Risk factors for appendicular osteosarcoma occurrence in large and giant breed dogs in western Canada

Abstract

Objective

Risk factors for the development of canine appendicular osteosarcoma (OSA) have been investigated in numerous studies, but with contradictory results. The aim of this study was to analyze weight, age, breed, sex, neuter status, body condition score, and previous lameness in a population of large and giant breed dogs in western Canada with and without appendicular OSA.

Animals and procedure

Medical records of 227 large or giant breed dogs diagnosed with appendicular OSA were compared to records from a control population of 454 large and giant breed dogs from the years 2000 to 2020.

Results

Gonadectomized dogs, body condition score (BCS), and a history of lameness condition(s) (other than OSA) were associated with increased odds for presentation with OSA. Breeds shown to have increased odds for appendicular OSA occurrence included Rottweilers and Great Danes relative to Labrador retrievers.

Conclusion and clinical relevance

Obesity and lameness appear to be independently associated with appendicular osteosarcoma. This study demonstrated that spayed females had the greatest risk compared to other sex and neuter status combinations; further investigation of these factors would be beneficial.

Introduction

Osteosarcoma is the most often diagnosed primary bone tumor in the dog, representing an estimated 85% of all skeletal malignancies, and unfortunately carries a generally poor prognosis (1). At the time of diagnosis, most dogs with appendicular OSA have pulmonary micro-metastasis present, so the amputation or limb-sparing procedures commonly employed to remove the primary tumor do palliate pain but do not address suspected metastasis (1). Although combination therapy using chemotherapeutics in addition to surgical removal of the primary tumor has demonstrated an ability to prolong lifespan, compared to surgery alone, most animals do not survive 1 y beyond diagnosis (1).

Numerous studies have attempted to characterize risk factors for canine OSA. Perhaps the best described predisposing factor is large size, whereby increasing height and weight have been strongly associated with increased OSA risk (2–4). There also appears to be a genetic predisposition to the condition, with Rottweilers, greyhounds, golden retrievers, Labrador retrievers, German shepherds, and Irish wolfhounds consistently over-represented for developing OSA (2,5,6). More contradictory are claims that sex and neuter status influence OSA development, with some studies indicating a bias towards male dogs or gonadectomized animals, and others finding no significant difference between the sexes or between gonadectomized and intact animals (2,3,5,7,8). Similarly, the reported age distribution of canine OSA has some discrepancies, with some indications of a bimodal distribution, and other findings suggesting a generalized increase in risk with age, without bimodal distribution (2,3,5).

To the authors’ knowledge, obesity and historical lameness conditions have not been explored as possible risk factors for the development of appendicular OSA. Notably, body condition score as a measure of obesity has been previously assessed as a prognostic factor for dogs with OSA and lymphoma, but risk of OSA development was not examined (9). Although an etiology for OSA has not been established, chronic inflammation, chronic trauma, and rapid growth have all been implicated as possible causes or influences since they promote a state of osteoblast proliferation (3). Therefore, it is plausible that obesity, which promotes a pro-inflammatory state and has been linked to an overall increased risk of cancer, should be investigated for an association in dogs with OSA (10). In addition, with evidence of increased risk of appendicular OSA correlated to increased weight across breeds, obese body condition likely also promotes osteoblast proliferation in the limbs as they remodel in response to increased forces placed upon them (2–4). Similarly, we predict that historic lameness conditions alter osteoblast activity either by promoting inflammation in the painful limb, or by altering weight distribution off the lame limb, thereby increasing the forces acting upon the remaining sound limbs (6).

Interestingly, despite being the most common primary bone tumor in dogs, risk factors for canine appendicular OSA have not been specifically assessed in dogs in western Canada to the authors’ knowledge (1). Furthermore, obesity and historic lameness conditions have not been investigated for their potential associations with appendicular OSA development. Therefore, the objective of this study was to assess these potential risk factors in large and giant breed dogs in western Canada through description of characteristics of weight, age, breed, sex, neuter status, body condition, and historical lameness in dogs affected or unaffected with OSA. A secondary objective was to characterize potential risk factors through multivariable analysis.

Materials and methods

Study design

Using a retrospective case-control approach, medical records of dogs were searched and retrieved from the Western College of Veterinary Medicine-Veterinary Medical Centre (VMC) data base. Search criteria included dogs that were admitted between January 1, 2000 and January 1, 2020 that were diagnosed with OSA, weighed at least 18 kg at time of diagnosis (4), and those with large or giant breed lineage (according to American Kennel Club classifications). Control subjects were recruited based on the same criteria (except that there was no diagnosis of OSA in their record) at a 2:1 ratio to the OSA cases. As the primary objective of the study was descriptive, sample size was determined by including all available cases in the described time frame. The number of controls selected was based on a 2:1 ratio to increase the power of the study.

