Antimicrobial resistance trends among canine Escherichia coli isolated at a New York veterinary diagnostic laboratory between 2007 and 2020

Abstract

Dogs are a potential source of drug-resistant Escherichia coli, but very few large-scale antimicrobial resistance surveillance studies have been conducted in the canine population. Here, we assess the antimicrobial susceptibility patterns, identify temporal resistance and minimum inhibitory concentration (MIC) trends, and describe associations between resistance phenotypes among canine clinical E. coli isolates in the northeastern United States. Through a retrospective study design, we collected MICs from 7709 E. coli isolates from canine infections at the Cornell University Animal Health Diagnostic Center between 2007 and 2020. The available clinical data were limited to body site. Isolates were classified as resistant or susceptible to six (urinary) and 22 (non-urinary) antimicrobials based on Clinical and Laboratory Standards Institute breakpoints. We used the Mann-Kendall test (MKT) and Sen’s slope to identify the presence of a significant trend in the percent of resistant isolates over the study period. Multivariable logistic regression (MLR) models were built with ceftiofur, enrofloxacin, or trimethoprim-sulfamethoxazole resistance as the outcome and either body site and isolation date, or resistance to other antimicrobials as predictors. MIC trends were characterized with survival analysis models, controlling for body site and year of isolation. Overall, 16.4% of isolates were resistant to enrofloxacin, 14.3% to ceftiofur, and 14% to trimethoprim-sulfamethoxazole. The MKT and Sen’s slope revealed a significant decreasing temporal trend for gentamicin and trimethoprim-sulfamethoxazole resistance among non-urinary isolates. No significant temporal resistance trends were detected by MKT for other antimicrobials. However, controlling for body-site in MLR models identified a decrease in resistance rates to enrofloxacin and trimethoprim-sulfamethoxazole after 2010. Similarly, survival analysis data confirmed these findings and showed a decrease in MIC values after 2010 for gentamicin and trimethoprim-sulfamethoxazole, but an increase in cephalosporin MICs. MLR showed that non-urinary isolates were significantly more likely than urinary isolates to demonstrate in vitro resistance to ceftiofur, enrofloxacin, and trimethoprim-sulfamethoxazole after controlling for year of isolation. We identified a higher level of ceftiofur resistance among enrofloxacin resistant isolates from urinary and non-urinary origins. Our findings confirmed that dogs are still a non-negligeable reservoir of drug-resistant E. coli in the northeastern United States. The increase in extended-spectrum cephalosporin MIC values in 2018–2020 compared to 2007–2010 constitutes a particularly worrying issue; the relationship between ceftiofur and enrofloxacin resistance suggests that the use of fluoroquinolones could contribute to this trend. Trimethoprim-sulfamethoxazole may be a good first-line choice for empiric treatment of E. coli infections; it is already recommended for canine urinary tract infections.

Introduction

Escherichia coli is a ubiquitous Gram-negative bacterium that is considered a natural inhabitant of humans’, dogs’ and other mammals’ guts (Martinson and Walk, 2020). Such commensal bacteria may also cause opportunistic infections outside their normal niche, particularly the urinary tract where E. coli is the most common pathogen (Day et al., 2019), and constitutes a common reason for antimicrobial prescription (Weese et al., 2019). Indeed, E. coli contains many pathotypes that cause a variety of canine diseases including urinary tract infections (UTIs), enteric diseases, wound infections, ear and respiratory infections, bacteremia, and meningitis (Kaper et al., 2004).

Although numerous antimicrobials remain effective for treating E. coli infections, the inappropriate use of these drugs contributes to the enrichment and selection of resistant strains (Brown et al., 2019). After an antimicrobial treatment, E. coli strains present in the intestine of hosts may acquire antimicrobial resistance (AMR) by either de novo mutation or through horizontal gene transfer of foreign resistance genes on mobile genetic elements from other bacteria; therefore E. coli may also act as a reservoir for AMR genes (Pouwels et al., 2019). Resistant E. coli are a significant One Health concern since they can spread between various animal species, people, and the environment, and readily share AMR determinants with other bacterial pathogens (Ogura et al., 2020). Therefore, E. coli is widely used as an indicator organism for monitoring AMR and multidrug resistance prevalence in commensal bacteria in the intestinal flora of animals (AbuOun et al., 2020). Unfortunately, multidrug-resistant (MDR) E. coli isolates have been commonly reported among dogs and other companion animals in the last decade throughout the world (Cummings et al., 2015; Yousfi et al., 2016; Kidsley et al., 2020; Valat et al., 2020; Marchetti et al., 2021). Recently, Hewitt and colleagues (Hewitt et al., 2020) described an alarming escalation in the prevalence of MDR bacterial pathogens among dogs with ulcerative keratitis from 5% in 2016 to 34% in 2020. Moreover, the percentage of MDR pathogens, including E. coli isolates, in dogs with complicated UTIs was twice as high in those with uncomplicated infections (Wong et al., 2015). This worrisome increasing trend of MDR is associated with more complicated illnesses and more deaths caused by bacterial infections (Dadgostar, 2019). MDR limits antimicrobial treatment options and may require the use of antimicrobials that are critically important for human and veterinary medicine (Weese et al., 2019).

Resistant E. coli in canine populations can be transmitted to humans through direct and indirect routes (Bourély et al., 2019). Several previous reports observed a co-carriership of MDR E. coli strains between dogs and their owners (Ljungquist et al., 2016; Lei et al., 2017; Valat et al., 2020). Moreover, the first confirmed case of dog to human transmission of a carbapenemase-producing (New Delhi metallo-beta-lactamase-5) E. coli isolate was recently described in Europe (Grönthal et al., 2018). The frequency of pet ownership, notably dog ownership, has been increasing in many high-income countries, including the United States, over the last few years. A recent National Pet Owners Survey showed that 63% of US households own a pet and the US pet dog population was estimated at nearly 77 million (AVMA, 2018; Overgaauw et al., 2020). This increases the risk of dog to human transmission of resistant or MDR pathogens and illustrates the need for regular surveillance of AMR in companion animal populations.

Even though surveillance studies provide critical data to assist clinicians with antimicrobial use decisions and can help policymakers to guide antimicrobial use and public health practices, very few large-scale studies have been conducted in the canine population in the US (Cummings et al., 2015; Wong et al., 2015). In this context, the main aims of this study were to assess the antimicrobial susceptibility patterns, identify temporal resistance and minimum inhibitory concentration (MIC) trends, and describe associations between resistance phenotypes among canine clinical E. coli isolates in the northeastern United States.

Material and methods

Study design

Retrospective clinical and antimicrobial susceptibility data from 7709 E. coli strains isolated from canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC) between July 15, 2007, and December 31, 2020, were collected. Variables collected from the laboratory information system included the date of the isolation, origin of clinical sample (body site), and MIC value for each antimicrobial agent.

Microbiologic procedure for E. coli detection

E. coli was cultured from clinical samples by the Cornell University AHDC using standard bacteriologic culture methods. Briefly, sample material was inoculated onto Columbia agar with 5% sheep blood and onto Eosin Methylene Blue (EMB) agar. Individual colonies were then chosen as presumptive for E. coli based upon morphology. Identity of isolates was confirmed as E. coli using either the Sensititre Automated Microbiology System (TREK Diagnostic Systems, Cleveland, Ohio, USA) or Matrix-Assisted Laser Desorption/Ionization Time-Of-Flight mass spectrometry (MALDI Biotyper; Bruker, Bellerica, MA, USA). All procedures were performed in accordance with accreditation by the American Association of Veterinary Laboratory Diagnosticians (AAVLD).

Antimicrobial susceptibility testing

Antimicrobial susceptibility testing of E. coli isolates was carried out using the broth microdilution method as previously described (Cummings et al., 2015). Two Thermo Scientific^™ Sensititre^™ MIC susceptibility plates were used for canine non-urinary E. coli isolates from 2007 to November 2016 (COMPAN1F and COMPAN2F panels), after which another plate (COMPGN1F panel) was adopted. Urinary isolates were tested with a different plate (CMV1BURF panel). Quality control was performed weekly using E. coli ATCC 25922, Staphylococcus aureus 29213, Enterococcus faecalis 29212, and Pseudomonas aeruginosa 27853. The MIC ranges for quality control recommended by the Clinical and Laboratory Standards Institute (CLSI) were used, and results were accepted if the MIC values were within expected ranges for these bacterial strains.

