- Email: [email protected]
A breed-matched case-control study of potential risk-factors for canine pyometra
Available online at www.sciencedirect.com
Theriogenology 75 (2011) 1251–1257 www.theriojournal.com
A breed-matched case-control study of potential risk-factors for canine pyometra Ragnvi Hagmana,*, Anne-Sofie Lagerstedta, Åke Hedhammara, Agneta Egenvallb a
Department of Clinical Sciences, Division of Small Animals, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden b Department of Clinical Sciences, Division of Ruminant Medicine and Veterinary Epidemiology, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden Received 18 June 2010; received in revised form 3 November 2010; accepted 29 November 2010
Abstract The objective was to evaluate plausible risk-factors for pyometra, a common disease affecting almost 25% of all (unspayed) female dogs before 10 years of age. Because of the strong breed-predilection, an age- and breed-matched case-control study was undertaken on 87 pairs (pyometra-cases and healthy controls) from five breeds (Rottweiler (n ⫽ 13), Collie (n ⫽ 8), Golden retriever (n ⫽ 24), Labrador retriever (n ⫽ 16) and German shepherd dog (n ⫽ 26)). The mean age was 7.9 y (range 0.8 –13.8 y). Variables analyzed included pseudopregnancy, age at first oestrus, length of and regularity of the interoestrus interval, hormone treatments, nulliparity, number of parities, age at first whelping, previous urinary tract infections and mammary tumours. Data were modelled multivariably using matched-pair conditional logistic regression. Analysing interactions with breed, previous pregnancy was statistically associated with pyometra. When amalgamated, in three breeds previous pregnancy was protective (Rottweiler, Collie, Labrador retriever) and in one breed statistically intermediate (German shepherd dog) when compared to the baseline (Golden retriever). Previous pregnancy was a statistically significant factor that had a protective effect against pyometra in some breeds but not in the Golden retriever breed. These findings indicate that protective- and risk-factors may vary between different breeds. The obvious problem with low power and limited possibility for extrapolation, using few dogs in few breeds, is acknowledged. However, it is suggested that failure to control for the confounding effect of breed, especially in epidemiological studies on dog diseases, may lead to potentially erroneous conclusions. © 2011 Elsevier Inc. All rights reserved. Keywords: Infection; Dog; Epidemiology; Risk factors; Pair-matched case-control study; Pyometra; Confounding
1. Introduction Pyometra is a common disease in countries such as Sweden, where spaying/neutering is not generally practised and is diagnosed in nearly 25% of all intact (not spayed) female dogs before 10 y of age [1]. The potentially life-threatening disease is characterised by * Corresponding author. Tel.: ⫹46 18 672918; fax: ⫹46 18 673534. E-mail: [email protected] (R. Hagman). 0093-691X/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.theriogenology.2010.11.038
uterine infection and inflammation in combination with systemic illness [2]. Hormonal factors as well as bacterial infection are involved in the development of the disease [3– 4]. Large age-and breed-related differences with respect to occurrence have been reported [1,5]. Oestrogen therapy and nulliparity have been associated with increased risk for pyometra whereas overt pseudopregnancy has been linked with a reduced risk of developing the disease [5–7].With the purpose of further investigating the effect of some of these known and
1252
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
other new plausible factors the aim of this study was to perform a prospective case-control study. Because of previously shown age- and breed related differences in the risk for developing pyometra, matching on both breed and age was undertaken in order to be able to remove the influence of those risk-factors. Variables explored included pseudopregnancy, age at first oestrus, length of and regularity of the inter-oestrus interval, hormone treatment, parity, age at first whelping, previous urinary tract infections and mammary tumours amongst others. 2. Materials and methods 2.1. Study design A prospective case-control study matched on breed and age was designed. Five breeds with a high incidence of pyometra (Rottweiler, Collie, Golden retriever, Labrador retriever and German shepherd dog) were included in the study [1,5]. Approximately 40 pyometra cases of the selected breeds are admitted every year to the University Animal Hospital (UDS), Swedish University of Agricultural Sciences, Uppsala, Sweden. Sampling calculations for a case-control study showed that 80 cases were needed to detect differences of pseudopregnancy (unmatched design, 95% confidence and 80% power (http://www.open-epi.com)) based on a previous report of 20% prevalence differ-
ence in dogs with pyometra compared with healthy control dogs [6]. It was decided to sample data during approximately two years full-time knowing that all eligible cases could not be included due to time constraints, and continue until ⬃80 cases were accrued. All telephone interviews were made by one of the authors (RH). 2.2. Case selection Pyometra cases admitted to the UDS were intermittently included from 2001 through to 2007 (part-time collection due to employment reasons). During the time periods of employment, every diagnosed pyometra case of the chosen breeds was included to avoid possible selection bias. Dog-owners signed an informed consent. Complete physical examination was performed. Case history and physical examination data were registered and the diagnosis pyometra confirmed by identification of a pus-filled enlarged uterus during ovariohysterectomy. A telephone interview with the owner was performed the same day by one of the authors, and a questionnaire filled in. Demographic data including breed, Swedish Kennel Club (SKC) registration number, gender, date of birth, and contact details of owner were registered. The history questionnaire was designed to retrieve information regarding parameters presented in Tables 1– 4.
Table 1 Various clinical signs in number (No) and proportion (%) of the studied 87 female dogs with pyometra. Parameter Thirst Increased Decreased Appetite Increased Decreased General condition Moderately depressed Severely depressed Hydration status Mild dehydration Moderate dehydration Severe dehydration Circulatory abnormalities* Mammary gland tumour/s Lymph node/s enlarged Vaginal discharge Slight Moderate Abundant
Normal
Normal
Normal
Normal
Absent (i.e., closed cervix)
* Evaluated by auscultation of heart sounds, pulse quality.
