Materials and Methods

Animals—All cats that were brought to any Banfield Pet Hospital in the United States from January 1 through December 31, 2010, and for which complete signalment information (eg, breed, reproductive status, and age) was available were eligible for inclusion in the study. The Banfield hospital network at the time included 755 full-service, primary care veterinary hospitals in 43 states with a mean of approximately 121,000 dog and cat office visits/wk. All hospitals used the same proprietary practice management software, which was used to create electronic health records that were uploaded nightly to a central database.

Cats were included in the case group when the diagnosis of CKD was first made in 2010, serum creatinine concentration and USG were measured within 30 days before or after the CKD diagnosis, serum creatinine concentration was > 1.6 mg/dL (the lower cutoff suggested by the IRIS guidelines⁹ for classification of stage 2 CKD), and USG was < 1.035. For this group of cats, the study entry date was considered the date that the diagnosis of CKD was first made. Cats were eligible for inclusion in the control group when they had been brought to the hospital at least once before the qualifying 2010 visit, CKD had not been diagnosed in the past, serum creatinine concentration was not > 2.2 mg/dL (the upper reference limit for serum creatinine concentration provided by the laboratory) within 30 days before or after the 2010 visit, and USG was not < 1.035 within 30 days before or after the 2010 visit. These cats were not required to have had their serum creatinine concentration or USG measured at any previous visit. From the population of cats that met these control criteria, a number equal to the number of case cats were randomly selected to match the case cats by age at study entry (± 6 months). For the group of control cats, the study entry date was considered the date of the hospital visit when the cats were deemed to have met the inclusion criteria and were selected. All included cats were required to have had at least 1 visit on record prior to the date of study entry.

Study design—Data extracted from the medical record of each cat that met the study criteria included age at study entry; body weight at study entry and at the latest visit during the period 6 to 12 months prior to the entry date, when available; breed; sex and reproductive status; body condition at study entry; type of diet fed within 1 year prior to study entry (wet, dry, or wet and dry food), when available; number and administration rate of FVRCP vaccinations; frequency of hospital visits between the first visit on record and the last visit during the study period (presumed to contribute to detection bias, with more frequent evaluations increasing the likelihood of disease detection); US region of residence; history of general anesthesia or dehydration within 1 year prior to study entry (yes or no); clinical signs of CKD (polyuria, polydipsia, vomiting over a period > 48 hours, energy loss, decrease in appetite, or halitosis) at the latest visit between 12 and 6 months prior to study entry; status of wellness plan (a prepaid medical care plan offered by Banfield Pet Hospital that typically includes preventive care and unlimited hospital visits) at any time in the past (never had one or did have one) and total wellness plan duration; diagnosis of predisposing or concurrent diseases made by a Banfield veterinarian at any point on or before the date of study entry (ie, arthritis, leptospirosis, toxoplasmosis, diabetes mellitus, hypertension, periodontal disease, overweight or obese condition, hyperthyroidism, heart disease [undifferentiated], upper respiratory tract infection, or cystitis); and clinical signs of CKD (vomiting, polyuria or polydipsia, loss of appetite or energy, and halitosis) at study entry.

At hospital evaluations prior to June 2010, body condition had been subjectively assessed by veterinarians as thin, normal, or heavy. At hospital evaluations after June 2010, body condition had been classified by use of a 5-point scale standardized body condition scoring system (1 = very thin, 2 = thin, 3 = ideal weight, 4 = overweight, and 5 = markedly obese). Because of these differences in techniques, a decision was made to use the less refined subjective classifications for study purposes and to treat any cat with a body condition score of 1 or 2 as thin and any cat with a body condition score of 4 or 5 as heavy. Regions of residence were defined as northwest (Idaho, Montana, Oregon, and Washington), northeast (Connecticut, Delaware, Indiana, Kentucky, Massachusetts, Maryland, Michigan, New Hampshire, New Jersey, New York, Ohio, Pennsylvania, Rhode Island, and Virginia), north central (Iowa, Illinois, Minnesota, Nebraska, South Dakota, and Wisconsin), south central (Arkansas, Colorado, Kansas, Louisiana, Missouri, New Mexico, Oklahoma, and Texas), southeast (Alabama, Florida, Georgia, Mississippi, North Carolina, South Carolina, and Tennessee), and southwest (Arizona, California, Nevada, and Utah). Diagnoses that comprised the undifferentiated heart disease variable included cardiomyopathy of any type (eg, hypertrophic, taurine-responsive, dilated, or restrictive), heart failure (eg, left- or right-sided or congestive), and valvular insufficiency of any type (eg, mitral or tricuspid). Other diagnoses such as hypertension, diabetes mellitus, or cystitis were accepted as recorded and were not validated through examination of test results.