The drugs selected for this study (Table 1) have pharmacologic activity against E. coli and are clinically relevant to canine medicine, either through therapeutic use or as markers for susceptibility to commonly used agents. Although 23 antimicrobials were used by the AHDC for non-urinary isolates, the susceptibility of urinary isolates was systematically assessed using a narrow antimicrobial susceptibility testing panel, including ampicillin, amoxicillin-clavulanate, ceftiofur, tetracycline, enrofloxacin, and trimethoprim-sulfamethoxazole. Except for tetracycline, these five antimicrobials were regularly tested in all isolates throughout the study period, including in different non-urinary body sites. In the case of non-urinary isolates, the antimicrobial susceptibility testing panel included other antimicrobials, including ticarcillin, ticarcillin-clavulanate, piperacillin-tazobactam, cefazolin, cephalothin, cefoxitin, cefpodoxime, cefovecin, ceftazidime, imipenem, gentamicin, amikacin, doxycycline, chloramphenicol, marbofloxacin, orbifloxacin, and pradofloxacin. In rare cases of MDR E. coli urinary tract infections, the susceptibility was assessed using the extended panel (COMPAN1F, COMPAN2F, or COMPGN1F). We do not report percentages of resistance against these antimicrobials among urinary isolates, because the available AMR data only reflect a subpopulation of MDR urinary E. coli isolates. Also, we do not report the tetracycline resistance rate among non-urinary isolates because the susceptibility to this drug was not tested in E. coli non-urinary isolates between 2010 and 2016.

Table 1.

Prevalence of antimicrobial resistance among Escherichia coli clinical isolates from dogs stratified by sample source, from canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020.

		Non-susceptibility rate
		Total		Urinary tract		Non-Urinary tract
Antimicrobial category 1	Antimicrobial agent	N = 6765	%	N = 3715	%	N = 3050	Invasive	Skin and soft tissues	Reproductive system	Intestinal	Unspecified location
Penicillins (PEN)	Ampicillin *	3641	30.3	3641	30.3	–	–	–	–	–	–
	Ticarcillin 4	1960	35.0	113	NS	1847	37.2	37.3	24.9	46.3	44.5
Penicillins with β-lactamase inhibitor (PI)	Amoxicillin-clavulanate *	3641	17.9	3641	17.9	–	–	–	–	–	–
	Ticarcillin-clavulanate 4	1943	22.9	113	NS	1830	26.8	25.3	15.9	26.9	28.1
	Piperacillin-tazobactam	1214	4.5	63	NS	1151	6.9	1.8	1.4	9.0	9.4
First-generation cephalosporins (FGC)	Cefazolin or cephalothin	2932	38.4	176	NS	2756	37.2	40.4	28.2	50.8	45.2
Cephamycins (FOX)	Cefoxitin 4	1944	17.7	113	NS	1831	20.5	21.7	10.9	18.3	23.0
	Ceftiofur 2	5526	14.3	3646	12.8	1880	21.6	20.3	9.4	20.8	20.0
Extended-spectrum cephalosporins (ESC)	Cefpodoxime	3159	23.2	176	NS	2983	24.9	25.3	13.2	33.9	28.2
	Cefovecin	2745	23.1	145	NS	2600	23.2	24.2	14.1	34.5	25.7
	Ceftazidime	1215	20.3	63	NS	1152	18.4	19.6	12.7	30.0	18.9
Carbapenems (IMP)	Imipenem 3	3160	0.4	176	NS	2984	0.3	0.6	0.0	0.7	0.5
Aminoglycosides (AMG)	Gentamicin	3178	9.4	176	NS	3002	12.9	9.6	4.0	14.8	11.6
	Amikacin	3157	3.6	176	NS	2981	3.5	4.5	3.0	3.4	3.2
Tetracyclines (TET)	Tetracycline 3	3627	16.3	3627	16.3	1567	NS	NS	NS	NS	NS
	Doxycycline 3	2752	21.7	145	NS	2607	24.1	21.8	11.1	35.8	23.7
Phenicols (CHL)	Chloramphenicol 3	3177	14.0	176	NS	3001	12.9	12.4	7.8	27.7	13.7
	Enrofloxacin	6723	16.4	3709	14.3	3014	21.9	20.6	6.6	35.6	17.5
Fluoroquinolones (FQ)	Marbofloxacin	3159	17.8	176	NS	2983	20.3	19.4	5.5	34.2	17.6
	Orbifloxacin	1627	23.3	94	NS	1533	27.6	23.5	9.7	39.0	23.0
	Pradofloxacin	1217	21.3	63	NS	1154	23.5	18.3	6.6	39.1	22.6
Folate pathway inhibitors (SXT)	Trimethoprim-sulfamethoxazole	6741	14.0	3707	12.2	3034	18.1	15.2	9.0	26.4	23.5

N: number of tested E. coli isolates.

NS: Resistance data are not shown because most isolates tested for their antimicrobial susceptibility does not reflect the real resistance rate (e.g., a subpopulation of multidrug-resistant E. coli isolates). These isolates were also excluded from the total resistance rate.

^*As reported in the Clinical and Laboratory Standards Institute (CLSI) guidelines, E. coli canine isolates should be interpreted as resistant to ampicillin and amoxicillin-clavulanate, except in urinary tract infection cases.

¹The antimicrobial categories were adopted from (Magiorakos et al., 2012).

²Clinical breakpoints were adopted from those related to cattle (CLSI VET01S ED5:, 2020) (CLSI, 2020).

³Clinical breakpoints were adopted from those related to humans (CLSI VET01S ED5:, 2020) (CLSI, 2020).

⁴Clinical breakpoints were adopted from those related to humans (M100 – 27th edition – 2017) (CLSI, 2017).

The CLSI guidelines (CLSI-VET01S ED5:2020 (CLSI, 2020) and CLSI-M100–27–2017 (CLSI, 2017) were used to interpret MIC values and classify isolates as resistant or susceptible to each agent (Weinstein and Lewis, 2020). Human breakpoints were used when veterinary breakpoints were not available. Regardless of isolation year, all MIC values were interpreted using the same set of current guidelines. We excluded the rare cases of historical MIC values that could not be interpreted with the current CLSI clinical breakpoints. Since all breakpoints of the CLSI guidelines are dose dependent, the few isolates with intermediate susceptibility were categorized as being resistant. Most clinicians ignore an intermediate result as impossible to interpret or categorize it as a surrogate for resistant and proceed to look for another active compound for which the antibiogram result was sensitive (Kahlmeter et al., 2019).

Definition of MDR isolates

As they were tested by different panels, we divided our isolates into two main groups, urinary and non-urinary isolates. To define MDR isolates, we adopted the standardized international definition of resistance to three or more antimicrobial categories (Magiorakos et al., 2012). Before defining categories, we predicted some missing in vitro resistance results to homogenize the output in both urinary and non-urinary groups. All urinary isolates resistant to amoxicillin-clavulanate or extended-spectrum cephalosporin were categorized as resistant to ampicillin. Moreover, non-urinary isolates were categorized as ticarcillin-resistant if they had resistance to ticarcillin-clavulanate, piperacillin-tazobactam, extended-spectrum cephalosporin, or imipenem. Imipenem resistant isolates were also categorized as resistant to ticarcillin-clavulanate and first-generation cephalosporins. Similarly, any isolate resistant to cefoxitin or an extended-spectrum cephalosporin was categorized as resistant to first-generation cephalosporins. Then, we defined eleven categories: penicillins (PEN; ampicillin and ticarcillin), penicillins with β-lactamase inhibitors (PI; amoxicillin-clavulanate, ticarcillin-clavulanate, and piperacillin-tazobactam), first generation cephalosporins (FGC; cefazolin or cephalothin), cephamycins (FOX; cefoxitin), extended-spectrum cephalosporins (ESC; cefpodoxime, cefovecin, ceftiofur, and ceftazidime), carbapenems (IMP; imipenem), aminoglycosides (AMG; gentamicin and amikacin), tetracyclines (TET; tetracycline and doxycycline), phenicols (CHL; chloramphenicol), fluoroquinolones (FQ; enrofloxacin, marbofloxacin, orbifloxacin, and pradofloxacin), and folate pathway inhibitors (SXT; trimethoprim-sulfamethoxazole) (Table 1).

