Frederic Lusoli 1, Celine Pitou 2, Leon Aarons 1, Kayode Ogungbenro 1
1 Centre for Applied Pharmacokinetic Research - University Of Manchester (Manchester, United Kingdom), 2 Global PK/PD & Pharmacometrics - Eli Lilly and Company (Bracknell, United Kingdom)
Introduction: Abemaciclib is an orally administered drug targeting the cyclin-dependent kinase 4 and 6 (CDK4/6), responsible for tumour growth. It showed a complex absorption with the formation of two active metabolites by CYP3A4, M2 and M20, that account for over 10% of the overall drug-related exposure. A number of complex absorption models have been published in the literature, with mixed results [1,2]. In this study, we investigated the absorption process through a deconvolution approach and explored the potential use of the IV data at the initial stage in the development process.
Objectives:
– To characterise the individual drug absorption through a deconvolution method
– To translate the deconvolution information in a population approach to capture the parent-metabolite exposure
Methods: The dataset contained 11 healthy individuals from an absolute-bioavailability study in phase I (NCT02327143). The study involves a semi-simultaneous design with a single oral dose (200 mg) at time 0 on day 1, followed 6 hours later by an intravenous (IV) injection (0.4 mg for 15 minutes) of a 13C tracer in the same individual. The plasma concentration-time profile following each individual IV and oral administration were fitted in R using a non-linear least square method with the Levenberg-Marquardt algorithm [3] using a tri-exponential function. The fitted profile parameters were implemented in a deconvolution algorithm [4] that derives the input rate and percentage of bioavailable-dose that reach the systemic circulation. A transition towards a population input rate was initiated with the selection of a Weibull distribution function [5]. To restrain the Weibull function within an acceptable time span, a finite absorption time [6] of 50 hours was included in the popPK model, based on the time required to cover 95% of the population absorption profile, as determined by the percentage of bioavailable dose results. Different models were tested and developed in Monolix [7]: 1-, 2- or 3- compartment disposition model, normal or lognormal distribution of the error model. The best model was selected based on the parameter estimates, the goodness-of-fits and the visual predictive check (VPC).
Results: The population deconvoluted input rate profiles showed a median lag time of 1.3h [range: 0-1.8], a maximum input rate of 12.925 µmol/h [range: 3.30-42.76] and a time to reach the maximum input rate of 5 h [range: 0-7] after the dose administration. The absorbed percentage of bioavailable-dose matched the bioavailability obtained by non-compartmental analysis with a median of 41.02 % [range: 12.8-53.75]. The Weibull distribution function was integrated in a 3-compartment distribution model and adequately captured the parent plasma concentration-time profiles. As no metabolites intravenous data was available, the metabolites’ parameters are apparent, assuming the parent central volume of distribution (Vcentral) equals the metabolite central volume of distribution (Vcentral, metabolite). The sequential fitting of the metabolites’ model showed a sufficient agreement with the metabolites’ plasma concentrations with a 2-compartment disposition model.
Conclusion: A generic input rate was selected from the observation of the individual deconvoluted profile and implemented successfully in a population approach. The proposed parent-metabolites PopPK model described the plasma concentration time-profiles observed in healthy participants after a single oral 200mg dose. The model information will be translated to a PBPK model to optimise the metabolites’ exposure.
References:
References:
[1] Chigusta et al. CPT:PSP (2020) 9, 523-533.
[2] Posada et al. J Clin Pharmacol. (2020) 60, 915-930.
[3] Elzhov et al. R package minpack.lm. (2023).
[4] Veng-Pedersen P., J Pharmacokinet Biopharm. (1980) 8, 463-481.
[5] Piotrovskii et al. J Pharmacokinet Biopharm. (1987) 15, 681 686.
[6] Macheras, P., Tsekouras, A.A. J Pharmacokinet Pharmacodyn (2023) 50, 5–10
[7] Monolix 2023R1, Lixoft SAS, a Simulations Plus company
Reference: PAGE 34 (2026) Abstr 12231 [www.page-meeting.org/?abstract=12231]
Poster: Drug/Disease Modelling - Oncology