Methicillin-Resistant and -Susceptible Staphylococcus aureus Infections in Dogs

Risk factors for MRSA include intravenous catheterization and receipt of certain antimicrobial drugs.

D uring the past 2 decades, methicillin-resistant Staphylococcus aureus (MRSA) has gained global attention as a human pathogen in hospitals and in communities. Recent reports of MRSA infection and colonization of dogs and cats (1)(2)(3)(4)(5) indicate that MRSA has apparently emerged as a pathogen of animals as well. Most reported MRSA infections in dogs have involved wound and postoperative infections (2), but evaluation is lacking regarding specifi c types of infections, clinical outcomes, and risk factors associated with such MRSA infections in dogs. No current evidence points to whether MRSA infections, in terms of location of infection, severity of disease, or clini-cal outcome, differ from methicillin-susceptible S. aureus (MSSA) infections.
The literature about human medicine has compared MRSA-infection risk factors (6,7), mortality rates (6,(8)(9)(10), and clinical features (6) with those for MSSA infections. Results from a meta-analysis of 31 cohort studies showed that for patients with MRSA bacteremia, mortality rates were signifi cantly higher than for patients with MSSA bacteremia (11). This mortality rate difference between MRSA and MSSA infections might result from treatment with inappropriate antimicrobial drugs or the restricted number of antimicrobial drugs available for treatment (12). With respect to animals, however, limited data are available; only 1 study has evaluated these MRSA and MSSA infections. Morris et al. (13) compared MRSA and MSSA infections in cats but were unable to detect signifi cant differences in signalment, clinical presentations, or outcomes. Two other studies have reported potential risk factors associated with MRSA colonization in horses admitted to a veterinary referral hospital (14,15). Accurate data are needed to identify the differences between MRSA and MSSA infections in dogs as well as the clinical relevance of MRSA beyond concerns associated with antimicrobial drug resistance. To gain those data, epidemiologic research is required. Research is also required for proper medical treatment of dogs with MRSA infections, for counseling of clients of infected animals, and for elucidation of possible reasons for the emergence of MRSA in pets. Our study objectives were to compare the types of infections, clinical outcomes, and risk factors for MRSA and MSSA infections in dogs.

Selection of Case-Patients and Controls
From 2001 through 2007, we conducted a retrospective, secondary-base, case-control study at 3 veterinary re-

Data Collection
To collect information, we used medical records of all case-patient and control dogs to answer a pretested, standardized questionnaire. Data were collected on signalment, medical and surgical history, infection, hospital duration, and clinical outcome. Signalment data included breed, age, and sex. Medical and surgical history was limited to a 90day period before admission to the referral hospital and included antimicrobial drug treatment, hospitalization, and surgical procedure. MRSA or MSSA infection data comprised site of infection and procedures performed before onset of infection, such as surgery, endoscopy, colonoscopy, intravenous catheterization, and urinary catheterization. Clinical outcome data covered whether surgery was required because of the infection and whether the animal was discharged, was euthanized, or died.
Because of a large number of categories, we recategorized several variables. Breed was categorized according to size based on weight: small (1-10 kg), medium (>10-25 kg), or large (>25 kg). Age was categorized as young (<2 years), middle aged (>2-8 years), or old (>8 years). Site (where MRSA or MSSA was cultured) was categorized as skin, ear, urinary, skeletal, and other (abdominal fl uid, thoracic fl uid, blood, oral cavity, lymph node, vagina, transtracheal wash fl uid, and milk). The number of days dogs were hospitalized was categorized as short (<2 days), medium (3-7 days), or long (>7 days). Finally, specifi c antimicrobial drugs were grouped according to classes, i.e., aminoglycoside, β-lactam, chloramphenicol, fl uoroquinolone, lincosamide, nitroimidazole, and tetracycline. For analytical purposes, dogs were subcategorized into 4 groups: did not receive any antimicrobial drug in the previous 90 days, received that specifi c class of antimicrobial drug, received another class of antimicrobial drug, and unknown status.

