Development of an estimated glomerular filtration rate formula in cats

Background Estimated glomerular filtration rate (eGFR) formulas are routinely used in human patients to provide a more accurate evaluation of GFR compared to serum creatinine concentration alone. Similar formulas do not exist for cats. Objectives To validate a prediction formula for eGFR in cats based on adjusting serum creatinine concentration. Animals Client‐owned cats with various levels of renal function. Methods The study was cross‐sectional. Glomerular filtration rate was determined by iohexol clearance. Variables including signalment, biochemical markers, and noninvasive measurements considered to represent surrogate markers of muscle mass were evaluated with the reciprocal of serum creatinine concentration in a multivariable regression model. The derived eGFR formula was subsequently tested in another group of cats and agreement with GFR assessed. Results The formula was developed in 55 cats. Only a single morphometric measurement (pelvic circumference) along with the reciprocal of serum creatinine concentration (creatinine−1) independently predicted GFR in the final multivariate model. The derived eGFR formula was 0.408 + (243.11 × creatinine−1 [μmol/L]) ‐ (0.014 × pelvic circumference [cm]). When the formula was tested in another 25 cats it was not found to offer any advantage over creatinine−1 alone in its relationship with GFR (eGFR, R 2 = 0.44, P < .001 vs reciprocal of creatinine, R 2 = 0.45, P < .001). Furthermore, agreement between eGFR and GFR was poor. Conclusions and Clinical Importance An eGFR formula for cats that adjusted serum creatinine concentration for a marker of muscle mass was developed. The formula did not provide a reliable estimate of GFR, and therefore, its routine use cannot be recommended.


| INTRODUCTION
Detection of chronic kidney disease (CKD) before development of azotemia in cats is desirable because implementation of therapeutic interventions at this stage may delay or prevent progression of CKD.
Limitations of serum creatinine and urea concentrations and urine specific gravity for assessing renal function, particularly in early stage CKD, are recognized. 1 Measurement of glomerular filtration rate (GFR) is considered to provide the most accurate estimation of renal function. Serum creatinine concentration is the most routinely used Abbreviations: BCS, body condition score; BL, body length; BSA, body surface area; BW, body weight; CKD, chronic kidney disease; CKD-EPI, chronic kidney diseases epidemiology collaboration; eGFR, estimated glomerular filtration rate; ECFV, extracellular fluid volume; FFM, fat free mass; FM, fat mass; FLH, forelimb height; GFR, glomerular filtration rate; HLH, hindlimb height; LFLC, left forelimb circumference; LHLC, left hindlimb circumference; LPHLC, left proximal hindlimb circumference; MDRD, modification of diet in renal disease; PC, pelvic circumference; RFLC, right forelimb circumference; RHLC, right hindlimb circumference; RPHLC, right proximal hindlimb circumference; TBW, total body water; TC, thoracic circumference marker of GFR, but it not only reflects renal function but also other factors including muscle mass. Creatinine is generated endogenously from creatine and creatine phosphate in skeletal muscle cells. Therefore, methods that correct plasma creatinine concentration for a patient's muscle mass may give a more accurate estimation of GFR.
Estimated GFR (eGFR) formulas offer the advantage of more accurately reflecting actual GFR than serum creatinine concentration in human patients. By incorporating demographic and clinical variables that may affect physiological processes contributing to serum creatinine concentration, a more accurate measurement of renal function may be ascertained. Of most relevance are factors that contribute to muscle mass such as age, sex, and race. It is now mandatory for eGFR to be reported with every serum creatinine concentration measurement performed in human patients in several states in the United States, in the United Kingdom, and in Australia, 2 highlighting the importance of such formulas. The most widely used prediction formulas for GFR in human patients are the Cockcroft-Gault, 3 Modification of Diet in Renal Disease (MDRD), 4 and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formulas 5 (Table 1). These formulas estimate GFR from serum creatinine concentration by means of prediction equations that take into account of factors such as age, sex, race, and body size. [3][4][5] After introduction of reporting of eGFR alongside serum creatinine concentration, recognition of CKD by doctors has increased. 6 Identifying useful methods that correct serum creatinine concentration to account for muscle mass would be important in developing an eGFR formula for cats. Methods to directly measure muscle mass cannot be readily applied to clinical patients. 7 Skeletal muscle mass is the largest component of fat-free mass (FFM) reported to be 0.49 × FFM in human patients. 8 Determination of FFM therefore may provide an estimate of muscle mass. In addition, morphometric measurements, body condition score (BCS), and body weight (BW) also may provide surrogate markers of muscle mass. A formula to predict FFM in cats based on BW (kg) and various morphometric measurements has been reported (Finch et

