J Vet Intern Med 2006;20:1136–1142
Comparison of Canine Cardiac Troponin I Concentrations as Determined by 3 Analyzers Darcy B. Adin, Mark A. Oyama, Margaret M. Sleeper, and Rowan J. Milner Background: Recent interest in cardiac biomarkers has led to the validation of several commercial analyzers for cardiac troponin I (cTnI) evaluation in dogs; however, these analyzers have not been standardized. Hypothesis: It was hypothesized that canine plasma cTnI concentrations as determined by 3 different analyzers would be similar. Animals: Twenty-three dogs with cardiac disease were studied. Methods: Reconstituted purified canine free cTnI was diluted with canine plasma to 8 concentrations (0.01, 0.1, 0.78, 1.56, 3.13, 6.25, 12.5, and 25 ng/mL), for analysis by 3 analyzers, the Biosite Triage Meter, the Dade-Behring Stratus, and the Beckman-Coulter Access AccuTnI. Plasma samples from 23 dogs with cardiac disease were also analyzed for cTnI concentrations on all analyzers. Results: Troponin I concentrations in sick dogs were ,0.05–5.72 ng/mL (Biosite), 0.02–11.1 ng/mL (Access), and 0.02– 9.73 ng/mL (Stratus). Analyzer results were highly correlated with each other (r 5 0.97 to 1.0 for purified dilutions, r 5 0.61 to 0.89 for samples from dogs); however, the limits of agreement were wide for both purified dilutions and clinical samples. Recovery was highest for the Access (334–1467%) and lowest for the Biosite (38–60%); Stratus 52–233%. Analyzer variability was lowest for the Access (1.2–10.4%) and highest for the Stratus (4.8–33.6%); Biosite 2.8–16.5%. Conclusions and Clinical Importance: Results from this study suggest that although canine cTnI values obtained from the Biosite, Stratus, and Access analyzers are closely correlated, they cannot be directly compared with each other. In the absence of a gold standard none of the analyzers can be considered more correct than the others. Key words: Access; Biosite; Cardiac biomarker; Method bias; Recovery; Stratus.
ardiac troponin I (cTnI) is a highly sensitive and specific blood marker for the noninvasive diagnosis of increased cardiomyocyte permeability and, as such, has become very important for the early detection of myocardial infarction in human beings.1–6 Although myocardial infarction is uncommon in veterinary medicine, some dogs and cats with conditions involving myocardial damage, such as hypertrophic cardiomyopathy, unclassified cardiomyopathy, dilated cardiomyopathy, subaortic stenosis, mitral valve degeneration, gastric dilatation-volvulus, blunt thoracic trauma, and babesiosis have been found to have high cTnI concentrations.7–12 Cardiac troponin I analyzers have not been standardized and because each company can use a potentially different and proprietary target amino acid in the assay, values obtained on 1 analyzer may not be directly comparable to values obtained on a different analyzer.13–16 In addition, assays may have different reference ranges because of assay differences in the antibody specificity for free and complexed cTnI. The pre-
C
From the Department of Small Animal Clinical Sciences, College of Veterinary Medicine, University of Florida, Gainesville, FL (Adin); the Department of Small Animal Clinical Sciences, College of Veterinary Medicine, University of Illinois, Urbana, IL (Oyama); Department of Clinical Studies, School of Veterinary Medicine, University of Pennsylvania, Philadelphia, PA (Sleeper); and Department of Small Animal Clinical Sciences, College of Veterinary Medicine, University of Florida, Gainesville, FL (Milner). Reprint requests: Darcy B. Adin DVM, Veterinary Specialists of Rochester, 825 White Spruce Blvd, Rochester, NY 14623; e-mail:
[email protected]. Submitted January 26, 2006; Revised March 6, 2006; Accepted March 29, 2006. Copyright E 2006 by the American College of Veterinary Internal Medicine 0891-6640/06/2005-0011/$3.00/0
dominant circulating form of cardiac troponins in human beings is cTnI complexed to cardiac troponin C; however, the predominant circulating form in dogs is not known.17–19 Several studies in human medicine have compared analyzers, and although results obtained on different analyzers correlate well, a large (up to 20-fold) difference in the absolute values obtained between some machines has been observed.13–15,17,20,21 The use of and the potential applications for cTnI analysis are growing in veterinary medicine; however, the inability to compare values obtained on different analyzers has limited collaborative studies between institutions to this point. Reference ranges for dogs have been developed for several different analyzers using the 95th percentile as the upper end of the range, and the established ranges are similar for 3 of these analyzers (Biosite Triage Metera ,0.05–0.12 ng/mL, Dade-Behring Stratusb ,0.03–0.07 ng/mL, and Beckman-Coulter Access AccuTnIc 0.01–0.11 ng/mL).11,22,23 Each of these analyzers uses a 2-site sandwich immunoassay to detect free and complexed cTnI. They all use murine monoclonal antibodies, while the Biosite Triage Metera also uses goat polyclonal antibodies. The target regions for antibody binding for the Beckman-Coulter Access AccuTnI are the 24–40 and 41–49 amino acid sequences of the stable N-terminal region of cTnI. Target amino acid sequences are considered proprietary information for the Dade-Behring Stratus and the Biosite Triage Meter. The similarity among established reference ranges for dogs suggests that the target amino acid for each of these assays may be similar enough to allow meaningful comparison of samples among analyzers. The objective of this study was to compare canine cTnI concentrations as determined by 3 analyzers: the Biosite Triage Meter, the Dade-Behring Stratus, and the Beckman-Coulter Access AccuTnI.