Statistical Analyses
All descriptive statistics, model building, and analyses were performed by using Stata 10.0 (StataCorp, College Station, TX, USA) and by using exact logistic regression. In Stata, the score method was used for calculating p values, and the group option was applied to account for matching (16). All tests were 2 sided, and signifi cance was based on p<0.05. For predictor variables, odds ratios (ORs) and 95% confi dence intervals (CIs) were calculated. For descriptive variables listed in the categories of signalment, medical and surgical history, and infection, the outcome was defi ned as having MRSA or MSSA infection. Variables in the clinical outcome category were modeled as dependent variables, and MRSA or MSSA infection was the independent variable. To avoid problems associated with colinearity, we performed a correlation analysis to identify pairs of predictor variables that had high colinearity (|r|>0.8). We did not construct a multivariable model because of the relatively small sample size, resultant concerns of stability, and problems associated with overfi tting. Consequently, we constructed only univariable models.

Results
A total of 40 MRSA case-patients and 80 MSSA controls were eligible for inclusion. From each hospital, the number of case-patients and controls were, respectively, 7 and 14 (Ontario Veterinary College), 20 and 40 (Matthew J. Ryan Veterinary Hospital), and 13 and 26 (Angell Animal Medical Center).
Breed distribution was categorized according to weight ( Table 1). Ages of MRSA case-patients ranged from 1 to 13 years (mean 5.6 and median 5.0 years). Ages of MSSA control dogs ranged from 6 months to 16 years (mean 6.8 and median 7.0 years). No distinction was made between intact or sterilized dogs. Overall, no signifi cant differences appeared between case-patients and controls with respect to breed (p = 0.18), age (p = 0.50), or sex (p = 0.29).
Regarding previous hospitalization or surgical procedures, no overall signifi cant differences appeared between the MRSA and MSSA groups (p = 0.62 and 0.40, respectively). Results from the univariable analysis (Table 2) indicate that receipt of antimicrobial drugs (OR 3.84, 95% CI 1.21-14.74, p = 0.02), β-lactams (OR 3.58, 95% CI 1.04-14.79, p = 0.04), or fl uoroquinolones (OR 4.61, 95% CI 1.08-27.37, p = 0.02), within 90 days before admission was signifi cantly associated with a MRSA infection. Furthermore, when fl uoroquinolones and β-lactams were included in the "other classes" category, the odds of a dog having MRSA versus MSSA infection increased over odds for dogs that had not received antimicrobial drugs (Table 2).
Most MRSA and MSSA infections were located on the skin. Overall, with regard to infection site, we found no signifi cant difference between the MRSA and MSSA groups (p = 0.50) ( Table 3). Before onset of the MRSA and MSSA infections, the most common procedure was intravenous catheterization-a signifi cant risk factor for a MRSA in-fection (OR 3.27, 95% CI 1.14-10.65, p = 0.02). Neither colonoscopy nor endoscopy was performed on any animal. Dogs with MRSA infection were hospitalized 0-29 days (mean 3.4 and median 1.5 days), whereas dogs with MSSA infection were hospitalized 0-13 days (mean 2.0 and median 0 days). Overall, in terms of duration of hospitalization, we found no signifi cant difference between case-patients and controls (p = 0.49).
Surgery was required for treatment of 16 (40.0%) of 40 dogs with MRSA infection and 34 (42.5%) of 80 dogs with MSSA infection. Most dogs with MRSA and MSSA infections were discharged from the hospital (Table 4). For all dogs in the MRSA and MSSA groups that were euthanized, the infection was reported as the attributed cause of death. Overall, no signifi cant differences were noted between case-patients and controls with regard to surgery (p = 0.79) or outcome (p = 0.64).