| Measurement of GFR
Food was withheld for 12 hours before performing the measurements. Glomerular filtration rate was determined by a previously described iohexol clearance method. 9 Briefly, a bolus dose of iohexol (Omnipaque 300 [647 mg/mL; 300 mg of iodine/mL, GE Healthcare, Wauwatosa, Wisconsin) was administered IV (1 mL/kg). Blood samples were collected at 120, 180, and 240 minutes postinjection.
Iohexol concentrations were determined at an external commercial laboratory by a high performance liquid chromatography (HPLC) was applied to correct for the 1-compartment assumption. 9 In addition, serum creatinine, urea, albumin, and total protein concentrations were determined from a sample collected at the same time as GFR measurement.

| Development of eGFR formula
The following variables were considered for inclusion in a multivari- were analyzed at an external laboratory (Institute of Child Health, London) by isotope ratio mass spectrometry (IRMS). Dilution space of 18 O was calculated by the standard equation 12 : where T is the mass of tap water diluent in which a is diluted, A is dose of 18 O administered, a is portion of dose administered of 18 O that was retained for mass spectrometer analysis, and δ a , δ t , δ s , and δ p are isotopic enrichment in delta units of the portion of dose administered, tap water diluent, postdose serum sample, and predose serum sample, respectively. Delta units express isotopic enrichment relative to 2 standard waters (standard mean ocean water and standard light arctic precipitate). Total body water then was calculated as: 18 O dilution space=1:01 where 1.01 is the correction factor to correct for 1% over-exchange with nonaqueous compartments.
Predicted muscle mass was determined by the equation 13 : where BW is body weight (kg), FLH is forelimb height (cm), RFLC is right forelimb circumference (cm), and HLH is hindlimb height (cm).  Table 3).
Except for the reciprocal of serum creatinine concentration, variables describing signalment and serum parameters (listed in Table 3) were not predictive of GFR in the final multivariable regression model.