No
%
29 56 2 24 0 63 12 52 23 50 26 11 0 3 10 4 13 18 30 26
33 64 2 27 0 72 14 60 26 57 30 13 0 3 11 5 15 21 34 30
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
1253
Table 2 Data regarding oestrus cycles and previous health history in pyometra cases and healthy control dogs illustrated in numbers (No) and proportions (%) of the total number of pairs evaluated. If data was missing in one dog (as for oestrus cycle interval regularity in 9 dogs and pseudopregnancy in 1 dog), the entire pair was removed from the analysis. The P-values display the statistical difference between the cases and controls from Bowkers/Mc Nemar tests. NT, Not tested; NE, Not estimable. Case history parameters
No pairs
Pseudopregnancy No Mild Moderate Severe Moderate to severe Oestrus interval Regular Nulliparity Hormone treatment Oestrus postponement, continuous Oestrus postponement, single Abortion Urinary tract infections Mammary tumour/s Skin disease Gastro-intestinal disease Respiratory disease
Pyometra cases
Control dogs
P
No
%
No
%
86 86 86 86 86
55 15 12 4 16
64 17 14 5 16
53 18 9 6 15
62 21 10 7 17
0.74 NT NT NT 0.85
78 87 87 87 87 87 87 87 87 87 87
67 75 9 4 3 1 9 8 9 7 0
86 86 10 6 3 1 10 9 10 8 0
73 67 8 3 0 5 8 7 9 4 0
94 77 9 3 0 6 9 8 10 5 0
0.11 0.12 0.62 0.71 NE 0.10 0.78 0.80 0.81 0.32 NE
2.3. Control dog selection
2.4. Definition of selected variables
For each pyometra case a female control was selected from data from the SKC registry. The control dog was matched by breed, geographical region (through postal address of the owner) and age (born ⱕ 3 mo before or after the case) by assistance of SKC staff. An identical questionnaire form as for the pyometra cases was filled in during a telephone interview, excluding questions related to the disease. The interview with the owner of the selected control dogs was performed within three weeks of the admittance of the pyometra case. Since no physical examination was performed in the control dogs, clinical variables are absent in this group. Siblings, dogs with previous pyometra or neutered, were excluded from the control group.
Oestrus intervals that had a variation in duration of maximum ⫾ 1 mo were accepted as regular. If the intervals were prolonged at older age, this was considered regular as long as they were uniform [6]. Clinical signs of pseudopregnancy were graded as mild, moderate or severe. Bitches that did not show any signs of altered behaviour or physiology as compared to anoestrus were defined as not pseudopregnant. Mild pseudopregnancy was defined as minor signs of inappetence, depression, decreased willingness to exercise or decreased activity. Moderate pseudopregnancy was defined as clearly noticeable clinical signs of altered behaviour as above with the addition of nesting and/or carrying around fictive pups. Severe pseudopregnancy
Table 3 Reproductive case history data in dogs with pyometra and control dogs. No* Beginning of last heat (weeks ago) Age at first heat (years) Interoestrus interval (years)
74 68 76
Pyometra
Control
P†
Mean
SD
Range
Mean
SD
Range
7.5 0.9 0.6
8.4 0.3 0.2
1–56 0.5–2.5 0.08–1.08
11.5 0.9 0.6
9.8 0.9 0.2
1–40 0.5–3.0 0.29–1.02
* Number of dog-pairs. † Paired Wilcoxon test. a Comparison biased since oestrus did not begin on the same day.
⬍0.002a 0.95 0.97
1254
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
Table 4 Presence of selected case history variables in the studied dog-pairs, as analysed by breed. Variable
Case
Control
Collie (8)
Golden Retriever (24)
Labrador retriever (16)
Rottweiler (13)
German shepherd dog (26)
All (87)
Previous pregnancy
Pos Pos Neg Neg Pos Pos Neg Neg Pos Pos Neg Neg Pos Pos Neg Neg
Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg
0 0 2 6 0 0 1 7 0 0 1 7 7 0 1 0
1 5 1 17 1 3 3 17 0 1 0 23 18 0 4 0
1 0 4 11 1 2 4 9 0 0 0 16 14 1 0 0
1 1 6 5 0 2 4 7 0 0 2 11 7 2 2 1
0 3 4 19 0 7 1 17 0 0 2 24 17 1 3 0
3 9 17 58 2 14 13 57 0 1 5 81 63 4 10 1
Pseudo-Pregnancy*
Abortion
Regular Oestrus cycles†
* Moderate to severe pseudopregnancy. Missing data in 1 dog-pair (German shepherd dog), 86 dog-pairs analysed. † Missing data in 9 dog-pairs (2 Golden retriever, 1 Labrador retriever, 1 Rottweiler, 5 German shepherd dog), 78 dog-pairs analysed.
was defined as major behavioural changes, as above, in combination with lactation.
2-way interactions, as was interactions with the matching variables breed and age.
