Materials and Methods

Study overview—Fecal samples were collected over a 2-year period (December 2002 through February 2005) from 3 groups of horses. The NTSEC isolates recovered from the fecal samples were evaluated to determine their susceptibility to a standardized panel of AMDs. Susceptibility profiles of the isolates were then compared among the 3 groups of horses.

Study population—Horses enrolled in the 3 study groups represented a convenience sample of horses that were examined at the James L. Voss Veterinary Teaching Hospital during the study period or that were owned by Veterinary Teaching Hospital staff members, students, or their friends. The horses were identified as being in one of the following groups: HOSP–AMD, HOSP–NOAMD, or community. The HOSP–AMD group consisted of horses that were hospitalized at the Veterinary Medical Center and had been hospitalized and treated with AMDs for at least 3 days prior to sample collection. The HOSP–NOAMD group consisted of horses that were hospitalized at the Veterinary Medical Center for at least 5 days and had not received any AMDs for at least 4 days prior to sample collection. The community group consisted of horses without apparent illness that were not hospitalized and had not received any AMDs for at least 5 days prior to sample collection (most horses in this group had not received any AMDs for several months prior to sample collection). Horses in the HOSP–NOAMD group were subcategorized as horses that had no history of treatment with AMDs during the current hospitalization period, horses that had a history of recent AMD treatment (but had not received any AMDs for at least 4 days), and healthy mares that had been hospitalized with their foals and whose foals were being treated with AMDs.

Sample collection—During the study period, horses hospitalized at the Veterinary Medical Center were monitored several days each week, and horses eligible for enrollment were identified on the basis of information obtained from the medical records. Fresh fecal samples were collected from stall floors or by means of rectal retrieval from horses enrolled in the study. Samples were placed in a sterile container and processed immediately or refrigerated overnight. In general, a single fecal sample was collected from each horse enrolled in the study. However, multiple samples were collected from horses that were hospitalized for prolonged periods.

Information regarding each horse enrolled in the study was gathered from the Veterinary Medical Center's database system (HOSP–AMD and HOSP–NOAMD groups) or from the horse's owner (community group). Additionally, for hospitalized horses, admission date and information regarding treatment with AMDs were recorded.

Culture methods—All fecal samples were processed within 24 hours after collection. Standard bacterial culture procedures were used to recover NTSEC from the samples.²⁴ In brief, each fecal sample was plated directly on a MacConkey agar plate^a that was incubated for 18 to 24 hours at 37°C. Colonies were presumptively identified as NTSEC on the basis of colony morphology, lactose fermentation, and a positive indole reaction after subculture on tryptic soy agar with 5% sheep blood.^a Up to 3 NTSEC colonies were selected by convenience from each culture. These colonies were inoculated into tryptic soy broth and frozen at −80°C for future antimicrobial susceptibility testing. A subset of isolates (n = 40) was subjected to further identification with 1 or 2 commercially available microbial identification systems^b,c according to the manufacturers' instructions.

Antimicrobial susceptibility testing—All isolates were tested for susceptibility to 16 AMDs (amikacin, amoxicillinclavulanate, ampicillin, cefoxitin, ceftiofur, ceftriaxone, cephalothin, chloramphenicol, ciprofloxacin, gentamicin, kanamycin, nalidixic acid, streptomycin, sulfamethoxazole, tetracycline, and trimethoprim-sulfamethoxazole). Minimum inhibitory concentrations of the AMDs were determined by use of a broth microdilution method in a semiautomated antimicrobial susceptibility testing system^d in accordance with the manufacturer's instructions and guidelines published by the Clinical and Laboratory Standards Institute.²⁵ The panel of AMDs^e used for this testing was that established for use by the National Antimicrobial Resistance Monitoring System for enteric bacteria, included drugs important to veterinary and human medicine, and included representatives of the major AMD classes.²⁶ Isolates were categorized as susceptible, of intermediate susceptibility, or resistant to each AMD on the basis of interpretive guidelines published by the Clinical and Laboratory Standards Institute or on the basis of interpretive criteria established by the National Antimicrobial Resistance Monitoring System for gram negative bacterial species. Reference strains of E coli (American Type Culture Collection 25922),^f Enterococcus faecalis (American Type Culture Collection 29212),^f Staphylococcus aureus (American Type Culture Collection 29213),^f and Pseudomonas aeruginosa (American Type Culture Collection 27853)^f were used as reference isolates for quality control during antimicrobial susceptibility testing.