Subjects

The OSA group was then limited to dogs diagnosed with appendicular-located OSA. Radiographic interpretations were only included in the study if the radiographs were interpreted with OSA as the primary differential by a radiologist at the VMC. Microscopic diagnoses made up most cases and were determined by fine-needle aspiration and cytopathology, or biopsy and histopathology performed either pre-treatment, or post-amputation.

The control group was recruited from the Ophthalmology Service, as dogs presented to this service were expected to have a similar likelihood of being brought to the VMC on a referral, emergency, or routine care basis compared with the OSA cases. By sharing this trait, case and control populations had similar amounts of data available in terms of the information sent by their primary veterinarian as a part of their referral record, or data collected by VMC clinicians as a part of an emergency visit or general wellness record. These controls are likely to represent the same underlying population as for the oncology service. This service also provided enough potential controls for the designated study period and was less likely to be biased towards the variables of interest. For reference, the only other service with an adequate number of dogs during this period was the surgery service, which was not selected due to the bias towards lameness cases. The control group was also restricted to dogs with large or giant breed lineage. For the remainder of the study report the cases will be referred to as the “OSA group” and controls as the “Non-OSA group.”

For both OSA and Non-OSA dogs, weight at diagnosis, breed, sex, neuter status, age at initial presentation, body condition at diagnosis, and historical lameness were extracted from the case record. All breeds were recorded, but only breeds represented by ≥ 20 dogs were included as categories for analysis, with all other breeds categorized as “other.” Mixed breeds were included in the category of their dominant breed as per the record description for greater power. Body condition score had been assessed by VMC clinicians using the American Animal Hospital Association Guidelines (11). Both 5-point and 9-point scales are included in these guidelines; therefore, measurements from the 5-point scale were converted to their 9-point counterpart for statistical analysis of obesity in this study (11). Dogs were categorized as ideal or underweight if the BCS was ≤ 5/9 and were categorized as overweight or obese if the BCS was ≥ 6/9. Lameness conditions were included if they had been diagnosed at least 1.5 y before presenting for OSA or ophthalmic consult to allow comparison with available force plate metric data on weight distribution patterns associated with chronic lameness (12). These conditions must have also induced lameness for a minimum of 2 wk to allow for the possibility that bone remodeling could occur in response to the condition. Based on this history, dogs were then classified either as those with a history of a lameness condition or those without. Data on the location and method of diagnosis were also included for the OSA group. All extracted data were recorded in a spreadsheet software package (Excel, Microsoft Office, 2010).

Statistics

Descriptive

Continuous variables (weight and age) are reported as an arithmetic mean with standard deviation (SD). Frequency of occurrence is reported for categorical variables. Odds ratios (OR) with a 95% confidence interval (CI) were used to assess effect of variables (age, weight, breed, sex, neuter status, excess body condition score, and history of lameness) on OSA occurrence. A commercial software package (Stata 16. StataCorp, 2019. Stata Statistical Software: Release 16. College Station, Texas, USA) was used for all statistical analyses.

Univariable logistic regression

Logistic regression was used to assess each independent variable for association with OSA outcome with logistic regression. A threshold of P < 0.2 was used for consideration in the final model. The linearity assumption was assessed for significant continuous variables with inclusion of a squared term. All variables considered for inclusion in the final model were evaluated for collinearity by calculating a Kendall’s tau-b value, with 0.8 or less used as a threshold for separate inclusion.

Multivariable logistic regression

In this hospital population, enough cases were identified to develop estimates of odds ratios in a multivariable analysis. Variables that were not independently associated with the outcome (P < 0.05) or important confounders were excluded from the final model. Model building was conducted by backwards stepwise elimination. Reference groups for variables were chosen based on the recommended practice to have a group of sufficient size and a low rate of occurrence of the outcome of interest. Variables were assessed for confounding effects with a change in coefficient > 20% considered significant for confounding. Model fit was assessed with McFadden’s R² value and output from a receiver operator’s curve. Residuals for the final model were examined for evidence of observations with high leverage or deviance.

