Abstract
Objectives
The aim of the study was to identify any dietary, signalment, geographic and clinical factors associated with hematuric struvite crystalluria (HSC) in a population of cats that visit general care veterinary hospitals in the USA.
Methods
In total, 4032 cats that had a first-time diagnosis of HSC and 8064 control cats with no history of hematuria or crystalluria were identified from medical records of all cats examined between 2007 and 2011 at 790 US veterinary hospitals. Extracted variables included age, sex, neuter status, breed, diet, urinalysis results and history of cystitis. Potential associations between these variables and HSC were estimated.
Results
Controlling for other factors, young cats fed a dry diet had an increased likelihood of HSC relative to young cats fed a non-dry diet. However, as age increased, the likelihood of HSC declined for cats fed a dry diet and increased for cats fed a non-dry diet. Moreover, the odds of HSC were significantly greater when cats were unneutered (vs neutered; odds ratio [OR] 45.52) or had a thin (vs heavy) body condition (OR 23.81), diagnosis of cystitis (OR 2.84), urine protein concentration >30 mg/dl (OR 4.72), alkaline (vs neutral) urine pH (OR 3.34), pyuria (OR 23.67) or bacteriuria (OR 2.24).
Conclusions and relevance
The present study provides estimates of the strengths of association between HSC and certain signalment and clinical characteristics of cats. This information could help clinicians to perform a more directed screening for struvite crystalluria in certain cat populations. Follow-up studies that build on the findings of this study could explore the clinical importance of HSC in cats.
Introduction
Crystalluria, defined as the presence of crystals in urine, results from an excessive super saturation of certain electrolytes. Although not a specific marker of a pathologic condition in cats, crystalluria is the best marker for predicting urolith reoccurrence in humans. 1 Specifically, struvite crystals (magnesium ammonium phosphate hexahydrate) are often found in the urine of cats with lower urinary tract diseases. In the past decade, struvite has been identified as the most common mineral in feline urethral plugs. 2 For over three decades, struvite urolithiasis has been recognized as a clinically important lower urinary tract condition in the US cat population. 3 The consequences of these conditions can vary from irritation to total obstruction of the urinary tract. Total obstruction can yield a fatal outcome if not managed promptly. Invariably, urolithiasis causes some degree of suffering for affected cats and reduces their quality of life. However, limited information is available on the possible association between struvite crystalluria and the development of feline lower urinary tract diseases such as struvite urolithiasis in cats.
Although occasional detection of struvite crystalluria in clinically healthy cats is not considered to be clinically important, more frequent detection of these crystals may have prognostic, diagnostic or therapeutic importance.4,5 Studies concerning clinical and signalment factors associated with the development of struvite crystalluria in cats have been infrequently reported. However, specially formulated diets have been shown to reduce the severity of struvite crystalluria in cats. 6 Based on these study results, it could be inferred that diet could also contribute to the formation of crystalluria in cats. Furthermore, hematuria is a non-specific pathologic condition in cats that could be a consequence of various lower urinary tract disorders. Specifically, nearly 30% of cats with struvite crystalluria experience hematuria due to the irritation of the bladder walls by the crystals. 6 Therefore, an evaluation of the presence of struvite crystals in the urine of cats with hematuria could provide useful information in our understanding of these common condition in cats.
The objectives of our retrospective case-control study were to identify any dietary, signalment, geographic and clinical factors associated with hematuric struvite crystalluria (HSC) in a population of cats that visit general care veterinary hospitals in the USA. In the wake of the pet food industry’s unclear messaging about the significance of crystalluria and the use of specialized pet foods that are meant to reduce the severity of crystalluria, the findings of this study are expected to contribute to the current body of knowledge on HSC in cats. The outcome could help clinicians to perform a more directed screening for struvite crystalluria in certain cat populations.