2.5. Statistical analysis Associations between categorical variables and case status were tested using Bowker=s test (i.e., a generalised McNemars test). In order to perform the pairwise analysis the entire pair was removed from the dataset if data were missing in one individual. Continuous variables were tested against case status using the paired Wilcoxon test. Data were managed with SAS (SAS Institute Inc., Cary, NC, 27513, US) and the procedure PHREG was used to develop a multivariable pairmatched conditional logistic regression. The following dichotomous variables were first screened univariably: pseudopregnancy, moderate to severe pseudopregnancy, regular oestrus interval, nulliparity, ever had hormone treatment, elective pregnancy termination (abortion); ever had urinary tract infections, ever had mammary tumour/s, skin problems, gastrointestinal or respiratory problems. Continuous variables analysed were age at first heat (oestrus) and interoestrus interval. To check for linearity relative to the outcome these were also entered as categories, based on percentiles of the data (0 ⬍ 25, 25 ⬍ 50, 50 ⬍ 75 and ⱖ 75%). Variables with a univariable P-value of 0.15 were entered into a primary multivariable model, which was further reduced to P ⬍ 0.05. Variables with a univariable P-value of 0.20 were deemed to be checked for
3. Results 3.1. Study population In total 87 pyometra cases and 87 female control dogs (87 pairs) were recruited to the current study during the study period. Included pairs were 26 German shepherd dogs, 24 Golden retrievers, 16 Labrador retrievers, 13 Rottweilers and 8 Collies. The mean age in both groups (cases and controls) was 7.9 y (range 0.75–14 y). 3.2. Case history and physical examination data Case history and physical examination data in the 87 female dogs with pyometra are illustrated in Tables 1–3. In 74 (85%) of the pyometra cases, purulent or purulent-hemorrhagic vaginal discharge was present (open cervix). Case history data in the control dogs are presented in Tables 2 and 3. Data showing number of dogs by the matching variable breed in pyometra cases and healthy control dogs are illustrated in Table 4. 3.3. Descriptive statistics Mean age ⫾ SD (Range) at first whelping in the ten pyometra cases and the nineteen control dogs that had whelped (12% vs 22%) was 4.2 ⫾ 0.7 years (4 – 6
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
years) and 4.7 ⫾ 1.5 years (2–7 years), respectively. Of these, six pyometra cases and eleven controls were uniparous, two pyometra cases and two controls were biparous, two pyometra cases and four control dogs were triparous whereas no pyometra case and two control dogs whelped four times or more. Caesarean section was performed in four pyometra cases and two control dogs. Nine dogs with pyometra and eight control dogs were treated with hormones for oestrus postponement or pregnancy termination. Oestrus postponement had been generated with medroxyprogesterone acetate (MPA) administered subcutaneously in therapeutic doses (3 mg/kg every 3⫺5 months) which during the study years was the only drug registered for oestrus postponement in Sweden. Pregnancy termination had been induced in one pyometra case and five control dogs with a progesterone receptor blocker, aglepristone, 10 mg/kg subcutaneously on two consecutive days. Silent heat was registered in two cases of pyometra and one control dog. Five of the cases in the pyometra group were diagnosed before two years of age. Of these, three were German shepherd dogs (ages: 9 mo/10 mo/1 y) and two were Rottweilers (ages: 1 y/1 y and 5 mo). None of them had received any previous hormone treatment. The mean heart rate and temperature of the pyometra cases was 103 beats/minute (range 60 –120) and 38.9°C (range 36.1– 40.2), respectively. The duration of depressed general condition in these dogs previous to admission was 8.5 d (range 0 – 60 d). Of all the selected variables studied (Tables 2, 3 and 4), only nulliparity was significantly associated with pyometra, and only when analysed as an interaction effect with the matching variable breed (Table 4). In order to yield interpretable result, due to low numbers in strata, the three breeds in which previous pregnancy was protective, according to the point estimates were amalgamated while the intermediate (German shepherd dog) and high-risk breed (Golden retriever ⫽ baseline) were kept separately (Table 4). This yielded a hazard ratio (HR) of 0.08 (95% CI 0.01– 0.64) in the “three breed group” and an intermediate HR in the German shepherd dog (HR ⫽ 0.75 (95% CI 0.17–3.4)). 4. Discussion Cases and controls in the present study were matched by breed to avoid confounding by the previously demonstrated breed-related differences in the occurrence of pyometra [1]. The inherent problem of controlling for breed which is found in many epidemiological dog-studies was thereby removed. In addition,
1255
only five breeds (high incidence) were included in the present study to further limit the effect of breed variation, however also limiting the ability for extrapolation to all dog breeds. We also matched by age because pyometra mainly affects middle-aged and older individuals. It may be argued that controls were selected from a wider population than the cases. However, because pyometra is a well-defined disease it is highly likely that the waste majority of pyometra-cases in this wider population, would seek medical attention, and also be treated, similar to the study-cases. Concluding from this, controls and cases actually may be considered to emanate from the same population and the potential problem with different base populations would have a very limited impact [8]. A protective association was found between previous pregnancy and pyometra in a clump of three breeds—Collie, Rottweiler and Labrador retriever— when data were pooled due to scarcity of data. That nulliparity was associated with an increased risk of the disease is in agreement with results from previous studies [5– 6]. One or just a few pregnancies do not appear to give life-long protection against pyometra since the disease was diagnosed in multiparous individuals in the present study, as has also been reported previously [9]. That the Golden retriever differed from the other breeds may depend on nulliparity not being a risk factor in this particular breed or due to chance because of relatively few dogs deciding this outcome (18/24 previously nonpregnant Golden retrievers with pyometra compared with 22/24 non-pregnant healthy controls). However, risk-factors for pyometra may differ depending on breed, and should be considered when studying the disease (i.e., interactions with breed should be included). None of the other potential risk-factors analysed in this study showed an association with the diagnosis of pyometra. These results may depend on the case selection or dog breeds included. The data were also analysed using un-matched logistic regression (data not shown), but relative to statistical significance the results were the same. As for hormone treatments, the compound that had been used for oestrus postponement was Medroxyprogesterone acetate (MPA). In accordance with the results of the present study (hormone treatments, MPA or aglepristone, not linked with pyometra), progestin in therapeutic doses has previously not been associated with an increased risk for developing cystic endometrial hyperplasia (pre-stage to pyometra) or pyometra [5,10]. In a long-term study of 22 dogs treated with aglepristone for inducing abortion, pyometra occurred in two dogs after two and four
1256
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