Statistical analysis

Data were analyzed using R software (R Core team, version 4.1.0; R Studio, version 1.4.1106). The dataset was imported for cleaning, variable coding, and analysis. Descriptive analysis was done on all variables using several packages (e.g., stringr, dplyr, tidyr, summarytools, prettyR, data.table, stats, icenReg) and obtained results were illustrated using the ggplot2 R package. All code necessary to replicate the analysis is publicly available (DOI: 10.5281/zenodo.7013921). According to the CLSI guideline regarding cumulative antibiograms reports, only one E. coli isolate per culture (our dataset lacked unique patient identifiers) was included in our study, regardless of the body site and antimicrobial susceptibility pattern (CLSI, 2020). Subsequent isolates were identified and removed from the database. The number of E. coli isolates was presented as the mean ± standard deviation. The categorical data were presented as frequencies and associated proportions. All statistical tests were two-sided, with a type I error set at α = 0.05. Since many regression models were created, as detailed below, we performed the Benjamini-Hochberg method to adjust the calculated P-values in each table, to control the false discovery rate to 0.05 (Benjamini and Hochberg, 1995).

Resistance trend analysis

We first used non-parametric statistical tests, the Mann–Kendall test (MKT) and Sen’s slope, t to detect monotonous temporal trends of antimicrobial monoresistance and multidrug resistance over the study period (2007–2020). The MKT identifies the significance of trends in a time series and Sen’s slope measures the steepness of a trend slope and is usually adopted to determine the slope of the Mann-Kendall trends over the study period (Verhoeven et al., 2022). The MKT and Sen’s slope were carried out separately on urinary and/or non-urinary isolates, except for the temporal trends of resistance to the three antimicrobials (ceftiofur, enrofloxacin, and trimethoprim-sulfamethoxazole) tested in all body sites; these antimicrobials had a single resistance breakpoint for urinary and non-urinary body sites.

Subsequently, we modeled resistance to these three antimicrobials, accounting for both body site and date. We divided the study years into four periods (period 1: 2007–2010, period 2: 2011–2014, period 3: 2015–2017, and period 4: 2018–2020). The differences in resistance rates across the antimicrobial agents tested in all specimen sources during the study period (ceftiofur, enrofloxacin, and trimethoprim-sulfamethoxazole) were first compared using the chi-squared test. Multivariable logistic regression (MLR) models were created for each abovementioned antimicrobial tested in all body sites. Resistance to the antimicrobial (yes/no) was the outcome and specimen source and study period were the explanatory variables. Two models were created for each antimicrobial: one (model 1) with each body site as a predictor and a second (model 2) comparing urinary isolates to all non-urinary isolates.

MIC trend analysis

We analyzed MIC distributions with Cox proportional hazards regression models (Stegeman et al., 2006; Otto, 2011). Briefly, the inhibition of bacterial growth was considered as the event; thus, we analyzed the concentration of antimicrobial required to achieve the event (i.e., MIC is the concentration at the event), instead of time to event. With survival analysis, resistance trends can be analyzed over an entire range of concentrations and no specific breakpoint value for resistance has to be applied. Previous approaches (Stegeman et al., 2006) to model MICs with Cox proportional hazards regression used the MIC as the event and only incorporated right censoring when the MIC was greater than the largest tested antimicrobial concentration. We coded the measured MICs as intervals that were open on the left and closed on the right to account for interval censoring. The growth inhibition event occurs within each interval, although the exact minimum concentration that could inhibit growth is unknown because growth is only assessed at two-fold dilutions of antimicrobial. When the MIC was recorded as “= ” to a specific concentration, the interval was from half the MIC (i.e., one dilution lower) to the MIC. For MICs recorded as “≤ ”, the interval was from 0 to the MIC, and for MICs recorded as “> ”, the interval was from the MIC to infinity. MICs were not log₂ transformed, as this would result in negative values for MICs less than 1 mg/L, and negative values cannot be used in survival analysis models. A separate model was created for each tested antimicrobial with specimen source and study period as the explanatory variables. A Hazard Ratio (HR) > 1 indicates a higher likelihood of growth inhibition (i.e., decreased survival in the presence of antimicrobial) of the studied E. coli group at each antimicrobial concentration compared to a reference E. coli group. This translates to lower MIC values among the E. coli group of interest compared to the reference group. The HR can also be represented as a probability (P = HR/(1 +HR)) (Spruance et al., 2004; Combescure et al., 2014). In the context of survival times, this represents the probability of a patient in one group surviving for a shorter time than a patient in the reference group. Considering bacterial growth inhibition, it is the probability of an isolate in a specific group having growth inhibition at a lower antimicrobial concentration (i.e., a lower MIC) than an isolate with the reference values; A hazard ratio of 2 corresponds to a 0.67 chance of an isolate in a specific group having a lower MIC value compared to an isolate in the reference group). For example, a HR of 3 corresponds to a 75% chance of a random E. coli isolate from a group of interest having a lower MIC compared to a random isolate from the reference group; thus, the group of interest is more sensitive to the antimicrobial than the reference group. We assessed the assumption of proportional hazards visually by examining the survival curves.

Prediction of antimicrobial resistance

A second group of MLR models was created to predict resistance to each of the regularly used antimicrobials in both urinary and non-urinary isolates. Resistance to other antimicrobial categories was considered as the explanatory variable. Antimicrobials within the same category were removed from the statistical models (e.g., cefovecin and cefpodoxime were excluded from models to predict ceftiofur resistance, and marbofloxacin was excluded from models to predict enrofloxacin resistance). Backward elimination was performed to identify the antimicrobial resistances that best predicted resistance to the outcome antimicrobial.

Results

A total of 7709 canine E. coli isolates collected at the Cornell University AHDC during a 14-year period (2007–2020) were included in this study. After limiting to one isolate per culture, 6765 isolates were further studied. Bacteria were mainly obtained from urine (N = 3715; 54.9%), followed by skin and soft tissues (N = 969; 14.3%), reproductive system (N = 936; 13.8%), intestinal tract (N = 556; 8.2%), and invasive locations (including, among others, blood, bone, joint, gallbladder, liver, lungs, lymph nodes, pancreas, peritoneum, and pleura) (N = 376; 5.6%). The remaining isolates (N = 213; 3.1%) were from unspecified sites. The mean number of E. coli isolated per year was 483 ± 105 (range: 222–637).

The prevalence of resistance to each antimicrobial across all years, stratified by body site, and antimicrobial class abbreviations are listed in Table 1. As the susceptibility of urinary and non-urinary isolates has been assessed using different testing panels throughout the study period, there are significant differences in the total numbers of isolates across antimicrobial agents. Antimicrobial susceptibility testing showed a low prevalence of resistance to regularly tested antimicrobials in E. coli isolates, including enrofloxacin (16.4%), ceftiofur (14.3%), and trimethoprim-sulfamethoxazole (14%). The resistance rates to these three antimicrobials were higher among non-urinary isolates compared to urinary isolates. This was also observed among antimicrobials belonging to the same antimicrobial category, such as the resistance to other third-generation cephalosporins among non-urinary isolates, which ranged between 20.3% and 23.2%, compared to 12.8% resistance to ceftiofur among urinary isolates. Ampicillin and amoxicillin-clavulanate showed higher percentages of resistance (30.3% and 17.9%, respectively) than other tested antimicrobials among urinary isolates. Although both antimicrobials were systematically tested in non-urinary isolates, they should be reported as resistant or not, regardless of their MIC values because ampicillin and amoxicillin-clavulanate concentrations achieved according to the dosage regimen used to determine clinical breakpoints are not high enough to reach the therapeutic target in non-urinary tract infections (CLSI, 2020). Carbapenem resistance was observed among all sample sources in this study, except reproductive system sites, although at a very low rate (0.4%). Regarding fluoroquinolones, 23.3%, 21.3%, and 17.8% of non-urinary isolates were resistant to orbifloxacin, pradofloxacin, and marbofloxacin, respectively. Lower resistance rates were observed for aminoglycosides such as gentamicin (9.4%) and amikacin (3.6%) (Table 1).

The susceptibility testing results showed most isolates were pan-susceptible (58.9%; 3985/6765). This wild-type phenotype was more common among urinary (64.8%) (Fig. 1 A) than non-urinary (51.7%) isolates (Fig. 2 A). The rates of monoresistance (7.9%) and biresistance (8.6%) patterns in urinary isolates were similar to those from non-urinary isolates (10.6% monoresistance, 5% biresistance). MDR, defined as in vitro acquired resistance to at least one drug in three or more antimicrobial categories, was observed in 18.7% and 32.7% of urinary and non-urinary isolates, respectively. The great majority (97.5%; 1648/1690) of MDR E. coli isolates showed resistance to at least one beta-lactam agent. Among urinary isolates, resistance to extended-spectrum cephalosporins was observed among the top three AMR patterns (Fig. 1 B). The most common multidrug resistance pattern among MDR urinary isolates was PEN-PI-ESC (18.6%; 129/693), followed by PEN-PI-ESC-TET-FQ-SXT (18.5%), PEN-PI-ESC-TET-FQ (7.4%), and PEN-TET-FQ-SXT (6.5%). Similar MDR patterns were observed among non-urinary isolates (Fig. 2 B): PEN-FGC-ESC was the most common (8.8%; 88/997), followed by PEN-PI-FGC-FOX-ESC (7.7%). Alarmingly, 128 urinary E. coli isolates showed non-susceptibility to the six tested antimicrobial categories and two non-urinary isolates were resistant to the 11 tested antimicrobial categories. Also, 27 non-urinary E. coli isolates showed resistance to 10 antimicrobial categories. We were unable to assess the presence of extremely- and pan-drug resistant isolates because an insufficient number of antimicrobial categories were included on the susceptibility testing plates (Magiorakos et al., 2012).