For cats that were brought to the hospital between 12 and 6 months prior to the study entry date, the percentage change in body weight between the latest visit within that 6-month period and the study entry date was calculated and data regarding clinical signs of CKD recorded during that period were also collected. For cats with a diagnosis of cystitis, data were extracted to identify whether results of urinalysis indicated the presence of WBCs in the urine.

Statistical analysis—All statistical analyses were performed with the aid of statistical software.^a Initially, the data were explored through calculation of summary statistics, including tests to determine whether values of continuous variables such as age or body weight were normally distributed (Shapiro-Wilk, Kolmogorov-Smirnov, Cramer-von Mises, and Anderson-Darling tests). Nonnormally distributed values, as identified by use of all 4 tests, are reported as median (range), and normally distributed values are reported as mean ± SD. For univariate analyses, the Mann-Whitney U test was used to determine whether the case and control groups differed significantly with respect to any nonnormally distributed data. Univariate logistic regression analysis by means of the Wald χ² test was also performed to identify variables with conditional associations with disease status (CKD vs no CKD). Factors hypothesized to be associated with CKD and potential confounders such as factors that would simply increase the likelihood of CKD detection (eg, visit frequency and enrollment in a wellness plan) were evaluated in this manner.

Variables that had liberal associations (ie, value of P < 0.20) with CKD identified in the univariate logistic regression analyses were included in multivariate logistic regression modeling, as were 2-way interaction terms that were deemed biologically plausible (eg, the interaction between sex and reproductive status or between age and body condition). Two models were constructed: one in which cats without body condition information were coded as though their body condition had not been thin, and another in which only cats with a recorded body condition were included. The purpose of this approach was to determine whether cats lacking a recorded body condition could be treated as though their body condition had been ideal (which presumably would cause veterinarians to overlook recording the body condition assessment). A forward stepwise approach to model building was used, with values of P < 0.05 being considered significant. Variables that were not deemed significant but were components of significant interaction terms were retained in the final model. Model goodness of fit was assessed with the Hosmer-Lemeshow test.

Descriptive statistics are summarized as median (range) or percentage, and results of logistic regression analysis are reported as OR and 95% CI, with a CI that excludes 1 used to indicate a significant association. Odds ratios (the odds of disease in one group vs another) can be used to quantify the magnitude of association between a factor and a disease. An OR of 1 indicates no association; a significant OR > 1 indicates that the factor, when present, is associated with an increased probability of disease; and a significant OR < 1 indicates that the factor is associated with a decreased probability of disease.¹⁰ Significant ORs only indicate an association (ie, the factor is more common [or less common] in the group with disease than the group without disease); they do not imply a causal relationship between a factor and the disease.

Results

Animals—A total of 425,745 cats were brought to Banfield hospitals in 2010. Of the 6,747 cats in which CKD was diagnosed during the study period, 1,230 (18.2%) met the criteria for inclusion in the case group. Typical reasons for exclusion other than not meeting the specific case or control definitions were errors in recording birth date or breed, unspecified reproductive status, or duplicate entries. Another 1,230 cats without CKD were randomly selected for the control group. The mean ± SD USG for the case group was 1.018 ± 0.006, and the median USG was 1.018 (range, 1.001 to 1.034). The median serum creatinine concentration was 3.1 mg/dL (range, 1.7 to 20.3 mg/dL). The distribution of IRIS CKD stages among the 1,230 case cats was as follows: 424 (34.5%) were at IRIS CKD stage 2 (serum creatinine concentration, 1.6 to 2.8 mg/dL), 544 (44.2%) were at IRIS CKD stage 3 (serum creatinine concentration, 2.8 to 5.0 mg/dL), and 262 (21.3%) were at IRIS CKD stage 4 (serum creatinine concentration, > 5.0 mg/dL).