Data analysis—Information regarding signalment of horses enrolled in the study, antimicrobial treatment history, and susceptibility results for isolates were entered into a computer spreadsheet and validated to ensure data integrity. Distributions of values for minimum inhibitory concentrations and susceptibility classification were summarized and evaluated. Log-linear modeling^g was used to estimate the least square mean prevalence of resistant isolates for each study group and the 95% CI for each prevalence estimate. Generalized estimating equation methods were used to correct estimates for correlations created by using multiple isolates from individual fecal samples, collecting multiple fecal samples from individual horses, and enrolling multiple horses from a single premise. For these analyses, study group was used as the exposure variable of interest and separate models were developed to estimate prevalences of isolates resistant to 15 of the 16 AMDs tested. It was not possible to estimate the prevalence of isolates resistant to amikacin, as the range of dilutions for that drug (0.5 to 4 mg/mL) that were included on the commercial plate used for susceptibility testing^d did not include the resistance breakpoint (16 mg/mL).

Potential confounding variables were screened as categorical covariates in log-linear models to determine whether they were associated with differences in resistance prevalence. These potential confounding variables included sex of the horses (sexually intact male, castrated male, female), breed (Quarter Horse–type [including Quarter Horse, Paint, Appaloosa, and crosses of these breeds], Arabian, Morgan Horse, Thoroughbred [including crosses], Warmblood breeds, and other breeds [including other mixed-breed horses]), and month that samples were collected. In addition, duration of hospitalization at the time of sample collection was compared between the 2 groups of hospitalized horses (HOSP–AMD vs HOSP–NOAMD) with the Wilcoxon rank sum and Kruskal-Wallis tests.

To evaluate potential associations between AMD use in horses and antimicrobial resistance in isolates, AMDs used to treat HOSP–AMD horses were categorized by class (aminoglycoside, penicillin, potentiated sulfonamide, cephalosporin, nitroimidazole, fluoroquinolone, tetracycline, and other) and analyzed as exposure variables in logistic regression models.^g Variables for exposure to each drug class (yes vs no) were screened in backward stepping logistic regression models for associations with resistance to each AMD. Odds ratios and their 95% CIs were obtained from the regression models. The critical a used for all statistical comparisons was 0.05.

Results

Distributions for age, sex, and breed were generally similar for the 3 groups of horses, and these variables were not found to be statistically associated with differences in prevalence of antimicrobial resistance (Table 1).

Similarly, differences in antimicrobial resistance prevalence were not associated with the month in which samples were obtained. Models that accounted for collection of multiple samples from individual horses and for testing of multiple isolates from individual samples were not substantially different from models that also accounted for enrollment of multiple horses from the same premises with an adequately robust number of data points to detect a difference. Thus, the simpler, 2-level correlation structure was used for all final log-linear and logistic regression models.

There was no significant (P = 0.15) difference in duration of hospitalization at the time of sample collection between the HOSP–AMD group (median, 7 days; interquartile range [ie, 25th to 75th percentile], 5 to 11 days) and the HOSP–NOAMD group (median, 7 days; interquartile range, 5 to 15 days). However, there were significant (P = 0.004) differences in duration of hospitalization among the 3 subgroups of HOSPNOAMD horses. Median duration of hospitalization was 19 days (interquartile range, 7 to 48 days) for horses with a history of previous AMD treatment, 7 days (interquartile range, 6 to 8 days) for mares housed with their foals, and 6 days (interquartile range, 5 to 12 days) for horses without any history of recent AMD treatment.