Troponin Analyzers
1137
Table 1. Correlation of actual results with expected results, recovery, and precision (interassay coefficient of variation) for each analyzer for purified canine free cTnI.
Biosite Stratus Access
Regression Equation
Comparison to Expected Results (P Value and r)
Recovery (% of Expected)
Interassay CoV (%)
Y 5 20.3086 + 0.6073X Y 5 0.3809 + 0.5342X Y 5 20.1378 + 3.5297X
P , .0001, r 5 1.0 P , .0001, r 5 0.99 P , .0001, r 5 1.0
38–60 52–233 334–1467
2.8–16.5 4.8–33.6 1.2–10.4
Materials and Methods Purified Canine cTnI Reconstituted purified free (not complexed to cardiac troponin T or C) canine cTnI with purity .98% as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGEd) was reconstituted with a urea-Tris buffer as recommended by the manufacturer and diluted to 8 concentrations: 25, 12.5, 6.25, 3.13, 1.56, 0.78, 0.1, and 0.01 ng/mL. Canine plasma with undetectable cTnI concentration (,0.05 ng/mL on the Biosite) was used as the diluent. Each diluted concentration was divided into 3 aliquots of 3 mL each, and aliquots were frozen at 270uC until analysis. Concentrated cTnI was diluted at the University of Florida (where the Biosite is housed), and aliquots were shipped on dry ice to the University of Pennsylvania (where the Stratus is housed) and the University of Illinois (where the Access is housed). Up to 5 freeze-thaw cycles have been shown not to have any significant effect on the concentration obtained.24 Three 1-mL samples from each aliquot were run separately on each of the 3 analyzers.
mean of measured values] 3 100) were reported for each analyzer. The fold difference between the results of 2 analyzers was calculated as the mean concentration for the analyzer returning the highest value divided by the mean concentration for the analyzer returning the lower value. This is expressed as a range for the fold differences over all the concentrations tested. Linear regression and Pearson’s correlation test (after log transformation) were also used to investigate relationships among analyzers by plotting each analyzer against the others for both purified dilutions and clinical samples. Method agreements were investigated with Bland-Altman analysis and expressed as percentage of the averages when an increase in variability of the differences was noted as the magnitude of the measurement increased.25 Because the lower limit of detection for the Biosite is reported as cTnI ,0.05 ng/mL, such readings were analyzed as .05 ng/mL for statistical purposes. P , .05 was considered significant and statistical software was used for analysis.e
Results Purified Canine cTnI
Dogs with Heart Disease Blood was collected from 23 dogs with advanced heart disease or arrhythmias in an attempt to obtain a wide range of naturally occurring cTnI elevations.11 Six milliliters of blood from each dog was divided into 3 blood collection tubes (1 Ethylenediaminetetraacetic Acid (EDTA) and 2 heparinized). Plasma was separated after centrifugation at 3,000 rpm for 8 minutes. Samples were frozen within 4 hours of collection at 270uC until they were shipped on dry ice to each institution for analysis. One aliquot from each dog was run on each of the 3 analyzers.