| Testing of the eGFR formula
Data regarding signalment, BW, and renal function are included in indicated that eGFR underestimated GFR ( Figure 2). The mean percent error in 25 cats in which the eGFR formula for cats was tested was −13.6%. The relationship between GFR and eGFR and GFR and the reciprocal of serum creatinine concentration is presented in Figure 3. The eGFR formula showed no advantage over the reciprocal   Bland-Altman agreement plot showing agreement between GFR (determined by iohexol clearance) and estimated GFR (eGFR). Bold line represents bias (mean difference between GFR and eGFR) and dashed lines represent upper and lower limits of agreement (mean difference between GFR and eGFR AE2 SD). The bias indicated eGFR underestimated GFR and limits of agreement were wide. Therefore, agreement was considered poor between GFR and eGFR predicted by the Cockcroft-Gault formula was 0.83 and mean percentage error (expressed as a percentage of true GFR) was 35% in 95% of patients. 3 In the study that developed the MDRD formula, the R 2 for true and predicted GFR was 0.9 and percentage error was 28.4% in 90% of the population. 4 The main difference between the 2 populations was that the MDRD formula included a more diverse population and studied additional factors such as age, sex, and ethnicity and included only patients with CKD. When the relationship between serum creatinine concentration and GFR was explored in the study in which the MDRD equation was developed, the R 2 was 0.8 (vs 0.9 for GFR and eGFR), 4 suggesting these formulas do offer some advantage over the use of serum creatinine concentration alone. However, it is clear from the mean percent errors that the formulas are not particularly precise predictors of GFR in humans. Mean percent error in 25 cats in which the eGFR formula for cats was tested was −13.6%, which, interestingly, is considerably lower than that obtained in some of the studies in humans.
An important conclusion from studies in humans that derived prediction equations for GFR 4,15 is that the population in which formulas are developed should represent the target population. Our study addressed this concern by inclusion of normal healthy cats and cats with decreased renal function in development and testing of the equation. It is disappointing that the eGFR formula for cats did not appear to offer any advantage over measurement of the reciprocal of serum creatinine concentration to predict GFR, which may reflect the small number of cats included in the development of the formula.
Until a more accurate formula can be developed in a larger population of cats, biomarkers such as serum creatinine or symmetric dimethylarginine concentrations remain the best surrogate markers of GFR in cats.
The primary source of creatinine generation is muscle mass. This factor is addressed in human patients by including coefficients in GFR prediction equations for factors affecting muscle mass such as age, sex, and race. It is likely that there are important differences in creatinine generation in cats. It has been reported that serum creatinine concentration is higher in Birman cats 16 although whether this observation relates to lower GFR or increased creatinine generation is unclear. Furthermore, chronic disease such as CKD can decrease muscle mass, and therefore cats with chronic disease with the same serum creatinine concentration as healthy cats will have lower GFR if measured. This leads to a circular argument in which serum creatinine concentration or adjusted serum creatinine concentration is used to predict GFR and determine if a patient has normal or abnormal renal function, but CKD itself will affect serum creatinine concentration and hence also affect accurate estimation of renal function by an eGFR formula. With this is mind, any eGFR formula that is validated for cats should serve only as a screening test. In addition, a useful eGFR formula should not only serve as a useful screening test but also provide reliability and accuracy for monitoring progression of CKD.
When using eGFR formulas it is important to ensure that identical methods of analysis and the same laboratory are employed as those used for deriving the formula. Differences in determination of serum creatinine concentration, for example, if a method that detected noncreatinine chromogens was used, would lead to errors in predicted GFR. In studies of humans, differences in creatinine assays at different clinical laboratories can cause errors in GFR estimation as high as 20%. 17 Differences in serum creatinine concentrations determined at different veterinary practices have been reported, making this factor a relevant consideration in cats as well. 18 Additionally, the same units of measurement of creatinine must be required. The current formula was derived by standard international (SI) units (μmol/L) and creatinine concentrations reported in mg/dl would require conversion to μmol/L before applying the formula.
The formula used to estimate FFM in our study does not predict muscle mass. Skeletal muscle mass is the largest component of FFM (reported to be 0.49 × FFM in human patients). 8 No ratio has been reported for cats. A method to determine skeletal muscle mass may be important in cats because of its relationship with serum creatinine concentration. In our study, it was not possible to measure muscle mass directly and therefore predicted FFM was determined. The formula for FFM in cats has been shown to provide good prediction of true FFM determined from total body water and the hydration con- In pediatric patients, correcting serum creatinine concentration for body surface area or body mass index was not found to improve the accuracy of GFR prediction. 19 Further studies to explore the relationships among muscle mass, serum creatinine concentration, and GFR are required in cats. Larger studies performed in a more varied population of cats to that included in our study may be needed before a reliable eGFR formula can be recommended. Doing so may involve direct measurement of muscle mass rather than FFM, although this may be difficult to achieve in clinical patients, and other factors such as sex, age, breed and disease state may influence GFR and serum creatinine concentration.
Development of an eGFR formula for cats to correct creatinine for body composition in our study did not provide a reliable estimate of GFR in cats, and therefore its routine use cannot be recommended.
Moreover, the formula does not appear to improve the accuracy of predicting GFR over serum creatinine concentration. Therefore, determination of GFR will remain important in the early identification and accurate assessment of the stage of CKD.