Horses in the HOSP–AMD group had been treated with 11 different AMDs prior to collection of fecal samples (Table 2), with the most commonly used AMDs being gentamicin, potassium penicillin, and trimethoprim-sulfamethoxazole (alone or in combination with other drugs). Multiple fecal samples were collected from 13 horses in this group. Of these horses, 8 were being treated with different combinations of AMDs when additional samples were collected. Twenty-seven fecal samples were collected from 22 horses that had been treated with a single AMD during the 5 days prior to sample collection, 40 samples were collected from 34 horses that had been treated with 2 AMDs, 19 samples were collected from 19 horses that had been treated with 3 AMDs, and 1 sample was collected from a horse that had been treated with 4 AMDs. The most common patterns of AMD use during the 5 days prior to fecal sample collection were penicillin and gentamicin (29 samples from 24 horses); trimethoprim-sulfamethoxazole alone (23 samples from 19 horses); penicillin, gentamicin, and trimethoprim-sulfamethoxazole (11 samples from 11 horses); gentamicin and ampicillin (5 samples from 4 horses); gentamicin alone (2 samples from 1 horse); gentamicin and cefazolin (2 samples from 2 horses); and penicillin, gentamicin, and metronidazole (2 samples from 2 horses). The remaining 13 samples were collected from 13 horses treated with other individual AMDs or other combinations.

Isolates of NTSEC were not recovered from 17 fecal samples, including 2 samples collected from horses in the HOSP–AMD group, 8 samples from horses in the HOSP–NOAMD group, and 7 samples from horses in the community group; data from these horses were removed from analyses. A total of 724 NTSEC isolates were obtained from the remaining 250 fecal samples (mean, 2.9 isolates/sample). Thirty-eight of the 40 isolates submitted for additional testing with a microbial identification system^b were identified as E coli. The remaining 2 isolates were not identified as E coli with this identification system, but were identified as E coli with the second microbial identification system.^c For 127 fecal samples, different colonies from the same fecal sample had identical susceptibility profiles. For 117 fecal samples, different colonies from the same fecal sample had 2 or 3 different susceptibility profiles. The remaining 6 fecal samples yielded only a single E coli colony.

Compared with isolates from horses in the community group, isolates from horses in the HOSP–AMD group (OR, 12.6; 95% CI, 6.93 to 22.77) and from horses in the HOSP–NOAMD group (OR, 4.6; 95% CI, 2.63 to 8.18) were significantly more likely to be resistant to at least 1 AMD (Table 3). Overall, 73% of NTSEC isolates from HOSP–AMD horses and 50% of NTSEC isolates from HOSP–NOAMD horses were resistant to ≥1 AMD, whereas only 17.7% of NTSEC isolates from community horses were resistant to ≥1 AMD. Similarly, 23% of NTSEC isolates from HOSP–AMD horses and 5% of NTSEC isolates from HOSP–NOAMD horses were resistant to ≥6 AMDs, but none of the isolates from community horses were resistant to this many AMDs. There were no significant (P > 0.10) differences in prevalences of resistance to any of the AMDs tested among the 3 subgroups of HOSP–NOAMD horses.

For 12 of the 15 AMDs, there were significant differences in the prevalence of antimicrobial resistance among the 3 groups (Figure 1), with prevalence being lowest among isolates from community horses, intermediate among isolates from HOSP–NOAMD horses, and highest among isolates from HOSP–AMD horses. For all 3 groups, substantial proportions of NTSEC isolates were resistant to sulfamethoxazole and trimethoprim-sulfamethoxazole. In addition, high proportions of NTSEC isolates from hospitalized horses were resistant to streptomycin, ampicillin, tetracycline, and gentamicin. In all 3 groups, the most common resistance phenotype was resistance to sulfamethoxazole and trimethoprim-sulfamethoxazole (Table 4).

View Full Size

Least square mean estimates of the prevalence of resistance to various AMDs among NTSEC isolates obtained from fecal samples from 3 populations of horses. Error bars represent 95% CI. *Least square mean prevalences of resistance varied significantly (P < 0.05) among groups. AMC = Amoxicillin-clavulanate. AMP = Ampicillin. CFX = Cefoxitin. CTF = Ceftiofur. CTR = Ceftriaxone. CPH = Cephalothin. CHL = Chloramphenicol. CIP = Ciprofloxacin. GEN = Gentamicin. KAN = Kanamycin. NAL = Nalidixic acid. STR = Streptomycin. SUL = Sulfamethoxazole. TET = Tetracycline. TMS = Trimethoprim-sulfamethoxazole.

