Antimicrobial Resistance and Pharmacodynamics of Canine and Feline Pathogenic E. coli in the United States

Introduction

Multidrug resistant (MDR) Escherichia coli (E. coli) is emerging as a major health concern, impacting choices for empirical therapy.^1–3 Public awareness of the potential role of resistant E. coli as a cause of disease is increasing.⁴ Although much attention has focused on the impact of enteropathogenic E. coli (e.g., O157:H7) in human medicine, “other” E. coli, particularly extraintestinal strains, are also associated with morbidity and mortality in humans as well as dogs.^5–7 Evidence that pets and owners share extraintestinal E. coli is increasing.^8–10 E. coli is a reasonable sentinel organism by which the emergence of MDR might be monitored in either dogs or cats. Ubiquitous in nature, and able to adapt to a multitude of environments or the presence of antimicrobials, E. coli is characterized by a high mutagenic capacity and is capable of rapid division, assuring opportunity for spontaneous mutations.^11,12 As such, E. coli rapidly develops resistance toward multiple drugs when exposed to selected antimicrobials.^13,14 In dogs or cats, MDR E. coli has been reported in association with nosocomial infections and urinary tract infections.^15–17 Further, MDR E. coli has been documented in fecal microflora in canine intensive care unit patients and experimentally in dogs receiving standard doses of fluoroquinolones.^18,19 Whereas several systems exist for monitoring antimicrobial resistance in either humans or food animals, no federally funded mechanism exists for monitoring either dogs or cats. Proactive surveillance of pathogenic (those that cause harm to the patient) E. coli is indicated to: (1) establish its current and emerging resistance in dogs and cats and potentially their owners; (2) provide a basis for evaluating the need for, and success of, antimicrobial use guidelines in animals; and (3) lay a foundation for the study of factors that might be associated with the transfer of MDR E. coli between the pet and owner.

The purpose of this study was to describe the pharmacodynamic (i.e., minimum inhibitory concentration, MIC) statistics, proportion of resistance versus susceptible isolates, and the level (high versus low) of susceptibility or resistance for a limited population of feline and canine E. coli pathogens to those antimicrobial drugs commonly used by veterinary practitioners. This information was reported with the intent of establishing a baseline for future surveillance and providing practitioners with evidence that there is a need to change their approach to empirical drug selection and dose design when treating infections associated with E. coli in dogs and cats.

Materials and Methods

Sample Collection and Handling

Isolates (n = 376) of E. coli were submitted by participating veterinary microbiology laboratories to the investigator’s laboratory. Samples had been submitted to participating laboratories from May 2005 to Sep 2005 by veterinary practitioners after collection from either dogs or cats with presumed spontaneous infections. Submitting laboratories had cultured and isolated presumed infecting organisms and were requested to send all E. coli isolates to the authors’ laboratory collected during this time period. Laboratories submitting samples to the authors included both commercial diagnostic laboratories (CDL n = 9; n = 4, laboratories A–D) and veterinary teaching hospital laboratories ([VTH], n = 5, laboratories E–I). All submitted samples (save those from feces or anal sacs) were processed. Samples were stratified into four geographical areas (the south, west, central, and north regions of the USA as shown in Figure 1) based on the state of the practitioner and thus sample origin. Isolates were submitted to the authors’ laboratory on trypticase agar slants^a shipped overnight at room temperature.

View Figure

• Download as Powerpoint
• Download Figure

Results

Descriptive Pharmacodynamic Statistics

Distributions of MIC, in general, were similar in dogs and cats and were bimodal for each antimicrobial drug (Figure 2). The second, higher mode for each drug occurred at or above the highest MIC recorded (i.e., the limitation of the E test strip), except for GM. As such, the MIC₉₀ of each drug exceeded the R-MIC_BP for each drug by at least eight-fold, indicating high level resistance. The exception was GM in which the MIC₉₀ exceeded R-the MIC_BP by only two-fold (Table 1).

Table 1 Pharmacodynamic Data by Region for Pathogenic E. coli (n = 376) Collected from Dogs and Cats in the USA

*

Clinical and Laboratory Standards Institute breakpoints for AMX (ampicillin is the model drug) and AMXC have been refined (and are lower) since the implementation of this study. These breakpoints reflect the current urine susceptible breakpoints for these two drugs.²²

†

Geometric mean

AMX, amoxicillin; AMXC, amoxicillin trihydrate/clavulanate potassium; CFP, cefpodoxime; DXY, doxycycline; ENR, enrofloxacin; E. coli Escherichia coli; GM, gentamicin; MIC, minimum inhibitory concentration; MIC₅₀, median MIC; MIC₉₀ MIC required to inhibit the growth of 90% of organisms; R-MIC_BP, resistant MIC breakpoint; S-MIC_BP, susceptible MIC breakpoint; TMS, trimethoprim-sulfamethoxazole.