Fig. 1.

Distribution of resistance by number of antimicrobial categories (A) and most common multidrug resistance patterns among urinary Escherichia coli isolates (B), in canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020. Antimicrobial category abbreviations are listed in Table 1.

Fig. 2.

Distribution of resistance by number of antimicrobial categories (A) and most common multidrug resistance patterns among non-urinary Escherichia coli isolates (B), in canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020. Antimicrobial category abbreviations are listed in Table 1.

The MKT and Sen’s slope showed a significant decreasing temporal trend for gentamicin (Z = −2.52, Sen’s = −0.546, P-value = 0.012) and trimethoprim-sulfamethoxazole (Z = −1.97, Sen’s = −0.574, P-value = 0.049) resistance among non-urinary isolates (Fig. 3 A). No significant temporal resistance trends were detected for other antimicrobials or the prevalence of MDR isolates either in urinary (P-value = 0.125) or non-urinary isolates (P-value = 0.584) (Fig. 3 B). Regarding body sites, reproductive system isolates showed the lowest frequency of resistance for all the tested drugs (Table 1). Considering the three regularly tested antimicrobials among different sample sites, isolates from reproductive system infections were always the most susceptible to ceftiofur (90.6% susceptible), trimethoprim-sulfamethoxazole (91%), and enrofloxacin (93.4%), followed by urinary tract infection isolates (Table 1).

Fig. 3.

Temporal trends in the prevalence of resistance to ceftiofur (CFT), gentamicin (GMN), enrofloxacin (EFX), and trimethoprim-sulfamethoxazole (SXT) (A), and multidrug resistance (B) among canine Escherichia coli urinary (UTI) and non-urinary (NUTI) isolates during the study period (2007–2020), in canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020. CFT, EFX, and SXT were tested among UTI and NUTI E. coli isolates, but GMN was only tested among NUTI E. coli isolates.

After dividing the study years into four periods and accounting for year of isolation, MLR analysis (Table 2) confirmed that non-urinary isolates were significantly more likely than urinary isolates to demonstrate in vitro resistance to ceftiofur (odds ratio (OR) = 1.40; 95% confidence interval (95% CI) = 1.19–1.65; P < 0.001), enrofloxacin (OR = 1.40; 95% CI = 1.23–1.59; P < 0.001), and trimethoprim-sulfamethoxazole (OR = 1.40; 95% CI = 1.22–1.61; P < 0.001). Intestinal E. coli isolates were the most likely to be resistant to enrofloxacin (OR = 3.31; 95% CI = 2.71–4.04; P < 0.001), and trimethoprim-sulfamethoxazole (OR = 2.68; 95% CI = 2.16–3.32; P < 0.001) (Tables 1–2). Isolates from invasive infections showed the highest rates of ceftiofur resistance (OR = 1.82; 95% CI = 1.33–2.47; P < 0.001), followed by intestinal infections (OR = 1.80; 95% CI = 1.27–2.51; P = 0.001) (Tables 1–2). Some decreases in resistance over time were observed after accounting for body site. E. coli strains isolated between 2011 and 2017 showed a lower level of resistance to both enrofloxacin and trimethoprim-sulfamethoxazole compared to those isolated between 2007 and 2010. Also, isolates from 2011 to 2014 showed a lower resistance rate to ceftiofur compared to peers isolated between 2007 and 2010 (OR = 0.69; 95% CI = 0.57–0.83; P < 0.001). No significant difference was observed in AMR rates after 2018 compared to the 2007–2010 period, except for a decrease in trimethoprim-sulfamethoxazole resistance (OR = 0.70; 95% CI = 0.58–0.86; P = 0.001).

Table 2.

Determinants of resistance to ceftiofur, enrofloxacin, and trimethoprim-sulfamethoxazole including specimen source and study period among Escherichia coli isolates using multivariable logistic regression models, in canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020.

	Model 1a				Model 2b
	adj. OR	95% CI	P-value	adj. P-value§	adj. OR	95% CI	P-value	adj. P-value§
Resistance to ceftiofur (N = 5526 isolates)
Urinary tract 1
Intestinal	1.80	(1.27–2.51)	0.001	0.001
Invasive	1.82	(1.33–2.47)	< 0.001	< 0.001
Unspecified site	1.70	(1.12–2.51)	0.010	0.014	1.40	(1.19–1.65)	< 0.001	< 0.001
Reproductive system	0.70	(0.51–0.94)	0.020	0.025
Skin and soft tissues	1.74	(1.39–2.17)	< 0.001	< 0.001
Isolation date (2007–2010) 1	0.69	(0.57–0.83)	< 0.001	< 0.001	0.69	(0.57–0.83)	< 0.001	< 0.001
Isolation date (2011–2014)
Isolation date (2015–2017)	0.83	(0.68–1.02)	0.078	0.086	0.83	(0.68–1.02)	0.074	0.086
Isolation date (2018–2020)	0.79	(0.61–1.02)	0.079	0.086	0.80	(0.62–1.03)	0.087	0.092
Resistance to enrofloxacin (N = 6723 isolates)
Urinary tract 1
Intestinal	3.31	(2.71–4.04)	< 0.001	< 0.001
Invasive	1.65	(1.26–2.14)	< 0.001	< 0.001
Unspecified site	1.27	(0.87–1.82)	0.198	0.204	1.40	(1.23–1.59)	< 0.001	< 0.001
Reproductive system	0.43	(0.32–0.56)	< 0.001	< 0.001
Skin and soft tissues	1.56	(1.30–1.87)	< 0.001	< 0.001
Isolation date (2007–2010) 1
Isolation date (2011–2014)	0.67	(0.56–0.82)	< 0.001	< 0.001	0.69	(0.57–0.83)	< 0.001	< 0.001
Isolation date (2015–2017)	0.77	(0.64–0.93)	0.007	0.011	0.80	(0.67–0.97)	0.020	0.025
Isolation date (2018–2020)	0.84	(0.70–1.01)	0.068	0.082	0.93	(0.78–1.11)	0.426	0.426
Resistance to trimethoprim-sulfamethoxazole (N = 6741 isolates)
Urinary tract 1
Intestinal	2.68	(2.16–3.32)	< 0.001	< 0.001
Invasive	1.55	(1.16–2.04)	0.002	0.004
Unspecified site	2.19	(1.56–3.04)	< 0.001	< 0.001	1.40	(1.22–1.61)	< 0.001	< 0.001
Reproductive system	0.72	(0.56–0.91)	0.007	0.011
Skin and soft tissues	1.29	(1.05–1.58)	0.013	0.017
Isolation date (2007–2010) 1	0.68	(0.56–0.82)	< 0.001	< 0.001	0.68	(0.56–0.83)	< 0.001	< 0.001
Isolation date (2011–2014)
Isolation date (2015–2017)	0.62	(0.51–0.76)	< 0.001	< 0.001	0.63	(0.52–0.77)	< 0.001	< 0.001
Isolation date (2018–2020)	0.70	(0.58–0.86)	< 0.001	0.001	0.76	(0.62–0.91)	0.004	0.006

^aIn model 1, origin of clinical sample and date of isolation (divided into four periods: 2007–2010, 2011–2014, 2015–2017, and 2018–2020) were entered in the model.

^bIn model 2, origin of clinical sample (divided into UTI and non-UTI) and date of isolation (divided into the same four periods defined in model 1) were entered in the model.

¹Reference group.

^§P-values were adjusted according to (Benjamini and Hochberg, 1995).