Citation: Journal of the American Veterinary Medical Association 228, 12; 10.2460/javma.228.12.1909

• Download Figure
• Download figure as PowerPoint slide

Use of a potentiated sulfonamide, aminoglycosides, and cephalosporins was positively associated with increased odds that isolates would be resistant to AMDs from the same class (Table 5). In addition, use of cephalosporins or metronidazole was positively associated with an increased risk of resistance to other classes of AMD. Overall, use of cephalosporins appeared to have the greatest effect on selection of resistant phenotypes, as this was the only variable associated with resistance to ≥6 AMDs (OR, 39.5; 95% CI, 4.5 to 347.3). Use of cephalosporins (OR, 17.7; 95% CI, 2.1 to 153.0; P = 0.02) and use of metronidazole (OR, 22.9; 95% CI, 3.5 to 152.2; P = 0.04) were positively associated with resistance to ≥4 AMDs. Use of penicillin was not associated with resistance to any of the AMDs tested.

Discussion

Results of the present study suggest that both hospitalization and AMD use were associated with prevalence of antimicrobial resistance among E coli strains isolated from the feces of horses. It has generally been accepted that AMD use selects for resistant bacteria, and similar conclusions have been reached by other investigators.^7,10–12 The association between AMD use and an increased prevalence of antimicrobial resistance among NTSEC isolates was further supported by the similarity in hospitalization time for the HOSP–AMD group and HOSP–NOAMD group prior to fecal sample collection. Thus, NTSEC isolates recovered from HOSP–AMD horses were not more likely to be resistant to AMDs solely because of longer exposure of the horses to the hospital environment. Also, for horses that had been treated with AMDs, use of a potentiated sulfonamide, aminoglycosides, cephalosporins, or metronidazole was associated with increased odds of antimicrobial resistance among NTSEC isolates; use of cephalosporins or metronidazole was positively associated with resistance to ≥4 AMDs; and use of cephalosporins was positively associated with resistance to ≥6 AMDs. Our findings may have important clinical implications because they suggest that use of cephalosporins had the greatest potential to promote selection of multidrug resistance in NTSEC isolates. In a similar study²⁷ performed in a human hospital, use of ticarcillin-clavulanate, trimethoprim-sulfamethoxazole, and ceftazidime were most commonly associated with decreased susceptibility to other AMDs. Also, an increased rate of infections caused by multidrug-resistant Enterobacter strains has been identified in human patients who received prior treatment with third-generation cephalosporins.²⁸

In contrast, use of penicillin or ampicillin did not appear to promote resistance to any of the AMDs evaluated, and NTSEC isolates from horses treated with penicillin or ampicillin were less likely to be resistant to trimethoprim-sulfamethoxazole, cefoxitin, amoxicillin-clavulanate, or ceftiofur than were NTSEC isolates from horses treated with other AMDs. This apparent protective effect is puzzling and may be a reflection of relatively stronger associations between use of other AMDs and resistance. In contrast with the usual paradigm that AMD use selects for antimicrobial resistance, some AMD use seems to be associated with decreased antimicrobial resistance.²⁷ It is also worth pointing out that the presence of any of these associations does not necessarily imply causation. As such, it is possible that some unidentified factors (such as the severity of illness, diet, or environmental or health history prior to admission) may have influenced results of the present study. The possibility of horizontal transfer of resistant bacteria may be another confounding factor when trying to identify relationships between AMD use and antimicrobial resistance.²⁹

The complicated relationship between AMD use and antimicrobial resistance is further highlighted by the fact that, with the exception of ampicillin, prevalences of resistance to b-lactam AMDs and gentamicin were lower than prevalences of resistance to tetracycline, streptomycin, sulfamethoxazole, and trimethoprim-sulfamethoxazole, despite the fact that horses in the study were frequently treated with penicillin and gentamicin. With the exception of the high prevalence of resistance to trimethoprim-sulfamethoxazole, these results are similar to findings from studies^30–32 of clinical E coli isolates of human and animal origin, for which resistance to streptomycin, simple sulfonamides, and tetracycline were most common. Similarly, among equine E coli isolates collected from diagnostic laboratories in the 2004 National Antimicrobial Resistance Monitoring System study,²⁶ resistance prevalence was highest for tetracycline (44.4%), sulfamethoxazole (44.4%), streptomycin (40.7%), ampicillin (40.7%), and gentamicin (37.0%). The most frequent resistance phenotype among E coli isolates from all animal sources in that study²⁶ was combined resistance to gentamicin, streptomycin, sulfamethoxazole, and tetracycline.