Materials and methods
Study population
All included cats came from a population of cats routinely (at least one visit before the study period) examined from 1 October 2007 to 31 December 2011 at 790 general care veterinary hospitals located in 43 states of the USA. The seven states without hospitals were Alaska, Hawaii, Maine, North Dakota, Vermont, West Virginia and Wyoming. All hospitals, operated and managed by Banfield Pet Hospital, utilized a uniform proprietary health data entry software program (PetWare) and electronic health records. These records were uploaded daily to a central data warehouse (Oracle Enterprise Software, Oracle Database 11g Enterprise Edition Release 11.2.0.2.0 – 64bit Production, Redwood City, CA, USA) for storage. Veterinary personnel recorded information on each patient’s signalment, diets, urinary results and diagnosis. Information was recorded by selecting from a list of options provided in drop-down menus. In a subset of the records, medical notes containing detailed descriptions of the composition of urine crystals were entered as free text as part of the electronic record.
Criteria for selection of cases and controls
The medical records of cats in this study population were evaluated to identify case and control cats. First, the entire medical record database was searched to identify eligible cats by means of a proprietary programming language (SQL and PL/SQL, Toad for Oracle Base Version 10.6.1.3, Quest Software). Among the eligible cats, a medical record code for the diagnosis of crystalluria was used to select only those with crystalluria for consideration of inclusion in the case group. Because factors that are associated with the first episode of crystalluria may differ from those of the recurrent episodes, cats at the visit at which crystalluria was first identified were selected. Hence, cats were automatically excluded electronically from selection if they had a prior diagnosis of either hematuria (defined as ⩾1+ red blood cells [RBCs] identified on urine dipstick testing), crystalluria or chronic kidney disease at any time in the past, lacked a hospital visit before the diagnosis of hematuria or crystalluria (cats with only one hospital visit record), or had a record of any prescription diet marketed to promote urinary health prior to diagnosis of either hematuria or crystalluria. Records of subsequent diagnosis of hematuria or crystalluria within the study period were deleted because cats could only be included as a case once during the study. From the selected records, case cats were isolated. These were cats with a record of hematuria and crystalluria, a condition heretofore referred to as HSC. The inclusion date referred to the date of the visit at which cats were first identified as having HSC within the study period.
Control cats were randomly identified and selected from the electronic records of all remaining cats. A rate-based selection method was used to ensure that the exposure distributions of the cats selected as controls matched the exposure distributions of cats in the source population because cats could enter or leave the study population at any time during the risk period. 7 In this selection process, control cats were randomly selected from a list of all cats that visited the hospitals within the study period and were not already identified as cases; the probability of a cat being selected was proportional to the frequency of its hospital visits. In an effort to control for detection bias, control cats were required to have records of a urinalysis (unrelated to urolithiasis) performed on the visit at which cats were selected for inclusion (ie, inclusion date). These urinalyses had been performed primarily as part of annual comprehensive examinations for cats enrolled in a wellness plan or, less commonly, because of a suspected urinary tract infection. Other search criteria for selection included no record of crystalluria, hematuria, chronic kidney disease or prescription diet for urinary health recorded at any previous time. Control cats were selected at a ratio of two controls to every one case, to increase the statistical power of the study.
Medical records review
Variables selected for evaluation included age, sex, reproductive status, breed, body weight, body condition, primary type of food or treats fed (when recorded, at the inclusion date or the closest visit to that date), number of dogs and cats in the household, inclusion date (quarter of the year, year), location of the hospital (hospital number, zip code, state and region [north-east, north-central, north-west, south-east, south-central and south-west]), and urinalysis results and history of certain disease conditions (eg, diabetes mellitus) when applicable.
The primary type of food or treats consumed was initially categorized based on the moisture content (dry, semi-dry and moist canned food) and was subsequently categorized as dry or non-dry. The dry diet category was reserved for cats that were fed dry food or treats exclusively, whereas the non-dry diet category was reserved for cats that were fed moist canned food and treats alone or in combination with semi-dry or dry food and treats. The number of cats and dogs in the household was recorded because of the possibility that a cat may have routine access to another pet’s food.