Discussion
The identifi cation of receipt of antimicrobial drugsspecifi cally β-lactams and fl uoroquinolones-as risk factors for a MRSA infection was not unexpected. Data from human medicine and a logical hypothesis each indicate that antimicrobial drug use in animals would increase the likelihood of selection for multidrug-resistant bacteria such as MRSA. The case and control dogs included in this study were from veterinary referral hospitals; that is, tertiary care facilities that manage complicated medical and surgical cases referred from other veterinary facilities where treatment, surgery, or both might have been initiated. This study identifi ed the highest prevalence of MRSA and MSSA infections from the skin (pyoderma) and ears (otitis), which in dogs are frequently treated with β-lactams and fl uoroquinolones, respectively (17). Moreover, these conditions can become chronic and can result in repeated or prolonged antimicrobial drug treatments that might select for the development of antimicrobial drug resistance (18). Before ad-mission to the referral hospitals, >50% of dogs with MRSA infection were given antimicrobial drugs from the β-lactam family. Methicillin resistance in staphylococci involves the mecA gene, which encodes for the penicillin-binding protein 2a and results in reduced affi nity for all β-lactam antimicrobial drugs. Thus, medical management of MRSA cases can become complicated and can result in the administration of various classes of antimicrobial drugs (some of which can be ineffective), especially when culture and susceptibility testing have not been conducted.
In small animal medicine, fl uoroquinolones are commonly used because of their activity against a wide range of bacteria and their ability to be given orally (19). In humans, administration of antimicrobial drugs, including macrolides (7), β-lactams (20), and fl uoroquinolones (7,21), has been associated with increased risk for development of nosocomial MRSA infections compared with nosocomial MSSA infections. Specifi cally, use of fl uoroquinolones has been positively correlated with the incidence of hospital-associated MRSA (22) infections. In addition to the direct effect of antimicrobial drugs on selection for antimicrobial drugresistant organisms, other mechanisms could facilitate emergence of MRSA during fl uoroquinolone treatment. Research performed by Bisognano et al. (23) demonstrated that fl uoroquinolone-resistant MRSA and MSSA isolates exposed to subinhibitory levels of ciprofl oxacin resulted in increased production of binding proteins, leading to higher levels of bacterial attachment. Thus, exposure to fl uoroquinolones might promote the attachment of S. aureus while eradicating MSSA strains and might therefore promote acquisition of MRSA strains (24).
In our study, information pertaining to antimicrobial drug exposure in the 90 days before admission to the veterinary referral hospital was selected because that period was suffi cient for adequate review of medical charts. In the literature, periods for antimicrobial drug exposure as a risk factor for MRSA acquisition have ranged from 1 to  12 months (25). Future studies with larger datasets might investigate the effect of varying time frames with respect to antimicrobial drug administration. In dogs, identifi cation of intravenous catheterization as a risk factor for MRSA infection was not unexpected. Intravenous catheterization has been associated with increased rates of MRSA infections in humans (26) and has been signifi cantly associated with death of horses with MRSA infections (27). In previous studies as well as ours, however, intravenous catheterization might refl ect a consequence of MRSA infection rather than a risk factor for development of MRSA infection.
Overall and with respect to outcome (discharged vs. euthanized), no signifi cant differences between MRSA and MSSA infections were found. This fi nding is relevant for counseling clients, particularly considering the publicity regarding MRSA and the possible perception that MRSA infections are untreatable or carry a poor prognosis. Numerous studies in human medicine have compared mortality rates associated with MRSA and MSSA infections, but the results have been confl icting (6,8,10,11). Wang et al. (6) were unable to detect an association between higher mortality rates in patients with community-associated MRSA infections than in those with community-associated MSSA infections. Melzer et al. (10) were unable to demonstrate that mortality rates for patients with hospital-associated MRSA infections were signifi cantly higher than those for patients with hospitalassociated MSSA infections. Conversely, results from a retrospective cohort study conducted by Wang et al. (8), indicated that the mortality rate for patients with hospitalassociated MRSA bacteremia was 1.78× higher than that for hospital-associated MSSA bacteremia.
