Antimicrobial resistance among Escherichia coli isolates from dogs presented with urinary tract infections at a veterinary teaching hospital in South Africa

Background This study investigated the burden and predictors of canine E. coli urinary tract infections (UTI) and antimicrobial resistance among dogs presented at a veterinary teaching hospital in South Africa, 2007–2012. Methods The Cochran-Armitage trend test was used to investigate temporal trends while logistic regression models were used to investigate predictors (age, sex, breed, year) of E. coli infections and antimicrobial resistance (AMR). Results A total of 22.3% (168/755) of the urinary specimens tested positive for E. coli. A significant (p = 0.0004) decreasing temporal trend in the percentage of E. coli positive isolates was observed over the study period. There were high levels of AMR to penicillin-G (99%), clindamycin (100%), tylosine (95%), cephalothin (84%) but relatively low levels of resistance to enrofloxacin (16%), orbifloxacin (21%). Almost all (98%, 164/167) the isolates exhibited multidrug resistance (MDR), while only 11% (19/167) and 2% (4/167) exhibited extensive drug resistance (XDR) and pan-drug resistance (PDR), respectively. Conclusions Although, the risk of E. coli UTI declined during the study period, the risk of AMR increased. The high levels of AMR and MDR as well as the presence of XDR and PDR is concerning as these have the potential of affecting prognosis of UTI treatments.


Background
Although recent studies show that Enterococcus spp. and Pseudomonas spp. are increasingly becoming prominent in urinary tract infections (UTI) in dogs, Escherichia coli remains the most common cause of UTI in dogs [1][2][3][4]. These infections are caused by uropathogenic E. coli (UPEC), which differ from intestinal E. coli strains as they contain extra virulence genes, allowing a successful transition from the intestinal tract to the urinary tract [5]. The perianal and genital areas are the principal reservoirs of the E.coli organism known to cause UTI [5]. Animals with compromised immune systems are at higher risks of UTI than those that are not immune-compromised. In these animals, the organisms are able to multiply and persist in a portion of the urinary tract resulting in clinical disease [6,7]. Urinary tract infections can be divided into upper and lower tract infections. The former affects the kidneys and ureters while the latter affects the bladder, urethra and vagina [2,3,[8][9][10]. Clinical signs of E. coli UTI in dogs may include acute cystitis, pyelonephritis and urosepsis. These three clinical signs are distinct indicators of UTI syndromes [1].
Antimicrobial agents such amoxycillin-clavulanic acid, tetracyclines, trimethoprim-potentiated sulphonamides and cephalexin are reported to be effective against E. coli UTI [11][12][13]. However, there are concerns of increased antimicrobial resistance among E. coli isolates to fluoroquinolones in dogs with UTI [1]. Unfortunately, there is limited information on the burden and predictors of antimicrobial resistance (AMR), multidrug resistance (MDR) and extensive drug resistance (XDR) among companion animals in South Africa. This is despite evidence of transfer of resistance between animals and their owners. Therefore, the objective of this study was to investigate the burden of E. coli infections and antimicrobial drug resistance among dogs presented with UTI at a veterinary teaching hospital in South Africa.

Data source and management
This study used retrospective data from the bacteriology laboratory of a veterinary teaching hospital in South Africa. Client owned dogs from Gauteng Province with suspected UTI whose specimens were tested at the laboratory between January 2007 and December 2012 were included in the study. Since this study used retrospective laboratory records, it did not directly involve animals and thus posed no risk to client animals. The data were assessed for duplicates and missing information. Only complete records were selected for inclusion in this study. The following variables were extracted from the records: age (in months), sex, breed and the date of specimen submission. The breed classification used in the study was adapted from the American Kennel Club (AKC) and included the following categories: working, sporting, herding, hound, toy, terrier, nonsporting and mixed breeds [14].

Escherichia coli identification and antimicrobial susceptibility testing
Escherichia coli were isolated and identified using standard bacteriological methods and suspected E. coli colonies were subjected to different biochemical tests as described by Quinn et al. [15]. The E. coli reference strain (ATCC 25922) was used for quality control.

Data analysis Descriptive analysis
Crude and factor-specific proportions of E. coli UTI and AMR as well as their 95% confidence intervals were computed. The factors assessed were age, sex, breed and year. Associations between UTI and AMR and each of the above factors were assessed using the Chi-square or Fishers Exact tests as appropriate. The temporal trends in the proportions of E. coli UTI and AMR between 2007 and 2012 was assessed using Cochran-Armitage trend tests. Significance was set at α = 0.05 for all statistical tests.

Predictors of infection
The predictors of E. coli UTI were assessed using logistic regression models. A simple binary logistic regression was first fit to assess the association between infection status (yes/no) and covariates age, sex, breed and year. Predictors with a p-value less than 0.20 were considered for inclusion in the multivariable logistic regression model. A manual backwards elimination approach was then used to build a multivariable logistic regression model containing variables that had potential univariable associations (p < 0.2) with the outcome. At this stage significance was set at α = 0.05. To assess for confounding, the changes in parameter estimates of the predictors in the model with and without the suspected confounding variable were compared. Changes of 20% in the estimates were considered indicative of significant confounding and hence the suspected confounding variables were retained in the final model. Adjusted odds ratios and 95% confidence intervals were calculated for all predictors retained in the final model. Statistical significance was assessed using Wald Chi-Squared Test at α = 0.05. Hosmer-Lemeshow test was used to assess goodness-of-fit of the final model.

Risks of Escherichia coli UTI
Twenty two percent (22%; n = 168/755) of the urinary specimens tested positive for E. coli. Based on simple association assessments, there was no significant association between risk of E. coli UTI and breed (p = 0.283). On the other hand, there was a significant association between risk of E. coli UTI and both year (p < 0.001) and sex (p = 0.054) ( Table 2).