Results

A total of 227 client-owned dogs met inclusion criteria for the OSA group, with an age range of 6 mo to 15 y, and a mean age at diagnosis of 7.9 y (SD = 2.9 y). The average weight of dogs recruited in this population was 41.9 kg (SD = 12.5 kg), with a range of 18.3 to 88.5 kg. Diagnosis of OSA was achieved microscopically (by cytopathology or histopathology) in 62% of subjects, with the remaining 32% based on radiographic interpretation; the most frequent tumor locations observed were the proximal humerus and the distal radius (Table 1). A total of 454 client-owned dogs were recruited for the Non-OSA group according to inclusion criteria and had a mean age of diagnosis of 7.9 y (SD = 2.8 y). Weight in the Non-OSA group ranged between 18.3 and 83 kg with a mean weight of 33.5 kg (SD = 10.1 kg).

Table 1.

Frequency of tumor location in large or giant breed dogs in western Canada with appendicular osteosarcoma by bone involved.

Location	Number of cases	% cases
Humerus	73	32.16
Radius	64	28.19
Tibia	42	18.50
Femur	32	14.10
Ulna	6	2.64
Scapula	4	1.76
Metatarsal	2	0.88
Fibula	2	0.88
Calcaneus	1	0.44
Metacarpal	1	0.44
Total cases	227

When assessed with univariable regression, neither age nor weight was significant (P = 0.78 and P = 0.81, respectively) and therefore were not included for assessment in the final model. However, the remaining categorical variables of breed, sex/neuter status, historic lameness, and obesity were all determined to be independently associated (P < 0.05) with appendicular OSA development and were retained in the final model. The final model was run with and without radiographically diagnosed cases, and since no significant difference was demonstrated in models or associations, results from the model with these cases included are presented.

Breeds appearing the most frequently in the OSA group included the Rottweiler, Labrador retriever, golden retriever, German shepherd, and Great Dane. Prominent breeds amongst the Non-OSA group included the Labrador retriever, German shepherd, and golden retriever (Table 2). A total of 83 mixed-breed animals were present in the study population: 27 from the OSA group, and 56 from the Non-OSA group. The effect on the analysis of including mixed-breed dogs with their dominant breed (i.e., Labrador retriever cross with ‘Labrador retriever’) was tested by running a separate analysis excluding all mixed-breed dogs from the analysis. No significant difference in OR for any variable was detected; therefore, the final model results presented here include mixed-breed dogs with their dominant breed type. Since Labrador retrievers were well-represented in both groups and demonstrated low odds for OSA occurrence in univariable modeling, this group was selected as the reference category in the final model. This study showed Rottweilers and Great Danes to have increased odds for presenting with appendicular OSA, with an OR of 6.84 (95% CI: 3.08 to 15.2) and OR of 3.7 (95% CI: 1.25 to 11) compared to Labrador retrievers, respectively (Table 3).

Table 2.

Frequency of various breeds in dogs in western Canada with appendicular osteosarcoma, and those without.

Breeda	Osteosarcoma group			Non-osteosarcoma group
	Number of cases	Crosses	% of total cases	Number of cases	Crosses	% of total cases
Alaskan malamute	7		3.08	2		0.44
American pitbull	2	2	1.76	4		0.88
Bernese Mountain dog	5		2.20	2		0.44
Boxer	1	2	1.32	26		5.73
Bulldog	6		2.64	25	1	5.73
Dalmatian	2	1	1.32	1		0.22
Doberman pinscher	4		1.76	1		0.22
German shepherd	12	9	9.25	52	4	12.33
Golden retriever	22	1	10.13	52	3	12.11
Great Dane	11		4.85	11		2.42
Great Pyrenees	6		2.64	4	4	1.76
Greyhound	9		3.96	3		0.66
Labrador retriever	32	3	15.42	101	2	22.69
Mastiff	3		1.32	8		1.76
Newfoundland	4		1.76	4	1	1.10
Rottweiler	43	6	21.59	19	2	4.63
Saint Bernard	7		3.08	2		0.44
Siberian husky	4		1.76	28	3	6.83
Total	200	27		425	29

^aOnly breeds represented by at least 3 dogs (including crosses of that breed) were included in this table.

Table 3.

Association between appendicular OSA and breed, sex, neuter status, body condition score, and history of lameness in dogs in western Canada in a multivariable logistic regression model. Only patients with an available BCS were included in the final model and are presented in this table (N = 526).

