactamase characterization in Escherichia coli isolates with diminished susceptibility or resistance to extended-spectrum cephalosporins recovered from sick animals in Spain. Microb. Drug Resist. 9, 201–209. Browne, W.J., 2009. MCMC Estimation in MLwiN, v2.13. Centre for Multilevel Modelling. University of Bristol. Bryan, J., Leonard, F.C., Fanning, S., Katz, L.M., Duggan, V., 2010. Antimicrobial resistance in commensal faecal Escherichia coli of hospitalised horses. Irish Vet. J. 63, 373–379. BSAC, 2007. BSAC Methods for Antimicrobial Susceptibility Testing (Version 6.1). BSAC.org.uk. Bucknell, D.G., Gasser, R.B., Irving, A., Whithear, K., 1997. Antimicrobial resistance in Salmonella and Escherichia coli isolated from horses. Aust. Vet. J. 75, 355–356. Cook, P.P., Catrou, P.G., Christie, J.D., Young, P.D., Polk, R.E., 2004. Reduction in broad-spectrum antimicrobial use associated with no improvement in hospital antibiogram. J. Antimicrob. Chemother. 53, 853–859. Damborg, P., Nielsen, S.S., Guardabassi, L., 2009. Escherichia coli shedding patterns in humans and dogs: insights into within-household transmission of phylotypes associated with urinary tract infections. Epidemiol. Infect. 137, 1457–1464. Dolejska, M., Duskova, E., Machkova, G., Kohoutova, L., Cizek, A., 2008. Detection of ESBL-producing Escherichia coli in horses, personel and the environment within a veterinary teaching hospital. In: Conference proceedings; Antimicrobial Resistance in Zoonotic Bacteria and Foodborne Pathogens , Copenhagen, Denmark, pp. 57–58. Dunowska, M., Morley, P.S., Traub-Dargatz, J.L., Hyatt, D.R., Dargatz, D.A., 2006. Impact of hospitalization and antimicrobial drug administration on antimicrobial susceptibility patterns of commensal Escherichia coli isolated from the feces of horses. J. Am. Vet. Med. Assoc. 228, 1909–1917. Enne, V.I., Delsol, A.A., Davis, G.R., Hayward, S.L., Roe, J.M., Bennett, P.M., 2005. Assessment of the fitness impacts on Escherichia coli of acquisition of antibiotic resistance genes encoded by different types of genetic element. J. Antimicrob. Chemother. 56, 544–551. Gillespie, S.H., McHugh, T.D., 1997. The biological cost of antimicrobial resistance. Trends Microbiol. 5, 337–339. Gualco, L., Schito, A.M., Schito, G.C., Marchese, A., 2007. In vitro activity of prulifloxacin against Escherichia coli isolated from urinary tract infections and the biological cost of prulifloxacin resistance. Int. J. Antimicrob. Agents 29, 679–687. Hall, R.M., Collis, C.M., 1995. Mobile gene cassettes and integrons—capture and spread of genes by site-specific recombination. Mol. Microbiol. 15, 593–600. Harihara, H., Barnum, D.A., 1973. Drug resistance among pathogenic enterobacteriaceae from animals in Ontario. Can. J. Publ. Health 64, 69–169. Hart, W.S., Heuzenroeder, M.W., Barton, M.D., 2006. A study of the transfer of tetracycline resistance genes between Escherichia coli in the intestinal tract of a mouse and a chicken model. J. Vet. Med. Ser. B-Infect. Dis. Vet. Publ. Health 53, 333–340. Hirsh, D.C., Kirkham, C., Wilson, W.D., 1993. Characteristics of Escherichia coli isolated from septic foals. Vet. Microbiol. 34, 123–130. Holland, R.E., Schmidt, A., Sriranganathan, N., Grimes, S.D., Wilson, R.A., Brown, C.M., Walker, R.D., 1996. Characterization of Escherichia coli isolated from foals. Vet. Microbiol. 48, 243–255. Hopkins, K.L., Batchelor, M.J., Liebana, E., Deheer-Graham, A.P., Threlfalla, E.J., 2006. Characterisation of CTX-M and AmpC genes in human isolates of Escherichia coli identified between 1995 and 2003 in England and Wales. Int. J. Antimicrob. Agents 28, 180–192. Johnson, J.R., Clabots, C., Kuskowski, M.A., 2008a. Multiple-host sharing. long-term persistence, and virulence of Escherichia coli clones from human and animal household members. J. Clin. Microbiol. 46, 4078–4082. Johnson, J.R., Owens, K., Gajewski, A., Clabots, C., 2008b. Escherichia coli colonization patterns among human household members and pets, with attention to acute urinary tract infection. J. Infect. Dis. 197, 218– 224.