Statistical Analysis Kolmogorov-Smirnov testing for normality was performed and repeated after log transformation if raw data failed normality testing. Linear regression and Pearson’s correlation test were performed on log transformed data for investigation of relationships between expected purified free canine cTnI concentrations (dilution values) and measured cTnI concentrations for each of the 3 analyzers. Recovery (mean cTnI concentration reported/cTnI value expected based on the calculated dilution 3 100) as well as interassay precision (coefficient of variation 5 [standard deviation/
All purified cTnI dilutions were readable on all 3 analyzers except the 25 ng/mL dilution for which the reported value was too high to read on the Access. Interpretation of data for all 3 analyzers revealed normal distribution. Measured values of purified canine free cTnI were highly correlated with expected concentrations based on the dilutions for all 3 analyzers (Table 1). Recovery was highly variable, with the Biosite returning the lowest recoveries and the Access returning the highest recoveries (Table 1). Analyzer precision, as assessed by the interassay coefficient of variation (CoV) was lowest for the Access and highest for the Stratus. The relationships among analyzer results determined by linear regression are shown in Figure 1. Analyzer results for purified cTnI dilutions were highly correlated with each of the other analyzers (Table 2). Table 2 also shows the bias (mean difference among analyzers) and limits of agreement (61.96 standard deviations) among analyzers as well as the fold difference for each analyzer compared with the others. Although all correlation
Table 2. Correlation, measures of agreement, and fold difference for each analyzer as compared with the other 2 for purified cTnI dilutions.
Stratus versus Biosite Access versus Stratus Access versus Biosite
P Value
r Value
95% CI for r
Bias (%)
Limits of Agreement (61.96 SD) (%)
Fold Difference
P , .0001 P , .0001 P , .0001
0.9734 0.9966 0.984
0.85–0.99 0.98–1.0 0.89–1.0
23.7 133.4 146.4
269.9–117.3 114.5–152.2 103.7–189.1
0.8–2.2 3.7–6.3 2.9–8.9
CI, confidence interval; SD, standard deviation.
1138
Adin et al
Fig 1. Linear regression plots for each analyzer as compared with the other analyzers by means of dilutions of purified canine free cTnI. Axis values represent ng/mL of cTnI. Ninety-five percent confidence intervals are shown by the hatched lines on either side of the regression lines. (A) Access versus Biosite, (B) Access versus Stratus, and (C) Stratus versus Biosite.
coefficients were high, the limits of agreement and fold difference were variably large when comparing the analyzers. These measures of agreement were best between the Stratus and the Biosite. Because standard Bland-Altman plots indicated that the variability of the
Fig 2. Bland-Altman plots for purified free canine cTnI dilutions. The y-axis is expressed as the percentage of the averages. (A) Access versus Biosite, (B) Access versus Stratus, and (C) Stratus versus Biosite.
differences increased as the magnitude of the measurement increased, the differences were plotted as the percentage of the averages (Figure 2).25 Figure 3 illustrates the differences in cTnI concentrations obtained on each analyzer at each dilution and shows how the Access
Troponin Analyzers
1139
other analyzers; however, correlation coefficients were lower than for the purified dilutions (Table 3, Figure 4). Correlation and measures of agreement (expressed as the percentage of the averages of the 2 methods) were best between the Stratus and the Access; however, the limits of agreement and fold differences were large for all comparisons (Table 3). Because standard Bland-Altman plots indicated that the variability of the differences increased as the magnitude of the measurement increased, the differences were plotted as the percentage of the averages (Figure 5).25 The x-axes of Figure 5A–C (average of the 2 methods) were log transformed to visually spread data points. The Access and the Stratus were most closely correlated, with the Biosite returning lower values (Figure 6).
Discussion Fig 3. cTnI concentration obtained on each analyzer for each purified canine free cTnI dilution.
returned higher values than the Stratus and the Biosite, especially at higher concentrations.
