Concordance, variance, and outliers in 4 contemporary cardiac troponin assays: implications for harmonization.

BACKGROUND Data to standardize and harmonize the differences between cardiac troponin assays are needed to support their universal status in diagnosis of myocardial infarction. We characterized the variation between methods, the comparability of the 99th-percentile cutoff thresholds, and the occurrence of outliers in 4 cardiac troponin assays. METHODS Cardiac troponin was measured in duplicate in 2358 patient samples on 4 platforms: Abbott Architect i2000SR, Beckman Coulter Access2, Roche Cobas e601, and Siemens ADVIA Centaur XP. RESULTS The observed total variances between the 3 cardiac troponin I (cTnI) methods and between the cTnI and cardiac troponin T (cTnT) methods were larger than expected from the analytical imprecision (3.0%-3.7%). The between-method variations of 26% between cTnI assays and 127% between cTnI and cTnT assays were the dominant contributors to total variances. The misclassification of results according to the 99th percentile was 3%-4% between cTnI assays and 15%-17% between cTnI and cTnT. The Roche cTnT assay identified 49% more samples as positive than the Abbott cTnI. Outliers between methods were detected in 1 patient (0.06%) with Abbott, 8 (0.45%) with Beckman Coulter, 10 (0.56%) with Roche, and 3 (0.17%) with Siemens. CONCLUSIONS The universal definition of myocardial infarction should not depend on the choice of analyte or analyzer, and the between- and within-method differences described here need to be considered in the application of cardiac troponin in this respect. The variation between methods that cannot be explained by analytical imprecision and the discordant classification of results according to the respective 99th percentiles should be addressed.