Breed effects on biochemical variables have been reported

Transcription

1 J Vet Intern Med 2010;24: Breed Dependency of Reference Intervals for Plasma Biochemical Values in Cats B.S. Reynolds, D. Concordet, C.A. Germain, T. Daste, K.G. Boudet, and H.P. Lefebvre Background: Reference intervals (RI) are pivotal in clinical pathology. The influence of breed on RI has been poorly documented in cats. Hypothesis/Objectives: RI for plasma biochemistry variables are breed-dependent in cats. Animals: Five hundred and thirty-six clinically healthy, fasted, client-owned cats from 4 breeds: Holly Birman (n 5 132), Chartreux (n 5 129), Maine Coon (n 5 139), and Persian (n 5 136). Methods: Prospective observational study: Blood samples were collected from the cephalic vein into capillary tubes containing lithium heparin. Plasma glucose, urea, creatinine, total proteins, albumin, calcium, phosphate, sodium, potassium, chloride, and total CO 2 concentrations and the activities of alanine aminotransferase and alkaline phosphatase were assayed with a dry slide biochemical analyzer. RI were defined as central 95% intervals bounded by the 2.5th and 97.5th percentiles. Data were analyzed by a linear mixed effects model with type I error rate of Results: A significant (P o.05) breed effect was observed for 9/13 variables. The magnitude of the differences between breeds could be clinically relevant for creatinine, glucose, and total protein. Age, body weight, sex, and housing conditions had significant (P o.05) breed-related effects on different variables. Conclusions and Clinical Importance: Breed-specific RI should be considered for cats. Key words: Breeds; Clinical chemistry; Feline. Breed effects on biochemical variables have been reported for dogs. 1 8 The dog provides an unique model because there are 4300 breeds and the morphological differences between breeds can be large with body weight (BW) varying from o1 kg (Chihuahua dog) to 460 kg (Irish Wolfhound). Although interindividual phenotype differences are more limited in humans compared with those observed in dogs, there are racial differences. 9,10 In contrast, limited information is available for domestic cats. The number of breeds of cats is also much more limited than in dogs: 55 feline breeds are recognized for championship competition by the International Cat Association ( Moreover, about 60% of cat breeds are of recent lineage and were produced during the last 100 years. 11 In addition, breed barriers are not so strictly defined in cats as in dogs and outcrosses are permissible for some breeds of very recent origin. 11 Nevertheless, a breed effect on plasma biochemical values cannot be excluded for at least 3 reasons. Firstly, breed has been identified as a risk factor for predisposition to specific clinical settings, such as Holly Birmans for feline infectious peritonitis 12 or Burmese for diabetes mellitus. 13,14 Secondly, a breeddependent stratification of reference intervals (RI) in cats From the Department of Clinical Sciences (Reynolds, Germain, Daste, Boudet, Lefebvre) and the Unite Mixte de Recherche 181 Physiopathologie et Toxicologie expe rimentales, Institut National de la Recherche Agronomique (Concordet, Germain, Lefebvre), National Veterinary School of Toulouse, Toulouse, France. Results were presented in part at the American College of Veterinary Internal Medicine Congress, San Antonio, June 4 7, The study was performed at the National Veterinary School of Toulouse, France. Corresponding author: Herve P. Lefebvre, Department of Clinical Sciences, National Veterinary School of Toulouse, 23 chemin des Capelles, BP 87614, Toulouse cedex 03, France; h.lefebvre@envt.fr. Submitted September 3, 2009; Revised March 8, 2010; Accepted April 16, Copyright r 2010 by the American College of Veterinary Internal Medicine /j x Abbreviations: ALP ALT B BUN BW C CI CLSI DSH MC P RI alkaline phosphatase alanine aminotransferase Holly Birmans blood urea nitrogen body weight Chartreux confidence intervals Clinical and Laboratory Standards Institute domestic shorthair cats Maine Coons Persians reference intervals might be necessary at least for plasma creatinine which appears to be higher in the Holly Birman. 15 Thirdly, recent data have demonstrated that genetic diversity in cats is broad and different clusters can be identified. 16,17 The primary objective of the prospective study reported here was to assess the effect of breed on routine plasma analytes and their corresponding RI, in cats from 4 different breeds. A secondary objective was to assess the effects of age, BW, sex, and housing conditions on the biochemistry results. Materials and Methods This study was conducted between March 27 and November 28, Written consent was obtained from each owner of participating cats. Cat Selection Healthy purebred owned cats were prospectively recruited for this study. The participating owners/breeders were solicited by phone contact. Heparinized blood samples were obtained from the cats either in private homes, in a veterinary practice a or in 2 veterinary teaching hospitals. b,c Appointments were made with all participating owners/breeders, who were instructed to withhold