In our study, inadequate epidemiologic defi nitions and veterinary surveillance data prevented us from classifying MRSA and MSSA infections as hospital or community as-  sociated. Nevertheless, other possible explanations as to why mortality rates between case-patients and controls did not differ signifi cantly might in part be the predominant infection types and the retrospective aspect of the study. Most MRSA and MSSA infections were pyodermas and otitis externa or interna infections, which are superfi cial, rarely become invasive, and seldom result in death. Consequently, infection types for which death would be a more realistic possible outcome were limited, resulting in a corresponding limitation in statistical power. Comparison of mortality rates between patients with MRSA or MSSA infections would be best performed among only those with invasive infections and should be considered for future studies. Here, mortality rate information was obtained retrospectively and only recorded up to the time of discharge. Thus, whether dogs died from their infections after discharge from the referral hospital, causing an underestimate of deaths, is unknown. Although our study was larger than previous studies, the power was still limited despite enrollment of 2 MSSA controls per each MRSA case-patient. Additional limitations were enrollment of case-patients and controls from referral hospitals and the use of matching. Because dogs in this study were from referral hospitals, extrapolation of results to the general dog population might be biased. In general veterinary practice, antimicrobial drug use, hospitalization, surgical procedures, and specifi c medical and surgical cases might differ considerably from those in referral hospitals. The incomplete medical records that accompanied case-patients and controls from referral hospitals might have affected responses to questions regarding previous medical or surgical procedures and antimicrobial drug use, all of which might have affected the results. Other potential risk factors such as underlying illnesses, admitting service, hospitalization locations (i.e., intensive care unit vs. hospitalization ward), and treatment cost were not investigated but could play a role in the development and outcome of MRSA infections in dogs. Finally, by using matching to control for potential confounders, the matched factors-date of admission and referral hospital-precluded the investigation of these variables as potential risk factors for a MRSA infection.
Despite these limitations, however, we found no identifi able differences between MRSA and MSSA infections in dogs with regard to signalment, types of infections, and clinical outcome. The prognosis for a dog with a MRSA infection is reasonably good. However, when determining that prognosis and when counseling owners, veterinarians should focus on the location and severity of infection rather than the bacterium involved. Furthermore, administration of β-lactams and fl uoroquinolones were signifi cant risk factors for the development of a MRSA infection. This fi nding strengthens the need for veterinarians to consider prudent antimicrobial drug-use guidelines and to restrict the use of fl uoroquinolones as empirical or fi rst-line therapy. Guidelines should recommend identifi cation and susceptibility of the causal bacterial pathogen by performing a culture and susceptibility test. On the basis of susceptibility results, antimicrobial drugs should be dispensed at the proper dosage and duration for treatment and, in the absence of clinical disease, should not be prescribed.
Although only 4 risk factors were identifi ed as being signifi cantly associated with MRSA infection, results from the univariable analyses isolated several risk factors that have considerably large odds ratios or p values slightly greater than 0.05. With the exception of fl uoroquinolones and β-lactams, measure of association for all other antimicrobial drug classes was reasonably higher for those dogs given specifi c antimicrobial drugs compared with those that were not. Because of the small sample size, however, the power of these associations was limited.
Our study shows that MRSA is an emerging pathogen in dogs, and risk factors for MRSA infection are similar to those identifi ed in humans. Results from larger studies in the future might indicate that other classes of antimicrobial drugs, previous hospitalization and surgery, age, and the presence of a urinary catheter are also signifi cantly associated with MRSA infections. Only larger sample sizes will provide more information on MRSA and MSSA infections and will determine more accurately other risk factors associated with MRSA infections in dogs.
Dr Faires is a veterinarian and a PhD student in the Department of Population Medicine at the Ontario Veterinary College, University of Guelph. Her primary research interests include the epidemiology of MRSA in people and animals.