Materials and Methods

Sample population—Records of the Canadian Veterinary Urolith Centre were compiled for CaOx and struvite urinary calculi submitted from dogs that lived in Ontario between February 1, 1998, and December 31, 2006. Submissions that represented recurrent uroliths were excluded. A questionnaire that included demographic information, diet information, and owner and clinic addresses was submitted along with each urolith. Owner and clinic addresses were used in a spatial program^a to code dogs and clinics with regard to location (latitude and longitude coordinates).

Dog-level variables—Individual dog-level variables included demographic variables (age, sex, breed, and neuter status), body condition (thin, normal, or obese), diet (primary diet fed and duration diet was fed), and year the urolith submission was received. Breed was grouped into 3 categories (mixed-breed dogs, small-breed dogs, and medium- or large-breed dogs). The small-breed category included all the small or toy breeds listed by the American Kennel Club.²⁰ Small or toy breeds not listed by the American Kennel Club were included in the small-breed category if they were listed on a small-breed website²¹ and their size could be verified from pertinent breed-specific websites. The mixed-breed category included those listed as a mixed-breed dog on the questionnaire regardless of size. The medium- or large-breed category included all remaining dogs. Their classification was verified through a listing by the American Kennel Club and information from pertinent breed-specific websites and a large-breed website.²² Body condition of a dog was classified on the basis of the submitting veterinarian's assessment. Diet (considered the first diet listed or major diet consumed) was dichotomized into veterinary-exclusive diet fed for a minimum of 6 months or other diet, which comprised veterinary-exclusive diets fed for < 6 months or an unknown duration, commercial diets (ie, nonprescription diets), homemade diets, and unknown diets. The cutoff of 6 months was chosen to avoid any bias associated with short-term veterinary treatment diets. Veterinary-exclusive diets included any treatment, preventive, or maintenance diet that was only available by veterinarian prescription and unavailable through a retail outlet. Urinary diet was dichotomized into urolithiasis treatment diets fed for a minimum of 6 months or nonurinary diets, which comprised urinary diets fed for < 6 months or a nonurinary treatment diet. All urinary diets were veterinary-exclusive diets. One additional dietary variable was included: dietary moisture content was categorized as dry, canned, or both.

Community-level contextual variables—Canadian municipalities or communities deemed to be the equivalent of municipalities for statistical reporting purposes are classified as CSDs. Uroliths were submitted from dogs that lived in 302 CSDs. Geocoding of the owner location and linking to its Statistics Canada CSD allowed us to obtain information for the communitylevel socioeconomic variables median family income and SAC.²³ The CSD median family income was categorized in quartiles rounded to the nearest $1,000 (Canadian dollars). All CSDs were classified by SAC type on the basis of their degree of integration with a central urban area. The SAC is a measure of urban influence on a municipality, and SAC classification combines CSDs according to whether they are a component of a CMA (1 or more adjacent municipalities centered on an urban core with a population of ≥ 100,000), a CA (1 or more adjacent municipalities centered on an urban core with a population of 10,000 to 99,000), or a CMA- or CA-influenced zone based on the percentage of the resident employed labor force working in the urban core of these CMAs or CAs.²³ Two additional contextual variables (owner's location within a previously detected, significant spatial CaOx or struvite cluster¹¹ and distance between the geocoded owner and clinic location) were investigated to determine association with the outcome. The variable cluster was used to describe uroliths from dogs that were within the 122-km-diameter struvite spatial cluster centered in the municipality of Hastings Highlands, ON, Canada; the 42-km-diameter CaOx spatial cluster centered in Toronto; or neither cluster. Accessibility to a submitting veterinary clinic was determined by the distance between the owner and location of the clinic that submitted the urolith, as calculated by use of the Haversine equation.^b Year of urolith submission was also included as a variable.