years, respectively [11]. Pregnancy termination with the progesterone receptor blocker aglepristone was not associated with a diagnosis of pyometra in this study, but only few dogs were treated with this drug. Pyometra has been reported to occur less frequently in bitches that show clinical signs of pseudopregnancy [6,12–13]. In an early study of 68 cases and 245 healthy female dogs at least one occasion of overt pseudopregnancy was linked with reduced risk of developing the disease [6]. In the same study pseudopregnancy occurred in 43% of the control dogs and 23% of 68 pyometra cases [6]. The results of the present study did not, regardless of degree, demonstrate any protective effect of pseudopregnancy against pyometra. Pseudopregnancy was noted in 31 dogs with pyometra and 33 control dogs. Moderate to severe pseudopregnancy was noted in 16 of the dogs with pyometra dogs and 15 control dogs, respectively. The risk for pyometra was neither increased nor decreased in dogs that had developed mammary tumours despite reproductive hormone involvement in the pathogenesis of both diseases. These results are supported by the fact that the English springer spaniel was the only breed reported to have an increased risk of both pyometra and mammary tumours [14]. If the etiology of both diseases was similar the breed-predisposition would also possibly have been more alike in previous studies [1,5,14]. In pyometra, a bacterial infection is established during metoestrus when the uterine milieu is suitable for foetal development but also for bacterial growth [12– 13,15]. If every reproductive cycle (metoestrus) renders the uterus susceptible for pyometra, individuals with short interoestrus intervals would be at prone to develop the disease. Similarly, individuals with the first oestrus cycle at an early age would theoretically be susceptible for infection during longer time (every metoestrus) as compared to individuals that become fertile later. In this study, a short interoestrus interval, early fertility or irregular oestrus cycles were not associated with increased risk for pyometra. Escherichia coli (E. coli) are the most common bacteria isolated from both UTI and pyometra [16]. A subclinical urinary tract infection (UTI), as has been demonstrated previously in 25% of dogs with pyometra, may act as a bacterial reservoir since many virulence traits of E. coli from UTI and pyometra are similar [4,17]. In this study, it was for practical reasons not possible to explore the presence of subclinical UTI, but suffering from previous UTI or other diseases (in disease-prone individuals) was not associated with increased risk for pyometra. The clinical signs most commonly reported in the pyo-
metra group were vaginal discharge (in 84%), depressed general condition (in 83%), decreased appetite (in 72%), increased thirst (in 64%) and dehydration (in 43%) (Table 1). These results are similar to what has previously been reported in the disease and reflect a local uterine as well as generalised disease [2]. Data on time since the beginning of last heat (Table 3) is displayed to add information, but as noted, due to bias the difference between the two groups would only be relevant for comparison if oestrus started at the same time. The results of this study showed a significant effect only of nulliparity and not in all breeds. However, none of the earlier studies of variables related to pyometra controlled for breed, which is a strong risk factor [5– 6,12–13]. On the other hand the present study included a relatively small number of cases and with low power, hence in spite of reaching the goal of cases to include it is still possible that some of the anticipated other risk factors are true biological risk factors. If increasing the sample size to 180 pairs (unmatched design, 95% confidence and 80% power) it would have been possible to demonstrate statistical significance for nulliparity in the entire data-set given that the effect was the same as in this study (i.e., 89% in the case-group and 78% in the control-group) (http://www.open-epi.com). It was deemed possible to use population controls because pyometra is a disease which the owner is likely to have noted and remembered. Recall bias was difficult to avoid completely. Recall may be different among cases and controls, or here their owners. In this study, however we believe that having puppies is the only perceived protective factor that people would ruminate over relative to the occurrence of pyometra and that pregnancy is highly likely to be equally well remembered by both groups of owners. 5. Conclusion Previous pregnancy was a statistically significant factor that had a protective effect against pyometra in some breeds but not in the Golden retriever breed. These findings indicate that protective- and risk-factors may vary between different breeds.
Acknowledgments We thank A-M Darell and T. Wink at the SKC for their help with selecting control dogs and U. Emanuelsson for valuable input on the manuscript. Lennart Swensson is acknowledged for contribution to the initial study design. The Foundation for Research, Agria
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
Insurance Co and the Swedish Kennel Club and Thure F. and Karin Forsgren’s Foundation are acknowledged for financial support. The funding body had no role neither in or influence over study design, data collection, analysis, and interpretation of data, nor in the writing of the manuscript or in the decision to submit the manuscript for publication. RH carried out the interviews and data collections and drafted the manuscript. AE performed the statistical analyses and helped to draft the manuscript. All authors participated in the planning and design of the study and interpretation of the results. All authors read, revised and approved the final manuscript.
References [1] Egenvall A, R. Hagman BN Bonnett A Hedhammar P Olson and AS Lagerstedt. Breed risk of pyometra in insured dogs in Sweden. J Vet Intern Med 2001;15:530 – 8. [2] Borresen B. Pyometra in the dog. II—A pathophysiological investigation. II. Anamnestic, clinical and reproductive aspects. Nord Vet Med 1979;31:251–7. [3] Borresen B. Pyometra in the dog—a pathophysiological investigation. I. The pyometra syndrome, a review. Nord Vet Med 1975;27:508 –17. [4] Sandholm M, Vasenius H. AK Kivisto Pathogenesis of canine pyometra. J Am Vet Med Assoc 1975;167:1006 –10. [5] Niskanen M, MV Thrusfield. Associations between age, parity, hormonal therapy and breed, and pyometra in Finnish dogs. Vet Rec 1998;143:493– 8.
1257
[6] Fidler IJ, Brodey RS, Howson AE, Cohen D, Relationship of estrous irregularity, pseudopregnancy, and pregnancy to canine pyometra. J Am Vet Med Assoc 1966;149:1043– 6. [7] Whitehead ML. Risk of pyometra in bitches treated for mismating with low doses of oestradiol benzoate. Vet Rec 2008;162: 746 –9. [8] Dohoo I MW, Stryhn H. Validity in Observational Studies. In: Veterinary Epidemiologic Research. AVC Inc.: Charlottetown, PEI, Canada, 2008, p. 184. [9] Fukuda S. Incidence of pyometra in colony-raised beagle dogs. Exp Anim 2001;50:325–9. [10] Kooistra HS, Okkens AC, Mol JA, E van Garderen, Kirpensteijn J, Rijnberk A. Lack of association of progestin-induced cystic endometrial hyperplasia with GH gene expression in the canine uterus. J Reprod Fertil Suppl 1997;51:355– 61. [11] Pettersson CH Tidholm A. Safety and efficacy of mid-term pregnancy termination using aglepristone in dogs. J Small Anim Pract 2009;50:120 –3. [12] Dow C. The cystic endometrial hyperplasia-pyometra complex in the bitch. Vet Rec 1958;69:1409 –14. [13] Dow C. The cystic hyperplasia-pyometra complex in the bitch. J Comp Pathol 1959;69:237–50. [14] Egenvall A, Bonnett BN, Ohagen P, Olson P, Hedhammar A, H von Euler. Incidence of and survival after mammary tumors in a population of over 80,000 insured female dogs in Sweden from 1995 to 2002. Prev Vet Med 2005;69:109 –27. [15] Cox JE. Progestagens in bitches: a review. J Small Anim Pract 1970;11:759 –78. [16] Hagman R, Greko C. Antimicrobial resistance in Escherichia coli isolated from bitches with pyometra and from urine samples from other dogs. Vet Rec 2005;157:193– 6. [17] Chen YM, PJ Wright, Lee CS, Browning GF. Uropathogenic virulence factors in isolates of Escherichia coli from clinical cases of canine pyometra and feces of healthy bitches. Vet Microbiol 2003;94:57– 69.