	OSA	Non-OSA	Odds ratio	95% CI	P-value
Breeda,b					< 0.001
Labrador retriever	24	85	—	—	—
Rottweiler	33	18	6.84	3.08 to 15.2	< 0.001
Golden retriever	17	45	1.53	0.70 to 3.33	0.286
German shepherd	16	44	1.88	0.85 to 4.16	0.121
Great Dane	9	11	3.70	1.25 to 11.0	0.018
Siberian husky	3	25	0.48	0.12 to 1.87	0.287
Boxer	3	21	0.72	0.19 to 2.76	0.635
Other	54	124	1.77	0.96 to 3.28	0.069
Sex and neuter statusc					0.047
Intact female	2	48	—	—	—
Spayed female	74	152	6.95	1.60 to 30.1	0.010
Intact male	7	31	3.85	0.72 to 20.7	0.116
Castrated male	70	142	6.33	1.45 to 27.6	0.014
Body condition scored,e
Ideal or underweight	85	303	—	—	—
Overweight or obese	68	70	3.18	2.01 to 5.02	< 0.001
Historical lamenessf
Not present	100	329	—	—	—
Present	54	44	3.11	1.88 to 5.14	< 0.001
Total	153	373

^aBreeds that were represented by less than 20 dogs were included in the “Other” category.

^bLabrador retrievers were used as a reference category.

^cIntact females were used as a reference category.

^dDogs were considered “under or ideal weight” at scores between 1 and 5 out of 9, and “overweight or obese” at scores between 6 and 9 out of 9.

^eUnder or ideal weight dogs were used as the reference category.

^FDogs without historic lameness were used as the reference category.

Sex and neuter status were categorized into 4 groups: intact male, castrated male, intact female, and spayed female. Univariable analysis determined intact females to have the lowest OR of these categories within the study population, so thereafter intact females were used as the reference group for comparative purposes in the final model. This grouping was created to control for the interaction between sex and neuter status and allow for a complete overall model to be built. Within the Non-OSA group a total of 17 dogs presented for ophthalmic assessment for breeding soundness examination. To assess whether this biased the analysis towards increased intact animals within the Non-OSA group, the final model was run without these dogs. No appreciable change in odds, P-values, or interpretation was noted; therefore, these dogs were retained in the final model. Both spayed females and castrated males were shown to have increased odds for presenting with appendicular OSA (compared to intact females), with an OR of 6.95 (95% CI: 1.6 to 30.1) and an OR of 6.33 (95% CI: 1.45 to 27.6), respectively (Table 3). Overall, historic lameness conditions were noted in a total of 136 dogs; 98 dogs with a history of lameness, that also had BCS recorded, were included in the full analysis (N = 526) for the final model (Table 3). Within the OSA group, the previous lameness condition was most often seen in the eventual OSA limb (45% of dogs), followed closely by lameness in the limb ipsilateral to the eventual neoplastic limb (42% of dogs). The most common cause of previous lameness within the OSA population was cruciate disease (31% of dogs), followed by osteoarthritis and fracture (24% and 19% of cases, respectively). Dogs with historic lameness conditions had increased odds of presenting with appendicular OSA (3.11, 95% CI: 1.88 to 5.14) compared to those who did not (Table 3).

A total of 526 dogs had a BCS available in their record: 153 from the OSA group and 373 from the Non-OSA group (Table 3). A median score of 6/9 was calculated for the OSA group, and a median score of 5/9 was calculated for the Non-OSA group. Only dogs with an available BCS were included in the final analysis. This analysis showed that overweight or obese dogs had increased odds (OR 3.18, 95% CI: 2.01 to 5.02) for presentation with appendicular OSA compared to ideal or underweight dogs.

Fitness of the logistic regression model was determined to predict 77% of outcomes according to the logistic receiver operator’s curve. Other than the described interaction effect of sex and neuter status, no other significant interactions were noted for variables in the final model. No observations were noted with undue leverage or deviance.

Discussion

Risk factors

This study demonstrated that Rottweilers, Great Danes, gonadectomized dogs, overweight and obese dogs, and dogs with historic lameness to have a significantly increased OR for presenting with appendicular OSA compared to controls drawn from a referral population.

Similar to other studies, results of the present investigation determined that average weight in the Non-OSA group was below that calculated for the OSA group (2–4). However, due to the limitation of the study to large and giant breeds only, a significant effect for weight (which does not discriminate by body condition) as a risk factor for OSA could not be demonstrated. In addition, since body condition and not weight was not identified as a risk factor, weight separate from that may not be a risk factor.