2 810 Reynolds et al food from their cats hours before the scheduled blood collection. Breed, age, sex, fasted status, housing conditions, and medical history were recorded for each cat. It was then weighed and a physical examination performed just before blood collection. Two to 3 months later, the owners/breeders of participating cats were contacted by phone and asked about the apparent health status of their animal. Cats were not included in the study if they were o6 months old, in a nonfasted state, not purebred, had a history of any disease, had received any drug treatment during the 4 weeks before sample collection, if they showed clinical signs of illness on the day of blood collection, or were not reported to be healthy at phone follow-up. Sampling Procedure All blood samples were collected by the same investigator. Blood was obtained from a cephalic vein with a 25-G needle and two to three 200 ml capillary tubes that contained heparinate lithium (lithium heparinate Microvette, 200 ml), d as described elsewhere. 18,19 Sample Processing Blood samples were centrifuged (3,000 g for 10 minutes) within 30 minutes after collection. The plasma supernatant was then immediately harvested and stored at 201C until assayed within 5 days in the same laboratory. Sample stability in such conditions has been described previously. 19 On the days of testing, the plasma samples were allowed to thaw at room temperature for 30 minutes before being assayed. Sample quality was recorded after centrifugation at the time of collection and again at the time of the assay. Assays All plasma biochemical assays were performed with a dry-slide technology analyzer e (Table 1) in the clinical pathology laboratory of the National Veterinary School of Toulouse. The plasma analytes tested were glucose, urea, creatinine, total proteins, albumin, calcium, phosphate, sodium, potassium, chloride, and total CO 2, together with the activities of alanine aminotransferase (ALT) and alkaline phosphatase (ALP). A detailed description of the validation of these assays in the authors laboratory, under the same conditions, has been published previously (Table 1). 19 Laboratory values are reported in SI units, as recommended by the Clinical and Laboratory Standards Institute (CLSI). Conversion factors for US units are provided in Appendix 1. Statistical Analysis Identification of outliers and determination of RI were performed according to the current CLSI guidelines. 20 Native and Box-Cox transformed data in each breed were first tested for normality by use of the Anderson-Darling test. The data set was then examined for outliers in each breed separately. Tukey s criterion was applied when the distribution was Gaussian. When the data distribution remained non-gaussian after Box-Cox transformation, this concept was not applicable and visual inspection of values in both tails of the distribution was used. When a value was deemed questionable, the following criteria were used to decide to remove a cat from the study: any abnormality concerning medical history or clinical examination, an atypical value for more than one analyte Table 1. Test characteristics for analytes, a repeatability, and within-laboratory imprecision estimates for control solutions of plasma analytes b measured by use of the analyzer. Analyte (unit) Methodology Analytic Range Control Concentrations Repeatability Imprecision SD CV (%) SD CV (%) Glucose (mmol/l) Colorimetric (glucose oxidase) Urea (mmol/l) Colorimetric (urease) Creatinine (ímol/l) Enzymatic (creatinine amidinohydrolase) 4 1, Total proteins (g/l) Colorimetric (cupric tartrate) Albumin (g/l) Colorimetric (Bromocresol green) Calcium (mmol/l) Colorimetric (Arsenazo III) Phosphates (mmol/l) Colorimetric (molybdate) Sodium (mmol/l) Potentiometric (ion selective electrode) Chloride (mmol/l) Potentiometric (ion selective electrode) Potassium (mmol/l) Potentiometric (ion selective electrode) Total CO 2 (mmol/l) Enzymatic (phosphoenolpyruvate carboxylase) ALT (U/L) Enzymatic (L-alanine and a-acetoglutarate substrate) 3 1, ALP (U/L) Enzymatic (p-nitrophenyl phosphate substrate) 20 1, ALT, alanine aminotransferase; ALP, alkaline phosphatase. a From Instructions for use, Vitros chemistry products, Ortho-Clinical Diagnostics Inc. b From Reynolds et al. 19