Theriogenology 75 (2011) 1251–1257 www.theriojournal.com
A breed-matched case-control study of potential risk-factors for canine pyometra Ragnvi Hagmana,*, Anne-Sofie Lagerstedta, Åke Hedhammara, Agneta Egenvallb a
Department of Clinical Sciences, Division of Small Animals, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden b Department of Clinical Sciences, Division of Ruminant Medicine and Veterinary Epidemiology, Faculty of Veterinary Medicine and Animal Science, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden Received 18 June 2010; received in revised form 3 November 2010; accepted 29 November 2010
Abstract The objective was to evaluate plausible risk-factors for pyometra, a common disease affecting almost 25% of all (unspayed) female dogs before 10 years of age. Because of the strong breed-predilection, an age- and breed-matched case-control study was undertaken on 87 pairs (pyometra-cases and healthy controls) from five breeds (Rottweiler (n ⫽ 13), Collie (n ⫽ 8), Golden retriever (n ⫽ 24), Labrador retriever (n ⫽ 16) and German shepherd dog (n ⫽ 26)). The mean age was 7.9 y (range 0.8 –13.8 y). Variables analyzed included pseudopregnancy, age at first oestrus, length of and regularity of the interoestrus interval, hormone treatments, nulliparity, number of parities, age at first whelping, previous urinary tract infections and mammary tumours. Data were modelled multivariably using matched-pair conditional logistic regression. Analysing interactions with breed, previous pregnancy was statistically associated with pyometra. When amalgamated, in three breeds previous pregnancy was protective (Rottweiler, Collie, Labrador retriever) and in one breed statistically intermediate (German shepherd dog) when compared to the baseline (Golden retriever). Previous pregnancy was a statistically significant factor that had a protective effect against pyometra in some breeds but not in the Golden retriever breed. These findings indicate that protective- and risk-factors may vary between different breeds. The obvious problem with low power and limited possibility for extrapolation, using few dogs in few breeds, is acknowledged. However, it is suggested that failure to control for the confounding effect of breed, especially in epidemiological studies on dog diseases, may lead to potentially erroneous conclusions. © 2011 Elsevier Inc. All rights reserved. Keywords: Infection; Dog; Epidemiology; Risk factors; Pair-matched case-control study; Pyometra; Confounding
1. Introduction Pyometra is a common disease in countries such as Sweden, where spaying/neutering is not generally practised and is diagnosed in nearly 25% of all intact (not spayed) female dogs before 10 y of age [1]. The potentially life-threatening disease is characterised by * Corresponding author. Tel.: ⫹46 18 672918; fax: ⫹46 18 673534. E-mail: [email protected] (R. Hagman). 0093-691X/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.theriogenology.2010.11.038
uterine infection and inflammation in combination with systemic illness [2]. Hormonal factors as well as bacterial infection are involved in the development of the disease [3– 4]. Large age-and breed-related differences with respect to occurrence have been reported [1,5]. Oestrogen therapy and nulliparity have been associated with increased risk for pyometra whereas overt pseudopregnancy has been linked with a reduced risk of developing the disease [5–7].With the purpose of further investigating the effect of some of these known and
1252
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
other new plausible factors the aim of this study was to perform a prospective case-control study. Because of previously shown age- and breed related differences in the risk for developing pyometra, matching on both breed and age was undertaken in order to be able to remove the influence of those risk-factors. Variables explored included pseudopregnancy, age at first oestrus, length of and regularity of the inter-oestrus interval, hormone treatment, parity, age at first whelping, previous urinary tract infections and mammary tumours amongst others. 2. Materials and methods 2.1. Study design A prospective case-control study matched on breed and age was designed. Five breeds with a high incidence of pyometra (Rottweiler, Collie, Golden retriever, Labrador retriever and German shepherd dog) were included in the study [1,5]. Approximately 40 pyometra cases of the selected breeds are admitted every year to the University Animal Hospital (UDS), Swedish University of Agricultural Sciences, Uppsala, Sweden. Sampling calculations for a case-control study showed that 80 cases were needed to detect differences of pseudopregnancy (unmatched design, 95% confidence and 80% power (http://www.open-epi.com)) based on a previous report of 20% prevalence differ-
ence in dogs with pyometra compared with healthy control dogs [6]. It was decided to sample data during approximately two years full-time knowing that all eligible cases could not be included due to time constraints, and continue until ⬃80 cases were accrued. All telephone interviews were made by one of the authors (RH). 2.2. Case selection Pyometra cases admitted to the UDS were intermittently included from 2001 through to 2007 (part-time collection due to employment reasons). During the time periods of employment, every diagnosed pyometra case of the chosen breeds was included to avoid possible selection bias. Dog-owners signed an informed consent. Complete physical examination was performed. Case history and physical examination data were registered and the diagnosis pyometra confirmed by identification of a pus-filled enlarged uterus during ovariohysterectomy. A telephone interview with the owner was performed the same day by one of the authors, and a questionnaire filled in. Demographic data including breed, Swedish Kennel Club (SKC) registration number, gender, date of birth, and contact details of owner were registered. The history questionnaire was designed to retrieve information regarding parameters presented in Tables 1– 4.
Table 1 Various clinical signs in number (No) and proportion (%) of the studied 87 female dogs with pyometra. Parameter Thirst Increased Decreased Appetite Increased Decreased General condition Moderately depressed Severely depressed Hydration status Mild dehydration Moderate dehydration Severe dehydration Circulatory abnormalities* Mammary gland tumour/s Lymph node/s enlarged Vaginal discharge Slight Moderate Abundant
Normal
Normal
Normal
Normal
Absent (i.e., closed cervix)
* Evaluated by auscultation of heart sounds, pulse quality.