Also, a significant relationship between age and increasing risk for OSA was not observed in this study, possibly because of similarities in age of presentation for common ophthalmic disorders in the Non-OSA group when compared to age of presentation for OSA in this study. However, unlike previous research (3), no bimodal distribution was observed for age of diagnosis of appendicular OSA in dogs in this investigation.

Commonly implicated breeds in terms of predisposition to OSA were also observed to have increased odds in this study, most notably the Rottweiler and the Great Dane. Previous studies have demonstrated numerous highly prevalent and heritable OSA-associated loci in Rottweiler populations, as well as increased expression of genes which control bone differentiation pathways (13). In addition, altered expression of p53, a known tumor suppressor protein, has been observed in osseus tumors of Rottweilers compared with other breeds, further suggesting a genetic component to OSA susceptibility (14). However, unlike previous work, our study did not find significantly increased risk for appendicular OSA in golden retrievers, Labrador retrievers, or German shepherds; and Saint Bernards, greyhounds, and Irish wolfhounds were unable to be assessed due to low frequency in the sample population (2,5,15). A plausible reason for golden retrievers, Labrador retrievers, and German shepherds not demonstrating significant risk in this study could be that genes which predispose to OSA development are less prevalent in this population than in others (6). Also, for German shepherds in particular, breed-related factors which make them more likely to present to the ophthalmology service could potentially mask breed effects for OSA risk in this study due to the recruitment of control cases from this service (16).

Contradictory results have been reported regarding males having an increased risk for developing appendicular OSA (2,7,8), and in the current investigation no sex effect was demonstrated for the disease. This is contrary to medical literature, which consistently finds increased risk in young men for appendicular OSA, which otherwise presents similarly to canine OSA (3). A limitation to detecting this difference is the low number of intact males in the study.

Another inconsistent finding in the literature is that neuter status is influential in the development of OSA (2,3,7,8). Increased odds in the study herein was determined for gonadectomized animals compared to the intact female reference group, with spayed females demonstrating the highest odds overall. It has been theorized that sex steroids are pro-differentiation agents which discourage the malignant transformation of osteoblasts, and therefore have protective effects against the development of OSA and other cancers (17). Therefore, the deprivation of these hormones at an early age is proposed to predispose to OSA development and could explain the increased odds for gonadectomized dogs observed in the present study (18). The age at which gonadectomy is performed is also thought to be influential, with dogs neutered before 1 year-of-age demonstrating generally increased risk, but unfortunately date of surgery was not consistently available in this population for the purposes of analysis (17). More recent evidence suggests that age of gonadectomy has differing effects on future cancer risk depending on the breed, although no significantly increased risk for OSA specifically was determined for any of the well-represented breeds in the current investigation (18). Some studies have suggested that spayed female dogs appear to have increased risk for age-related disorders because gonadectomy has a more profound effect on prolonging life in females compared to males. It is proposed that by decreasing risk for diseases with high mortality such as pyometra, spayed female dogs are more likely to reach an age at which they are potentially at greater risk for developing appendicular OSA (19). In addition, numerous studies have shown an association between neutering and the development of overweight and obese conditions, which could further precipitate OSA in the limbs of gonadectomized dogs (10). However, our study suggests that even after accounting for the effect of obesity, gonadectomized dogs are at increased odds of being diagnosed with OSA.

Whether or not excess body condition score was driven by gonadectomy in this population, when studied as an independent variable, above-ideal body condition proved to be highly correlated to the occurrence of appendicular OSA. Although this finding is unique in the literature it is not surprising given the well-established concept that increasing weight is associated with increased risk of OSA (2,3). In previous work, weight had been adjusted according to breed or class (small, medium, large, or giant breeds), but obesity within those classes was not explored (4). In theory, obesity places increased forces on the limbs, encouraging bone remodeling in response, and the resultant osteoblast-proliferative state could be a precursor to oncogenesis in the bone. This is supported by the finding (in this study and others) that appendicular osteosarcoma is more commonly diagnosed in the forelimbs, as they proportionally bear more weight than the hind limbs in the dog (5). Furthermore, obesity has been linked to increased rates of cancer in general due to the ability of adipocytes to produce cytokines which promote a pro-inflammatory state, precipitating cellular damage and eventual oncogenesis (10). As obesity is a preventable condition, it is recommended that veterinarians provide appropriate client counselling in regard to increased odds for OSA development as well as numerous other conditions which are associated with excess body condition.