Dogs with Heart Disease The 23 dogs were 8 male intact, 8 male castrated, and 7 female spayed dogs. The dogs were 8.1 6 2.5 years of age and were 3 Labrador retrievers; 2 each of Rottweiler, Doberman Pinscher, Newfoundland, Maltese, and Golden Retriever; one each of English Bulldog, Dalmatian, American Pit Bull Terrier, Dachshund, German Shepherd Dog, Neapolitan Mastiff, Pug, American Cocker Spaniel, and Cavalier King Charles Spaniel; and 3 mixed-breed dogs. Diagnoses included myxomatous mitral valve degeneration and regurgitation (9), dilated cardiomyopathy (5), subaortic stenosis (3), heartworm disease (2), mitral valve dysplasia (1), Wolf-Parkinson-White Syndrome (1), uroabdomen with ventricular tachycardia (1), and postcardiopulmonary resuscitation (1). Eight dogs were in congestive heart failure, 3 dogs had ventricular tachycardia, and 2 dogs had atrial fibrillation. Concentrations of cTnI from clinical patients ranged from ,0.05–5.72 ng/mL on the Biosite, 0.02–11.1 ng/mL on the Access, and 0.02– 9.73 ng/mL on the Stratus. Data from the dogs for all 3 analyzers failed normality testing but had an approximate Gaussian distribution after log transformation. Results among analyzers were significantly correlated with each of the
This study is the first to compare 3 cTnI analyzers by means of both purified canine free cTnI and plasma obtained from clinical canine patients. Results from all 3 analyzers were significantly correlated with each other with the highest correlations found when purified free cTnI was used. Despite close correlations, however, measures of agreement were poor among all analyzers, and we found up to a 19-fold difference among analyzer results. Bias, as reported by Bland-Altman analysis, may be used as a correction factor if the limits of agreement (61.96 standard deviations) can be accepted as the potential error in each measurement. For both purified dilutions as well as clinical patient samples, the limits of agreement were considered too large to be clinically acceptable. Even for the smallest range obtained of 115– 152% (Access versus Stratus for purified dilutions), a possible error of up to 38% is considered too great for clinical comparative use. This conclusion is made in light of the narrow reference ranges that have been established in dogs (Biosite ,0.05–0.12 ng/mL, Stratus ,0.03–0.07 ng/mL, and Access 0.01–0.11 ng/mL).11,22,23 Although the exclusion of the few discrepant data points on these graphs would have made the standard deviations for Bland-Altman analysis smaller, these values were included because they represented clinical patients and the intent of the study was to provide clinically applicable information. These points were not obviously because of technical error and are assumed to be the result of biologic or analyzer variability. For purified dilutions, we found that the Biosite and Stratus produced similar cTnI values, with the Access
Table 3. Correlation, measures of agreement, and fold difference for each analyzer as compared with the other 2 for clinical canine patient samples.*
Stratus versus Biosite Access versus Stratus Access versus Biosite
P Value
r Value
95% CI for r
Bias (%)
Limits of Agreement (61.96 SD) (%)
Fold Difference
P , .0018 P , .0001 P , .0019
0.6157 0.8923 0.6114
0.27–0.82 0.76–0.95 0.26–0.82
53.6 19.7 70.1
2100.0–207 280.0–119.5 295.4–235.6
0.01–8.3 0.3–19.0 0.03–15.2
CI, confidence interval; SD, standard deviation. * Measures of agreement are expressed as the percentage of the averages.
1140
Adin et al
Fig 4. Linear regression plots after log transformation for each analyzer as compared with the other analyzers by means of plasma samples from diseased dogs. Ninety-five percent confidence intervals are shown by the hatched lines on either side of the regression lines. (A) Access versus Biosite, (B) Access versus Stratus, and (C) Stratus versus Biosite.
producing much higher values. For clinical patient samples, we found that the Access and the Stratus produced similar cTnI values, with the Biosite producing the lowest values. Because a gold standard is not
available for cTnI analysis, no analyzer can be considered more correct than others. The differences in results between purified free cTnI dilutions and clinical patient samples may be explained by differing degrees of affinity of each analyzer for free cTnI and cTnI complexed to cardiac troponin T or C. The predominant circulating form of cTnI after release from myocardial cells is not known in dogs, although it is suspected to be cTnI complexed to cardiac troponin C in human beings.17–19 The use of purified complexed cTnI (to cardiac troponin T or C) might have been useful in this study; however, similar differences in analytical methods between purified cTnI and patient cTnI values have been noted in human medicine.20 Although the lack of a common calibrator among analyzers may be a major reason for the large differences in results seen both in this study and in human studies, other factors may be at play. Additional explanations for the differences among analyzers as well as the differences among purified samples and clinical samples include different target amino acids for each analyzer, the variable complexed or noncomplexed forms of cTnI recognized by the antibodies in each assay, and modification (degradation or phosphorylation) of circulating cTnI after myocardial injury that variably influences the detection of cTnI by different analyzers.1,19,20 This latter point may explain why the fold differences among analyzers were greater at higher cTnI concentrations; perhaps the Access is better able to detect modified forms of cTnI after cardiac injury than the Biosite and Stratus. Such cTnI modifications have been shown in rat models and in human beings; however, this hypothesis has not been investigated in dogs. Although large differences in results were obtained when the 3 analyzers were compared with each other, all 3 analyzers were able to detect purified free canine cTnI and produced values that were highly correlated with expected cTnI concentrations in a linear manner. Assay precision (variability as assessed by coefficient of variation) was different for each analyzer with only the Access having #10% variability at all tested concentrations. Recovery was highly variable among analyzers, with the Biosite recovering approximately half of the expected cTnI, and the Access recovering up to 1,500% of expected cTnI. The reason for these findings is not known; however, possible explanations include differences in the target amino acid sequence, the use of monoclonal versus polyclonal antibodies resulting in under-recovery or over-recovery, and possible positive or negative interfering plasma substances.26 False positive cTnI concentrations have been reported in human medicine because of heterophilic (antimouse) antibodies, rheumatoid factor, and binding of modified cTnI to immunoglobulin G, resulting in delayed clearance.27 The cTnI elevations in these clinical cases were determined to be false positives based on the clinical scenario, cardiac troponin T testing, and use of several cTnI analyzers.27 Similar investigations into the possibility of false positive cTnI concentrations in dogs have not been performed, and therefore this remains a concern based on the overestimation and high
Troponin Analyzers
1141
Fig 6. cTnI concentration obtained on each analyzer for all 23 dogs.
