Background: The diagnosis of myocardial injury/infarction (MI) mainly relies on relative changes in cardiac troponin. However, absolute change cutoffs provide greater diagnostic sensitivity. We determined the absolute changes in high-sensitive cardiac troponin T concentrations (absΔhs-cTnT) corresponding to the main relative cutoffs (relΔhs-cTnT), using a quantile generalized additive model (qgam).
Methods: Plasma Δhs-cTnT from patients selected with a time variation of 1 to 6 hours were collected over a 6-year period. The absΔhs-cTnT-to-relΔhs-cTnT relationship was fitted using qgam, after ordered quantile-based normalization (OQN) to reduce the influence of extreme values.
Results: The qgam regression curve was nonlinear. Classifying patients (n = 9,753) above the recommended relΔhs-cTnT and predicted absΔhs-cTnT cutoffs as positive, the MI diagnosis rates were similar, and more reliable using the OQN-transformed data-based qgam, as compared to the untransformed data-based one.
Conclusions: Through an optimized qgam-based approach accounting for heavy-tailed distributions, absolute Δhs-cTnT are provided for the corresponding relative Δhs-cTnT cutoffs.