This study also demonstrated historic lameness conditions to have increased odds for the eventual diagnosis of appendicular OSA. Previous case series have shown increased susceptibility to OSA at sites of previous fractures repaired with metal implants (20). Although inconclusive at present, it is proposed that any combination of repair processes in response to the trauma, chronic inflammation in response to the implant, or carcinogenic properties of the implants themselves contribute to this phenomenon (20,21). Also, the high rate of appendicular OSA development in retired racing dogs is suggestive that a history of trauma and ongoing repair processes in the appendicular bones can predispose to OSA in the limbs (22). These mechanisms are consistent with our findings of previous lameness in a limb being highly associated with later development of OSA in that limb. However, this does not explain the relatively high amount of OSA cases in which OSA development occurred in a limb ipsilateral to that affected by a previous lameness condition. Extensive studies on dogs following repair of a ruptured cruciate ligament have demonstrated that although the affected limb is painful, the weight is disproportionally redistributed to the remaining limbs, with the forelimbs still bearing the most weight (12). In addition, force plate analysis performed on dogs following the amputation of a hind limb showed that the ipsilateral forelimb to the amputated leg was subject to the greatest amount of force during weight-bearing after surgery (23). These observations are supportive of the theory that lameness ipsilateral to a future site of OSA could contribute to that neoplastic development by increasing forces placed on the eventual OSA-affected limb, in an effort to shift weight off of the originally lame leg. For instance, dogs in this study which suffered a cruciate injury frequently developed OSA in the ipsilateral forelimb, possibly due to excess forces placed on that forelimb during the period of lameness. It is imperative, therefore, that painful conditions causing lameness be addressed as early as possible to prevent excessive remodeling processes in the limbs as they adjust to altered weight distribution.

Limitations

The retrospective nature of this study introduces some limitations. As in all retrospective studies, we are not able to rule out that evaluated risk factors that develop (i.e., BCS, lameness) occurred before the development of the evaluated outcome. Recruited dogs were seen by various clinicians, so there may be some variability involved in how each animal was diagnosed for OSA and lameness, and how they were scored for body condition. Ideally all OSA cases would have a histologic diagnosis, but to provide more power to the study and to better reflect the clinical reality, radiographic diagnoses were also included. Misdiagnosing OSA (misclassifying some dogs in the OSA group) would decrease our ability to detect risk factors. Not all records included a body condition score at time of diagnosis, so conclusions made from these measurements and the overall final multivariable analysis are only from a subset of our study subjects. This would be a limitation if non-recording was differentially distributed in the OSA and Non-OSA groups and related to BCS value. We also recognize that medical records in either group are subject to a certain amount of recall bias in that potentially not all lameness conditions were reported in the medical record. The difficulty of accurately genetically characterizing dogs based on client reported breed (especially for mixed-breed) means that breed-specific odds should be interpreted with caution; this is in common with all studies using client-reported breed. Breed-associated odds were not a focus for this study; they are included in the analysis to better explore other risk factors. Since results for lameness and BCS-related effects were similar regardless of how mixed-breeds were grouped, we have left them included with their dominant breed. Furthermore, by recruiting a control population from the ophthalmology service, this study has the potential to under or overestimate risk for considered variables, depending on their association with being presented. Selection of control populations in risk factor studies is complicated in veterinary medicine by the fragmented and voluntary nature of health care utilization.

Since our study was limited to dogs admitted to the VMC, these findings cannot be applied to the general dog population. A limitation our study shares with previous investigations is that demonstration of risk in gonadectomized versus intact animals is difficult, due to the high prevalence of neutered status amongst client-owned dogs in North America. Although the precision of the estimate of odds in intact animals was less [fewer animals: females (n = 50) and males (n = 38)] than in gonadectomized animals the analysis still demonstrated that compared to intact females, both spayed females and neutered males had higher odds of having developed OSA; intact male odds were intermediate and not significantly different. Our recommendation is that further studies be performed in other populations in which intact status is more prevalent. In addition, prospective cohort and case-control studies should be undertaken to further explore identified potential risk factors; determination of sample sizes in these hypothesis testing studies can be informed by the initial work of this study. This study was also limited to dogs with appendicular OSA, and therefore our findings are not applicable to non-appendicular OSA.

In conclusion, in agreement with and as an extension from previous studies, our conclusions were that certain breeds, spayed females, overweight or obese body condition, and dogs with a history of another lameness condition were at increased risk for developing appendicular OSA. Continued exploration of these factors is recommended so that veterinarians may advise clients in a manner that contributes towards reducing risk of appendicular OSA in dogs.