3 Reference Intervals for Cats 811 from the same cat and distance of an extreme value from the closest value observed. RI were defined as central 95% intervals bounded by the 2.5th and 97.5th percentiles. Upper and lower limits of RI with their 90% confidence intervals (CI) were determined in the global population and in each breed by a nonparametric approach. The potential relevance of a breed specific RI was further addressed. Lower and upper limits of the RI with corresponding 90% CI for each breed and the overall population were first graphically represented for visual assessment of between-breed discrepancies. For plasma creatinine, which appeared to be the most relevant variable in terms of breed differences, appropriate partitioning criteria were applied to each pair of breeds to assess the need for separate RI. Partitioning, by comparing the breeds 2 by 2, was indeed considered necessary if any of the 4 proportions (2 at the lower and 2 at the upper end of the distributions) outside the common reference limits were 4.1% or 0.9%. 21 Effects of breed and other covariables on plasma variables were tested using a statistical software package. 22 A value of P o.05 was considered significant. Age and BW in each breed were graphically assessed by boxplots and compared by ANOVA. The effects of breed, age, BW, sex, and housing conditions on plasma variables were tested by the following linear mixed effects model in which the owner appears as a random effects factor: Y ijklm ¼m þ Owner i þ Breed j þ a Age ijklm þ Sex k þ b BW ijklm þ Housing l þ a j Age þðbreed SexÞ jk þ b j BW þðbreed HousingÞ jl þ e ijklm where Y ijklm is the value expressed in SI units of plasma variable Y for Cat m, with Owner i, Breed j, Sex k, and Housing l; m is a constant term; Owner i is a random effect factor assumed to be N (0; 2 s owner ); Age and BW are continuous covariables, a and b are the slope coefficients for Age and BW irrespectively of the breed; Sex k is the differential effect of level k for the Sex factor and Housing l is the differential effect of level l for the Housing factor; a j (Breed Sex) jk, b j, BW, and (Breed Housing conditions) jl denote an interaction between Breed and Age, Breed and Sex, Breed and BW, and Breed and Housing, respectively; a 1 a j and b 1 b j are the slope coefficients for Age and BW in Breed j, respectively; e ijklm is the residual term of the model assumed to be N (0; s 2 ). When a significant interaction between a covariable and breed was evidenced, the effects of age, BW, sex, or housing on plasma variables were assessed breed by breed. Otherwise these covariables effects were assessed on the global sample. Posthoc comparison tests were not performed because all comparisons were a priori planned. Therefore, no correction for multiple comparisons was needed to control the overall type I error. All plasma variables moreover were analyzed separately as a single plasma analyte can be assayed and interpreted to confirm or infirm a clinical hypothesis. Results Reference Sample Group and Assays Five hundred and seventy-one purebred cats were sampled. Five hundred and thirty-six of these (Holly Birmans [B]: n 5 132; Chartreux [C]: n 5 129; Maine Coons [MC]: n 5 139, and Persians [P]: n 5 136) met the inclusion criteria. They belonged to 97 different owners/ breeders. Hemolysis, lipemia, or clot was not observed in any plasma sample. One plasma sample, belonging to a cat that was subsequently excluded, was icteric. Twenty-five batches, corresponding to the whole population of tested cats, were analyzed. Identification of Outliers, Distribution of Plasma Concentrations, and Determination of RI in Each Breed No abnormality concerning medical history or clinical examination could be found to warrant discarding any cat identified with possible outlying values in any of the breeds. Eleven cats were, however, excluded from the data set based on the other criteria: 4 (1 B, 3 C) with atypical values for 2 analytes, 7 (3 C, 2 MC, and 2 P) with an extreme value distant from the closest observed values. Consequently, 525 cats were used for further data analyses. Detailed descriptive statistics of the population for each of the 4 breeds are given in Table 2. Of the 52 (13 variables 4 breeds) breed-specific distributions of plasma variables tested for normality, 14 were Gaussian. Box-Cox transformation yielded a Gaussian distribution for 20 other variables. The remaining 18 could not be transformed to fit a Gaussian distribution (Table 3). Corresponding lower and upper limits of the RI for tested plasma variables with 90% CI are provided for each breed in Table 3. They are graphically represented in Figure 1 as well as those corresponding to the overall population for plasma glucose, creatinine, and total proteins. Results of the application of objective partitioning criteria to each pair of breeds to assess the need for separate RI according to breed for plasma creatinine are provided in Table 4. Table 2. Descriptive statistics of the reference sample group in each breed. Results Variable B C MC P Number of cats Number of owners/ breeders Sex (%) F 93 (71) 76 (62) 97 (71) 94 (70) M 38 (29) 47 (38) 40 (29) 40 (30) NF 17 (13) 18 (15) 19 (14) 17 (13) NM 24 (18) 21 (17) 12 (9) 11 (8) Housing conditions (%) I 55 (42) 36 (29) 49 (36) 91 (68) O 76 (58) 87 (71) 88 (64) 43 (32) Age (years) msd Median Min max BW (kg) msd Median Min max B, Holly Birman; C, Chartreux; MC, Maine Coon; P, Persian; F, total number of females; M, total number of males; NF, number of neutered females; NM, number of neutered males; I, strictly indoor; O, some access outside.