No
%
29 56 2 24 0 63 12 52 23 50 26 11 0 3 10 4 13 18 30 26
33 64 2 27 0 72 14 60 26 57 30 13 0 3 11 5 15 21 34 30
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
1253
Table 2 Data regarding oestrus cycles and previous health history in pyometra cases and healthy control dogs illustrated in numbers (No) and proportions (%) of the total number of pairs evaluated. If data was missing in one dog (as for oestrus cycle interval regularity in 9 dogs and pseudopregnancy in 1 dog), the entire pair was removed from the analysis. The P-values display the statistical difference between the cases and controls from Bowkers/Mc Nemar tests. NT, Not tested; NE, Not estimable. Case history parameters
No pairs
Pseudopregnancy No Mild Moderate Severe Moderate to severe Oestrus interval Regular Nulliparity Hormone treatment Oestrus postponement, continuous Oestrus postponement, single Abortion Urinary tract infections Mammary tumour/s Skin disease Gastro-intestinal disease Respiratory disease
Pyometra cases
Control dogs
P
No
%
No
%
86 86 86 86 86
55 15 12 4 16
64 17 14 5 16
53 18 9 6 15
62 21 10 7 17
0.74 NT NT NT 0.85
78 87 87 87 87 87 87 87 87 87 87
67 75 9 4 3 1 9 8 9 7 0
86 86 10 6 3 1 10 9 10 8 0
73 67 8 3 0 5 8 7 9 4 0
94 77 9 3 0 6 9 8 10 5 0
0.11 0.12 0.62 0.71 NE 0.10 0.78 0.80 0.81 0.32 NE
2.3. Control dog selection
2.4. Definition of selected variables
For each pyometra case a female control was selected from data from the SKC registry. The control dog was matched by breed, geographical region (through postal address of the owner) and age (born ⱕ 3 mo before or after the case) by assistance of SKC staff. An identical questionnaire form as for the pyometra cases was filled in during a telephone interview, excluding questions related to the disease. The interview with the owner of the selected control dogs was performed within three weeks of the admittance of the pyometra case. Since no physical examination was performed in the control dogs, clinical variables are absent in this group. Siblings, dogs with previous pyometra or neutered, were excluded from the control group.
Oestrus intervals that had a variation in duration of maximum ⫾ 1 mo were accepted as regular. If the intervals were prolonged at older age, this was considered regular as long as they were uniform [6]. Clinical signs of pseudopregnancy were graded as mild, moderate or severe. Bitches that did not show any signs of altered behaviour or physiology as compared to anoestrus were defined as not pseudopregnant. Mild pseudopregnancy was defined as minor signs of inappetence, depression, decreased willingness to exercise or decreased activity. Moderate pseudopregnancy was defined as clearly noticeable clinical signs of altered behaviour as above with the addition of nesting and/or carrying around fictive pups. Severe pseudopregnancy
Table 3 Reproductive case history data in dogs with pyometra and control dogs. No* Beginning of last heat (weeks ago) Age at first heat (years) Interoestrus interval (years)
74 68 76
Pyometra
Control
P†
Mean
SD
Range
Mean
SD
Range
7.5 0.9 0.6
8.4 0.3 0.2
1–56 0.5–2.5 0.08–1.08
11.5 0.9 0.6
9.8 0.9 0.2
1–40 0.5–3.0 0.29–1.02
* Number of dog-pairs. † Paired Wilcoxon test. a Comparison biased since oestrus did not begin on the same day.
⬍0.002a 0.95 0.97
1254
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
Table 4 Presence of selected case history variables in the studied dog-pairs, as analysed by breed. Variable
Case
Control
Collie (8)
Golden Retriever (24)
Labrador retriever (16)
Rottweiler (13)
German shepherd dog (26)
All (87)
Previous pregnancy
Pos Pos Neg Neg Pos Pos Neg Neg Pos Pos Neg Neg Pos Pos Neg Neg
Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg Pos Neg
0 0 2 6 0 0 1 7 0 0 1 7 7 0 1 0
1 5 1 17 1 3 3 17 0 1 0 23 18 0 4 0
1 0 4 11 1 2 4 9 0 0 0 16 14 1 0 0
1 1 6 5 0 2 4 7 0 0 2 11 7 2 2 1
0 3 4 19 0 7 1 17 0 0 2 24 17 1 3 0
3 9 17 58 2 14 13 57 0 1 5 81 63 4 10 1
Pseudo-Pregnancy*
Abortion
Regular Oestrus cycles†
* Moderate to severe pseudopregnancy. Missing data in 1 dog-pair (German shepherd dog), 86 dog-pairs analysed. † Missing data in 9 dog-pairs (2 Golden retriever, 1 Labrador retriever, 1 Rottweiler, 5 German shepherd dog), 78 dog-pairs analysed.
was defined as major behavioural changes, as above, in combination with lactation.
2-way interactions, as was interactions with the matching variables breed and age.
2.5. Statistical analysis Associations between categorical variables and case status were tested using Bowker=s test (i.e., a generalised McNemars test). In order to perform the pairwise analysis the entire pair was removed from the dataset if data were missing in one individual. Continuous variables were tested against case status using the paired Wilcoxon test. Data were managed with SAS (SAS Institute Inc., Cary, NC, 27513, US) and the procedure PHREG was used to develop a multivariable pairmatched conditional logistic regression. The following dichotomous variables were first screened univariably: pseudopregnancy, moderate to severe pseudopregnancy, regular oestrus interval, nulliparity, ever had hormone treatment, elective pregnancy termination (abortion); ever had urinary tract infections, ever had mammary tumour/s, skin problems, gastrointestinal or respiratory problems. Continuous variables analysed were age at first heat (oestrus) and interoestrus interval. To check for linearity relative to the outcome these were also entered as categories, based on percentiles of the data (0 ⬍ 25, 25 ⬍ 50, 50 ⬍ 75 and ⱖ 75%). Variables with a univariable P-value of 0.15 were entered into a primary multivariable model, which was further reduced to P ⬍ 0.05. Variables with a univariable P-value of 0.20 were deemed to be checked for
3. Results 3.1. Study population In total 87 pyometra cases and 87 female control dogs (87 pairs) were recruited to the current study during the study period. Included pairs were 26 German shepherd dogs, 24 Golden retrievers, 16 Labrador retrievers, 13 Rottweilers and 8 Collies. The mean age in both groups (cases and controls) was 7.9 y (range 0.75–14 y). 3.2. Case history and physical examination data Case history and physical examination data in the 87 female dogs with pyometra are illustrated in Tables 1–3. In 74 (85%) of the pyometra cases, purulent or purulent-hemorrhagic vaginal discharge was present (open cervix). Case history data in the control dogs are presented in Tables 2 and 3. Data showing number of dogs by the matching variable breed in pyometra cases and healthy control dogs are illustrated in Table 4. 3.3. Descriptive statistics Mean age ⫾ SD (Range) at first whelping in the ten pyometra cases and the nineteen control dogs that had whelped (12% vs 22%) was 4.2 ⫾ 0.7 years (4 – 6
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