Uroliths—Uroliths had been surgically removed or passed during natural or assisted voiding procedures and submitted to the Canadian Veterinary Urolith Centre for quantitative analysis. Only uroliths comprising a minimum of 70% of CaOx or magnesium ammonium phosphate (ie, struvite) mineral were included in the study. When the nidus of a urolith differed from its overlying layers, then the urolith was classified on the basis of the mineral type in the nidus. Mineral analysis involved the use of 2 techniques (polarized light microscopy and x-ray diffraction scanning electron microscopy). These quantitative analyses were provided as a complimentary service to veterinarians and their clients. For the purposes of the study, CaOx included CaOx monohydrate and CaOx dihydrate.

Statistical analysis—Adaptive quadrature was used to build multilevel logistic regression models with a random intercept for owner CSD.^c The outcome variable was dichotomous (ie, CaOx or struvite urolith). All tests were performed as 2-tailed tests with significance set at a value of P ≤ 0.05. All explanatory variables were initially assessed through descriptive statistics and univariable models. The assumption of linearity for age, year of urolith submission, distance between owner and clinic location, and CSD median family income was tested by use of 3 techniques: adding a squared term, categorizing the continuous variable to determine whether the coefficients increased in a uniform manner, and plotting the logarithmic odds of the outcome against the mean of categories of the variable divided into deciles. If the independent variable's relationship with the logarithmic odds of the outcome was not linear, the variable was categorized or, if there was a quadratic relationship, the independent variable's square term was added to the model. When the correlation between 2 variables was ≥ 0.7, only 1 of the variables was chosen for subsequent analyses on the basis of completeness of the data, biological plausibility, or both. Univariable analyses by use of multilevel logistic regression with owner's CSD as the random intercept were performed. The 2-way interactions among demographic variables and diet variables were tested by use of a multilevel model that included the 2 main effects and their interaction term. The initial multivariable multilevel model was built with the significant individual dog-level main effects and interaction terms. Subsequent models were built in a forward manner with the addition of community-level contextual variables. Explanatory variables and interactions not initially included were retested in the full model and retained in the model when they were significant or acted as a confounding variable. A confounding variable was defined as a nonintervening variable that changed the coefficient of a significant variable in the logarithmic odds scale by ≥ 20%.^24,25 Odds ratios were expressed in terms of the odds for developing CaOx urolithiasis, compared with the odds for developing struvite urolithiasis.

The intraclass correlation coefficient (ie, the correlation between 2 observations within a cluster) was calculated from the variance component for the owner's CSD in the multilevel logistic model by applying the latent variable approach (ie, fixing the error variance at π²/3) to calculate an intraclass correlation coefficient.^24,25

Results

Descriptive statistics—Dogs in Ontario accounted for 11,414 of 26,239 (43.5%) of all uroliths submitted to the Canadian Veterinary Urolith Centre from 1998 to 2006. Calcium oxalate and struvite uroliths comprised 10,478 (91.8%) urolith submissions from dogs in Ontario during the 9-year study period. Of these, 9,764 (93.2%) were successfully geocoded. There were 7,297 uroliths (3,867 [53.0%] CaOx uroliths and 3,430 [47.0%] struvite uroliths) with full covariate information (Table 1). Mean age of dogs for CaOx urolith submissions was 8.2 years, whereas the mean age of dogs for struvite uroliths was 5.7 years. Of the 7,297 dogs from which uroliths were submitted, 6,932 (95.0%) were between 0.1 and 12 years of age. Small breeds were the most commonly represented breed category for both urolith types, but CaOx uroliths comprised most of the submissions for small-breed dogs. The CaOx submissions were predominantly from patients that were male, neutered, obese, or a small-breed dog. Among dogs fed a urinary diet for a minimum of 6 months, there were more CaOx urolith submissions than struvite submissions. Most dogs were fed a dry diet; the fewest dogs were fed strictly a canned diet, and only 29 medium- or large-breed dogs were fed strictly a canned diet.

Households in the lowest community median family income group (< $54,000 [Canadian dollars]) had a greater proportion of struvite submissions, whereas all higher income groups (> $54,000) had a greater proportion of CaOx urolith submissions (Table 2). Households within the previously identified clusters of CaOx and struvite uroliths had a higher proportion of CaOx and struvite urolith submissions, respectively. The annual number of submissions peaked at 1,122 in 2004, and the highest proportion of CaOx urolith submissions was in 2006. Median distance between an owner and the submitting clinic was 3.9 km, and for 5,473 (75%) owners, it was < 10.3 km to the clinic where the urolith was removed from their dog. Households with the greatest metropolitan influence (ie, those from within a CMA) accounted for the greatest proportion of CaOx urolith submissions.