4 812 Reynolds et al Table 3. Reference intervals for plasma analytes in healthy Holly Birman (B), Chartreux (C), Maine Coon (MC), and Persian (P) cats. B C MC P Analyte Distribution LL (90% CI) LL (90% CI) LL (90% CI) LL (90% CI) UL (90% CI) Distribution UL (90% CI) Distribution UL (90% CI) Distribution UL (90% CI) Glucose (mmol/l) NG 3.4 ( ) BCG 4.2 ( ) NG 3.4 ( ) BCG 3.5 ( ) 6.1 ( ) 9.3 ( ) 9.1 ( ) 6.8 ( ) Urea (mmol/l) G 7.2 ( ) BCG 4.5 ( ) G 6.1 ( ) BCG 4.7 ( ) 12.8 ( ) 10.0 ( ) 11.2 ( ) 9.4 ( ) Creatinine (mmol/l) BCG 91 (88 98) G 80 (61 88) G 77 (66 83) BCG 80 (66 87) 239 ( ) 189 ( ) 192 ( ) 164 ( ) Proteins (g/l) BCG 64 (56 65) G 64 (62 65) G 59 (55 60) G 59 (55 62) 103 ( ) 88 (86 90) 85 (82 91) 83 (82 88) Albumin (g/l) G 26 (23 28) G 25 (23 26) G 23 (21 24) G 24 (21 25) 37 (37 42) 38 (37 39) 36 (35 40) 39 (38 40) Calcium (mmol/l) G 2.35 ( ) G 2.39 ( ) G 2.26 ( ) BCG 2.31 ( ) 2.84 ( ) 2.79 ( ) 2.86 ( ) 2.93 ( ) Phosphate (mmol/l) BCG 1.08 ( ) BCG 1.05 ( ) BCG 0.90 ( ) BCG 1.13 ( ) 2.51 ( ) 2.74 ( ) 2.79 ( ) 2.51 ( ) Sodium (mmol/l) NG 151 ( ) NG 152 ( ) NG 151 ( ) NG 151 ( ) 163 ( ) 162 ( ) 161 ( ) 164 ( ) Potassium (mmol/l) NG 3.4 ( ) NG 3.2 ( ) NG 3.4 ( ) NG 3.4 ( ) 4.8 ( ) 4.6 ( ) 4.5 ( ) 4.9 ( ) Chloride (mmol/l) NG 114 ( ) NG 117 ( ) NG 119 ( ) NG 117 ( ) 129 ( ) 129 ( ) 129 ( ) 128 ( ) CO2 (mmol/l) NG 14 (12 14) NG 15 (14 16) NG 14 (11 15) NG 15 (14 16) 22 (21 23) 24 (23 24) 22 (22 23) 23 (23 25) ALT (U/L) BCG 27 (21 35) BCG 25 (15 30) BCG 24 (16 26) BCG 25 (18 28) 148 ( ) 157 ( ) 103 (87 158) 147 ( ) ALP (U/L) BCG 29 (26 32) BCG 29 (26 32) BCG 28 (26 31) BCG 26 (20 29) 93 (79 123) 93 (82 133) 107 ( ) 105 (88 120) G, Gaussian; BCG, Gaussian after Box-Cox transformation; NG, nongaussian; LL, Lower limit; UL, upper limit; 90% CI, 90% confidence intervals; ALT, alanine aminotransferase; ALP, alkaline phosphatase.

5 Reference Intervals for Cats 813 A Glucose (mmol/l) B Breed Effect Significant differences between breeds were evident for BW (P o.001) but not for age (P 5.08) (Fig 2). Some interactions between covariables and breed were statistically significant for some analytes (Table 5). A breed effect was demonstrated for plasma glucose (P o.001), urea (P o.001), creatinine (P o.001), total protein (P o.001), albumin (P o.001), calcium (P 5.007), potassium (P 5.04), total CO 2 (P o.001), and activity of ALT (P 5.006) but not for plasma sodium (P 5.052), chloride (P 5.074), phosphate (P 5.435), and ALP activity (P 5.243) (Tables 6 8). Effects of Age, BW, Sex, and Housing Conditions on Plasma Variables Results are provided in Tables Creatinine (µmol/l) C Total proteins (g/ L) Figure 1. Lower and upper limits with 90% confidence intervals of reference intervals for the global population (white squares), Birmans (black triangles), Chartreux (white circles), Maine-Coons (black squares), and Persians (white triangles) for plasma glucose (A), creatinine (B), and total proteins (C). Dotted vertical lines represent the value for lower and upper limits of the reference interval for the global population. Discussion This study demonstrated a breed effect for 9/13 plasma variables tested. Comparison of rationally established RI further revealed discrepancies between breeds that could be of clinical relevance for some plasma analytes. A major advantage is that the large number of animals included in this study allowed RI to be calculated according to the current CLSI guidelines. 20 Moreover, for most of the published RI for cats, limited information is available regarding the reference sample group and the methods of calculation. As also stated in the CLSI guidelines, all the preanalytical and analytical steps (fasting, blood collection technique, specimen processing, and assays) were here standardized and controlled. However, it should be kept in mind that our results cannot be extrapolated to laboratory conditions other than those described. This study also draws attention to potential issues regarding RI establishment, especially the selection of Table 4. Partitioning of RI for plasma creatinine according to breed. Proportions (%) Outside the Common Reference Limits Combined Common RI Limits (mmol/l) B C MC P Decision B1C Lower NA NA Partition Upper NA NA B1MC Lower NA 4.4 NA Partition Upper NA 0 NA B1P Lower NA NA 3 Partition Upper NA NA 0 C1MC Lower 5 78 NA NA Combined Upper NA NA C1P Lower 5 81 NA 2.4 NA 2.2 Partition Upper NA 4.9 NA 0.7 MC1P Lower 5 78 NA NA Partition Upper NA NA A common RI is established for each pair of breed. Proportions of cats in each breed at the lower and upper end of the distributions outside the limits of the common RI are determined. Separate RI have to be considered when 1 of the 4 proportions observed is either 4.1 or 0.9%; Combined RI is appropriate if all the proportions fall between 1.8 and 3.2%. 21 B, Holly Birman; C, Chartreux; MC, Maine Coon; NA, not applicable; P, Persian; RI, reference interval.