Fig 5. Bland-Altman plots for diseased dogs. The x-axis has been log transformed and the y-axis is expressed as the percentage of the averages. (A) Access versus Biosite, (B) Access versus Stratus, and (C) Biosite versus Stratus.
recovery values for the Access in this study. Likewise, possible explanations for the low recovery of cTnI by the Biosite were not investigated and raise concern that the Biosite may produce false negative results. It has been documented that modification of cTnI either within the myocardium or in the circulation after myocardial infarction occurs because of degradation (cleavage of the C-terminal or N-terminal portions of the molecule) or phosphorylation, which may result in the inability of the antibody to bind to the targeted epitope. These modifications occur rapidly and are affected by the extent of complexing to troponin C and T. The rate of release from myocardium, extent of cTnI modification, and rate of degradation affect the ability of the antibody to bind to the troponin molecule, emphasizing the
biologic complexity that hinders analyzer standardization. This phenomenon is compounded by the different antibodies used by different manufacturers. Most analyzers target the more stable N terminal; however, differences in epitope selection as well as monoclonal or polyclonal differences likely exist among analyzers, resulting in differing specificities between analyzers. The amino acid sequence targeted by the Access has been published; however, the sequences targeted by the Biosite and Stratus are considered proprietary information.24 Further studies are indicated to investigate possible negative or positive interfering substances for each analyzer and to determine if cTnI circulates in a complexed form in dogs as occurs in human beings. Although we demonstrated that the Access, Biosite, and Stratus analyzers all detected canine cTnI, we found clinically significant differences in absolute concentrations obtained for purified dilutions of free cTnI as well as for clinical canine patients. Until cTnI analyzer manufacturers standardize the antibodies used in the assays, reference ranges should be developed for each analyzer used in dogs, and direct comparison of results from analyzers cannot be recommended.
Footnotes a
Biosite Triage Meter, Biosite Inc, San Diego, CA Dade-Behring Stratus, Dade-Behring, Newark, DE c Beckman-Coulter Access AccuTnI, Beckman Coulter Inc, Fullerton, CA d SDS-PAGE, Advanced Immunochemical Inc, Long Beach, CA e MedCalc Software, Version 7.5, Mariakerke, Belgium b
Acknowledgments The authors would like to thank Marc Salute and James Van Gilder for technical and statistical assistance.
1142
Adin et al
Funding for this study was provided by an intramural grant from the College of Veterinary Medicine, University of Florida, and an interdepartmental grant from the School of Veterinary Medicine, University of Pennsylvania.