T.W. Maddox et al. / Preventive Veterinary Medicine 100 (2011) 134–145 Karami, N., Martner, A., Enne, V.I., Swerkersson, S., Adlerberth, I., Wold, A.E., 2007. Transfer of an ampicillin resistance gene between two Escherichia coli strains in the bowel microbiota of an infant treated with antibiotics. J. Antimicrob. Chemother. 60, 1142–1145. Kolar, M., Urbanek, K., Vagnerova, I., Koukalova, D., 2006. The influence of antibiotic use on the occurrence of vancomycin-resistant enterococci. J. Clin. Pharm. Ther. 31, 67–72. Kuster, S.P., Ruef, C., Ledergerber, B., Hintermann, A., Deplazes, C., Neuber, L., Weber, R., 2008. Quantitative antibiotic use in hospitals: comparison of measurements. Literature review, and recommendations for a standard of reporting. Infection 36, 549–559. Landis, J.R., Koch, G.G., 1977. Measurement of observer agreement for categorical data. Biometrics 33, 159–174. Lanz, R., Kuhnert, P., Boerlin, P., 2003. Antimicrobial resistance and resistance gene determinants in clinical Escherichia coli from different animal species in Switzerland. Vet. Microbiol. 91, 73–84. Lavoie, J.P., Couture, L., Higgins, R., Laverty, S., 1991. Aerobic bacterial isolates in horses in a university hospital, 1986–1988. Can. Vet. J. 32, 292–294. Liebana, E., Batchelor, M., Hopkins, K.L., Clifton-Hadley, F.A., Teale, C.J., Foster, A., Barker, L., Threlfall, E.J., Davies, R.H., 2006. Longitudinal farm study of extended-spectrum beta-lactamase-mediated resistance. J. Clin. Microbiol. 44, 1630–1634. Livermore, D.M., 2003. Bacterial resistance: origins, epidemiology, and impact. Clin. Infect. Dis. 36, S11–S23. M’Zali, F.H., Chanawong, A., Kerr, K.G., Birkenhead, D., Hawkey, P.M., 2000. Detection of extended-spectrum beta-lactamases in members of the family Enterobacteriaceae: comparison of the MAST DD test, the double disc and the Etest ESBL. J. Antimicrob. Chemother. 45, 881–885. McDaniels, A.E., Rice, E.W., Reyes, A.L., Johnson, C.H., Haugland, R.A., Stelma, G.N., 1996. Confirmational identification of Escherichia coli, a comparison of genotypic and phenotypic assays for glutamate decarboxylase and beta-d-glucuronidase. Appl. Environ. Microbiol. 62, 3350–3354. McKay, K.A., Ruhnke, H.L., Barnum, D.A., 1965. The results of sensitivity tests on animal pathogens conducted over the period 1956–1963. Can. Vet. J. 6, 103–111. Meng, J.H., Zhao, S.H., Doyle, M.P., 1998. Virulence genes of Shiga toxinproducing Escherichia coli isolated from food, animals and humans. Int. J. Food Microbiol. 45, 229–235. Moro, M.H., Beran, G.W., Hoffman, L.J., Griffith, R.W., 1998. Effects of cold stress on the antimicrobial drig resistance of Escherichia coli of the intestinal flora of swine. Lett. Appl. Microbiol. 27, 251–254. Murray, A.C., Kuskowski, M.A., Johnson, J.R., 2004. Virulence factors predict Escherichia coli colonization patterns among human and animal household members. Ann. Intern. Med. 140, 848–849. Nataro, J.P., Kaper, J.B., 1998. Diarrheagenic Escherichia coli. Clin. Microbiol. Rev. 11, 403–1403. Ogeer-Gyles, J.S., Mathews, K.A., Boerlin, P., 2006. Nosocomial infections and antimicrobial resistance in critical care medicine. J. Vet. Emerg. Crit. Care 16, 1–18. Pakyz, A., Powell, J.P., Harpe, S.E., Johnson, C., Edmond, M., Polk, R.E., 2008. Diversity of antimicrobial use and resistance in 42 hospitals in the United States. Pharmacotherapy 28, 906–912. Paladino, J.A., Sunderlin, J.L., Price, C.S., Schentag, J.J., 2002. Economic consequences of antimicrobial resistance. Surg. Infect. (Larchmt) 3, 259–267. Payne, D.J., Marriott, M.S., Amyes, S.G.B., 1989. Mutants of the Tem-1 beta-lactamase conferring resistance to ceftazidime. J. Antimicrob. Chemother. 24, 103–110.