6 814 Reynolds et al A 10 9 Table 5. P value for the statistically significant interactions in the linear mixed effects model used for testing the plasma variables. 8 Interaction of the Breed Effect with the Following Covariable Effect 7 Tested Variables Age Sex Body Weight Housing Glucose Urea Creatinine Total proteins Albumin Calcium Phosphates Sodium Potassium Chloride Total CO ALT ALP BW (kg) B B C MC BREED P A significant interaction between a covariable and breed implies that the effects of this covariable on the plasma analyte tested changes is different between breeds. P values for statistically significant interactions appear in bold. ALT, alanine aminotransferase; ALP, alkaline phosphatase. AGE (y) B C MC P BREED Figure 2. Box-plots illustrating body-weight in kilograms (A) and age in years (B) for each of the 4 breeds. B, Holly Birman; C, Chartreux; MC, Maine Coon; P, Persian. reference individuals, treatment of outlying observations, and the partitioning criteria. B, C, MC, and P were selected as reference sample groups in this study for 2 main reasons. Firstly, recent studies have demonstrated that, in accordance with their phenotypic differences, these breeds are also clearly distinct on a genetic basis. 17 Secondly, the use of common breeds of cats was necessary to obtain sample groups of appropriate size and, in particular, to reach the minimum number of 120 subjects by breed recommended for establishing RI. 23 In addition, the choice of wellrepresented feline breeds was deemed relevant with regard to potential application of the results in clinical settings. Another advantage was that each tested breed population showed a wide range of age (from 24- to 30- fold) and BW (from 3.5- to 5.2-fold). It is worth noting that no difference in age between the 4 breeds was evidenced. By contrast, the observed differences in BW between some breeds were expected and unavoidable as they were intrinsic breed-dependent characteristics. Nevertheless, these differences in BW do not explain the breed-related variations, as both factors were taken into account in the statistical model. Selection of reference individuals is also a complex issue as the definition of health is not straightforward. As recommended by the guidelines of the CLSI, the subjects here were not ill and did not require hospitalization or treatment. 24 They underwent yearly veterinary examinations and were considered healthy by their owners/ breeders. Any history of clinical signs of illness or medication administration was defined as noninclusion criteria and only cats without any clinical sign were included. Moreover, all owners/breeders were contacted by phone 2 3 months after the sampling day to check that all the cats had remained healthy during the postsampling period. Finally, it can be helpful to identify any outlying values, as was done here. Eleven of the 536 (2.1%) included cats were considered to show aberrant values and were removed from the data set. The possibility of establishing separate RI, ie, partitioning, should also be considered before the process of analyzing the data. 20 When only 2 subclasses (eg, male and female) are compared, 2 methods (the Harris/Boyd approach for Gaussian distributions and an alternative approach for non-gaussian ones) are used. 20 However, no method has been proposed when more than 2 subclasses (here 4 breeds) need to be compared. Although not ideal, we therefore decided to use a partitioning criteria, as proposed previously, on each pair of breeds. 21 This partitioning was only performed for plasma creatinine,

7 Reference Intervals for Cats 815 Table 6. Slope coefficients for the continuous variables (age and body weight) and differential effects for the categorical variables (breed, sex, housing conditions) of the linear mixed effect model used for the statistical analysis of the plasma glucose, urea, creatinine, total proteins, and albumin with corresponding P values. Tested Variable m Breed Age Sex Body Weight Housing Glucose (mmol/l) 4.59 B: 0.32 P 5.20 F: P C: 0.02 M: P MC: P o.001 P: 0.20 P o.001 Urea (mmol/l) 7.39 B: F: 0.10 P P 5.16 C: 0.81 P o.001 M: MC: 0.48 P P: 0.44 P o.001 Creatinine (mmol/l) 87.5 B: F: B: P 5.66 C: 4.74 P o.001 M: 2.01 C: MC: 6.38 P o.001 MC: P: P: P o.001 P o.001 Total proteins (g/l) 70.2 B: P 5.90 P P 5.51 C: 0.92 P o.001 MC: 1.43 P: 3.41 P o.001 Albumin (g/l) 27.9 B: F: B: P 5.27 C: 0.06 P M: 0.36 C: MC: P MC: P: 0.52 P: P o.001 P o.001 The slope coefficients