Neonatal PBPK model of midazolam and its metabolites for perinatal asphyxia and therapeutic hypothermia.

BACKGROUND AND PURPOSE: Therapeutic hypothermia (TH) decreases morbidity and mortality in neonates with perinatal asphyxia (PA) and moderate-to-severe hypoxic-ischaemic encephalopathy. Midazolam is an antiepileptic and substrate of CYP3A4. Drug metabolism during PA/TH remains partially characterized, and disentangling effects of PA and TH is difficult, making precision dosing challenging. The I-PREDICT project aims to develop a framework for drug disposition in PA/TH neonates, using physiologically based pharmacokinetic (PBPK) modelling. The predictive performance of the model was explored for midazolam and its active metabolites, 1'-hydroxymidazolam and 1'-hydroxymidazolam-O-glucuronide, in neonates with PA/TH.

EXPERIMENTAL APPROACH: The model was established stepwise, starting with healthy adults, and extrapolated to paediatric (1 month-18 years) and neonatal (0-27 days, non-asphyxiated) populations. Finally, PA/TH-related physiological and metabolic changes were incorporated. For each population, performance was verified using clinical data.

KEY

RESULTS

Model predictive performance for midazolam and 1'-hydroxymidazolam was acceptable in adults and children, with AFE and AAFE values ranging from 0.87 to 1.22 and 1.15 to 1.23, respectively. The neonatal PA/TH PBPK model captured midazolam pharmacokinetics, with 71% of the observations within the 10th-90th percentiles and 46% within the 25th-75th percentiles of the prediction. Incorporating PA/TH physiology improved the neonatal PA/TH model by 6%-43% (within the 10th-90th percentiles), depending on the compound.

CONCLUSIONS AND IMPLICATIONS: This PA/TH-adjusted neonatal PBPK model may support future midazolam dose optimization through mechanistic understanding and provides a framework for other drugs in neonates with PA/TH. Future knowledge on the impact of TH on protein binding and renal clearance of 1'-hydroxymidazolam-O-glucuronide would further improve the model.