Comparison of troponin I (Abbott, Beckman Coulter, Siemens) and troponin T (Roche) high-sensitivity measurement results
Abstract
Objective: Our study is focused on comparing selected analytical characteristics of high-sensitivity methods for the measurement of cardiac troponin I (hs-cTnI) produced by Abbott, Beckman Coulter, and Siemens and cardiac troponin T (hscTnT) by Roche. Design: A prospective study aimed to validate analytical characteristics, estimate the 99th percentile, and compare the 99th percentile multiples.
Settings: Institute for Clinical and Experimental Medicine, Department of Laboratory Methods, Prague, Czech Republic Material and Methods: Analytical systems: Architect i2000 (Abbott), UniCel DxI 800 (Beckman Coulter), Cobas e801 (Roche), and ADVIA Centaur XP (Siemens). Diagnostic kits: STAT High Sensitive Troponin-I assay (Abbott; further hs-cTnI Abbott), ACCESS hsTnI (Beckman Coulter; further hs-cTnI Beckman Coulter), Elecsys hsTnT (STAT) (Roche; further hscTnT Roche), and ADVIA Centaur TNIH (Siemens; further hs-cTnI Siemens).
Linearity was assessed according to Kroll. The precision study was performed according to the EP05-A2 protocol.
Estimation of the 99th percentile was performed utilizing three following approaches: the first was a nonparametric estimation with 90% confidence interval and with elimination of outliers according to Reed; the second was a procedure according to Harrell and Davis (jackknifed method) with elimination of outliers according to Reed; the third was a nonparametric estimation of the 99th percentile (only the lower limit of 90% confidence interval was determined). The 99th percentile multiples comparability assessment, with respect to type 4a and 5 myocardial infarction definitions, was performed on clinical samples using kappa statistics.
Results: Total precision determined according to the EP05-A2 protocol did not exceed 7% over a broad concentration range for all four methods (hs-cTnI Abbott, hs-cTnI Beckman Coulter, hs-cTnT Roche, and hs-cTnI Siemens). Linearity was proved for all four methods of hs-cTn measurement; however, quadratic function was optimal for hs-cTnT Roche.
The estimation of the 99th percentile (performed for hs-cTnI Abbott, hs-cTnI Beckman Coulter and hs-cTnT Roche only) showed significant differences between men and women and complied with manufacturers' data. Determining the 99th percentile according to Harrell and Davis (outliers excluded) is a possible way to estimate the 99th percentile in reference groups with less than the recommended 299 individuals.
Comparability of the 99th percentile multiples was better between hs-cTnI Abbott and hs-cTnI Beckman Coulter; however, comparability was poor between hs-cTnI (Abbott, Beckman Coulter) and hs-cTnT Roche. Classification of the type 4a or 5 myocardial infarction depends on the diagnostic kit used.
Conclusion: We compared high-sensitivity measurement methods for troponin I (Abbott, Beckman Coulter, and Siemens) and troponin T (Roche) and proved sufficient precision of the measurement and linearity over a broad range of concentrations. We confirmed sex differences at the 99th percentile of healthy persons and tested different statistical methods for estimation of the 99th percentile and elimination of outliers.
For the classification of patients, with respect to the type 4a and 5 myocardial infarction definition, we found only a borderline comparability among results of the different types of cTn concentration measurement methods. Results obtained by the diagnostic kits are reliable from the analytical point of view, but they are not interchangeable.
The tested diagnostic kits offer reliable tools for the measurement of low concentrations of cardiac troponins enable, interpretation of small delta changes and strengthen clinical legitimacy of high-sensitivity methods of troponin concentration measurements.