Abstract
BACKGROUND: This study simulates how assay-specific bias influences the diagnostic performance of 0/1-h accelerated diagnostic protocols (ADPs) for 3 different high-sensitivity cardiac troponin (hs-cTn) assays. METHODS: We included 1493 patients presenting with chest pain. hs-cTnT (Roche Diagnostics), hs-cTnI from Abbott Diagnostics (hs-cTnI-A), and Siemens Healthineers (hs-cTnI-S) were measured at admission. The absolute total error observed in a state-of-the-art EQA study were added to the admission concentrations, producing 6 new variables being adjusted for maximum possible bias (if analytical variation is 0) (+biasmean, +biasmax95%CI, +biasmin95%CI, -biasmean, -biasmax95%CI, -biasmin95%CI). The influence of this "worst-case scenario" bias was compared after calculating sensitivity, specificity, negative and positive predictive values, and rule-out proportion for 30-day myocardial infarction or death for the observed and bias-adjusted hs-cTn concentrations. RESULTS: For 0-h rule-out, hs-cTnI-S and hs-cTnT had a sensitivity of >99.0%, compared to 97.7% for hs-cTnI-A. After adding the bias, sensitivity was unchanged for hs-cTnI-S (99.5%), but lower for hs-cTnT (95.5%), and hs-cTnI-A (96.2%). For the 0-/1-h algorithm, adding bias reduced sensitivity to 95.5% for hs-cTnT, while both hs-cTnI algorithms were unchanged (100.0%). Rule-out proportions for 0 h ranged from 0% to 60.0% for hs-cTnT, 28.2%-62.7% for hs-cTnI-A, and 3.5%-35.5% for hs-cTnI-S. For the 0-/1-h algorithm, ranges were 57.7%-75.8% (hs-cTnT), 52.8%-67.5% (hs-cTnI-A), and 45.7%-61.2% (hs-cTnI-S). CONCLUSION: Analytical bias of hs-cTn assays affects the clinical rule-out rate of the 0/1-h ADPs more than the diagnostic sensitivity. Bias may have a greater influence on the proportion of patients requiring hospital admission and may contribute to the heterogeneity of the reported rule-out rates of current ADPs. ClinicalTrials.gov Registration Number: NCT02620202.