Data were recorded for only a subset of all cats with respect to body condition (597 case and 435 control cats), type of diet (408 case and 424 control cats), and variables for which data obtained 6 to 12 months prior to the study entry date were available (507 case and 539 control cats). Among the 597 case cats with body condition information, the body condition was classified as thin for 396 (66.3%) cats, ideal for 45 (7.5%) cats, and heavy for 156 (26.1%) cats. Among the 435 control cats for which data were available, body condition was classified as thin for 167 (38.4%) cats, ideal for 31 (7.1%) cats, and heavy for 237 (54.5%) cats. A diagnosis of toxoplasmosis or leptospirosis was not recorded for any cat included in the study Pyuria (presence of WBCs in the urine) was identified in 175 of 243 (72.0%) case cats and in 29 of 81 (35.8%) control cats with cystitis for which urinalysis results were available. Results of bacterial culture of urine samples were unavailable for the study.

Unconditional associations—Cats in the case and control groups did not differ significantly (P > 0.05) in age at the time of CKD diagnosis or study entry, sex or reproductive status, number or rate of FVRCP vaccinations, type of diet fed within 1 year prior to study entry, or prior diagnosis of diabetes mellitus, overweight or obese condition, arthritis, or upper respiratory tract infection (Table 1). Distributions of various cat breeds were similar between groups, with 2 exceptions: domestic medium hair cats were more common (OR, 1.30; 95% CI, 1.02 to 1.64) and domestic shorthair cats were less common (OR, 0.77; 95% CI, 0.65 to 0.90) in the case group than in the control group. The number of hospital evaluations per year (OR, 1.20; 95% CI, 1.14 to 1.27), number of years of wellness plan enrollment (OR, 1.03; 95% CI, 1.00 to 1.07), and history of wellness plan enrollment (OR, 1.77; 95% CI, 1.51 to 2.07) were associated with increased odds of CKD detection.

A median weight loss of 10.8% was identified among cats with CKD, compared with a median weight loss of 2.1% among cats without CKD; the difference was significant (P < 0.001). Several other factors were overrepresented in the case group, including previous diagnoses of hypertension, periodontal disease, undifferentiated heart disease, and cystitis as well as living anywhere in the contiguous United States except the northeast. On the other hand, cats with CKD had a lower body weight and a lower frequency of hospital evaluations than did the other cats. When identified 6 to 12 months preceding the diagnosis of CKD, none of the factors hypothesized to predict a diagnosis of CKD (ie, vomiting, polyuria or polydipsia, loss of appetite or energy, or halitosis) had a significant relationship with that diagnosis, yet all were associated with increased odds of CKD when identified on the day of study entry (ie, date of diagnosis for cats with CKD). A post hoc review of the data revealed that cats with periodontal disease were more likely than cats without periodontal disease to have undergone general anesthesia in the preceding year (OR, 1.61; 95% CI, 1.25 to 2.10); however, when the data were stratified by study group, this association was significant for the control cats only.

Multivariate analysis—Two final models were developed to identify conditional associations between various factors and diagnosis of CKD in cats: one model that included all cats and another model in which only cats with a recorded body condition were included. Because the results of the second model were comparable to those of the first, the decision was made to use the second model because it included a larger number of cats and had no significant (P = 0.33) evidence of a lack of fit to the data on a goodness-of-fit test.

The final model, which included the number of hospital evaluations per year and number of years that cats were covered by a wellness plan as potential confounders, revealed several factors that were significantly associated with CKD diagnosis when controlling for other factors (Table 2). The odds of CKD development for cats residing in the northeast were lower than that for cats residing in other US regions. Factors that increased the likelihood of CKD diagnosis included thin body condition or dehydration at study entry (ie, time of diagnosis or selection for control group), being a neutered male rather than a spayed female, prior diagnosis of periodontal disease or cystitis, and undergoing anesthesia in the past year. For neutered males, the probability of CKD diagnosis did not differ from that for sexually intact males. Cats with greater body weight were less likely to have CKD, but only when the cats had no recorded episodes of dehydration. Other factors associated with a lower odds of CKD were domestic shorthair breed and diabetes mellitus.