Linearity and correlations—None of the continuous explanatory variables, including age, year of submission, CSD median family income, or distance between owner and clinic location, had a linear relationship with the outcome, as determined on the basis of plots of logarithmic odds. Transformations were made to address nonlinearity of the continuous explanatory variables. A significant squared term was included to determine the curvilinear relationship between age and outcome. Year was categorized on an annual basis. The CSD median family income was categorized by quartiles (< $54,000, $54,000 to $59,000, $59,000 to $70,000, and > $70,000 [Canadian dollars]), and distance was categorized by quartiles (< 1.6, 1.6 to 3.9, 3.9 to 10.3, and > 10.3 km). None of the correlations among the explanatory variables exceeded 0.7.

Individual dog-level risk factors and 2-way interactions—In the final multilevel model, the main effects included the following significant individual dog-level variables: age, age², sex, body condition, veterinary-exclusive diet, and urinary diet (Table 3). Age, sex, neuter status (sexually intact or neutered), breed (mixed-breed dog, small-breed dog, or medium- or large-breed dog), body condition (thin or normal vs obese), and dietary moisture content (dry, canned, or both) were all part of significant 2-way interactions with other individual dog-level variables. These interactions comprised sex with age (LR χ² = 32.89 [P < 0.001]), sex with neuter status (LR χ² = 17.21 [P < 0.001]), body condition with age (LR χ² = 12.49 [P < 0.001]), and dietary moisture content with breed (LR χ² = 19.71 [P = 0.003]).

Individual dog-level interaction terms that were not significant were breed with age (LR χ² = 3.82 [P = 0.15]), breed with sex (LR χ² = 3.12 [P = 0.21]), sex with body condition (LR χ² = 0.02 [P = 0.90]), breed with age (LR χ² = 3.82 [P = 0.15]), breed with neuter status (LR χ² = 4.20 [P = 0.12]), breed with body condition (LR χ² = 3.58 [P < 0.17]), neuter status with body condition (LR χ² = 0.61 [P = 0.44]), neuter status with age (LR χ² = 1.47 [P = 0.23]), veterinary-exclusive diet with age (LR χ² = 0.29 [P = 0.59]), veterinary-exclusive diet with sex (LR χ² = 0.91 [P = 0.34]), veterinary-exclusive diet with breed (LR χ² = 3.91 [P = 0.14]), veterinary-exclusive diet with neuter status (LR χ² = 0.06 [P = 0.81]), veterinary-exclusive diet with body condition (LR χ² = 0.78 [P = 0.38]), veterinary-exclusive diet with dietary moisture content (LR χ² = 4.15 [P = 0.39]), veterinary-exclusive diet with urinary diet (LR χ² = 0.41 [P = 0.52]), urinary diet with age (LR χ² = 3.05 [P = 0.08]), urinary diet with sex (LR χ² = 0.20 [P = 0.65]), urinary diet with breed (LR χ² = 1.34 [P = 0.51]), urinary diet with neuter status (LR χ² = 0.55 [P = 0.46]), urinary diet with body condition (LR χ² = 0.02 [P = 0.90]), urinary diet with dietary moisture content (LR χ² = 1.05 [P = 0.59]), dietary moisture content with age (LR χ² = 4.40 [P = 0.36]), dietary moisture content with sex (LR χ² = 4.60 [P = 0.33]), dietary moisture content with neuter status (LR χ² = 1.28 [P = 0.53]), and dietary moisture content with body condition (LR χ² = 5.78 [P = 0.06]).