References 1. Wu AHB, Feng YJ, Moore R, et al. Characterization of cardiac troponin sub-unit release into serum after acute myocardial infarct and comparison of assays for troponin T and I. Clin Chem 1998;44:1198–1208. 2. Ooi DS, Isotalo PA, Veinot JP, et al. Correlation of antimortem serum creatine kinase, creatine kinase-MB, troponin I, and troponin T with cardiac pathology. Clin Chem 2000; 46:338–344. 3. Antman EM, Tanasijevic MJ, Thompson B, et al. Cardiacspecific troponin I levels to predict the risk of mortality in patients with acute coronary syndromes. N Engl J Med. 1996;335: 1342–1349. 4. Babuin L, Jaffe AS. Troponin: The biomarker of choice for the detection of cardiac injury. CMAJ 2005;173:1191–1202. 5. Apple FS, Parvin CA, Buechler KF, et al. Validation of the 99th percentile cutoff independent of assay imprecision (CV) for cardiac troponin monitoring for ruling out myocardial infarction. Clin Chem 2005;11:2198–2200. 6. Hasic S, Kiseljakovic E, Jadric R, et al. Cardiac troponin I: The gold standard in acute myocardial infarction diagnosis. Bosn J Basic Med Sci 2003;3:41–44. 7. Herndon WE, Kittleson MD, Sanderson K, et al. Cardiac troponin I in feline hypertrophic cardiomyopathy. J Vet Intern Med 2002;16:558–564. 8. Connolly DJ, Cannata J, Boswood A, et al. Cardiac troponin I in cats with hypertrophic cardiomyopathy. J Feline Med Surg 2003;5:209–216. 9. Lobetti R, Dvir E, Pearson J. Cardiac troponins in canine babesiosis. J Vet Intern Med 2002;16:63–68. 10. Schober K, Cornand C, Kirbach B, et al. Serum cardiac troponin I and cardiac troponin T concentrations in dogs with gastric dilation-volvulus. J Am Vet Med Assoc 2002;221:382–388. 11. Oyama MA, Sisson DD. Cardiac troponin I concentration in dogs with cardiac disease. J Vet Intern Med 2004;18:831– 839. 12. Schober KE, Birbach B, Oechtering G. Noninvasive assessment of myocardial cell injury in dogs with suspected cardiac contusion. J Vet Card 1999;1:17–26.
13. Dasgupta A, Chow L, Nazareno L, et al. Performance evaluation of a new chemiluminescent cardiac troponin I assay. J Clin Lab Anal 2000;14:224–229. 14. Apple FS, Christenson RJ, Valdes R, et al. Simultaneous rapid measurement of whole blood myoglobin, creatine kinase MB and cardiac troponin I by the Triage Cardiac Panel for detection of myocardial infarction. Clin Chem 1999;45:199–205. 15. Saadeddin SM, Habbab MA, Siddieg HH, et al. Evaluation of 6 cardiac troponin assays in patients with acute coronary syndrome. Saudi Med J 2003;24:1092–1097. 16. Tate JR, Heathcote D, Rayfield J, et al. The lack of standardization of cardiac troponin I assay systems. Clin Chim Acta 1999;284:141–149. 17. Mo¨ckel M, Heller G, Berg K, et al. The acute coronary syndrome diagnosis and prognostic evaluation by troponin I is influenced by the test system affinity to different troponin complexes. Clin Chim Acta 2000;293:139–155. 18. Katrukha AG, Bereznikova AV, Filatov VL, et al. Degradation of cardiac troponin I: Implication for reliable immunodetection. Clin Chem 1998;44:2433–2440. 19. Labugger R, Organ L, Collier C, et al. Extensive troponin I and T modification detected in serum from patients with acute myocardial infarction. Circulation 2000;102:1221–1226. 20. Shi Q, Ling M, Zhang X, et al. Degradation of cardiac troponin I in serum complicates comparisons of cardiac troponin I assays. Clin Chem 1999;45:1018–1025. 21. Hafner G, Peetz D, Erbes H, et al. Comparison of diagnostic performance of cardiac troponin I on the IMMULITE system with other automated troponin I assays in minor myocardial damage. Scand J Clin Lab Invest 2001;61:227–236. 22. Sleeper MM, Clifford CA, Laster LL. Cardiac troponin I in the normal dog and cat. J Vet Intern Med 2001;15:501–503. 23. Adin DB, Berger K, Engel C, et al. Cardiac troponin I concentrations in normal dogs and cats using a bedside analyzer. J Vet Card 2005;7:27–32. 24. Oyama MA, Solter PF. Validation of an immunoassay for measurement of canine cardiac troponin I. J Vet Card 2004;6:17–24. 25. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8:135–160. 26. Eriksson S, Ilva T, Becker C, et al. Comparison of cardiac troponin I immunoassays variably affected by circulating autoantibodies. Clin Chem 2005;51:848–855. 27. Zaninotto M, Mion M, Altinier M, et al. Quality specifications for biochemical markers of myocardial injury. Clin Chim Acta 2004;346:65–72.