Discussion

The purpose of the present study was to identify clinical characteristics of cats with CKD that may be used to enhance early detection of CKD in cats in general practice. Although diagnosis of CKD requires determinations of serum creatinine concentration and USG, the clinical findings reported here have provided a basis for recommending that additional CKD screening tests be performed when clinical abnormalities in those 2 variables are detected. Clinical factors that increased the odds of diagnosis of CKD in cats included weight loss, thin body condition, dehydration, undergoing general anesthesia in the preceding year, and diagnosis of periodontal disease or cystitis in the preceding year. Clinical characteristics overrepresented among cats with CKD detected in univariate logistic regression analyses also included a history of hypertension and undifferentiated heart disease. In addition, owners of cats with CKD were more likely than owners of cats without CKD to report clinical signs typical of CKD such as polyuria or polydipsia, decline in appetite, vomiting, decline in activity, or halitosis at the hospital visit during which the diagnosis of CKD was made. Clinical factors that appeared to reduce the odds that a cat may have CKD included domestic shorthair breed and previous diagnosis of diabetes mellitus. The associations identified in our study should be viewed as potential indicators for facilitating earlier recognition and diagnosis of CKD and not necessarily as evidence of a cause-effect relationship between the risk factors and CKD in cats.

In the present study, the case group included cats with a broad range of serum creatinine concentrations consistent with (but not necessarily indicative of) IRIS CKD stages 2, 3, and 4. These cats would therefore be predicted to have a broad range of physical findings, such as mild clinical signs through overt uremia or even an absence of clinical signs, at the time of diagnosis. However, because cats were included in the case group on the basis of their diagnosis of CKD made by the attending veterinarians, serum creatinine concentration > 1.6 mg/dL, and USG < 1.035, their actual stage of CKD could not be assured because the IRIS staging process requires that preprandial blood samples are collected for serum creatinine concentration measurement on 2 occasions when the cats are clinically stable.

The principal potential benefits of early diagnosis of CKD include prolongation of survival time, with a high quality of life, through earlier intervention with treatments designed to slow the disease progression⁴; potential establishment of the initiating cause of CKD, provided that the cause remains present and active early in the course of CKD; and initiation of therapeutic diet feeding, which may be easier to achieve before a cat begins to develop gastrointestinal complications of CKD. Initiation of feeding a therapeutic diet to cats with CKD at IRIS stages 2 and 3 has been shown to result in delayed onset of uremic signs and prolonged survival time.^4,11,12

In the present study, 1 factor hypothesized to be associated with CKD—reproductive status—was significant only for neutered cats; neutered males were 30% more likely than spayed females to develop CKD (OR, 1.30). Because male cats are more likely to develop urethral obstruction than are female cats, renal damage and subsequent CKD resulting from an obstruction might contribute to this sex-related difference in susceptibility to CKD.^13,14 That particular hypothesis was not tested in the present study.

The study findings that a decrease in body weight over the preceding 6 to 12 months and a thin body condition at the time of study entry were associated with a diagnosis of CKD were not surprising. In our experience, these observations are typical of cats with CKD. However, the age-matched control group was comprised largely of geriatric cats with a median age of 14.2 years, among which weight loss and thin body score may be attributable to, at least in part, a negative correlation between age and nutrient digestibility in senior cats.¹⁵ Although the control group also had weight loss over the 6 to 12 months prior to study entry, the mean weight loss in cats with CKD was approximately 5 times as great on a percentage weight loss basis. This finding suggested that the greater loss of weight in the case group was likely associated more with the development of CKD rather than simply an effect of aging. Because owners of cats with CKD were approximately 3 times as likely to report a decrease in appetite at the time of CKD diagnosis as were owners of age-matched controls, the decrease in food consumption likely contributed to the weight loss and thin body condition. Because diet change or feeding of a therapeutic diet was not an exclusion criterion in the present study, we cannot exclude a role for diet in the weight loss observed in the cats with CKD.

A documented physical examination finding of dehydration in the preceding year also increased the odds that cats would have a diagnosis of CKD. Dehydration in cats with CKD is often attributed to the cats failing to consume adequate water to compensate for ongoing polydipsia. Indeed, owner reporting of polyuria or polydipsia at the time of study entry was 5 times as common for cats with CKD as for cats without CKD. However, distributions of the same clinical signs when documented within the last visit 6 to 12 months prior to study entry were similar for cats with and without CKD. The lack of predictive value of previous polyuria or polydipsia when considered in isolation from other clinical signs or diagnostic findings may reflect the diversity of causes for such signs in cats, most notably as a consequence of diabetes mellitus or hyperthyroidism, neither of which was associated with a CKD diagnosis in the present study.