As predicted, survival analysis models concurred with MKT and Sen’s slope findings, demonstrating a decrease in MIC values for trimethoprim-sulfamethoxazole (P ≤ 0.001; Table 3) and gentamicin (P < 0.05; Table 4) in the three study periods 2011–2014, 2015–2017, and 2018–2020 compared to the reference study period 2007–2010. Moreover, survival analysis models concurred with previous MLR models showing that MIC values for enrofloxacin were lower in the study period 2011–2017 compared to the reference study period (P < 0.05; Table 3). Surprisingly, even though MKT and Sen’s slope did not reveal any change in resistance rates to extended-spectrum cephalosporins after 2010, survival analysis data revealed an increase in cefpodoxime and cefovecin MIC values among non-urinary isolates in the study period 2018–2020 compared to the 2007–2010 reference period (P < 0.05; Table 4).

Table 3.

Multivariable Cox proportional hazard regression model of susceptibility of Escherichia coli isolates from urinary and non-urinary origin to different antimicrobials in this study, in canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020.

	Hazard ratio	Probability of lower MIC¶	P-value	adj. P-value§	95% CI
Ampicillin (N = 6724 isolates)
Urinary tract a
Intestinal	0.57	0.36	< 0.001	< 0.001	0.49–0.67
Invasive	0.78	0.44	0.007	0.015	0.65–0.93
Unspecified site	0.74	0.43	0.009	0.017	0.59–0.93
Reproductive system	1.07		0.222	0.278	0.96–1.19
Skin and soft tissues	0.83	0.45	0.002	0.004	0.73–0.93
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.09		0.095	0.122	0.99–1.20
Isolation date (2015–2017)	1.05		0.304	0.358	0.96–1.15
Isolation date (2018–2020)	0.98		0.608	0.657	0.90–1.06
Amoxicillin-clavulanate (N = 6721 isolates)
Urinary tract a
Intestinal	0.67	0.40	< 0.001	< 0.001	0.59–0.76
Invasive	0.75	0.43	< 0.001	0.001	0.65–0.88
Unspecified site	0.84		0.095	0.122	0.69–1.03
Reproductive system	1.16	0.54	0.016	0.026	1.03–1.30
Skin and soft tissues	0.90		0.077	0.110	0.80–1.01
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.11		0.071	0.109	0.99–1.24
Isolation date (2015–2017)	1.05		0.347	0.397	0.95–1.16
Isolation date (2018–2020)	1.02		0.637	0.671	0.94–1.11
Ceftiofur (N = 5526 isolates)
Urinary tract a
Intestinal	0.82	0.45	0.012	0.022	0.70–0.96
Invasive	0.81	0.45	0.009	0.017	0.69–0.95
Unspecified site	0.84		0.089	0.122	0.70–1.03
Reproductive system	1.09		0.076	0.110	0.99–1.21
Skin and soft tissues	0.81	0.45	< 0.001	< 0.001	0.73–0.89
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.09		0.034	0.054	1.01–1.18
Isolation date (2015–2017)	1.02		0.729	0.748	0.93–1.11
Isolation date (2018–2020)	1.02		0.777	0.777	0.92–1.12
Enrofloxacin (N = 6723 isolates)
Urinary tract a
Intestinal	0.54	0.35	< 0.001	< 0.001	0.48–0.60
Invasive	0.80	0.44	0.001	0.003	0.70–0.92
Unspecified site	0.91		0.270	0.327	0.76–1.08
Reproductive system	1.29	0.56	< 0.001	< 0.001	1.18–1.41
Skin and soft tissues	0.79	0.44	< 0.001	< 0.001	0.72–0.86
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.18	0.54	< 0.001	0.001	1.08–1.28
Isolation date (2015–2017)	1.12	0.53	0.009	0.017	1.03–1.23
Isolation date (2018–2020)	1.04		0.367	0.408	0.96–1.13
Trimethoprim -sulfamethoxazole (N = 6741 isolates)
Urinary tract a
Intestinal	0.63	0.39	< 0.001	< 0.001	0.56–0.70
Invasive	0.83	0.45	0.006	0.014	0.73–0.95
Unspecified site	0.67	0.40	< 0.001	< 0.001	0.57–0.79
Reproductive system	1.15	0.53	0.005	0.014	1.04–1.26
Skin and soft tissues	0.88	0.47	0.006	0.014	0.81–0.97
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.19	0.54	< 0.001	< 0.001	1.09–1.29
Isolation date (2015–2017)	1.23	0.55	< 0.001	< 0.001	1.12–1.35
Isolation date (2018–2020)	1.17	0.54	< 0.001	0.001	1.07–1.28

^aReference group

^¶Probability = HR/(1 + HR) − calculated if P-value< 0.05 (e.g., a hazard ratio of 2 corresponds to a 0.67 chance of an isolate at this condition having a lower MIC value compared to an isolate in the reference group).

^§P-values were adjusted according to (Benjamini and Hochberg, 1995).

Table 4.

Multivariable Cox proportional hazard regression model of susceptibility of Escherichia coli isolates from non-urinary origin to different antimicrobials in this study, in canine clinical infections at the Cornell University Animal Health Diagnostic Center (AHDC), 2007–2020.

	Hazard ratio	Probability of lower MIC¶	P-value	adj. P-value§	95% CI
Ticarcillin! (N = 1965 isolates)
Intestinal a
Invasive	1.40	0.58	0.004	0.011	1.11–1.76
Unspecified site	1.15		0.329	0.469	0.87–1.51
Reproductive system	1.83	0.65	< 0.001	< 0.001	1.49–2.24
Skin and soft tissues	1.32	0.57	0.007	0.016	1.08–1.62
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.28	0.56	< 0.001	0.001	1.13–1.45
Isolation date (2015–2017)	1.12		0.130	0.219	0.97–1.31
Ticarcillin-clavulanate! (N = 1830 isolates)
Intestinal a
Invasive	1.13		0.222	0.327	0.93–1.37
Unspecified site	1.06		0.601	0.747	0.85–1.32
Reproductive system	1.46	0.59	< 0.001	< 0.001	1.24–1.72
Skin and soft tissues	1.10		0.250	0.362	0.93–1.30
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.19	0.54	0.005	0.013	1.05–1.33
Isolation date (2015–2017)	1.01		0.903	0.947	0.88–1.15
Cefazolin (N = 2981 isolates)
Intestinal a
Invasive	1.28	0.56	0.004	0.011	1.08–1.51
Unspecified site	1.14		0.192	0.288	0.94–1.39
Reproductive system	1.80	0.64	< 0.001	< 0.001	1.60–2.03
Skin and soft tissues	1.29	0.56	< 0.001	< 0.001	1.14–1.46
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.27	0.56	< 0.001	0.001	1.12–1.45
Isolation date (2015–2017)	0.94		0.390	0.539	0.82–1.08
Isolation date (2018–2020)	0.77	0.44	< 0.001	< 0.001	0.68–0.87
Cephalothin! (N = 1534 isolates)
Intestinal a
Invasive	1.20		0.143	0.235	0.94–1.52
Unspecified site	0.90		0.509	0.681	0.65–1.24
Reproductive system	1.44	0.59	< 0.001	0.002	1.18–1.77
Skin and soft tissues	1.04		0.739	0.857	0.83–1.29
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.29	0.56	0.002	0.006	1.10–1.50
Isolation date (2015–2017)	1.02		0.780	0.881	0.87–1.21
Cefoxitin! (N = 1831 isolates)
Intestinal a
Invasive	1.01		0.938	0.961	0.82–1.25
Unspecified site	0.96		0.728	0.856	0.76–1.22
Reproductive system	1.27	0.56	0.006	0.015	1.07–1.50
Skin and soft tissues	0.93		0.406	0.551	0.78–1.11
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.10		0.148	0.239	0.97–1.24
Isolation date (2015–2017)	0.79	0.44	< 0.001	0.003	0.69–0.91
Cefovecin (N = 2600 isolates)
Intestinal a
Invasive	1.30	0.57	0.004	0.011	1.09–1.56
Unspecified site	1.15		0.187	0.288	0.94–1.41
Reproductive system	1.51	0.60	< 0.001	< 0.001	1.33–1.73
Skin and soft tissues	1.20	0.55	0.009	0.020	1.05–1.38
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.05		0.538	0.703	0.89–1.25
Isolation date (2015–2017)	0.82		0.028	0.057	0.69–0.98
Isolation date (2018–2020)	0.75	0.43	0.001	0.004	0.63–0.89
Cefpodoxime (N = 2983 isolates)
Intestinal a
Invasive	1.20		0.033	0.063	1.01–1.43
Unspecified site	1.15		0.173	0.274	0.94–1.41
Reproductive system	1.79	0.64	< 0.001	< 0.001	1.56–2.06
Skin and soft tissues	1.23	0.55	0.002	0.007	1.08–1.40
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.12		0.101	0.176	0.98–1.27
Isolation date (2015–2017)	0.89		0.089	0.158	0.77–1.02
Isolation date (2018–2020)	0.83	0.45	0.004	0.011	0.73–0.94
Imipenem (N = 2984 isolates)
Intestinal a
Invasive	1.19		0.794	0.886	0.32–4.45
Unspecified site	1.03		0.964	0.964	0.24–4.49
Reproductive system	4.04	0.80	< 0.001	< 0.001	2.48–6.58
Skin and soft tissues	1.06		0.776	0.881	0.70–1.61
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.04		0.950	0.961	0.32–3.36
Isolation date (2015–2017)	0.79		0.600	0.747	0.33–1.90
Isolation date (2018–2020)	1.08		0.886	0.947	0.36–3.26
Gentamicin (N = 3018 isolates)
Intestinal a
Invasive	1.18		0.031	0.061	1.02–1.38
Unspecified site	1.26		0.032	0.062	1.02–1.55
Reproductive system	1.74	0.64	< 0.001	< 0.001	1.54–1.96
Skin and soft tissues	1.32	0.57	< 0.001	< 0.001	1.17–1.49
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.18	0.54	0.017	0.036	1.03–1.35
Isolation date (2015–2017)	1.20	0.55	0.008	0.019	1.05–1.36
Isolation date (2018–2020)	1.30	0.57	< 0.001	< 0.001	1.14–1.48
Amikacin (N = 3015 isolates)
Intestinal a
Invasive	0.99		0.944	0.961	0.80–1.22
Unspecified site	0.96		0.804	0.886	0.67–1.37
Reproductive system	1.05		0.541	0.703	0.90–1.23
Skin and soft tissues	0.93		0.344	0.483	0.80–1.08
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.15		0.084	0.153	0.98–1.35
Isolation date (2015–2017)	1.04		0.648	0.794	0.89–1.22
Isolation date (2018–2020)	1.13		0.131	0.219	0.96–1.33
Doxycycline (N = 2615 isolates)
Intestinal a
Invasive	1.33	0.57	0.001	0.003	1.12–1.58
Unspecified site	1.34	0.57	0.004	0.012	1.10–1.65
Reproductive system	1.90	0.66	< 0.001	< 0.001	1.67–2.18
Skin and soft tissues	1.40	0.58	< 0.001	< 0.001	1.22–1.61
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.02		0.819	0.891	0.85–1.23
Isolation date (2015–2017)	0.88		0.190	0.288	0.73–1.06
Isolation date (2018–2020)	0.10		0.685	0.828	0.81–1.15
Chloramphenicol (N = 3013 isolates)
Intestinal a
Invasive	1.54	0.61	< 0.001	< 0.001	1.31–1.81
Unspecified site	1.32	0.57	0.006	0.015	1.08–1.61
Reproductive system	1.49	0.60	< 0.001	< 0.001	1.32–1.68
Skin and soft tissues	1.45	0.59	< 0.001	< 0.001	1.28–1.65
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.02		0.720	0.856	0.90–1.16
Isolation date (2015–2017)	0.86	0.46	0.011	0.023	0.76–0.97
Isolation date (2018–2020)	0.90		0.074	0.136	0.80–1.01
Marbofloxacin (N = 2983 isolates)
Intestinal a
Invasive	1.50	0.60	< 0.001	< 0.001	1.27–1.76
Unspecified site	1.64	0.62	< 0.001	< 0.001	1.34–2.00
Reproductive system	2.38	0.70	< 0.001	< 0.001	2.09–2.72
Skin and soft tissues	1.47	0.60	< 0.001	< 0.001	1.28–1.68
Isolation date (2007–2010) a
Isolation date (2011–2014)	1.17	0.54	0.019	0.039	1.03–1.34
Isolation date (2015–2017)	0.96		0.554	0.709	0.84–1.10
Isolation date (2018–2020)	0.99		0.903	0.947	0.88–1.13