Some isolates in the present study were resistant to some AMDs, such as streptomycin, tetracycline, and chloramphenicol, that were never, or only rarely, used in equine patients at the Veterinary Teaching Hospital during the study period. This suggests that factors other than frequency of use of these specific AMDs in the patient population likely affected resistance patterns seen in the NTSEC isolates. Factors that may have played a role include coselection with other genetic determinants or exposure to these specific AMDs outside the hospital environment.

At a genetic level, AMD resistance could be acquired by bacteria through mutation or by acquisition of extrinsic DNA elements, such as plasmids or phages.³³ The spread of resistance is greatly facilitated by incorporation of resistance genes into movable DNA elements, such as integrons or transposons, which allows several resistance genes to be physically linked and potentially coselected.³³ Coselection was thought to be the reason behind frequent chloramphenicol resistance observed in a recent survey⁸ of antimicrobial susceptibility of bacteria from food animals in various European countries. Interestingly, 29.6% of equine E coli isolates in the 2004 National Antimicrobial Resistance Monitoring System study²⁶ were resistant to chloramphenicol, which was as high as the prevalence of resistance to trimethoprim-sulfamethoxazole. A common sulfonamide resistance gene has been found to be linked to streptomycin resistance, which may explain the presence of streptomycin resistance in many E coli isolates in the present and previous studies^34–36 despite the fact that use of streptomycin in human and veterinary medicine was discontinued many years ago. Tetracycline resistance is widespread among pathogenic and commensal bacteria, probably because of extensive use of tetracyclines in human and veterinary medicine in combination with the large number of genes that code for tetracycline resistance.^37,38 An additional explanation for the persistence of streptomycin and tetracycline resistance is the fact that streptomycin and oxytetracycline are the only 2 AMDs approved for use in plant agriculture in the United States, increasing the amounts of these AMDs in the environment and the potential for bacterial exposure to these agents.³⁹

Interestingly, combined resistance to sulfamethoxazole and trimethoprim-sulfamethoxazole was the most common resistance phenotype among NTSEC isolates from all 3 groups in the present study. This could reflect the frequent use of trimethoprim-sulfamethoxazole in hospitalized horses and in the general equine population. Several authors have described a positive association between the presence of the sul1 gene coding for sulfamethoxazole resistance and the presence of class 1 integrons,^{23,31,34,40,41} which may contain genes responsible for resistance to other classes of AMDs.

In the present study, NTSEC strains resistant to at least 1 AMD were more likely to be isolated from hospitalized horses that had not received AMDs in the preceding 4 days (HOSP–NOAMD group) than from horses that had not been hospitalized (community group). It is possible that administration of AMDs to some HOSP–NOAMD horses > 4 days prior to collection of fecal samples may have affected these results. However, the lack of difference in susceptibility patterns among isolates from the 3 subgroups of HOSP–NOAMD horses suggests that resistant NTSEC strains were acquired simply through exposure to the hospital environment. If this hypothesis is true, then the fact that the median duration of hospitalization for HOSP–NOAMD horses with no history of AMD treatment throughout their hospital visit (6 days) was shorter than the median duration of hospitalization for HOSP–NOAMD horses with a history of previous AMD treatment (19 days) suggests that exposure to the hospital environment may affect antimicrobial susceptibility of commensal flora within the first week of hospitalization. Disinfectants and AMDs are used more commonly within hospitals than they are in the community. As such, the commensal flora from all patients is exposed to greater selective pressures when patients are hospitalized than when they are living in the community.