The five most prevalent cat breeds were domestic longhair (DLH), domestic medium hair (DMH), domestic shorthair (DSH), Siamese and Persian, whereas the remaining breeds were categorized as ‘other’. Because no difference was observed between of the domestic breeds (DSH, DMH and DLH) in the prevalence of HSC, these breeds were further grouped as ‘mixed’. The breed categories utilized in the assessment of statistical associations were mixed, Siamese, Persian and other. Body condition on the day of inclusion or at the most recent visit ⩽30 days prior to the inclusion date was recorded as thin, ideal or heavy. A diagrammatic 5-point scale for body condition scoring (1 = very thin, 2 = thin, 3 = ideal weight, 4 = overweight, and 5 = markedly obese) was introduced to the hospital staff and added to the electronic medical data recording system in June 2010. Before that time, body condition was recorded as thin, normal or heavy.
Diabetes mellitus and cystitis were recorded as present or absent. These conditions were identified solely on the basis of a diagnosis in the medical record obtained from the most recent visit ⩽1 year before the inclusion date. If no diagnosis was made, these conditions were presumed to be absent. For all other variables that were not captured as present or absent, missing values were not altered during data analysis.
Urinalysis
Urinalysis results were obtained for the most recent visit ⩽30 days before the inclusion date. In the hospitals, urine samples were routinely collected via cystocentesis and complete urinalysis was performed by use of commercially available dipsticks with manual interpretation, microscopic examination of centrifuged urine sediment, refractometry and an automated analyzer (IDEXX VetTest 8008, Chemistry Analyzer; IDEXX Laboratories). Urinalysis were routinely performed in-house and no standardized time frame existed between collection and urinalysis. Resulting data included hematuria (as defined in the case inclusion criteria [⩾1+ RBCs identified on urine dipstick testing]; present vs absent), pH, protein concentration, specific gravity, ketone concentration, casts (present vs absent), glucose (present vs absent), pyuria (defined as ⩾1+ white blood cells identified on urine dipstick testing; present vs absent) and bacteriuria (present vs absent). A urine pH of <7.0 was recorded as acidic, 7.0–7.5 as neutral and >7.5 as alkaline. A urine protein concentration of ⩽30 mg/dl was considered unremarkable and a concentration of >30 mg/dl was considered abnormal. A urine ketone concentration of <5 mg/dl was considered unremarkable and a concentration ⩾5 mg/dl was considered abnormal. Urine crystal composition was inferred by microscopic examination of urine sediment.
Statistical analysis
The data were analyzed using statistical software (SAS version 9.4). Data regarding the measured explanatory variables (putative risk factors) and HSC were summarized as descriptive statistics (frequencies, proportions, means, SDs, medians and quartiles). Associations between the putative risk factors and HSC were first tested by means of univariable logistic regression (PROC LOGISTIC, SAS version 9.4). Strengths of association was measured by estimated ORs and their 95% confidence intervals. A P value <0.05 was considered significant. In preparing for fitting of a multivariable logistic regression model, the assumption of linearity of age and body weight was tested by examining the significance of a squared term. 8 If a quadratic relationship was identified, then the squared term for the independent variable was considered for addition to the model. Because body weight and body condition represented similar characteristics, possible multicollinearity was assessed using variance inflation factor (VIF) to avoid modeling issues associated with multicollinearity. A VIF value >10 was suggestive of presence of multicollinearity.
Multivariable model building was performed in steps. First, we pre-screened variables to consider for inclusion in the multivariable model based on a significance level of 15% (ie, P = 0.15). Next, we performed both forward and backward automated stepwise selection procedures to identify putative risk factors for fitting a mixed multivariable model (PROC GLIMMIX, SAS version 9.4). Variables that were excluded from the automated selection process were manually included in the model to evaluate if they improved the model fit. The final model was fitted using an adaptive Gauss–Hermite quadrature likelihood approximation and possible effects of confounding were evaluated. Selection of the final random effects and fixed-effects model was performed based on the Bayesian information criteria value. The final multivariable model included only those variables with a P value <0.05, which acted as a confounding variable, or were part of a significant interaction term. A variable that changed the coefficient of a previously significant variable by ⩾20% was treated as a confounder and was retained in the model if it was not an intervening variable.9,10
Afterwards, effects of interactions selected based on biological plausibility were tested. These interactions included those between diet and sex, diet and body condition, diet and age, diet and pH, diet and proteinuria, diet and urine specific gravity, sex and urine pH, sex and age, sex and neuter status, and neuter status and age.