^aReference group

^¶Probability = HR/(1 + HR) − calculated if P-value< 0.05 (e.g., a hazard ratio of 2 corresponds to a 0.67 chance of an isolate at this condition to have a lower MIC value compared to an isolate in the reference group)

^!Not tested from 2018 to 2020.

^§P-values were adjusted according to (Benjamini and Hochberg, 1995).

MLR models revealed several associations between antimicrobial resistances. Some followed expected cross-resistance patterns; for instance, ampicillin (OR = 115.4; 95% CI = 24.7–2059; P < 0.001) and amoxicillin-clavulanate (OR = 20.3; 95% CI = 13.8–30.5; P < 0.001) were strong predictors of ceftiofur resistance among urinary isolates (Table 5). We also found relationships between resistances to different antimicrobial classes, broadly capturing the multidrug resistance patterns of beta-lactams, enrofloxacin, and trimethoprim-sulfamethoxazole resistance that were observed in the MDR tabulation (Fig. 1 B, Fig. 2 B).

Table 5.

Association between resistance to ceftiofur, enrofloxacin, or trimethoprim-sulfamethoxazole and other antimicrobial compounds among Escherichia coli urinary isolates using multivariable logistic regression models.

	Model 1a				Model 2a
	adj. OR	95% CI	P-value	adj. P-value §	adj. OR	95% CI	P-value	adj. P-value §
Resistance to ceftiofur (N = 3562 isolates)
Resistance to ampicillin	115.4	(24.7–2059)	< 0.001	< 0.001	115.4	(24.7–2059)	< 0.001	< 0.001
Resistance to amoxicillin-clavulanate	20.3	(13.8–30.5)	< 0.001	< 0.001	20.3	(13.8–30.5)	< 0.001	< 0.001
Resistance to tetracycline	1.32	(0.90–1.93)	0.149	0.155	1.32	(0.90–1.93)	0.149	0.155
Resistance to enrofloxacin	4.84	(3.32–7.14)	< 0.001	< 0.001	4.84	(3.32–7.14)	< 0.001	< 0.001
Resistance to trimethoprim-sulfamethoxazole	1.55	(1.03–2.35)	0.036	0.040	1.55	(1.03–2.35)	0.036	0.040
Resistance to enrofloxacin (N = 3562 isolates)
Resistance to ampicillin	5.68	(3.96–8.16)	< 0.001	< 0.001	5.68	(3.96–8.16)	< 0.001	< 0.001
Resistance to amoxicillin-clavulanate	0.60	(0.40–0.88)	0.011	0.013	0.60	(0.40–0.88)	0.011	0.013
Resistance to ceftiofur	4.78	(3.26–7.10)	< 0.001	< 0.001	4.78	(3.26–7.10)	< 0.001	< 0.001
Resistance to tetracycline	3.36	(2.50–4.50)	< 0.001	< 0.001	3.36	(2.50–4.50)	< 0.001	< 0.001
Resistance to trimethoprim-sulfamethoxazole	5.20	(3.81–7.09)	< 0.001	< 0.001	5.20	(3.81–7.09)	< 0.001	< 0.001
Resistance to trimethoprim-sulfamethoxazole (N = 3562 isolates)
Resistance to ampicillin	6.83	(4.71–9.98)	< 0.001	< 0.001	6.91	(4.76–10.1)	< 0.001	< 0.001
Resistance to amoxicillin-clavulanate	0.54	(0.36–0.80)	0.003	0.004	0.61	(0.43–0.84)	0.003	0.004
Resistance to ceftiofur	1.23	(0.81–1.86)	0.326	0.326
Resistance to tetracycline	7.85	(5.85–10.6)	< 0.001	< 0.001	7.93	(5.92–10.7)	< 0.001	< 0.001
Resistance to enrofloxacin	5.48	(4.01–7.48)	< 0.001	< 0.001	5.74	(4.27–7.73)	< 0.001	< 0.001

^aIn model 1, selected antimicrobials regularly tested on urinary Escherichia coli isolates were entered in the model as explanatory variables.

ⁱⁱIn model 2, a backward logistic regression model was created including only complete cases.

^§P-values were adjusted according to (Benjamini and Hochberg, 1995).