The odds for a submission of a CaOx urolith, compared with the odds for submission of a struvite urolith, were increased by the following factors: increasing age; being a male; being a mixed-breed or small-breed dog, compared with being a medium- or large-breed dog; having an obese body condition, compared with having a thin or normal body condition; being fed a veterinary-exclusive diet for a minimum of 6 months, compared with being fed a veterinary-exclusive diet for < 6 months or for an unknown duration or being fed a commercial, homemade, or unknown diet; and dietary moisture content (canned vs dry; Table 3). The odds for submission of a CaOx urolith, compared with the odds for submission of a struvite urolith, were decreased for a urinary diet fed for a minimum of 6 months, compared with a nonurinary diet or with a urinary diet fed for < 6 months. In the multivariable model, the odds for submission of a CaOx urolith increased with age, but this effect was far greater in males than in females (Figure 1). In males, the odds were higher at a younger age and increased at a more rapid rate with each successive age increment up to a plateau between 9 and 10 years of age. Sex also interacted with neuter status; neutering increased the odds of developing CaOx urolithiasis, but more so in males. Obesity increased the odds of developing CaOx urolithiasis, but more so for younger dogs.

View Full Size

Predicted values for the probability of a CaOx urolith submission, compared with the probability of a struvite urolith submission, among dogs in Ontario, Canada, from 1998 to 2006 on the basis of age, sex, neuter status (neutered males [squares], sexually intact males [circles], spayed females [triangles], and sexually intact females [diamonds]), and body condition. For each variable, all other variables were held constant at their referent values.

Citation: American Journal of Veterinary Research 71, 9; 10.2460/ajvr.71.9.1045

• Download Figure
• Download figure as PowerPoint slide

Dietary moisture content (canned vs dry) interacted with breed. In the small-breed category, eating a canned diet, compared with eating a dry diet, had a significant sparing effect on the submission of CaOx uroliths; however, this effect was not observed for the mixed-breed category. Feeding both canned and dry diets, compared with feeding only a dry diet, did not have a significant effect for either the small- or mixed-breed categories. Feeding a canned diet, compared with feeding only a dry diet or with feeding both canned and dry diets, resulted in a significant increase in the odds for submission of a CaOx urolith in medium- or largebreed dogs.

Community-level contextual risk factors—In the final multivariable model, community-level main effects that had a significant association with outcome for submission of a CaOx urolith versus submission of a struvite urolith were year of submission, cluster, and CSD median family income (Table 4). Compared with 1998, only 2006 had a significant increase in the odds for submission of a CaOx urolith versus submission of a struvite urolith. The odds for submission of a CaOx urolith were greater in dogs whose owners resided in the CaOx spatial cluster, compared with the odds for those whose owners resided outside a cluster or that resided in the struvite spatial cluster. The risk for submission of a CaOx urolith from dogs whose owners resided within the struvite spatial cluster was less than that for dogs whose owners resided in neither spatial cluster. In the multivariable model, median family income quartiles for the highest income group (ie, > $70,000 [Canadian dollars]), compared with the lowest CSD median family income group, was associated with a significant increase in the odds for submission of a CaOx urolith. There was no significant association between outcome and SAC type (LR χ² = 3.34 [P = 0.65]) or between outcome and distance of owner from the submitting clinic (LR χ² = 0.06 [P = 0.99]).

The intraclass correlation coefficient in the intercept-only model, full model excluding the cluster variable, and full model that included the cluster variable was 2.95%, 1.38%, and 0.75%, respectively. The final model was not affected by the removal of extreme Pearson residuals, and extreme residuals did not appear to be a result of errors in data entry.

Discussion

Results for the study reported here reflected findings in other studies^3–9 in that female and younger dogs were at greater risk of struvite urolith submissions, whereas male and older dogs were at greater risk of CaOx uroliths submissions. However, we found complex interactions among the demographic variables. The interaction between age and sex resulted in the effect of age differing between the sex categories. In other studies^9,10 conducted to investigate the interaction of age and sex, variables were categorized into groups. In our study, by treating age as a continuous variable, we had more power to evaluate the complex interactions between age and the other variables, including sex and body condition. Investigators in a descriptive study¹ reported an interaction between age group and sex that resulted in the median age for dogs with CaOx uroliths being older than that for dogs with struvite uroliths, but this effect was more marked in males. We found that the effect of age began earlier and peaked when dogs were 9 to 10 years old in males.

