Population and Bayesian kinetic modelling of necrosis biomarkers to assess the effect of conditioning therapies on infarct size
David Ternant(1), Stéphanie Chadet(2), Theodora Angoulvant(1), Fabrice Prunier(3), Nathan Mewton(4), Gilles Paintaud(1), Michel Ovize(4), Denis Angoulvant(2), Fabrice Ivanes(2)
(1) Université de Tours, GICC, CHU de Tours, department of medical pharmacology, Tours, France; (2) Université de Tours, EA 4245 CDG & FHU SUPORT, Tours, France; (3) Université d’Angers, EA 3860 CRT, Angers, France; (4) Université Claude Bernard Lyon 1, INSERM U1060 CarMeN, Lyon, France
Objectives: Infact size is a major predictor of subsequent cardiovascular event following ST-segment elevation myocardial infarction (STEMI). It is frequently used in clinical trials focused on cardioprotection. Imaging techniques, as cardiac magnetic resonance imaging, accurately measure infarct size but are of limited availability. Therefore, prediction of infarct size is often assessed using repeated blood sampling and the estimation of area under the concentration curve (AUC) of biomarkers, as total creatine phosphokinase (CK), myocardal band CK (CK-MB) and troponin I (cTnI). However, the performance of AUC to estimate infarct size is limited, because is blurred by interindividual variability in input, distribution and elimination rates of biomarkers. This work aimed at (i) building models allowing the estimation of the necrose biomarker amount released by lesion using population compartimental analysis, (ii) investigating the relevance of biomarker amount and (i) developing limited sampling strategies (LSS) allowing accurate biomarker amount estimates using Bayesian anlysis.
Methods: Three population kinetic biomarker amount estimators were developed for CK, CK-MB and cTnI biomarkers. Repeated biomarker concentrations measurements were obtained from five clinical trials evaluating the impact of conditioning therapies in STEMI between 2005 and 2013 [1-5] . For each patient, 13 to 15 biomarker repeated measurements between hours 0 and 72 after inclusion were available. Patients were randomly assigned to learning (2/3) or validation (1/3) subsets. Using learning subset: 1 or 2 compartment population kinetic models with 1 or 2 gamma distribution functions for biomarker input, and zero and/or first-order transfer distribution and elimination rate constants were built; Bayesian LSS estimators including 1, 2 and 3 samples were developed. Predictive performances of LSS estimators were compared using both learning and validation subsets.
Results: Among clinical trials, 132 patients were evaluable for CK and cTnI, and 49 patients for CK-MB. CK and cTnI kinetics were best described using 2 compartment models, whereas 1 compartment was sufficient for CK-MB. Two gamma distribution functions were necessary for cTnI input rate, whereas 1 distribution was sufficient for CK and CK-MB input. Our kinetic models provided accurate estimations of biomarker release input and powerful assessment of conditioning treatment. Short sampling (within 24 hours after inclusion) was possible for LSS estimators. Three-sample LSS provided best prediction accuracies (R2 > 95%). For CK-MB, sampling times 8, 16 and 20 hours was the best LSS.
Conclusions: Accurate estimations for biomarker input and powerful assessment of conditioning therapies were obtained using a limited number of samples taken within 24 hours after inclusion. This « more powerful and less expensive » strategy will certainly be a useful add-on to future studies in the fild of STEMI and cardioprotection.
References:
[1] Thuny, F. et al. Post-conditioning reduces infarct size and edema in patients with ST-segment elevation myocardial infarction. J Am Coll Cardiol 59, 2175-2181, doi:10.1016/j.jacc.2012.03.026 (2012).
[2] Staat, P. et al. Postconditioning the human heart. Circulation 112, 2143-2148, doi:10.1161/CIRCULATIONAHA.105.558122 (2005).
[3] Piot, C. et al. Effect of cyclosporine on reperfusion injury in acute myocardial infarction. N Engl J Med 359, 473-481, doi:10.1056/NEJMoa071142 (2008).
[4] Mewton, N. et al. Postconditioning attenuates no-reflow in STEMI patients. Basic Res Cardiol 108, 383, doi:10.1007/s00395-013-0383-8 (2013).
[5] Prunier, F. et al. The RIPOST-MI study, assessing remote ischemic perconditioning alone or in combination with local ischemic postconditioning in ST-segment elevation myocardial infarction. Basic Res Cardiol 109, 013-0400 (2014).