Cats undergoing anesthesia within the preceding year and prior diagnosis of periodontal disease were also associated with a diagnosis of CKD. In our experience, dental procedures directed at managing periodontal disease are among the more common reasons for anesthesia of mature adult or older cats. Although periodontal disease was associated with anesthesia in the control group, this association was not found in the case group, even though periodontal disease was more common among the cats with CKD. The reasons for anesthesia were not identified in either group; however, it appeared that more nondental procedures were performed in the case group than in the control group and that anesthesia within the preceding year and prior diagnosis of periodontal disease were likely discrete risk factors in cats with CKD. Although an association between episodes of anesthesia and later diagnosis of CKD does not support a cause-effect relationship, anesthesia has often been considered a risk factor for acute kidney injury because of the risk of renal hypoperfusion during and after surgical procedures, and acute kidney injury may result in development of CKD in some instances.^16–19 However, crystalloid fluids are routinely administered IV during surgeries requiring anesthesia at Banfield hospitals in an attempt to mitigate this risk of hypoperfusion. Because of the design of the present study, it could not be established whether periodontal disease preceded CKD development, and no conclusion could be made regarding the observed association between the 2 diseases other than that they coexisted. That said, high concentrations of urea and other nitrogenous wastes in saliva have been hypothesized to promote development of dental tartar by providing a substrate for oral bacterial growth.²⁰

Another association detected in the present study was that between prior diagnosis of cystitis and diagnosis of CKD. In cats, cystitis may develop as a result of any inflammatory or irritative condition of the urinary bladder, such as bacterial cystitis, idiopathic cystitis, urocystoliths, or bladder neoplasia. The retrospective nature of the present study and the complexity of the electronic medical record data did not allow exploration or confirmation of the specific nature of the disease in the cats with cystitis; those diagnoses were made by hundreds of veterinarians who undoubtedly varied in their understanding of and diagnostic threshold for cystitis and represent a limitation of the study. Although the uncertainty of the cystitis diagnoses was a limitation of the study, we believe that the findings nonetheless reflect what might be encountered in general veterinary practice. Of the conditions that might prompt a diagnosis of cystitis, bacterial UTI is most likely associated with CKD because bacterial UTI is a fairly common complication in cats with CKD.^1,2 A diagnosis of bacterial cystitis can be confirmed only through bacterial culture of urine samples, but such data were not available for most of the study cats. Instead, we used pyuria as a surrogate for bacterial UTI in an attempt to ascertain whether bacterial or nonbacterial causes were underlying the clinical diagnosis of cystitis in cats with CKD. In that analysis, 72.0% of case cats with a diagnosis of cystitis also had pyuria, whereas only 35.8% of control cats with a diagnosis of cystitis also had pyuria. This finding was consistent with the recognized association between CKD and bacterial UTI.³ Because identification of pyuria is not strong evidence for positive results of bacterial culture of urine, bacterial culture of urine samples should be performed whenever a presumptive diagnosis of cystitis is made for geriatric cats.

Overt clinical signs typical of CKD alert a clinician to the possibility of CKD but do not appear useful as early indicators of CKD. Not surprisingly, clinical signs typical of moderate to advanced CKD, including polyuria or polydipsia, vomiting, halitosis, and decrease in appetite and activity, were predictive of a diagnosis of CKD on the day of the hospital evaluation at which CKD was diagnosed. However, these signs were not predictive at evaluations made 6 to 12 months before the date of CKD diagnosis. Presumably, these clinical signs, with the possible exception of polyuria or polydipsia, become apparent in cats with more advanced stages of CKD; in those cats, azotemia and impaired urine concentrating ability are already well established and a diagnosis of CKD is likely to be made when those problems are present. Thus, these clinical markers are unlikely to be useful as early indicators of CKD.

Diagnoses of arterial hypertension and undifferentiated heart disease, as made by various veterinarians but not confirmed through results of diagnostic tests for purposes of the present study, were overrepresented in the case group of the present study. Arterial hypertension in cats is most often secondary rather than primary hypertension, and the most common cause for secondary hypertension in cats is CKD.^1,2 Thus, the finding of hypertension in cats should prompt consideration of CKD. The conditions included in the study definition of undifferentiated heart disease were a variety of clinical entities. However, arterial hypertension can be associated with cardiac enlargement, murmurs, and gallop rhythms in some cats and may explain at least some of this association.

