Abstract BACKGROUND We compared the 99th percentile reference intervals with 3 modern cardiac troponin assays in a single cohort and tested the hypothesis that the same individuals will be identified as above the cutoff and that differences will be explained by analytical imprecision. METHODS Blood was collected from 2005 apparently healthy blood donors. Cardiac troponin was measured with Abbott Architect STAT high sensitive troponin I, Beckman Coulter Access AccuTnI+3, and Roche Elecsys troponin T highly sensitive assays. RESULTS The 99th percentile cutoff limits were as follows: Abbott cardiac troponin I (cTnI) 28.9 ng/L; Beckman Coulter cTnI 31.3 ng/L; and Roche cardiac troponin T (cTnT) 15.9 ng/L. Correlation among the assays was poor: Abbott cTnI vs Beckman Coulter cTnI, R2 = 0.18; Abbott cTnI vs Roche cTnT, R2 = 0.04; and Beckman Coulter cTnI vs Roche cTnT R2 = 0.01. Of the results above the cutoff 50% to 70% were unique to individual assays, with only 4 out of 20 individuals above the cutoff for all 3 assays. The observed differences among assays were larger than predicted from analytical imprecision. CONCLUSIONS The 99th percentile cutoff values were in agreement with those reported elsewhere. The poor correlation and concordance amongst the assays were notable. The differences found could not be explained by analytical imprecision and indicate the presence of inaccuracy (bias) that is unique to sample and assay combinations. Based on these findings we recommend less emphasis on the cutoff value and greater emphasis on δ values in the diagnosis of myocardial infarction.