In the present study, it is also possible that resistant bacteria were transferred from horses that had been treated with AMDs to their nontreated neighbors by means of routine management procedures, such as cleaning, brushing, feeding, and treatment. However, if these factors had a major influence on results reported here we would have expected to recover isolates with higher resistance prevalences from the subgroup of healthy mares stabled together with their sick foals when compared to the remaining 2 subgroups of HOSP–NOAMD horses, which was not the case. Finally, it is possible that the general difference in health between hospitalized horses and horses in the community meant that hospitalized horses were more likely to have been treated with AMDs prior to admission to the Veterinary Teaching Hospital, so that even horses classified as having had no exposure to AMDs while at the Veterinary Teaching Hospital could have been treated with AMDs within a period of time that would have affected antimicrobial susceptibility test results. Although some of the HOSP–NOAMD horses were known to have been previously treated with AMDs, none of the community horses had a history of AMD treatment during the months or sometimes years prior to collection of fecal samples. It is not known how long the effect of AMD treatment influences the composition of the microflora of the equine gut. Such influence will likely depend on many variables, such as the type of AMD used, the dosage, and the route of administration. In a previous study,⁴² administration of quinupristin-dalfopristin had an effect on the fecal microflora of human volunteers for up to 12 weeks after the end of treatment. Thus, it is possible that administration of AMDs > 4 days prior to collection of fecal samples in HOSP–NOAMD horses in the present study might have affected our results. After examining patient records from previous years, the 4-day nontreatment period was chosen because it was the maximum period without treatment that would still allow enrollment of similar numbers of hospitalized horses into the HOSP–AMD and HOSP–NOAMD groups during the study period. In addition, the required 3-day treatment period was chosen because review of patient records showed that equine patients at the Veterinary Medical Center receiving AMDs were most commonly treated for 3 to 5 days.

Infection with multidrug-resistant bacteria can have serious clinical consequences. In humans, such infections increase the overall cost of treatment and duration of hospitalization and have detrimental effects on patient survival rates.⁴³ In addition, multidrug-resistant bacteria may be more virulent because of coselection of antimicrobial resistance genes and genes coding for virulence factors.⁴³ Treatment of veterinary patients infected with multidrug-resistant bacteria may be particularly challenging because of the limited number of AMDs to which such bacteria are susceptible, concerns that use of newer classes of AMDs in animals may have an adverse effect on human health, and potential adverse reactions related to AMD treatment. For example, it has been shown that for horses infected with some resistant Salmonella strains, prior treatment with AMDs is an important risk factor for development of clinical salmonellosis. In these horses, exposure to AMDs can actually promote the development of clinical disease, likely through alterations in the numbers of commensal intestinal bacteria and in the species distribution of the intestinal microflora.^9,44–46

The importance of commensal microflora in disseminating and maintaining the pool of resistance genes has been recently emphasized by several investigators.^3,5–7,47 In a previous study,⁴⁸ for instance, a higher percentage of fecal commensal E coli isolates from calves on dairy farms that had experienced recent outbreaks of infection with multidrugresistant Salmonella strains were resistant to selected AMDs than were E coli isolates from calves on control farms. Another study⁴⁹ demonstrated transmission of antimicrobial resistance between commensal E coli strains, E coli O157, and Salmonella spp through the use of a validated in vitro simulation of the porcine ileum. In another study,⁴⁷ conjugative R plasmids were transferred from a donor E coli strain to E coli O157:H7 with similar efficiencies in Luria Bertani broth and rumen fluid, indicating that the rumen environment provides favorable conditions for exchange of genetic material. Additional evidence that genetic exchange occurs between commensal and pathogenic bacteria in the gastrointestinal tract is provided by the fact that identical resistance genes are present in diverse bacterial species from different hosts.⁶

For nearly half the fecal samples in the present study, isolates obtained from a single fecal sample differed in their resistance patterns. This suggests that the commensal fecal microflora in horses is not homogeneous, which is consistent with data for other species.⁵⁰ One possible consequence of this heterogeneity is that development of antimicrobial resistance cannot be monitored by testing only a single bacterial colony from each animal for antimicrobial susceptibility.

In conclusion, our findings reinforce the suggestion that AMD treatment in animals should be carefully considered.^14,26,33 Administration of AMDs may allow for selection of resistant bacteria not only in the treated animal but also in other hospitalized patients. However, not all AMD use has the same potential for selection of resistant phenotypes, as exemplified by the difference between the penicillins and cephalosporins in the present study. Unnecessary or inappropriate administration of AMDs may increase the risk of transmission of resistant bacteria to humans and contributes to the increase in a global pool of resistant genes.^35,51–53 Equally important to judicious use of AMDs is attention to infection control strategies, such as frequent hand hygiene, barrier nursing, and proper cleaning and disinfection. Properly implemented, these practices should minimize the spread of infectious agents between patients and the spread of resistant commensal bacteria in the hospital environment.