years) and 4.7 ⫾ 1.5 years (2–7 years), respectively. Of these, six pyometra cases and eleven controls were uniparous, two pyometra cases and two controls were biparous, two pyometra cases and four control dogs were triparous whereas no pyometra case and two control dogs whelped four times or more. Caesarean section was performed in four pyometra cases and two control dogs. Nine dogs with pyometra and eight control dogs were treated with hormones for oestrus postponement or pregnancy termination. Oestrus postponement had been generated with medroxyprogesterone acetate (MPA) administered subcutaneously in therapeutic doses (3 mg/kg every 3⫺5 months) which during the study years was the only drug registered for oestrus postponement in Sweden. Pregnancy termination had been induced in one pyometra case and five control dogs with a progesterone receptor blocker, aglepristone, 10 mg/kg subcutaneously on two consecutive days. Silent heat was registered in two cases of pyometra and one control dog. Five of the cases in the pyometra group were diagnosed before two years of age. Of these, three were German shepherd dogs (ages: 9 mo/10 mo/1 y) and two were Rottweilers (ages: 1 y/1 y and 5 mo). None of them had received any previous hormone treatment. The mean heart rate and temperature of the pyometra cases was 103 beats/minute (range 60 –120) and 38.9°C (range 36.1– 40.2), respectively. The duration of depressed general condition in these dogs previous to admission was 8.5 d (range 0 – 60 d). Of all the selected variables studied (Tables 2, 3 and 4), only nulliparity was significantly associated with pyometra, and only when analysed as an interaction effect with the matching variable breed (Table 4). In order to yield interpretable result, due to low numbers in strata, the three breeds in which previous pregnancy was protective, according to the point estimates were amalgamated while the intermediate (German shepherd dog) and high-risk breed (Golden retriever ⫽ baseline) were kept separately (Table 4). This yielded a hazard ratio (HR) of 0.08 (95% CI 0.01– 0.64) in the “three breed group” and an intermediate HR in the German shepherd dog (HR ⫽ 0.75 (95% CI 0.17–3.4)). 4. Discussion Cases and controls in the present study were matched by breed to avoid confounding by the previously demonstrated breed-related differences in the occurrence of pyometra [1]. The inherent problem of controlling for breed which is found in many epidemiological dog-studies was thereby removed. In addition,
1255
only five breeds (high incidence) were included in the present study to further limit the effect of breed variation, however also limiting the ability for extrapolation to all dog breeds. We also matched by age because pyometra mainly affects middle-aged and older individuals. It may be argued that controls were selected from a wider population than the cases. However, because pyometra is a well-defined disease it is highly likely that the waste majority of pyometra-cases in this wider population, would seek medical attention, and also be treated, similar to the study-cases. Concluding from this, controls and cases actually may be considered to emanate from the same population and the potential problem with different base populations would have a very limited impact [8]. A protective association was found between previous pregnancy and pyometra in a clump of three breeds—Collie, Rottweiler and Labrador retriever— when data were pooled due to scarcity of data. That nulliparity was associated with an increased risk of the disease is in agreement with results from previous studies [5– 6]. One or just a few pregnancies do not appear to give life-long protection against pyometra since the disease was diagnosed in multiparous individuals in the present study, as has also been reported previously [9]. That the Golden retriever differed from the other breeds may depend on nulliparity not being a risk factor in this particular breed or due to chance because of relatively few dogs deciding this outcome (18/24 previously nonpregnant Golden retrievers with pyometra compared with 22/24 non-pregnant healthy controls). However, risk-factors for pyometra may differ depending on breed, and should be considered when studying the disease (i.e., interactions with breed should be included). None of the other potential risk-factors analysed in this study showed an association with the diagnosis of pyometra. These results may depend on the case selection or dog breeds included. The data were also analysed using un-matched logistic regression (data not shown), but relative to statistical significance the results were the same. As for hormone treatments, the compound that had been used for oestrus postponement was Medroxyprogesterone acetate (MPA). In accordance with the results of the present study (hormone treatments, MPA or aglepristone, not linked with pyometra), progestin in therapeutic doses has previously not been associated with an increased risk for developing cystic endometrial hyperplasia (pre-stage to pyometra) or pyometra [5,10]. In a long-term study of 22 dogs treated with aglepristone for inducing abortion, pyometra occurred in two dogs after two and four
1256
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
years, respectively [11]. Pregnancy termination with the progesterone receptor blocker aglepristone was not associated with a diagnosis of pyometra in this study, but only few dogs were treated with this drug. Pyometra has been reported to occur less frequently in bitches that show clinical signs of pseudopregnancy [6,12–13]. In an early study of 68 cases and 245 healthy female dogs at least one occasion of overt pseudopregnancy was linked with reduced risk of developing the disease [6]. In the same study pseudopregnancy occurred in 43% of the control dogs and 23% of 68 pyometra cases [6]. The results of the present study did not, regardless of degree, demonstrate any protective effect of pseudopregnancy against pyometra. Pseudopregnancy was noted in 31 dogs with pyometra and 33 control dogs. Moderate to severe pseudopregnancy was noted in 16 of the dogs with pyometra dogs and 15 control dogs, respectively. The risk for pyometra was neither increased nor decreased in dogs that had developed mammary tumours despite reproductive hormone involvement in the pathogenesis of both diseases. These results are supported by the fact that the English springer spaniel was the only breed reported to have an increased risk of both pyometra and mammary tumours [14]. If the etiology of both diseases was similar the breed-predisposition would also possibly have been more alike in previous studies [1,5,14]. In pyometra, a bacterial infection is established during metoestrus when the uterine milieu is suitable for foetal development but also for bacterial growth [12– 