Detecting patients with PMI post-CABG based on cardiac troponin-T profiles: A latent class mixed modeling approach

Ruben Deneer (Corresponding author), Astrid. G. M. van Boxtel, Arjen-Kars Boer, Mohamed A. Soliman Hamad, Natal A. W. van Riel, Volkher Scharnhorst

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

Background
Diagnosis of perioperative myocardial infarction (PMI) after coronary artery bypass grafting (CABG) is fraught with complexity since it is primarily based on a single cut-off value for cardiac troponin (cTn) that is exceeded in over 90% of CABG patients, including non-PMI patients. In this study we applied an unsupervised statistical modeling approach to uncover clinically relevant cTn release profiles post-CABG, including PMI, and used this to improve diagnostic accuracy of PMI.

Methods
In 624 patients that underwent CABG, cTnT concentration was serially measured up to 24 h post aortic cross clamping. 2857 cTnT measurements were available to fit latent class linear mixed models (LCMMs).

Results
Four classes were found, described by: normal, high, low and rising cTnT release profiles. With the clinical diagnosis of PMI as golden standard, the rising profile had a diagnostic accuracy of 97%, compared to 83% for an optimally chosen cut-off and 21% for the guideline recommended cut-off value.

Conclusion
Clinically relevant subgroups, including patients with PMI, can be uncovered using serially measured cTnT and a LCMM. The LCMM showed superior diagnostic accuracy of PMI. A rising cTnT profile is potentially a better criterion than a single cut-off value in diagnosing PMI post-CABG.
Original languageEnglish
Pages (from-to)23-29
Number of pages7
JournalClinica Chimica Acta
Volume504
DOIs
Publication statusPublished - May 2020

Keywords

  • Cardiac troponin
  • Coronary artery bypass grafting
  • Growth mixture models
  • Kinetics
  • Latent class linear mixed models
  • Perioperative myocardial infarction
  • Profiles
  • Serial measurements
  • Unsupervised statistical learning

Cite this