Steve Choy (1), Fran Stringer (1), Shaun Kumar (1), Allan Rae (1), Meera Tugnait (2), Vivek Kadambi (2), Bruce Green (1)
(1) Model Answers R&D, Australia (2) Blueprint Medicines Corporation, USA
Introduction: Avapritinib (approved by the U.S. Food and Drug Administration (FDA) as AYVAKIT™) is a potent and selective inhibitor of v-kit feline sarcoma viral oncogene homolog (KIT) and platelet-derived growth factor receptor alpha (PDGFRA) mutant kinase used in the treatment of unresectable or metastatic gastrointestinal stromal tumors (GIST).
The pharmacokinetic (PK) properties of avapritinib in patients with GIST following once daily dosing (QD) were determined using a non-compartmental analysis (NCA) approach from a single Phase 1 study[1]. However, conventional NCA was challenging due to the limited sampling scheme at steady state (samples only collected to 8 hours post-dose) and the long half-life of the drug (approx. 32-57 hours). In order to derive exposure metrics such as area under the plasma concentration-time curve from time 0 post-dose to the end of the dosing interval at steady-state (AUC0-t,ss), maximum plasma concentration at steady state (Cmax,ss), and plasma concentration at 24 hours post-dose at steady state (C24), a population PK model of avapritinib was developed and subsequently the individual predictions from the model were used to derive the PK exposure metrics of interest.
Objectives:
- Develop a population PK model for avapritinib in healthy volunteers and patients with GIST.
- Derive AUC0-t,ss, Cmax,ss, and C24 for patients with GIST using the population PK model.
Methods: Data from 4 Phase I clinical studies in healthy volunteers (n=122) and patients with GIST (n=221) were used to develop the population PK model in NONMEM[2]. The predictive performance of the final model was assessed on both an individual and population level using various model diagnostics such as relative bias and simulation-based posterior predictive checks (PPC)[3]. Following model evaluation, dummy time points on Day 15 were inserted into the dataset as follows: 0, 0.5, 1, 2, 4, 8, and 24 hours post-dose (equivalent to the nominal sample collection times from the study on Day 1). Individual post-hoc PK parameter estimates from the model were used to predict AUC0-t,ss, Cmax,ss, and C24 at steady state.
Results: The population PK of avapritinib was best described by a two-compartment model with absorption modelled via 4 or 5 transit compartments dependent on tablet or capsule formulation. Disposition of avapritinib was modelled in terms of apparent clearance (CL/F), apparent central and peripheral volume of distribution (Vc/F and Vp/F), and apparent inter-compartment clearance (Q/F). Between subject variability (BSV) was included on the absorption rate constant (KTR), CL/F, and Vc/F. Between occasion variability (BOV) was included on KTR and relative bioavailability (F). Statistically significant covariate effects identified in the model were lean body weight (LBW) on Vc/F, and formulation on KTR, patient population on F, and concomitant proton pump inhibitor (PPI) use on F. Overall, covariate effects are expected to have <0.5-fold change on exposure in terms of AUC0-t,ss and Cmax,ss. When comparing the observed and predicted area under the plasma concentration-time curve from first observation on Day 15 to the time of last quantifiable concentration above the lower limit of quantification on Day 15 (AUClast), 95% of the individual predicted AUClast were within ±15% compared to their observed values. The distribution of the p-values for PPC (pPPC) for AUClast based on 1000 simulated datasets indicated that the 95% prediction interval (PI) encompasses 0.5, with a median pPPC of 0.46. After the predictive performance of the model was verified, individual post-hoc PK parameter estimates from the model were then used to predict AUC0-t,ss, Cmax,ss, and C24 for patients with GIST.
Conclusions: A population PK model for avapritinib was developed using data from healthy volunteers and patients with GIST. The model was subsequently utilized to provide PK exposure metrics using an NCA approach to overcome sampling time limitations. This model-based approach was used in support of the regulatory filing for avapritinib which was approved by the FDA in January 2020[4].
References:
[1] Blueprint Medicines Corporation. 2015-2020. ClinicalTrials.gov Identifier: NCT02508532.
[2] Beal SL et al. 1989-2011. NONMEM Users Guides. Icon Development Solutions, Ellicott City, Maryland, USA.
[3] Yano Y et al. 2001. J Pharmacokinet Pharmacodyn.
[4] U.S. Food and Drug Administration. Prescribing Information for AYVAKIT™. January 9 2020.
Reference: PAGE () Abstr 9346 [www.page-meeting.org/?abstract=9346]
Poster: Drug/Disease Modelling - Oncology