and differential effects are only given when the effect is statistically significant. For example, the general regression equation for plasma creatinine in male Birman cats will be: Plasma creatinine (mmol/l) Age (years) BW (kg) Age (years) BW (kg). For a male Birman cat (10.0 years, 4.0 kg), the estimated plasma creatinine concentration will be: Plasma creatinine (mmol/l) mmol/l (or 2.0 mg/dl). m, overall mean for the global population; B, Holly Birman; C, Chartreux; MC, Maine Coon; P, Persian; F, female; M, male. Table 7. Slope coefficients for the continuous variables (age and body weight) and differential effects for the categorical variables (breed, sex, housing conditions) of the linear mixed effect model used for the statistical analysis of the plasma calcium, phosphate, sodium, potassium, chloride, and total CO 2, with corresponding P values. Tested Variable m Breed Age Sex Body Weight Housing Calcium (mmol/l) 2.60 B: P 5.12 P 5.27 P 5.32 C: P o.001 MC: P: P Phosphates (mmol/l) 2.17 P F: B I: P o.001 M: P o.001 O: 0.12 P o.001 C I: 0.05 O: MC I: O: 0.06 P I:10.04 O: 0.04 P Sodium (mmol/l) P P 5.51 P 5.94 P 5.13 P 5.71 Potassium (mmol/l) 3.87 B: P 5.15 P 5.21 P 5.44 B I: C: 0.06 O: 0.14 MC: 0.12 C I: P: O: 0.03 P MC I: 0.02 O: P I:10.02 O: 0.02 P 5.001

8 816 Reynolds et al Table 7. (Continued). Tested Variable m Breed Age Sex Body Weight Housing Chloride (mmol/l) P F: P P 5.48 P M: 0.30 P o.001 Total CO 2 (mmol/l) 18.2 B: 0.83 P F: I: C: 0.31 M: P O MC: 1.04 P o.001 P P: 3.41 P o.001 The slope coefficients and differential effects are only given when the effect is statistically significant. For example, for a male Maine Coon cat (9.0 years, 8.0 kg) living strictly indoors, the estimated plasma phosphate concentration will be: Plasma phosphate (mmol/l) mmol/l. m, overall mean for the global population; B, Holly Birman; C, Chartreux; MC, Maine Coon; P, Persian; F, female; M, male; I, strictly indoor; O, some access outside. Table 8. Slope coefficients for the continuous variables (age and body weight) and differential effects for the categorical variables (breed, sex, housing conditions) of the linear mixed effect model used for the statistical analysis of the plasma alanine aminotransferase (ALT) and alkaline phosphatase (ALP) with corresponding P values. Tested Variable m Breed Age Sex Body Weight Housing ALT (U/L) 77.7 B: B: F: P 5.27 C: C: 0.63 M: 1.41 P o.001 MC: 13.8 MC: P P: 3.33 P: 3.47 P P ALP (U/L) 78.6 P F: P 5.17 P o.001 M: P o.001 P The slope coefficients and differential effects are only given when the effect is statistically significant. For example, for a female Persian cat (3.0 years old, 3.5 kg), the estimated plasma ALT activity will be: Plasma ALT (U/L) U/L. m, overall mean for the global population; B, Holly Birman; C, Chartreux; MC, Maine Coon; P, Persian; F, female; M, male. which was the variable that appeared to be the most relevant to propose breed-specific RI. A need for partitioning was clearly demonstrated for all the 2 by 2 breed combinations except between MC and C. Moreover, differences assessed by visual inspection (Fig 1) of 90% CI for upper limits of breed-specific RI appear large enough to be considered clinically relevant for 3 plasma variables, namely creatinine, total proteins, and glucose, in pure-bred cats. This could be especially relevant when plasma creatinine is used as a first-line diagnostic test for screening early chronic kidney disease. The same issue exists in canine species where plasma creatinine concentration is known to be higher in Greyhounds than in other dogs. 8 Similar differences have also been observed in humans where plasma creatinine concentration has been reported to be higher in non- Hispanic black Americans, lower in non-hispanic whites and lowest in Mexican Americans. 10 Race-related differences in serum proteins have also been described in humans. 25 By contrast, the albumin RI were similar from one breed to another. A possible explanation for the observed differences in plasma glucose would be a stress-related increase, as described previously in felines. 26 The level of stress is quite unpredictable and difficult to assess by observation. 27 However, it was likely to be the same in all the breeds tested here because the sampling procedure was strictly identical. Moreover, to our knowledge, C and MC have not been reported to be more susceptible to stress than B and P. Higher fasting plasma glucose concentrations were reported in human ethnic groups that also showed a higher susceptibility to diabetes mellitus 9,28 and a predisposition of Burmese cats to diabetes mellitus has also been described. 