145

Quale, J., Landman, D., Saurina, G., Atwood, E., DiTore, V., Patel, K., 1996. Manipulation of a hospital antimicrobial formulary to control an outbreak of vancomycin-resistant enterococci. Clin. Infect. Dis. 23, 1020–1025. Roupas, A., Pitton, J.S., 1974. R factor-mediated and chromosomal resistance to ampicillin in Escherichia-coli. Antimicrob. Agents Chemother. 5, 186–191. Russell, C.M., Axon, J.E., Blishen, A., Begg, A.P., 2008. Blood culture isolates and antimicrobial sensitivities from 427 critically ill neonatal, foals. Aust. Vet. J. 86, 266–271. Saenz, Y., Brinas, L., Dominguez, E., Ruiz, J., Zarazaga, M., Vila, J., Torres, C., 2004. Mechanisms of resistance in multiple-antibiotic-resistant Escherichia coli strains of human, animal, and food origins. Antimicrob. Agents Chemother. 48, 3996–4001. Schnellmann, C., Gerber, V., Rossano, A., Jaquier, V., Panchaud, Y., Doherr, M.G., Thomann, A., Straub, R., Perreten, V., 2006. Presence of new mecA and mph(C) variants conferring antibiotic resistance in Staphylococcus spp. isolated from the skin of horses before and after clinic admission. J. Clin. Microbiol. 44, 4444–4454. Schrag, S.J., Perrot, V., Levin, B.R., 1997. Adaptation to the fitness costs of antibiotic resistance in Escherichia coli. Proc. R. Soc. Lond. Ser. B-Biol. Sci. 264, 1287–1291. Seguin, J.C., Walker, R.D., Caron, J.P., Kloos, W.E., George, C.G., Hollis, R.J., Jones, R.N., Pfaller, M.A., 1999. Methicillin-resistant Staphylococcus aureus outbreak in a veterinary teaching hospital: potential human-to-animal transmission. J. Clin. Microbiol. 37, 1459– 1463. Singh, M., Chaudhry, M.A., Yadava, J.N.S., Sanyal, S.C., 1992. The spectrum of antibiotic-resistance in human and veterinary isolates of Escherichia-coli collected from 1984–86 in Northern India. J. Antimicrob. Chemother. 29, 159–168. Smith, H.W., 1969. Transfer of antibiotic resistance from animal and human strains of Escherichia coli to resident E. coli in alimentary tract of man. Lancet 1, 1174-&. van Duijkeren, E., Moleman, M., Sloet van Oldruitenborgh-Oosterbaan, M., Multem, J., Troelstra, A., Fluit, A.C., van Wamel, W.J.B., Houwers, D.J., de Neeling, A.J., Wagenaar, J.A., 2009. Methicillin-resistant Staphylococcus aureus in horses and horse personnel: an investigation of several outbreaks. Vet. Microbiol., doi:10.1016/j.vetmic.2009.08.09. van Duijkeren, E., van Asten, A., Gaastra, W., 2000. Characterization of Escherichia coli isolated from adult horses with and without enteritis. Vet. Q. 22, 162–166. Vo, A.T.T., van Duijkeren, E., Fluit, A.C., Gaastra, W., 2007. Characteristics of extended-spectrum cephalosporin-resistant Escherichia coli and Klebsiella pneumoniae isolates from horses. Vet. Microbiol. 124, 248– 255. Ward, M.P., Brady, T.H., Couetil, L.L., Liljebjelke, K., Maurer, J.J., Wu, C.C., 2005. Investigation and control of an outbreak of salmonellosis caused by multidrug-resistant Salmonella typhimurium in a population of hospitalized horses. Vet. Microbiol. 107, 233–240. Weese, J.S., Anderson, M.E.C., Lowe, A., Monteith, G.J., 2003. Preliminary investigation of the probiotic potential of Lactobacillus rhamnosus strain GG in horses: fecal recovery following oral administration and safety. Can. Vet. J. 44, 299–302. Weese, J.S., Anderson, M.E.C., Lowe, A., Penno, R., Da Costa, T.M., Button, L., Goth, K.C., 2004. Screening of the equine intestinal microflora for potential probiotic organisms. Equine Vet. J. 36, 351– 355. WHO, 2007. Guidelines for ATC classification and DDD assignment 2007. WHO Collaborating Centre for Drugs Statistics Methodology available at: http://www.whocc.no/atcdd (accessed 10 October 2007).