13,15]. If every reproductive cycle (metoestrus) renders the uterus susceptible for pyometra, individuals with short interoestrus intervals would be at prone to develop the disease. Similarly, individuals with the first oestrus cycle at an early age would theoretically be susceptible for infection during longer time (every metoestrus) as compared to individuals that become fertile later. In this study, a short interoestrus interval, early fertility or irregular oestrus cycles were not associated with increased risk for pyometra. Escherichia coli (E. coli) are the most common bacteria isolated from both UTI and pyometra [16]. A subclinical urinary tract infection (UTI), as has been demonstrated previously in 25% of dogs with pyometra, may act as a bacterial reservoir since many virulence traits of E. coli from UTI and pyometra are similar [4,17]. In this study, it was for practical reasons not possible to explore the presence of subclinical UTI, but suffering from previous UTI or other diseases (in disease-prone individuals) was not associated with increased risk for pyometra. The clinical signs most commonly reported in the pyo-
metra group were vaginal discharge (in 84%), depressed general condition (in 83%), decreased appetite (in 72%), increased thirst (in 64%) and dehydration (in 43%) (Table 1). These results are similar to what has previously been reported in the disease and reflect a local uterine as well as generalised disease [2]. Data on time since the beginning of last heat (Table 3) is displayed to add information, but as noted, due to bias the difference between the two groups would only be relevant for comparison if oestrus started at the same time. The results of this study showed a significant effect only of nulliparity and not in all breeds. However, none of the earlier studies of variables related to pyometra controlled for breed, which is a strong risk factor [5– 6,12–13]. On the other hand the present study included a relatively small number of cases and with low power, hence in spite of reaching the goal of cases to include it is still possible that some of the anticipated other risk factors are true biological risk factors. If increasing the sample size to 180 pairs (unmatched design, 95% confidence and 80% power) it would have been possible to demonstrate statistical significance for nulliparity in the entire data-set given that the effect was the same as in this study (i.e., 89% in the case-group and 78% in the control-group) (http://www.open-epi.com). It was deemed possible to use population controls because pyometra is a disease which the owner is likely to have noted and remembered. Recall bias was difficult to avoid completely. Recall may be different among cases and controls, or here their owners. In this study, however we believe that having puppies is the only perceived protective factor that people would ruminate over relative to the occurrence of pyometra and that pregnancy is highly likely to be equally well remembered by both groups of owners. 5. Conclusion Previous pregnancy was a statistically significant factor that had a protective effect against pyometra in some breeds but not in the Golden retriever breed. These findings indicate that protective- and risk-factors may vary between different breeds.
Acknowledgments We thank A-M Darell and T. Wink at the SKC for their help with selecting control dogs and U. Emanuelsson for valuable input on the manuscript. Lennart Swensson is acknowledged for contribution to the initial study design. The Foundation for Research, Agria
R. Hagman et al. / Theriogenology 75 (2011) 1251–1257
Insurance Co and the Swedish Kennel Club and Thure F. and Karin Forsgren’s Foundation are acknowledged for financial support. The funding body had no role neither in or influence over study design, data collection, analysis, and interpretation of data, nor in the writing of the manuscript or in the decision to submit the manuscript for publication. RH carried out the interviews and data collections and drafted the manuscript. AE performed the statistical analyses and helped to draft the manuscript. All authors participated in the planning and design of the study and interpretation of the results. All authors read, revised and approved the final manuscript.
References [1] Egenvall A, R. Hagman BN Bonnett A Hedhammar P Olson and AS Lagerstedt. Breed risk of pyometra in insured dogs in Sweden. J Vet Intern Med 2001;15:530 – 8. [2] Borresen B. Pyometra in the dog. II—A pathophysiological investigation. II. Anamnestic, clinical and reproductive aspects. Nord Vet Med 1979;31:251–7. [3] Borresen B. Pyometra in the dog—a pathophysiological investigation. I. The pyometra syndrome, a review. Nord Vet Med 1975;27:508 –17. [4] Sandholm M, Vasenius H. AK Kivisto Pathogenesis of canine pyometra. J Am Vet Med Assoc 1975;167:1006 –10. [5] Niskanen M, MV Thrusfield. Associations between age, parity, hormonal therapy and breed, and pyometra in Finnish dogs. Vet Rec 1998;143:493– 8.
1257
[6] Fidler IJ, Brodey RS, Howson AE, Cohen D, Relationship of estrous irregularity, pseudopregnancy, and pregnancy to canine pyometra. J Am Vet Med Assoc 1966;149:1043– 6. [7] Whitehead ML. Risk of pyometra in bitches treated for mismating with low doses of oestradiol benzoate. Vet Rec 2008;162: 746 –9. [8] Dohoo I MW, Stryhn H. Validity in Observational Studies. In: Veterinary Epidemiologic Research. AVC Inc.: Charlottetown, PEI, Canada, 2008, p. 184. [9] Fukuda S. Incidence of pyometra in colony-raised beagle dogs. Exp Anim 2001;50:325–9. [10] Kooistra HS, Okkens AC, Mol JA, E van Garderen, Kirpensteijn J, Rijnberk A. Lack of association of progestin-induced cystic endometrial hyperplasia with GH gene expression in the canine uterus. J Reprod Fertil Suppl 1997;51:355– 61. [11] Pettersson CH Tidholm A. Safety and efficacy of mid-term pregnancy termination using aglepristone in dogs. J Small Anim Pract 2009;50:120 –3. [12] Dow C. The cystic endometrial hyperplasia-pyometra complex in the bitch. Vet Rec 1958;69:1409 –14. [13] Dow C. The cystic hyperplasia-pyometra complex in the bitch. J Comp Pathol 1959;69:237–50. [14] Egenvall A, Bonnett BN, Ohagen P, Olson P, Hedhammar A, H von Euler. Incidence of and survival after mammary tumors in a population of over 80,000 insured female dogs in Sweden from 1995 to 2002. Prev Vet Med 2005;69:109 –27. [15] Cox JE. Progestagens in bitches: a review. J Small Anim Pract 1970;11:759 –78. [16] Hagman R, Greko C. Antimicrobial resistance in Escherichia coli isolated from bitches with pyometra and from urine samples from other dogs. Vet Rec 2005;157:193– 6. [17] Chen YM, PJ Wright, Lee CS, Browning GF. Uropathogenic virulence factors in isolates of Escherichia coli from clinical cases of canine pyometra and feces of healthy bitches. Vet Microbiol 2003;94:57– 69.