Finally, random intercepts representing hospital location at various hierarchical levels (ie, hospital identity, zip code, state and region) were assessed in the model, starting with the hospital number and later nesting within zip code, state or region. Random effects were dropped from the model if they did not improve the fit of the model or if the model failed to converge.
A statement was included in the multivariable model command to obtain standard multiple pairwise comparisons. The best linear unbiased predictors from the final model were assessed visually using a normal quantile plot and by graphing them against the predicted outcome to make certain they met the assumptions of normality and homogeneity of variance, respectively. Deviance residuals were also plotted against the predicted outcome to identify potential outliers.
Results
Descriptive statistics
A total of 13,237 records of cats with crystalluria were identified between 1 October 2007 and 31 December 2011. Of these, 7621 (57.6%) cats had crystals inferred to be struvite, 800 (6.0%) had crystals inferred to be calcium oxalate, 63 (0.5%) had crystals inferred to be cystine, and the remainder had crystals inferred to be of mixed content (struvite and calcium oxalate) or other less prevalent compositions such as calcium or urate. Of the 7621 cats with struvite crystalluria, 4032 (52.9%) had hematuria and therefore met the case definition. A total of 8064 control cats were included in the study.
Included cats originated from all 790 veterinary hospitals that were operational during the study period. The case cats originated from 717 hospitals, averaging approximately six cats per hospital (range 1–56 cats per hospital), and the 8064 control cats originated from 768 hospitals, averaging approximately 11 cats/hospital (range 1–77 cats per hospital). Overall, 22 hospitals contributed only case cats, 73 hospitals contributed only control cats and 695 hospitals contributed both case and control cats.
Information regarding diet was available for only 1394 (34.6%) case cats and 3220 (39.9%) control cats. Information regarding body condition was available for only 1320 (32.7%) case cats and 1622 (20.1%) control cats. Less than 1.5% of case and control cats had a diagnosis of diabetes mellitus within 1 year prior to the study inclusion date. With respect to urinalysis results, data on bacterial culture and glucose test results were available for ⩽4.0% of case and control cats, so these variables were excluded from further analyses.
Univariable statistics
Univariable comparisons between the case and control groups identified several variables associated with HSC in cats (Table 1). These variables included age, body weight, number of cats and dogs in the household, sex, neuter status, breed, body condition, diagnosis of cystitis and diabetes mellitus. Other variables related to urinalysis were specific gravity, protein concentration (>30 mg/dl vs ⩽30 mg/dl), ketone concentration (⩾5 mg/dl vs <5 mg/dl), pH (alkaline, acidic vs neutral), and the presence of pyuria, bacteriuria or casts. The odds of HSC changed significantly with both age and body weight in a curvilinear manner. Male cats, unneutered, mixed breed or those fed a dry diet were more likely to have HSC than were females, neutered cats, a breed of either Siamese, Persian, or those classified as ’other’, or those cats fed a non-dry diet. Cats with a previous diagnosis of cystitis were more likely to have HSC. However, those with a previous diagnosis of diabetes mellitus were less likely to have HSC. No association was identified between HSC and the specific gravity of urine or the time (quarter of the year) of the visit at which HSC was diagnosed.