Enrofloxacin-resistant E. coli isolates showed a higher level of ceftiofur resistance (and vice versa), whether from urinary (OR = 4.78; 95% CI = 3.26–7.10; P < 0.001) (Table 5) or non-urinary (OR = 2.57; 95% CI = 1.35–5.04; P = 0.007) origins (Table 6). Amoxicillin-clavulanate resistance was found to be associated with a decrease in the probability of resistance to enrofloxacin among urinary isolates (OR = 0.60; 95% CI = 0.40–0.88; P = 0.013), after controlling for other resistances, including ampicillin resistance (Table 5). Also, a similar association was observed in the MLR model predicting trimethoprim-sulfamethoxazole resistance in urinary isolates (amoxicillin-clavulanate OR = 0.54; 95% CI = 0.36–0.80; P = 0.004) (Table 5). As reported in the CLSI guidelines, E. coli canine isolates should be interpreted as resistant to ampicillin and amoxicillin-clavulanate in non-urinary infections; thus, the respective information about these potential negative associations was not available in non-urinary isolates. In addition, among non-urinary isolates, cefoxitin resistance was associated with a decrease in the probability of trimethoprim-sulfamethoxazole resistance (OR = 0.47; 95% CI = 0.25–0.87; P = 0.025), but not enrofloxacin resistance (P = 0.340) (Table 6). Overall, ticarcillin resistance was the strongest predictor of resistance to ceftiofur (OR = 13.4; 95% CI = 4.78–41.7; P < 0.001), enrofloxacin (OR = 3.98; 95% CI = 2.15–7.38; P < 0.001), and trimethoprim-sulfamethoxazole (OR = 17; 95% CI = 9.41–31.8; P < 0.001) among non-urinary isolates (Table 6).

Table 6.

Association between resistance to ceftiofur, enrofloxacin, or trimethoprim-sulfamethoxazole and other antimicrobial compounds among Escherichia coli non-urinary isolates using multivariable logistic regression models.

	Model 1				Model 2a
	adj. OR	95% CI	P-value	adj. P-value §	adj. OR	95% CI	P-value	adj. P-value §
Resistance to ceftiofur (N = 1439 isolates)
Resistance to ticarcillin	13.4	(4.78–41.7)	< 0.001	< 0.001	14.9	(5.47–45.5)	< 0.001	< 0.001
Resistance to ticarcillin-clavulanate	3.75	(1.90–7.64)	< 0.001	< 0.001	3.73	(1.90–7.55)	< 0.001	< 0.001
Resistance to cefoxitin	37.2	(20.4–71.0)	< 0.001	< 0.001	36.1	(20.1–67.5)	< 0.001	< 0.001
Resistance to amikacin	1.95	(0.78–4.92)	0.154	0.193	1.96	(0.80–4.84)	0.143	0.185
Resistance to doxycycline	1.47	(0.79–2.77)	0.230	0.279
Resistance to chloramphenicol	0.84	(0.42–1.66)	0.615	0.648
Resistance to enrofloxacin	3.45	(1.77–6.85)	< 0.001	< 0.001	4.07	(2.31–7.38)	< 0.001	< 0.001
Resistance to trimethoprim-sulfamethoxazole	1.15	(0.61–2.20)	0.661	0.678
Resistance to enrofloxacin (N = 1439 isolates)
Resistance to ticarcillin	3.98	(2.15–7.38)	< 0.001	< 0.001	4.43	(2.57–7.73)	< 0.001	< 0.001
Resistance to ticarcillin-clavulanate	1.35	(0.75–2.45)	0.322	0.368
Resistance to cefoxitin	0.70	(0.36–1.33)	0.289	0.340
Resistance to ceftiofur	2.57	(1.35–5.04)	0.005	0.007	2.33	(1.48–3.71)	< 0.001	< 0.001
Resistance to amikacin	2.86	(1.30–6.37)	0.009	0.013	3.18	(1.48–6.96)	0.003	0.005
Resistance to doxycycline	3.71	(2.41–5.71)	< 0.001	< 0.001	3.63	(2.36–5.56)	< 0.001	< 0.001
Resistance to chloramphenicol	3.60	(2.21–5.89)	< 0.001	< 0.001	3.58	(2.21–5.84)	< 0.001	< 0.001
Resistance to trimethoprim-sulfamethoxazole	3.89	(2.48–6.14)	< 0.001	< 0.001	3.99	(2.56–6.25)	< 0.001	< 0.001
Resistance to trimethoprim-sulfamethoxazole (N = 1439 isolates)
Resistance to ticarcillin	17.0	(9.41–31.8)	< 0.001	< 0.001	16.2	(9.44–29.0)	< 0.001	< 0.001
Resistance to ticarcillin-clavulanate	0.95	(0.54–1.69)	0.871	0.871
Resistance to cefoxitin	0.47	(0.25–0.87)	0.019	0.025	0.40	(0.25–0.62)	< 0.001	< 0.001
Resistance to ceftiofur	0.79	(0.41–1.52)	0.474	0.527
Resistance to amikacin	1.24	(0.60–2.59)	0.563	0.609
Resistance to doxycycline	2.85	(1.83–4.45)	< 0.001	< 0.001	2.85	(1.83–4.43)	< 0.001	< 0.001
Resistance to chloramphenicol	3.91	(2.41–6.38)	< 0.001	< 0.001	3.89	(2.40–6.34)	0.001	< 0.001
Resistance to enrofloxacin	3.91	(2.49–6.17)	< 0.001	< 0.001	3.83	(2.48–5.93)	< 0.001	< 0.001

^aIn model 1, selected antimicrobials regularly tested on non-urinary Escherichia coli isolates were entered in the model as explanatory variables.

ⁱⁱIn model 2, a backward logistic regression model was created including only complete cases.

^§P-values were adjusted according to (Benjamini and Hochberg, 1995).

Discussion

AMR poses one of the most pressing public health threats worldwide, and resistance in E. coli is now considered a critical threat to human and animal health. Indeed, AMR E. coli ranks among the top three antimicrobial-resistant priority pathogens reported by the World Health Organization (Cassini et al., 2019). This study analyzed trends in resistance and MIC distributions for antimicrobial agents used to treat canine E. coli urinary and non-urinary infections in the northeastern United States, isolated at the Cornell University AHDC between 2007 and 2020. As E. coli is the main pathogen isolated from UTIs in dogs, this infection site was the most commonly identified, followed by skin and soft tissues and reproductive system infections (Gómez-Beltrán et al., 2020; Hernando et al., 2021).

In the United States and worldwide, most dogs with an uncomplicated UTI are commonly treated empirically with a first-line antimicrobial, typically a penicillin or folate-pathway antagonist (Weese et al., 2019). Overall, we found that canine urinary isolates were significantly more likely to be pan-susceptible and less likely to be MDR than were non-urinary isolates. Surprisingly, this finding was not in agreement with the previous study carried out in the same setting (AHDC), which revealed that canine E. coli urinary isolates from 2004 to 2011 were significantly more likely to demonstrate in vitro resistance to the majority of antimicrobials and to be MDR than were non-urinary isolates (Cummings et al., 2015). We found higher resistance rates and MICs against ceftiofur, enrofloxacin, and trimethoprim-sulfamethoxazole among non-urinary isolates compared to urinary isolates (Tables 2–3). Ampicillin and amoxicillin-clavulanate MIC values were also higher in the non-urinary isolates (Table 3). Similar findings were also observed among other antimicrobial families, including tetracyclines and extended-spectrum cephalosporins (Table 1). The inconsistency of urinary vs non-urinary results between the two studies is probably due to different strategies of data analysis. Of note, some antibiotics reported in the previous study (Cummings et al., 2015) were not analyzed in this study because the available AMR data only reflect a subpopulation of MDR urinary E. coli isolates. For example, the previous study examined susceptibility to first-generation cephalosporins, cefoxitin, gentamicin, and chloramphenicol among all urinary isolates; here we noted that these antimicrobials were likely only tested among MDR urinary isolates, which does not reflect the true prevalence of resistance against these antimicrobials in this body site. However, both studies revealed similar resistance rates against these antimicrobials among non-urinary isolates. The previous study also showed that non-urinary isolates have higher resistance rates to other antimicrobials such as ampicillin, amoxicillin-clavulanate, enrofloxacin, and trimethoprim-sulfamethoxazole compared to urinary isolates (Cummings et al., 2015).

This study confirms that all of the tested first-line antimicrobial treatments for UTIs, except ampicillin, showed a relatively low resistance rate and are likely still effective for treating E. coli causing UTIs in dogs in the northeastern United States. Although aminopenicillins achieve high concentrations in the bladder, these antimicrobial agents were the least effective (resistance rate of 18–30%) among the first line choices. Aminopenicillin resistance is mainly associated with the production of class A and class C beta-lactamases. While this prevalence of ampicillin resistance is significantly lower than that observed in developing countries (Chang et al., 2015; Qekwana et al., 2018), it remains higher than studies from developed countries such as Canada (8.8–15%) (Courtice et al., 2016, 2021) and France (27.5%) (Valat et al., 2020). It is also noteworthy that ampicillin resistance (30% in our study) is lower than that previously reported in other studies in the United States (34.3–47%) (Cummings et al., 2015; KuKanich et al., 2020). Unsurprisingly, aminopenicillins were reported to be the most commonly prescribed antimicrobials at veterinary teaching hospitals (Wayne et al., 2011; Robbins et al., 2020; Goggs et al., 2021). This prescribing practice is also similar to those previously reported for primary care veterinary hospitals in the United States (Wayne et al., 2011; Fowler et al., 2016).