A previous diagnosis of diabetes mellitus decreased the odds of diagnosis of CKD in the cats of the present study. Diabetes mellitus is generally not believed to have a protective effect on CKD, and it cannot be ascertained from the study design whether diabetes preceded development of CKD. However, the disease might be considered a condition that predisposes cats to CKD development because of the well-recognized association between diabetes mellitus and CKD in humans.^21–23 Although other species may also develop renal lesions typical of diabetic nephropathy, cats do not typically develop CKD as a consequence of diabetes mellitus.²⁰ Whether diabetes mellitus actually protects against CKD development is unclear, but the osmotic diuresis associated with glucosuria may promote phosphaturia, which could have a protective effect by limiting phosphate retention and its renal consequences.

The finding that a diagnosis of CKD was more likely to be made for cats in all regions of the United States other than the northeast was not expected. Region was originally considered for inclusion in the model in an attempt to control for any biases associated with veterinarians in the 5 regions having different levels of awareness of or diagnostic thresholds for various diseases. However, it is also possible that 1 or more unknown factors of an environmental or other nature might explain the observed protective effect of living in the northeast.

An additional finding that warrants discussion is the lack of an association between the cats' type of diet (dry, wet, or a combination of both) and CKD. One hypothesis that has yet to be scientifically supported is that the lack of moisture in the diets of cats fed only kibble taxes the kidneys, causing damage that leads to CKD. However, the study cats with a record of having been fed kibble were no more likely to develop CKD than those fed wet food, providing no support for this hypothesis.

The present study had several limitations that need to be considered when interpreting the results. First, the diagnoses of CKD among the cats were made by many veterinarians and were corroborated for study purposes by the documented presence of azotemia and impaired urine concentrating ability. Although we had every reason to trust the clinical judgment used in making the diagnosis of CKD, we were unable to confirm the criteria used by the clinicians to distinguish chronic from acute kidney disease. However, the prevalence of CKD in cats, particularly among the age group we studied, is dramatically greater than the prevalence of acute kidney injury.²⁴ As a consequence, errors in diagnosis, if any, were probably negligible. Second, some of the clinical characteristics, such as body condition and hydration status, are subjective and were assessed by a large number of veterinarians rather than by a single individual. Nonetheless, these are commonly determined subjective assessments that general practice veterinarians are expected to be competent at performing. Another limitation was that the serum creatinine concentration ranges used to identify cats for the case and control groups overlapped (> 1.6 mg/dL vs ≤ 2.2 mg/dL, respectively). The 1.6 mg/dL cutoff used to identify cats with CKD was based on IRIS guidelines,⁹ whereas the 2.2 mg/dL cutoff used to select the control group was based on the reference limits for serum creatinine concentration used in Banfield clinics. Although these 2 study groups may appear to have overlapped, cats defined as having CKD were required to have a USG < 1.035 at the time of diagnosis, whereas cats selected for the control group had a USG ≥ 1.035; thus, the supposition that any serum creatinine concentrations between 1.6 and 2.2 mg/dL did not represent renal azotemia was considered valid. However, the possibility remains that some nondehydrated cats had renal azotemia but maintained a USG > 1.035 (ie, glomerulotubular imbalance).

The present study provided evidence that clinical characteristics obtained from the signalment, medical history, and physical examination of cats can be used to assess the odds that a given cat is likely to have CKD. Results indicated that clinical signs such as thin body condition, polyuria and polydipsia, anorexia, vomiting, lethargy, and halitosis suggest a cat—one that is male and neutered—may currently have CKD, and this likelihood increased with historical factors such as weight loss ≥ 10% during the preceding 6 to 12 months, prior periodontal disease or cystitis, and dehydration detected or anesthesia in the preceding year. Recognition of these clinical findings and characteristics in cats should prompt clinicians to pursue additional diagnostic testing to determine whether CKD can be ruled out. The wide range of factors identified as associated with CKD suggests the importance of annual to semiannual measurement of serum creatinine concentration and USG in aging cats.