Predictors of Escherichia coli infection and antimicrobial resistance
Based on the multivariable logistic model, age (p = 0.465), sex (p = 0.318) and breed (p = 0.300) all showed no evidence of significant association with odds of E. coli UTI. However, there was a significant association between the odds of E. coli UTI and time (years) with the odds of infection significantly (p < 0.001) decreasing (OR = 0.78, 95% CI: 0.68-0.89) during the study period.
None of the variables assessed: age (p = 0.972), sex (p = 0.282), breed (p = 0.309) and year (p = 0.394) had a significant association with the odds of multi-drug resistance among E. coli isolates.

Discussion
In this study, we investigated the burden and predictors of E. coli UTI and their antimicrobial resistance patterns from dogs presented at a veterinary teaching hospital in South Africa. The results of this study will support management and treatment of UTIs in dogs presented at the veterinary hospital.
In this study, we observed a lower proportion of E. coli positive samples (22%) than the 27% reported by Stiffler et al. [17], 56% reported by Seguin et al. [13], 44% reported by Johnson et al. [2] in the USA and 62% reported by Gibson et al. [1] in Australia. The differences in the results may be due to differences in study designs. The studies by Stiffler et al. [17] and Seguin et al. [13] were longitudinal studies investigating risk before hospitalization and after surgery, while our study investigated E. coli UTI among hospitalized dogs in a veterinary teaching hospital. It is also possible that the presence of underlying disorders in the study by Stiffler et al. [17] and Seguin et al. [13], which our study did not investigate, could explain the differences in the proportions of E. coli positive samples. Dogs with underlying medical conditions such as diabetes mellitus, hyper-adrenocorticism and pre-existing urinary tract diseases have a weaker immune system making them prone to E. coli infections [13]. Thompson et al. [4] reported an increase in the prevalence of UTI in dogs over the period of their study. However, in our study, we observed a decrease in the proportion of E. coli positive samples. This could be due to improved health and welfare of the dogs visiting the hospital.
The results of this study show that age, breed and sex were not significantly associated with the odds of E. coli UTI among dogs presented at the hospital in South Africa. This is consistent with findings by Stiffler et al. [17] who reported no significant association between E. coli UTI and age, breed or weight of the dog. However, Stiffler et al. [17] reported that female dogs were 3 times more likely to contract E. coli related UTI compared to male dogs. By contrast, Johnson et al. [2] observed that the risk of E. coli UTI was higher in males compared to female dogs. These findings suggest that there may be sex predisposition for E. coli related UTI although our study did not identify such a relationship. Sex related risk of E. coli UTI has been shown to be related to the differences in the anatomic structure between male and female dogs. This makes manually expressing of female bladder easier for urine sample collection than intermittent catheterization needed in male dogs [17]. Moreover, indwelling urinary catheters during diuresis or administration of corticosteroid are also major risk factors of UTI in dogs. These may be due to conformational changes, altered normal flora, or decreased immune response [7,17,18]. Contrary to the findings of our study, Stiffler et al. [17] reported that dogs ≤3 years were more likely to present with UTI compared to dogs > 3 years.
While resistance to penicillin-G, clindamycin, tylosin, cephalothin, amoxycillin, ampicillin, doxycycline and lincospectin were common in the majority of the E. coli isolates in this study, a study in Sweden reported low levels of resistance to ampicillin (17%), and tetracycline (7%) among E. coli UTI [12]. The reason for the higher resistance levels observed in this study is unclear and requires further investigation. However, low levels of resistance to  fluoroquinolones, sulpha trimethoprim and chloramphenicol observed in our study are similar to the findings of other studies [11,13,19]. Although our findings are contrary to those of Wedley et al. [20] who observed a low proportion (18%) of E. coli that were MDR-among dogs with urinary tract infections, the high proportion of MDR (98%) in our study is not unusual. For example, Gibson et al. [1] and Wagner et al. [21] also observed high proportions of MDR among E.coli isolates from dogs with UTIs. This observation has serious implications for clinical outcomes during treatment given that multiple drug resistance has a negative effect on the prognosis of E. coli UTI in veterinary medicine [22][23][24]. Of even greater concern among the findings of the present study is the presence of XDR (11%) and PDR (2%) E. coli isolates. The proportion of XDR isolates observed in this study is higher than the 2% reported by Thungrat et al. [25] in the USA. Furthermore, the ability of E. coli isolates to transfer resistance genes among themselves and other species of bacteria [26], makes the levels of XDR and PDR observed in this study of great veterinary public health concern.
In the present study, no previous history of antimicrobial usage among the dogs tested was available. Moreover, some cases might have been treated empirically prior to culture and susceptibility testing. In addition, isolates that exhibited intermediate resistance in this study were re-classified as resistant. Therefore, it is possible that this could have slightly biased the findings towards higher estimates of resistance levels among the E. coli isolates observed in this study. Nonetheless, the results of this study support previous studies that showed that E. coli is a common cause of UTI in dogs and contributes to the understanding of antimicrobial resistance patterns among E. coli UTI in the dog population presented at the veterinary teaching hospital in South Africa.

Conclusions
This study shows that the proportion of cases of E. coli UTI among dogs presented at the veterinary teaching hospital declined over the study period. However, high levels of E. coli isolates exhibiting MDR, XDR and PDR are of clinical and a veterinary public health concern. Therefore, urgent action needs to be taken to tackle the development of antimicrobial drug resistant E. coli infections in dogs. This may require development of antimicrobial stewardship programme at the teaching hospital and in the country.