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

When considered by region, the MIC₉₀ for isolates in the north (CDL samples only) and south (VTH samples only) were above the R-MIC_BP for isolates for each drug. The MIC₉₀ values of isolates from the west were lower than the R-MIC_BP for ENR, GM, and TMS (Table 1). Comparisons among regions revealed that mean MICs were higher in the south compared with other regions for each drug (Table 1). When considered by laboratory type, the MIC₉₀ for VTH isolates were also above the resistant MIC_BP for each drug. In contrast, the MIC₉₀ for CDL isolates were lower than the MIC_BP for ENR, GM, and TMS (Table 2). Comparisons between laboratory types revealed that the mean MICs were higher in samples from the VTHs than CDLs for all drugs (Table 2) (P < 0.001).

Table 2 Pharmacodynamic Data for CDLs and VTHs for Pathogenic E. coli (n = 376) Collected from Dogs and Cats in the USA

*

The upper and lower MIC limits refer to the ranges tested by the E strip of the respective drug.

†

Clinical and Laboratory Standards Institute breakpoints for AMX (ampicillin is the model drug) and AMXC have been refined (and are lower) since the implementation of this study.²²

§

The geometric mean was lower (P < 0.001) for each drug for isolates submitted by the CDLs compared with the VTHs.

AMX, amoxicillin; AMXC, amoxicillin trihydrate/clavulanate potassium; CFP, cefpodoxime; CDL, commercial diagnostic laboratory; DXY, doxycycline; ENR, enrofloxacin; GM, gentamicin; MIC, minimum inhibitory concentration; MIC₅₀, median MIC; MIC₉₀, MIC required to inhibit the growth of 90% of organisms; R-MIC_BP, resistant MIC breakpoint; S-MIC_BP, sensitive MIC breakpoint; TMS, trimethoprim-sulfamethoxazole; VTH, veterinary teaching hospital laboratory.

View Fullscreen

Proportion of Resistant Versus Susceptible Isolates

The proportion of all isolates designated as resistant to at least one of the seven antimicrobials was 55% ± 2.6%. The order of proportion of resistance was AMX > AMXC > CFP = DXY = ENR = TMS > GM. The resistance for AMX was 46% ± 2.6% and 12% ± 1.7% for GM (P < 0.001; Table 3). A significant difference could not be detected in the proportion of either resistance between species (Figure 3) or gender status. Resistance among tissue sources ranged from 41% in skin to 13% in ears. Resistance in isolates from the ear and reproductive tract was lower than abdominal fluid, respiratory tract, urinary tract, and skin (P < 0.001) (Figure 4).

Table 3 Proportion of Resistant Pathogenic Canine or Feline E. coli by Region for Each Drug

*

Resistance in the south was significantly higher than all other regions for each drug.

†

Resistance in the central region was greater than the west.

§

The proportion of overall resistance differed among means.

**

The proportion of overall resistance differed among means.

††

The proportion of overall resistance differed among means.

§§

The proportion of overall resistance differed among means.

AMX, amoxicillin; AMXC, amoxicillin trihydrate/clavulanate potassium; CFP, cefpodoxime; DXY, doxycycline; ENR, enrofloxacin; GM, gentamicin; TMS, trimethoprim-sulfamethoxazole.

View Fullscreen

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

When considered by region, the proportion of isolates resistant to at least one drug was greatest in the south (VTH samples only) compared with all other regions. Resistance was greater in the central region compared with the west (P < 0.001; Table 4). When considered by laboratory type, the proportion resistance was greater in samples from the VTHs than the CDLs for each drug (P < 0.001; Table 3).

Table 4 Pharmacodynamic Data for Susceptible Pathogenic E. coli (n = 376) Collected from Dogs and Cats in the USA

*

The upper and lower limits refer to the ranges tested by the E strip of the respective drug.

†

Geometric mean

§

Susceptible MIC breakpoint to the MIC₉₀, indicating how susceptible the isolate is to the drug of interest

AMX, amoxicillin; AMXC, amoxicillin trihydrate/clavulanate potassium; CFP, cefpodoxime; DXY, doxycycline; ENR, enrofloxacin; GM, gentamicin; MIC, minimum inhibitory concentration; MIC₅₀, median MIC; MIC₉₀, MIC required to inhibit the growth of 90% of organisms; R-MIC_BP resistant MIC breakpoint; S-MIC_BP, sensitive MIC breakpoint; TMS, trimethoprim-sulfamethoxazole.