13,14 It could be of interest to compare the RI established in the present study with those available in the veterinary literature. The only original publication regarding RI for the same variables is available for domestic shorthair cats (DSH) cats. 19 The upper limit of the RI for plasma creatinine, glucose, and total proteins was 207 mmol/l (2.3 mg/dl), 8.2 mmol/l (148 mg/dl), and 85 g/l. For example, if the DSH upper limit for plasma creatinine had been applied, 13/131 (10%) of B tested in the present study would have been considered azotemic. Such a comparison, however, should be performed with caution because of the lower number (n 5 95) of animals and the different methodology for RI calculation. A linear mixed effects model was used here to assess the effect of covariables on the tested plasma analytes. The owner was considered as a random effects factor. The owner effect should indeed be treated as a random effects factor as a sample of owners has been drawn

9 Reference Intervals for Cats 817 within a population of owners. It is obvious that the owner may impact the environmental conditions of cats and that response variables for cats belonging to 1 owner could be more related than those of cats from different owners. As the breed effect is adjusted on the owner effect, the difference of number of owners between breeds may not lead to bias. Indeed, the statistical results demonstrated that the breed effect was evidenced for 9/13 variables when the owner effect was taken into account. In other words, the breed effect was not confounded with the owner effect. Another objective of this study was to assess the effects of other covariables (ie, age, BW, sex, and housing conditions) on plasma values. Interestingly, the linear mixed effects model also showed statistically significant interactions between the breed effect, and the effects of age, BW, and housing conditions, but not sex. A statistically significant interaction between the breed and a covariable means that the effect of this covariable on the analyte tested is dependent on the feline breed. For example, for plasma creatinine concentration, when all the other covariables are unchanged, a difference of 2 kg of BW between 2 individuals will lead to a difference in the basal plasma creatinine concentration of about 40 (0.45) and 8 (0.09) mmol/l (mg/dl) in B and P, respectively. A statistically significant effect of age, sex, BW, and housing conditions was observed for 9/13, 9/13, 7/13, and 3/13 plasma variables, respectively. However, when the slope coefficients and the differential effects were inspected, it could be seen that these effects would be clinically irrelevant except for the effect of BW on plasma creatinine in B. In conclusion, breed-dependent RI should be considered in feline clinical pathology, at least for some of the plasma analytes tested, such as creatinine, total proteins, and glucose. Age, sex, and housing conditions appear to be minor factors of variations of plasma variables in the 4 breeds tested. Conversely, BW was shown to affect plasma creatinine, especially in B. Further investigations are needed to understand the mechanisms underlying these differences. Footnotes a Veterinary Clinic Quai d Orleans, Tours, France b National Veterinary School of Toulouse, Toulouse, France c National Veterinary School of Nantes, Nantes, France d Sarstedt, Nu mbrecht, Germany e Vitros 250 chemistry system, Ortho-Clinical Diagnostics, Raritan, NJ Acknowledgments The study was supported in part by Royal Canin SAS, Centre de Recherches, Aimargues, France. The authors thank Mrs C. Moncelon and A. Dutech, Drs P. Brillard, A. Gambier, and C. Arpaillange for technical assistance. References 1. Moesgaard SG, Holte AV, Mogensen T, et al. Effects of breed, gender, exercise and white-coat effect on markers of endothelial function in dogs. Res Vet Sci 2007;82: Harper EJ, Hackett RM, Wilkinson J, et al. Age-related variations in hematologic and plasma biochemical test results in Beagles and Labrador Retrievers. J Am Vet Med Assoc 2003;223: Tryfonidou MA, Holl MS, Vastenburg M, et al. Hormonal regulation of calcium homeostasis in two breeds of dogs during growth at different rates. J Anim Sci 2003;81: Sato K, Agoh H, Kaneshige T, et al. Hypercholesterolemia in Shetland Sheepdogs. J Vet Med