Although obesity has been recognized as a factor contributing to CaOx urolithiasis,⁹ the nature of interactions between obesity and other factors has not been clear. In our study, the interaction between age and body condition resulted in the effect of age differing between body conditions (obesity increased the odds of developing CaOx urolithiasis in younger males, compared with the odds in older males).

The study reported here also revealed an interaction between sex and neuter status, similar to that observed in another study²⁶ of dogs that formed uroliths. However, males in our study were more affected by neutering than were females, whereas investigators in that other study²⁶ found that females, rather than males, appeared to be more affected by neutering with regard to the risk of developing struvite urolithiasis and CaOx urolithiasis, as determined on the basis of univariable analyses to compare dogs with a particular urolith type with those that had any other urolith type.

In another study²⁷ in which investigators evaluated urine variables in healthy Miniature Schnauzers (n = 8 dogs) and healthy Labrador Retrievers (8), the urine frequency and urine volume per kilogram of body weight of the Miniature Schnauzers were both reduced, compared with results for the Labrador Retrievers. The findings in that study²⁷ may reflect general differences in urine variables between small- and large-breed dogs and could partly explain the overrepresentation of small-breed dogs among those that form struvite and CaOx uroliths. We detected significantly increased odds for submission of CaOx uroliths, compared with the odds for submission of struvite uroliths, in small-breed dogs, compared with the odds for mixed-breed and medium-or large-breed dogs; however, this may not necessarily reflect the risk in all small and toy breeds. The breed categories in the present study can only crudely represent risk differences between the 2 urolith types. Some medium and large purebred dogs, such as Standard Schnauzers, may have an inherent risk for developing CaOx urolithiasis, as determined in 1 study²⁶ in which investigators evaluated urolith risk within breed while adjusting for breed popularity.

Analysis of results of our study suggested that feeding only canned diets to small-breed dogs decreased the risk for submission of CaOx uroliths, compared with the risk for submission of struvite uroliths. This protective effect of canned food did not extend to feeding regimens that included both canned and dry diets. In contrast to small-breed dogs, high dietary moisture content (canned vs dry or canned vs both canned and dry) was a significant risk factor for development of CaOx uroliths in medium- and large-breed dogs. This effect of dietary moisture content in medium- and large-breed dogs fed exclusively a diet of canned food was based on a small sample size (n = 29 dogs); consequently, the impact of dietary moisture content in these breeds requires further investigation. Our results for small-breed dogs are consistent with results from the challenge trial conducted by other investigators who reported²⁸ that Miniature Schnauzers, a small-breed dog, had a reduction in CaOx relative supersaturation (ie, a measure of the potential for forming CaOx uroliths) when fed a diet with a high moisture content, whereas this effect was absent in Labrador Retrievers, a large-breed dog.²⁸

In the study reported here, we found that dogs fed a veterinary-exclusive diet for > 6 months had increased odds for submission of a CaOx urolith, compared with the odds for submission of a struvite urolith. Other investigators⁸ have alluded to the dilemma in formulating diets that control risk factors for development of both CaOx and struvite uroliths. However, it is important to mention that because we compared the relative odds for the submission of 1 urolith type with that of another, our study cannot estimate the absolute risk of developing either urolith type. It is possible that by comparing cases of both CaOx and struvite urolithiasis to a true control group we might find different patterns concerning the risk of urolith formation associated with different diet types. In addition, we have no information as to why these dogs were fed these veterinary-exclusive diets, so we cannot control for predisposing conditions in our analyses. However, it is worth considering further epidemiological investigations into the components of these diets because formulation of diets that result in a reduction of urine pH, phosphorus concentrations, and magnesium concentrations to control the risk of developing struvite uroliths may predispose dogs to formation of CaOx uroliths.^7,8 In addition, 2 challenge studies^28,29 have revealed that increased dietary sodium content can reduce the CaOx relative supersaturation. Veterinary-exclusive diets and nonveterinary diets may differ in their sodium, magnesium, and phosphorus content and hence their effect on urinary pH. In addition, if veterinary-exclusive diets contain more meat protein, compared with that in nonveterinary diets, they may differ with regard to the risk for development of CaOx urolithiasis.