View Fullscreen

The level of both resistance and susceptibility were also examined by comparing the relationship of MIC₉₀ to the S-MIC_BP or R-MIC_BP for each drug. The proportion of all isolates and the proportion of resistant isolates expressing high level resistance (i.e., MIC₉₀:R-MIC_BP ≥ 8) for each drug was, respectively, was 38% and 99% for AMX, 14% and 66% for AMXC, 16% and 83% for CFP, 15% and 98% for ENR, 16% and 81% for DXY, 08% and 68% for GM, and 16% and 98% for TMS. Statistics were also determined for susceptible isolates (rather than all isolates) so levels of susceptibility could be examined (Table 4). For susceptible isolates (based on the S-MIC_BP:MIC₉₀), the drugs to which isolates were least susceptible were AMX and AMXC (S-MIC_BP:MIC₉₀ = 1) and the drug to which isolates were most susceptible was TMS (S-MIC_BP:MIC₉₀ > 10.

Discussion

As with previous studies, the authors of this report identified a high proportion of resistance in canine and feline pathogenic E. coli toward commonly used therapeutic drugs, particularly for AMX and AMXC. Another goal of this study was to address the level of both resistance and susceptibility for the isolates. A low level resistance cannot be discriminated from a high level resistance using standard commercial susceptibility testing cards because the highest concentration tested generally is very close to the R-MIC_BP. Such discrimination could be advantageous, however, because antimicrobials characterized by low-level resistance might yet be used to treat an infected patient either at an increased dose or in combination with another antimicrobial drug with a complementary mechanism of action. This study demonstrates that proportion of antimicrobial resistance expressed by E. coli is often high, ranging from 65% of isolates expressing resistance to AMXC to ≥ 97% expressing resistance to ENR, AMX, and TMS.

This study also focused on differences in susceptibility among susceptible drugs. In contrast to standard commercial susceptibility cards, the E test includes concentrations well below the S-MIC_BP. Commercial tests provide little to no indication of how susceptible an isolate might be to a drug; however, in this study, TMS had the highest S-MIC_BP:MIC₉₀. A S-MIC_BP:MIC₉₀ of 10 indicates that the population isolate MIC for TMS was at least 10 times lower than the upper criteria upon which CLSI bases susceptibility. In contrast, the ratios for AMX and AMXC were only 1. This suggests that empirically, TMS would be a better first choice treatment of E. coli infections than either AMX or AMXC.

This study also demonstrates the potential use of surveillance data for the design of dosing regimens. Ideally, such dosing regimens are based on a combination of pharmacokinetics and pharmacodynamics.^23,24,26 The MIC₉₀ of susceptible E. coli reported herein (8 μg/mL) for the time-dependent drugs AMX and AMXC was higher than the peak AMX plasma concentration (4–6 μg/mL) achieved for AMX at the labeled dose (10–11 mg/kg) (Table 1).²⁷ For ENR, although the susceptible E. coli MIC₉₀ (0.19 μg/mL) was well below the peak drug concentration achieved at the labeled oral low dose of 5 mg/kg (1.24 ± 0.39 μg/mL ENR), as a concentration-dependent drug, efficacy is facilitated if the peak drug concentration to MIC ratio is ≥ 8–10.^24,26,28 The more appropriate target of 1.5–1.9 μg/mL (i.e., 8–10 times the MIC₉₀ of susceptible isolates) would not be reached at the lower label dose, indicating a higher dose would be more prudent. As stated above, the β-lactams (AMC and AMXC) are time-dependent drugs. For such drugs, drug concentrations ideally should exceed the MIC of the infecting organism for at least 50% of the dosing interval.^24,26 As such, efficacy of AMX and AMXC may be hindered by the very short elimination half-life of AMX (≤ 1.5 hr), particularly if dosed twice daily.²⁷ In contrast, the elimination half-lives of potentiated sulfonamides, which are also time-dependent drugs, are generally much longer (≥ 8 hr), favoring efficacy at a more convenient dosing interval.²⁹

The majority of isolates in this study were collected from urine. The CLSI determination of breakpoints, the criteria on which isolate susceptibility or resistance was based in this study, are based on plasma concentrations; however, an exception is made for AMX and AMXC. For those drugs, isolate susceptibility and resistance are based on urine concentrations if the E. coli is an uropathogen.²⁴ In general, urine concentrations of several of the drugs studied here (e.g., AMX, ENR) would exceed that achieved in plasma. Accordingly, a designation of resistance based on plasma drug concentrations might not equate to a “resistant” designation based on urine concentrations. That said, assuming a lack of relevance of plasma-based breakpoints to isolates causing infection in the urine may not be prudent.²⁴ Indeed, the assumption that a urinary isolate is exposed to higher concentrations of drug compared with one in plasma assumes that urine is being concentrated, that is, renal function is normal and that the urine concentration is not influenced by drugs, fluid therapy or high water intake, or nonrenal diseases. Further, it assumes that the infection is limited to the urine and does not involve the bladder wall, that biofilm is not protecting the isolate, and (particularly for time-dependent drugs) that urine remains in the bladder for the appropriate proportion of the dosing interval (i.e., 50% of the dosing interval).^24,26 Finally, for isolates not located in the urine, the converse may be true, particularly for nonlipid soluble drugs such as β-lactams or aminoglycosides. That is, plasma drug concentrations may overestimate concentrations achieved in tissues.²⁴ Although further studies are needed to determine if detection of differences in susceptibility will be translated to improved antimicrobial efficacy in the infected patient, this study supports the utility of MIC as a basis for antimicrobial therapy of E. coli and the need for standard susceptibility that incorporate drug concentrations substantially lower than the CLSI S-MIC_BP.