Summary data and results of univariable logistic regression for cats with hematuric struvite crystalluria and control cats evaluated at general care veterinary practices in the USA, 2007–2011
Distribution values reported for case and control cats are (%) with factor for categorical variables and median (interquartile range) for quantitative variables. The case group consisted of cats with urinalysis results indicative of hematuric struvite crystalluria at the point these results were recorded for the first time at the hospitals at which they were evaluated. The control group consisted of cats with no history of crystalluria or hematuria that had a urinalysis on record during the study period. In the body condition classification, thin represented very thin and thin cats, ideal was for those within an ideal weight for the cat’s breed and age, while heavy represented cats that were either overweight or markedly obese
Parameter estimates
OR = odds ratio; CI = confidence interval; NA = not applicable
Multivariable statistics
The final model was fitted from data for 1025 cats (411 cases and 614 controls). Variables significantly associated with HSC – controlling for other variables – were age, diet, neuter status, body condition, urine protein concentration, urine pH, presence of pyuria or bacteriuria, or a previous diagnosis of cystitis (Tables 2 and 3). Young cats fed a dry diet were at increased odds of HSC relative to cats fed a non-dry diet. However, with increasing age, the odds of HSC declined with cats fed a dry diet and increased for cats fed a non-dry diet (Figure 1). Also, the odds of HSC were significantly greater in unneutered cats, thin-bodied cats, those with a previous diagnosis of cystitis, or cats with pyuria, bacteriuria, alkaline pH or a high protein concentration in their urine (Tables 2 and 3). Of all location hierarchies, only the unique hospital number was a significant random intercept (variance estimate = 1.1864, P = 0.012) tested in the model. The best linear unbiased predictors met the assumptions of normality and homogeneity of variance, and no outliers were identified. Based on VIF of 1, no multicollinearity was observed, but body weight was not significant in the final fitted model.
Coefficients from multivariable analysis for factors associated with hematuric struvite crystalluria in cats in the USA, 2007–2011
Calculated as log of the odds ratio of hematuric struvite crystalluria for dry (vs non-dry) diets averaged over years of cats. This model was estimated from a total of 1025 cats, of which 411 were case
cats CI = confidence interval
Results of multivariable analysis of factors associated with hematuric struvite crystalluria in cats in the USA, 2007–2011
The intercept estimate for the model was −2.49 (−4.27 to −0.70), and the variance estimate of hospital was 1.1864 (P = 0.01). The final model was estimated from a total of 1025 cats, of which 411 were case cats. Ages shown for diet represent the 10th, 25th, 50th, 75th, and 90th quantiles, respectively
OR = odds ratio; CI = confidence interval
Probability of hematuric struvite crystalluria among cats in the USA between 2007 and 2011 on the basis of age and diet after controlling for all other variables in the model. The solid and dashed lines represent dry and non-dry diets, respectively
Discussion
The study reported here describes for the first time signalment factors and other clinical characteristics associated with HSC in cats evaluated at general care veterinary hospitals. The odds of HSC in the present study was greater in male cats at the univariable analysis but not in the final multivariable analysis. Similarly, the quadratic effect of age (age 2 ) was significant in the univariable model but not in the multivariable model after other variables were controlled for. This disparity was potentially caused by the reduced number of cats included in the multivariable model or because a large number of variables were evaluated, which could have introduced a type II error. As an example, variables such as diet and body condition had fewer observations and their presence in the final model reduced the number of observations utilized in the final analysis.
Relative to older cats fed a non-dry diet, young cats (⩽4 years of age) fed a dry diet were at increased odds of HSC, but this effect of diet reversed as cats got older. Moist foods are known to reduce urine concentration of urolith-forming metabolites. 6 The observed reversal of the effect of moisture content of diet is unclear. Possibly, dry diets could be a predisposing factor for HSC up to a stage in life. Unneutered cats (regardless of sex) were at significantly greater (and not lower) odds for HSC than were neutered cats in the present study. In dogs, struvite uroliths are significantly more likely to be recovered from spayed vs unneutered female dogs,11,12 and hormonal changes associated with spaying have been hypothesized to underlie this finding. 11 A potential role of hormones in crystal development in cats remains unclear, because sex was not significant in the final model.