Sci 2000;62: Pedersen HD, Olsen LH, Arnorsdottir H. Breed differences in the plasma renin activity and plasma aldosterone concentration of dogs. Zentralbl Veterinarmed A 1995;42: Downs LG, Bolton CH, Crispin SM, et al. Plasma lipoprotein lipids in five different breeds of dogs. Res Vet Sci 1993;54: Reimers TJ, Lawler DF, Sutaria PM, et al. Effects of age, sex, and body size on serum concentrations of thyroid and adrenocortical hormones in dogs. Am J Vet Res 1990;51: Feeman WE III, Couto CG, Gray TL. Serum creatinine concentrations in retired racing Greyhounds. Vet Clin Pathol 2003;32: Whincup PH, Gilq JA, Owen CG, et al. British South Asians aged years have higher fasting glucose and insulin levels than Europeans. Diabet Med 2005;22: Jones CA, McQuillan GM, Kusek JW, et al. Serum creatinine levels in the US population: Third National Health and Nutrition Examination Survey. Am J Kidney Dis 1998;32: Menotti-Raymond M, David VA, Pflueger SM, et al. Patterns of molecular genetic variation among cat breeds. Genomics 2008;91: Pesteanu-Somogyi LD, Radzai C, Pressier BM. Prevalence of feline infectious peritonitis in specific cat breeds. J Feline Med Surg 2006;8: McCann TM, Simpson KE, Shaw DJ, et al. Feline diabetes mellitus in the UK: The prevalence within an insured cat population and a questionnaire-based putative risk factor analysis. J Feline Med Surg 2007;9: Lederer R, Rand JS, Jonsson NN, et al. Frequency of diabetes mellitus and breed predisposition in domestic cats in Australia. Vet J 2009;179: Gunn-Moore DA, Dodkin SJ, Sparkes AH. An unexpectedly high prevalence of azotemia in Birman cats (letter). J Feline Med Surg 2002;4: Driscoll CA, Menotti-Raymond M, Roca AL, et al. The Near Eastern origin of cat domestication. Science 2007;27: Lipinski MJ, Froenicke L, Baysac KC, et al. The ascent of cat breeds: Genetic evaluations of breed and worldwide randombreed populations. Genomics 2008;91: Reynolds BS, Boudet KG, Faucher MR, et al. Comparison of a new device for blood sampling in cats with a vacuum tube collection system plasma biochemistry, haematology and practical usage assessment. J Feline Med Surg 2007;9: Reynolds BS, Boudet KG, Germain CA, et al. Determination of reference intervals for routine plasma biochemistry in clinically normal adult domestic shorthair cats using a dry-slide technology analyzer. Am J Vet Res 2008;69: Clinical and Laboratory Standards Institute (CLSI). Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory; Approved Guidelines, CLSI document C28-A3, Vol. 28, No. 3, 3rd ed. Wayne, PA: Clinical and Laboratory Standard Institute; 2008.

10 818 Reynolds et al 21. Lahti A, Petersen PH, Boyd JC, et al. Partioning of non- Gaussian-distributed reference data into subgroups. Clin Chem 2004;50: R Development Core Team. R: A Language and Environment for Statistical Computing. ISBN Vienna, Austria: R Foundation for Statistical Computing; Available at: Solberg HE. Establishment and use of reference values. In: Burtis CA, Ashwood ER, eds. Tietz Textbook of Clinical Chemistry, 3rd ed. Philadelphia, PA: WB Saunders Co; 1999: Clinical and Laboratory Standard Institute. How to Define and Determine Reference Intervals in the Clinical Laboratory; Approved Guideline, CLSI document C28-A2, 2nd ed. Wayne, PA: Clinical and Laboratory Standard Institute; Kelley-Hedgepeth A, Lloyd-Jones DM, Colvin A, et al. Ethnic differences in C-reactive protein concentrations. Clin Chem 2008;54: Rand JS, Kinnaird E, Baglioni A, et al. Acute stress hyperglycemia in cats is associated with struggling and increased concentrations of lactate and norepinephrine. J Vet Intern Med 2002;16: Sparkes AH. Cats, diabetes and stress! J Feline Med Surg 1999;1: Borrell LN, Crawford ND, Dallo FJ, et al. Self-reported diabetes in Hispanic subgroup, non-hispanic black and non-hispanic white populations: National Health Interview Survey, Public Health Rep 2009;124: Conversion factors for SI units to conven- Appendix 1. tional units. Analyte SI Unit Multiply By For Conventional Units Glucose mmol/l mg/dl Urea a mmol/l 5.99 mg/dl Creatinine mmol/l mg/dl Total proteins g/l 0.1 g/dl Albumin g/l 0.1 g/dl Sodium mmol/l 1 meq/l Potassium mmol/l 1 meq/l Chloride mmol/l 1 meq/l Total CO 2 mmol/l 1 meq/l Calcium mmol/l 4 mg/dl Phosphate mmol/l 3.1 mg/dl ALT U/L 1 U/L ALP U/L 1 U/L ALT, alanine aminotransferase; ALP, alkaline phosphatase. a Multiply urea in SI units (mmol/l) by 2.8 to obtain blood urea nitrogen (BUN) in conventional units (mg/dl).