As with other studies, this study found that resistance to ENR was substantial (approximately 20%).^{11,16–18,30} Previously, the authors of this study reported that only 3% of the 84 E. coli isolates that expressed single drug resistance were resistant to ENR. This is much lower than the 82% of E. coli isolates expressing single drug resistance toward β-lactams.² Thus, ENR resistance in E. coli is commonly high-level and associated with MDR. That the lowest incidence of resistance in E. coli was expressed toward GM was not surprising considering that concerns for nephrotoxicity and the need for injection tends to limit its general use in both canine and feline populations. Likewise, the relatively low incidence of resistance for DXY and TMS may reflect a general decline in their use over the past decade by veterinarians. Presumably, widespread use by DXY is limited due to resistance and the use of TMS is limited due to adverse events.

This study provides limited information regarding potential risk factors for resistance. Neither species nor gender appear to be a risk factor associated with E. coli resistance, but tissue of origin does seem to be a risk factor because isolates from the respiratory tract, abdomen, skin, and urine had the highest resistance.

In this study it was not possible to discriminate infection (i.e., confirming the isolates as pathogens) from colonization.³¹ This reflects, in part, the lack of antimicrobial use history. One might assume that samples studied were collected from animals that had failed initial empirical antimicrobial therapy. As such, the populations sampled in this study may have been more likely to reflect infecting rather than colonizing isolates. However, previous treatment, including exposure to antimicrobials, may also decrease the applicability of information in this study to drug naïve E. coli isolates in dogs or cats. When only susceptible isolates were considered in this study, the sample population mode, MIC₅₀, and MIC₉₀ for AMXC (4 μg/mL, 4 μg/mL, and 8 μg/mL, respectively) are the same as reported for a population of pathogenic E. coli (n = 223) collected from antimicrobial naïve dogs or cats.³² Separation of antimicrobial naïve versus previously treated animals is likely to be important when assessing patterns of resistance and susceptibility. This study lacks other surveillance information that might be used in the prevention and control of emerging resistance.^33,34

Most problematic in this study is the authors’ inability to separate the impact of region and laboratory type on resistance. The authors’ options were to either ignore potential differences or report the categories separately while acknowledging that a potential predominant confounding effect had not been identified. The high proportion of resistance found in the south may reflect the VTH origin of samples. Treatment failure may be more likely in patients from the south because VTH patients tend to be referred patients, often with recurrent infections previously exposed to antimicrobials. Nonetheless, the impact of laboratory type should be balanced by both the fact that up to 50% of samples received by the VTHs were submitted by practitioners and up to 30% of CDL samples were submitted by referral and specialty practices. Further, differences occurred between the western and central regions (Table 4), each represented by both VTHs and CDLs (Table 3). Finally, although small sample size likely limited detection of significant differences, numerical differences in resistance among both CDLs (laboratory B versus C) and the VTHs (laboratory E versus F) suggests clinically, albeit not statistically, relevant regional differences (Table 3). The potential for regional differences underscores the importance of a robust survey for understanding emergent E. coli antimicrobial resistance.

Conclusion

Based on the results of this study, the proportion of canine and feline E. coli pathogens resistant to AMX and AMXC is high. Although this study could not separate out all confounding factors, proportion and level of resistance varies regionally, with potential influencing factors being region and laboratory type. Resistance, when it occurs, tends to be high-level, especially for AMX, ENR, and TMS. For susceptible isolates, the MICs for AMX and AMXC are close to the susceptible breakpoint, whereas the MIC for TMS is well below the susceptible breakpoint. Results of this study suggest that empirical use of selected antimicrobials to treat E. coli in dogs and cats should be based on culture and MIC susceptibility testing, and design of dosing regimens should take into account proximity of isolate MICs to the MIC breakpoints. Results of this study suggest the need for broader MIC ranges as part of standard susceptibility testing and a robust surveillance program for antimicrobial resistance in dogs and cats.