Veronika Voronova1, Victor Sokolov1, Holly Kimko2, Marcella Petrone3, Lars Hansen4, Joseph Grimsby4, Kirill Peskov1, 5, Rosalinda Arends2
1 – M&S Decisions LLC, Moscow, Russia 2 – Clinical Pharmacology & Safety Sciences, BioPharmaceuticals R&D, AstraZeneca, Gaithersburg, USA 3 – Clinical Pharmacology & Safety Sciences, BioPharmaceuticals R&D, AstraZeneca, Cambridge, UK 4 – Research and Early Development, Cardiovascular, Renal and Metabolic Diseases, BioPharmaceuticals R&D, AstraZeneca Ltd, Gaithersburg, MD, USA 5 – WCRC “Digital biodesign and personalized healthcare”, I.M. Sechenov First Moscow State Medical University, Moscow, Russia
Objectives: MEDI7219 is a novel orally-delivered GLP-1 peptide receptor agonist (RA) developed for the type II diabetes indication and characterized by significant between-subject and between-occasion variability (BSV, BOV). Immediate and controlled-release (IR, CR) formulations, assuming drug release in the upper and middle parts of the gastrointestinal tract, respectively, were investigated in healthy volunteers (N=60) in a phase 1 study. To mitigate nausea, MEDI7219 was titrated up every week following dose steps of 6, 8, 12, 16 and 20 mg (IR formulation) and 6, 8, 10 and 12 mg (CR formulation), given once-daily. In this work we aimed to evaluate the link between MEDI7219 exposure and probability of nausea as the most common class side effect of GLP-1 RAs to predict and compare safety profiles across various drug formulations and treatment regimens.
Methods: The analysis consisted of two steps representing (1) the development of a population pharmacokinetic (PK) model for IR and CR formulations of MEDI7219 and (2) the evaluation of quantitative relationship between MEDI7219 PK and observed nausea incidence via a logistic regression model. Patient-level data from the multipart phase I study NCT03362593 were used for the analysis. Monolix 2019R1 was applied for the PK model development; exploratory data analysis and logistic modeling was performed using R 3.5.1. Optimal models were selected based on Akaike information criterion, visual inspection of diagnostic plots and the analysis of model identifiability.
A two-compartment model with linear clearance and delayed absorption was used to characterize drug PK. The differences between MEDI7219 formulations were accounted for through the implementation of covariates on the absorption parameters. BOV and BSV were evaluated in different combinations to achieve optimal balance between model identifiability and quality of data description. The final population PK model was used to simulate individual MEDI7219 levels and estimate maximal drug concentration (Cmax) for the days with matching PK and nausea measurements. Covariate search was performed to evaluate the differences in exposure-safety relationships between the IR and CR MEDI7219 formulations and assess impact of the treatment duration on nausea probability in order to investigate tolerability development.
Results: The population PK analysis revealed (1) large BOV in bioavailability, absorption rate and delay, as well as (2) large BSV in volume of distribution and clearance rate responsible for up to 30-fold variation in the observed MEDI7219 plasma levels. Consequently, predicted individual steady-state drug concentrations overlapped across the tested doses given once daily: predicted Cmax for CR formulation was 1.61 [90% confidence interval (CI): 0.25-8.97], 2.49 [90% CI: 0.37-12.74], 2.96 [90% CI 0.51-19.03] and 3.71 [90% CI: 0.59-22.45] ng/ml for 6, 8, 10 and 12 mg respectively. For IR formulation model-predicted Cmax values were 1.87 [90% CI: 0.30-10.72], 2.80 [90% CI: 0.44-15.10], 3.93 [90% CI: 0.68-24.32], 5.37 [90% CI: 0.92-34.03] and 6.75 [90% CI: 1.22-39.55] ng/ml for 6, 8, 12, 16 and 20 mg respectively. CR formulation was characterized by a longer delay in absorption and slower absorption rate compared to the IR formulation, resulting in smoother PK profile with lesser difference between peak and trough levels.
The exposure-safety analysis successfully established the quantitative relationship between MEDI7219 Cmax and the risks of nausea incidence. Cmax values associated with 30% probability of nausea onset were 12.7 and 20.7 ng/ml for IR and CR formulations respectively. No significant time-dependent decrease in the risks of nausea was observed during the model evaluation, possibly due to the relatively short observation period.
Conclusions: This analysis enabled the quantitative assessment of BOV and BSV in the PK of an orally-delivered GLP-1 peptide and was subsequently applied to determine the impact of the variability in PK on the probability of nausea incidence. Better tolerability and smoother PK profiles were shown for CR compared to IR formulation of the compound.
Reference: PAGE 29 (2021) Abstr 9606 [www.page-meeting.org/?abstract=9606]
Poster: Methodology - Covariate/Variability Models