Unlike in developed countries, the use of antimicrobials in animals in developing countries is extensive and frequently observed without veterinary prescription or indication of infection (Kassem et al., 2019; Geta and Kibret, 2021). The inappropriate use of antimicrobials in veterinary medicine is considered one of the main drivers of the emergence of MDR bacteria fecal carriage and opportunistic infections, including extended-spectrum cephalosporin-resistant E. coli, in dogs and their owners (van den Bunt et al., 2019; Salgado-Caxito et al., 2021). We detected an increase in MIC values in all extended-spectrum cephalosporins and cefazolin among non-urinary isolates in 2018–2020 compared to the reference period. These results suggest that current antimicrobial use practices in canine medicine may be driving an increase in the emergence of third-generation cephalosporin-resistant E. coli in the northeastern United States. Extended-spectrum cephalosporin compounds are categorized as the highest priority critically important antimicrobials for human and animal health; thus, there are very few indications for the use of these antimicrobials as first-line antimicrobial treatment, and they are only recommended in specific situations where other treatments are not possible (Weese et al., 2019). For instance, cefovecin should only be used to treat bacterial UTIs in dogs where oral treatment is not possible (Weese et al., 2019). Although cephalosporins are widely used drugs in Europe (Buckland et al., 2016; Singleton et al., 2017; Schmitt et al., 2019), extended-spectrum cephalosporins only represented 4% of all prescriptions at Cornell University Hospital for Animals emergency and critical care services (Robbins et al., 2020). However, a recent report showed that extended-spectrum cephalosporins are more used in canine urinary (20–27%) and respiratory (7–15%) infections treated in United States hospitals (Aja and Bohn, 2017).

Even if the consumption of extended-spectrum cephalosporins is low in the northeastern United States, genetic co-resistance (i.e., resistance genes that are present on the same mobile genetic elements, conferring phenotypic resistance to multiple antimicrobials) could play a key role in promoting resistance to these antibacterial agents. For instance, our MLR models showed that enrofloxacin is significantly associated with ceftiofur resistance among both urinary (Table 5) and non-urinary isolates (Table 6). Several previous studies reported strong and significant associations between plasmid-mediated quinolone resistance (PMQR) genes and extended-spectrum beta-lactamase (ESBL) and AmpC genes such as bla_CTX-M-1, bla_CTX-M-15, and bla_CMY-2 among Enterobacterales in humans and dogs (Dupouy et al., 2019; Yassine et al., 2019). Furthermore, fluoroquinolones were among the top four antimicrobials prescribed by veterinarians to canine and feline patients in the emergency and critical care service of veterinary teaching hospitals in the United States (Aja and Bohn, 2017; Robbins et al., 2020; Goggs et al., 2021). Our MLR data showed that enrofloxacin is also a strong predictor for resistance to trimethoprim-sulfamethoxazole in both urinary and non-urinary isolates. Moreover, it is important to note that fluoroquinolone resistance rates were around 20% among all non-urinary E. coli isolates and reached 40% in intestinal infections. Hence, although our work did not detect an increasing resistance trend to enrofloxacin, we suggest reducing the use of fluoroquinolones to conserve their therapeutic effects and prevent the co-selection of resistance to other antimicrobials.

Interestingly, E. coli isolated from intestinal sample sites showed the highest percentages of resistance and/or MIC values for ticarcillin, cefazolin, cefovecin, cefpodoxime, gentamicin, doxycycline, chloramphenicol, enrofloxacin, marbofloxacin, and trimethoprim-sulfamethoxazole, after accounting for year of isolation. Around 50% of intestinal isolates showed resistance to ticarcillin and first-generation cephalosporins. Moreover, high rates of resistance were observed against third-generation cephalosporins (20.8–34.5%), doxycycline (35.8%), fluoroquinolones (34.2–39.1%), and trimethoprim-sulfamethoxazole (26.4%). Our findings suggested that antimicrobial use may have promoted the selection of the resistant E. coli of intestinal origin in the canine population. Antimicrobial treatment severely affects microbial diversity in the intestine and exerts selective pressure on bacteria, leading to the compromise of the exposed gut microbiota and the rise of MDR bacteria, including pathogenic E. coli (Wallace et al., 2020). Virulent E. coli are important pathogens related to public health concerns worldwide, which can cause gastrointestinal diseases in dogs and their owners. The current findings highlight an emergent public health issue—highly resistant intestinal E. coli strains that could lead to extra-intestinal infections and be transmitted to other animals, humans, and the environment (Denamur et al., 2021).

The rate of pet ownership in USA households is increasing. Currently, there are approximately 77 million pet dogs in the United States, living in 38% of households (Overgaauw et al., 2020). Since dogs are in close contact with their humans, the risk of transmission of drug-resistant E. coli and AMR determinants between animals and humans and to the environment is considered a potential threat to public health. Although only a few studies have evaluated the prevalence of AMR in the canine population, evidence-based data indicates the transfer of MDR bacteria and AMR genes between dogs and humans occurs through direct and indirect contact (Zhang et al., 2016; Derakhshandeh et al., 2018; Abbas et al., 2019). Therefore, there is an urgent need to tackle the burden of AMR bacterial infections through a One Health approach aimed at curbing transmission of AMR bacteria around the globe. This includes promoting better hygiene in communities following companion animal contact, particularly among vulnerable populations such as immunocompromised individuals, elderly adults, and children under the age of two years. Moreover, ending the inappropriate use of antimicrobials, implementing rapid and reliable antimicrobial susceptibility testing, and increasing awareness and knowledge of zoonotic diseases, antimicrobials, and AMR are all effective interventions to tackle the AMR threat in both human and veterinary settings.

This study has a few limitations. Due to the retrospective design of this study, we were unable to assess the susceptibility of antimicrobials in all body sites, or to collect more behavioral and clinical data that could be associated with resistance phenotypes. Moreover, as we do not have dog-identifying information, we are incapable of identifying isolates at different time points during an episode of illness. Carbapenem resistance was underestimated since ertapenem was not tested. Furthermore, since the isolates were tested against a limited number of beta-lactam agents, we were unable to identify whether extended-spectrum cephalosporin resistance is due to the production of an ESBL or an overproduced AmpC. Since isolates were not stored, we could not perform molecular analysis to determine the AMR gene profiles and identify the E. coli clones circulating in the northeastern United States. Molecular typing is essential to better understand the local epidemiology of E. coli from a One Health approach and, thus, to enhance the effectiveness of existing antimicrobials and develop potential interventions to prevent and treat MDR bacterial infections in healthcare and community settings.

In conclusion, this study provides a relevant update and an epidemiological evidence base for AMR associated with canine E. coli infections in the northeastern United States. To our knowledge, this investigation reported for the first time the prevalence of AMR among E. coli canine isolates from invasive and reproductive system infections in the United States. Although our work suggested that resistance to gentamicin and trimethoprim-sulfamethoxazole in E. coli is decreasing, resistance to extended-spectrum cephalosporins is on an upward trajectory compared to AMR data from the last decade. Enrofloxacin may play a key role in co-resistance to extended-spectrum cephalosporins and trimethoprim-sulfamethoxazole; thus, there is a need to review prescribing practices and antimicrobial stewardship programs to reduce the use of fluoroquinolones and the number of inappropriate prescriptions. For a better understanding of the local epidemiology of drug-resistant E. coli, further work based on a One Health approach is needed to explore AMR determinants, identify the circulating clones, and suggest interventions tackling AMR trends.

Abbreviations:

E. coli

Escherichia coli

AMR

Antimicrobial resistance

MDR

Multidrug-resistant

MIC

Minimum inhibitory concentration

MKT

Mann-Kendall test

MLR

Multivariable logistic regression

OR

Odds ratio

HR

Hazard ratio

CLSI

Clinical and Laboratory Standards Institute

PEN

Penicillins

PI

Penicillins with β-lactamase inhibitor

FGC

First-generation cephalosporins

FOX

Cephamycins

ESC

Extended-spectrum cephalosporins

IMP

Carbapenems

AMG

Aminoglycosides

TET

Tetracyclines

CHL

Phenicols

FQ

Fluoroquinolones

SXT

Folate pathway inhibitors

AHDC

Animal Health Diagnostic Center