The first or major diet listed for each dog was considered to be the main diet and was used for analysis in the present study. However, the diet variable lacked specificity for dogs that were fed > 1 diet, and there was inadequate information from the current registry to assess the composition of these diets. Increased efforts to record specific dietary information and the components of these diets should be considered to improve the epidemiological quality of these data for future studies. We found that dogs fed urinary diets for a minimum of 6 months had lower odds for submission of a CaOx urolith, compared with the odds for submission of a struvite urolith. However, our data lacked information concerning the reasons these dogs were being fed urinary diets, which makes it difficult to interpret this finding. It is possible that some dogs may have been incorrectly classified as incident cases rather than recurrent cases, and overinterpretation of these results should be avoided.

Dogs of owners residing in CSDs with the highest community income, as measured by CSD median family income, had an increased risk for submission of a CaOx urolith, compared with the risk for submission of a struvite urolith, even after controlling for a previously identified spatial cluster. Community income level can reflect the influence of personal income or the access of a community member to treatment options or diet types. Dogs of owners residing in the high-density-population area of Toronto had an increased risk for the submission of CaOx uroliths, compared with the risk for dogs whose owners resided in the struvite cluster or outside either cluster, even after controlling for community income. These owners may have preferentially chosen to treat their dogs medically rather than surgically for suspected struvite uroliths. Similarly, owner location and community income levels may influence the medical management given to dogs with urinary tract infections, thereby reducing their risk for developing struvite uroliths. To further investigate the effect of these contextual variables, we examined 2 variables: distance between owner and submitting clinic location (as a surrogate for convenient access to preferred medi- cal care) and owners’ CSD-SAC type (as a measure of the degree of urban influence on the owner's location); however, neither variable had significant effects in our model. Additional case-control studies that also include data from patients medically treated at community clinics would be needed to fully evaluate the impact of contextual factors on submission patterns.

We detected a significant difference in outcome in 2006, compared with outcome in 1998, but there was no evidence of a linear or consistent increase in the odds for submission of a CaOx urolith in the intervening years. In contrast, other investigators have reported^2,10 a linear pattern toward an increase in the incidence of CaOx uroliths and a decrease in the incidence of struvite uroliths. However, our model included covariates whose distribution could change over time, and this may have accounted for the fact that the findings of our study differed from the results of the 2 aforementioned studies.^2,10 Reporting of dogs with presumptive struvite uroliths that receive medical treatment may provide additional insight regarding the incidence of struvite urolithiasis and submission patterns.

Use of multilevel data allow for an estimation of the relative importance of variations in risk at the various levels of organization.³⁰ In our intercept-only model, with owner CSD as the random intercept, the proportion of the total variation that could be attributed to variation among the owner CSD groups was small. This decreased even further with the inclusion of the individual-dog- and community-level fixed effects. In our model, the community-level variables CSD median family income and cluster had significant effects and indicated areas of interest for future studies; however, the greatest amount of variation and thus the greatest potential for intervention was detected at the individual-dog level.

Although registry data are commonly used to understand the epidemiological aspects of urolithiasis and have been useful in identifying important risk factors for urolith formation,^5,6,10,26 it is important to mention some inherent limitations of these data. Comparisons between 2 urolith types do not provide a true control group, so the distribution of exposure among actual control dogs in the population is unlikely to be represented by those associated with another urolith type.^24,25 In addition, admission bias is an inherent part of working with any secondary database, and this is especially true when struvite uroliths are being examined. The decision to treat dogs with struvite uroliths medically or surgically, which determines whether these animals will enter the registry, can be related to a number of potential factors, including choice of diet, income-related issues, and the nature of veterinary practice in various regions. Consequently, the findings in any registry-based study of uroliths should be used as a foundation for designing primary-based studies that evaluate the nature of urolith formation.