Alkaline urine presents a favorable environment for the formation of struvite crystals. Hence, the calculolytic diets for treatment or prevention of struvite crystalluria or urolithiasis are geared towards acidifying the urine content of the animals on the diet.13–15 Such lowering of the urine pH has been effective in dissolving sterile struvite crystals and urolithiasis.13–15
The observed association between urine protein concentration and HSC in the present study should be interpreted cautiously. The presence of hematuria was defined as a urine dipstick test result of ⩾1+ RBCs, the presence of crystalluria on the basis of microscopic detection of crystals in urine sediment, and the presence of proteinuria on the basis of results of an automated analyzer (>30 mg/dl). The degree of hematuria or crystalluria was not considered, nor was the degree of proteinuria, and it is possible that had different criteria or cut-offs been used, the case group or results would have been different.
In addition, no inferences should be made about causal associations among these variables given the retrospective study design. In one study, the presence of soluble proteins in feline urine was reported to act as a promoter of struvite crystallization, leading the authors to conclude that reducing urinary protein excretion would be helpful in decreasing struvite crystallization and urolithiasis formation. 16 In another study, proteinuria was shown to be a consequence of hematuria and pyuria. 17 Because, lower urinary tract inflammation (eg, such as that caused by crystalluria) can result in both hematuria and proteinuria, it is difficult to identify the actual cause of the observed proteinuria in the present study.
No urine protein:creatinine ratios were available for the cats in the present study, and this variable would have been useful for assessing the potential cause of the observed proteinuria. Similarly, we could not conclude whether the presence of pyuria and cystitis were risk factors for, or consequences of, HSC because of the nature of the data. Given how the cystitis condition was captured, it is more likely that the cystitis could have occurred before the HSC. Establishing such temporal relationship could be explored by conducting a cohort study in which the cases and controls in the present study were considered the exposed and the unexposed groups, respectively, and cystitis or pyuria was used as the outcome variable.
Diabetes mellitus was not found to be associated with HSC but was evaluated for possible associations because it presents with excessive urination in cats. Similarly, there was no significant association between HSC and the quarter of the year, as well as the year of crystalluria development or selection of controls. Any such association would help evaluate possible seasonality or temporal changes in HSC. However, our study noted clustering at the hospital level but not higher levels. Hence, no additional variance in HSC was explained at the zip code (region where the hospital was located) or state levels.
Because the present study did not address any clinical signs, no conclusions can be drawn on the clinical or pathologic importance of HSC in cats. Characterization of concurrent diagnoses and clinical signs in affected cats was beyond the scope of the present study but could have yielded useful information in this regard. Crystalluria can exist without disease,14,18 and urinary tract disease can exist without crystalluria. 14 Diseases with which crystalluria can coexist include feline idiopathic cystitis, urolithiasis, urinary tract infection and chronic kidney disease.
A limitation of the present study was that cats were classified as having HSC on the basis of results of urinalysis. First, a diagnosis of HSC could be an artifact in many patients because of the unknown time from collection to analysis, and because samples were collected by cystocentesis which can cause blood contamination. These urinalyses were performed in clinical practice, which have standard but uncontrolled conditions from a research perspective. In addition, studies have shown that storage of urine samples can lead to in vitro crystal formation and misdiagnosis of crystalluria.19,20 It is recommended that urine samples be analyzed within 60 mins of collection to avoid influences of storage factors on in vitro crystal formation, and it is unlikely that these conditions were met in the present study. However, any HSC misclassification (measurement bias) that might have arisen from uncontrolled analytic procedures or any hematuria introduced during cystocentesis would have been non-differential, given that there would be no reason to suspect that certain cats would be more likely to have a mismeasurement of any urinary factor than other cats would.
Conclusions
The present study provided estimates of the strengths of association between HSC and certain signalment and clinical characteristics of cats. The immediate clinical importance of the findings is unclear, given that crystalluria can exist without clinical disease, and whether the included cats with HSC had clinical signs was not explored. However, this information could help clinicians to perform a more directed screening for struvite crystalluria in certain cat populations. Follow-up studies that build on the findings of this study could explore the clinical importance of HSC in cats.
Footnotes
Acknowledgements
The authors are grateful to Banfield Pet Hospital for providing access to data used in this study.
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
