Revisiting myocardial necrosis biomarkers: assessment of the effect of conditioning therapies on infarct size by kinetic modelling

Figures

• 

  Figure 1. Diagnostic plots of creatine phosphokinase (CK, left), cardiac troponin I (cTnI, middle) and creatine phosphokinase muscle-brain (CK-MB) kinetic models. For each biomarker: observed versus model-predicted for population parameters (top-left) and for individual parameters (top-right). The line is the first bisector line. White and black circles are observed/predicted biomarker level couples for learning and validation subsets, respectively. population weighted residuals versus population predicted biomarker levels (bottom-left); individual weighted residuals versus individual predicted biomarker levels (bottom-right). White and black circles are predicted biomarker level couples and weighted residuals of learning and validation subsets, respectively. Typical parameters describe covariate-explained variability, whereas individual parameters describe both explained and unexplained variability. Population plots of CK and cTnI show a bias for population parameters, which disappears for individual parameters. The strong similarity of results obtained from learning and validation subsets demonstrate the good predictive performance of biomarker kinetic models.

• 

  Figure 2. Prediction interval with 90% of kinetic profiles for creatine phosphokinase (CK), cardiac troponin I (cTnI) and creatine phosphokinase muscle-brain (CK-MB) kinetic models versus time. White and black circles are observed biomarker levels of learning and validation subsets, respectively.

• 

  Figure 3. Individual parameter estimates for creatine phosphokinase (CK, top), cardiac troponin I (cTnI, middle) and creatine phosphokinase muscle-brain (CK-MB, bottom) kinetic model parameters versus conditioned arm (left) and area at risk (AAR, right). For control arm, whisker boxes display, from bottom to top, 5th, 25th, 50th, 75th and 95th individual parameter value percentiles. White and grey boxes are learning and validation subset results, respectively. For AAR, circles are covariate/individual parameter couples and line is regression line with equation and coefficient of determination (R2). White and black circles are AAR/biomarker levels of learning and validation subsets, respectively.

• 

  Figure 4. Linear correlations between model-predicted area under the biomarker curve (AUC) and trapezoidal rule AUC for creatine phosphokinase (CK, left), cardiac troponin I (cTnI, middle) and creatine phosphokinase muscle-brain (CK-MB). All correlation coefficients are above 99%. White and black circles are trapezoid/model-estimated AUC of learning and validation subsets, respectively.