Clinical validation of an assay-independent algorithm to rule-out and rule-in acute myocardial infarction

N. Thießen (Hamburg)¹, J. Pickering (Christchurch)², C. Kellner (Hamburg)¹, P. Haller (Hamburg)¹, J. Lehmacher (Hamburg)¹, B. Toprak (Hamburg)¹, R. Twerenbold (Hamburg)¹, N. A. Sörensen (Hamburg)³, C. Pemberton (Christchurch)⁴, R. Troughton (Christchurch)⁴, M. Richards (Christchurch)⁴, S. Sharif (Hamilton)⁵, A. Worster (Hamilton)⁵, M. Than (Chrischurch)⁶, P. Kavsak (Hamilton)⁷, J. T. Neumann (Hamburg)^1
1Universitäres Herz- und Gefäßzentrum Hamburg Klinik für Kardiologie Hamburg, Deutschland; ²Christchurch Hospital, Department of Medicine University of Otago Christchurch and Emergency Department Christchurch, Neuseeland; ³Universitäres Herz- und Gefäßzentrum Hamburg Klinik und Poliklinik für Kardiologie Hamburg, Deutschland; ⁴University of Otago, Department of Medicine Christchurch Heart Institute Christchurch, Neuseeland; ⁵McMaster University Division of Emergency Medicine Hamilton, Kanada; ⁶Christchurch Hospital, University of Otago Christchurch and Emergency Department Department of Medicine Chrischurch, Neuseeland; ⁷McMaster University Department of Pathology and Molecular Medicine Hamilton, Kanada

Background: The established European Society of Cardiology (ESC) 0/1h and 0/2h algorithms for triage of patients with suspected myocardial infarction (MI) share the limitation of requiring assay-specific high-sensitivity cardiac troponin (hs-cTn) cut-off concentrations and change criteria. We sought to evaluate the diagnostic performance of a common change criteria algorithm (3C) for hs-cTn in comparison to the ESC algorithms.

Methods: We applied the 3C algorithm in two prospective cohort studies (Biomarkers in Acute Cardiac Care (BACC), Christchurch cohort) with 3 different hs-cTn assays of patients presenting to the emergency department with suspected MI who had serial hs-cTn results available. Diagnostic metrics included sensitivity, specificity, predictive values and likelihood ratios for MI for both 3C (change criteria >|3| for under 10 ng/L, >|30|% between 10-100 ng/L and >|15|% for above 100 ng/L) and the ESC algorithms for rule-in and rule-out as well as Confusion matrices, net reclassification improvement (NRI), and effectiveness (percentage rule-in and rule-out) analyses.

Results: In total, 5011 patients with a MI prevalence of 16.12% (n = 811) were included. Both the 3C and ESC algorithms demonstrated comparable diagnostic performance in terms of sensitivity, specificity, and predictive values. Direct comparison of the algorithms via net reclassification improvement (NRI) revealed no decisive advantage for either of the two algorithms. Confusion matrices for all three assays for both 0/1h and 0/2h sampling indicated that 3C identified more more patients with a MI as rule-in who were ruled-out with the ESC. Effectiveness was higher for 3C (83.2-88.8%) versus ESC (64.4-74.5%) for hs-cTnI but not for hs-cTnT (3C 64.5-71.8% versus 72.4-80.6%).

Interpretation: The 3C algorithm provides a uniform, assay-independent alternative to the ESC algorithms, offering flexibility in timing of serial sampling.