Predictive performance of two hepatic CYP3A4 ontogeny within a paediatric PBPK model.
Marc Codaccioni (1), Roz Southall (1), Jean Dinh (1), Trevor Johnson (1).
(1) Certara UK Limited (Simcyp Division), Sheffield, UK.
Introduction: Paediatric physiologically based pharmacokinetic models have been applied for various aims such as dose selection, new formulation development, and trial design optimization. Between 2005 and 2020, there has been a more than 33-fold growth in their use [1]. This mechanistic framework integrates information regarding organ development and ontogeny of enzyme expression or activity involved in drug disposition, enabling the prediction of drug pharmacokinetics in this vulnerable group. CYP3A4 is a major contributor to the elimination of drugs [2] and understanding its ontogeny is a crucial step toward robust prediction of the PK of CYP3A4 substrates across the paediatric age range. Currently, two ontogeny profiles, based on an in vivo clearance deconvolution approach, are available within Simcyp [3,4]. The fraction of adult values between the two are different, especially at younger ages; leading to discrepancies in the pharmacokinetic outcomes and dose recommendations predicted for the CYP3A4 substrates in children.
Objective: To compile evidence confirming a single CYP3A4 ontogeny profile in Simcyp Paediatric populations.
Methods: The Simcyp Paediatric module was used to simulate several published paediatric clinical studies for CYP3A4 compounds (midazolam, alfentanil) to compare the predictive performance obtained using the two ontogeny profiles. The analysis includes low-to-intermediate extraction ratio types of compounds (based on adult values) that show a fraction metabolized through the CYP3A4 route above 80% (i.e., fmCYP3A4 > 80%), aiming to maximize the change in metabolic intrinsic clearance on total clearance with a sufficient level of specificity. Only clinical scenarios following intravenous dosing were considered to focus on the liver’s contribution to metabolism. A curated dataset of 19 single-dose clinical scenarios was built covering ages ranging from neonates to adolescents. The assessment of prediction accuracy is based on predictive scores (i.e., GMFE: Geometric Mean Fold Error, PwBE%: Predictions were within the BioEquivalent range, Pw2FE%: Predictions were within 2-Fold Error) computed with the mean exposure up to the last measurable concentration (AUClast), as well as concentration versus time profiles visual inspection. Each scenario has been allocated to an age bin according to its age distribution (i.e., < 4 years old, between 4 and 12 years old and > 12 years old groups) allowing stratification of the analysis by age.
Results: Based on the current dataset, the ontogeny profile published by Upreti and Wahlstrom [4] presents good predictive capacity; showing 1.17, 84% and 100% for GMFE, PwBE% and Pw2FE%, respectively. The same scores for the ontogeny profile published by Salem et al. [3] were 1.41, 42% and 84%. Across the different clinical studies, the simulations integrating the Upreti and Wahlstrom ontogeny profile exhibit better exposure predictive performance in 74% of cases and from simulated concentration versus time profiles better capture the late elimination kinetic phase. These latter results agree with the slight trend of exposure overprediction only observed with the simulations integrating the Salem ontogeny profile. The age-stratified analysis shows similar predictive performance, with GMFE score discrepancies between the tested ontogenies decreasing with age gain, which is in line with the fact that the two ontogeny profiles are closer at older ages.
Conclusion: These findings suggest that the simulations integrating the CYP3A4 ontogeny profile of Upreti and Wahlstrom perform better than those integrating the Salem profile for two CYP3A4 substrates in children. Those conclusions are based on a limited dataset and will be extended to include other drugs. This work will serve as a basis to derive an optimal ontogeny profile using the ‘Simultaneous fitting tool’ included within the Simcyp V22 software.
References:
[1] Johnson et al. (2021). CPT Pharmacometrics Syst Pharmacol, 11(3): 373-383.
[2] Saravanakumar et al. (2019). Clin Pharmacokinet, 58(10): 1281-1294.
[3] Salem et al. (2014). Clin Pharmacokinet, 53(7): 625-636.
[4] Upreti and Wahlstrom. (2016). Clin Pharmacol, 56(3): 266-283.