Iasonas Kapralos (1,2) and Aristides Dokoumetzidis (1,2)
(1) Laboratory of Biopharmaceutics-Pharmacokinetics, Department of Pharmacy, National and Kapodistrian University of Athens, Athens, Greece, (2) Athena Research and Innovation Center in Information, Communication and Knowledge Technologies, Athens, Greece
Objectives: The introduction of a long-acting repeatable (LAR) formulation of octreotide offered distinctive benefits to patients with acromegaly and gastro-entero and pancreatic neuroendocrine tumors (GEP-NETs) regarding quality of life and compliance, allowing a single, once per month intramuscular administration. Octreotide is released from the poly- (lactic-co-glycolic acid) (PLGA) microparticles in which it is encapsulated, by a slow, complex process controlled by the interplay between the drug, the formulation and the host.
The aim of the study is to develop a population pharmacokinetic (PPK) model of Octreotide LAR in healthy volunteers which describes the highly variable, multiple peak absorption pattern of the pharmacokinetics of the drug, in individual and population levels. The characterization of the PPKs is needed to provide insights for the mechanistic understanding and may be useful for the future development of sustained release formulations based on PLGA. [1]
Methods: PK data: Data were obtained from a densely sampled bioequivalence (BE) PK study in 120 healthy volunteers following a single 30 mg intramuscular injection of Sandostatin® LAR Depot (octreotide acetate for injectable suspension, Novartis).
Population PK model: An empirical absorption model, coupled with a one-compartment distribution model with linear elimination, was developed to describe the typical PK profile, comprising the rapid initial burst, followed by up to three release phases with different delays. The initial burst phase was modeled as a first order process, defined by the absorption rate constant ka. Three parallel delayed processes, using the transit compartment model, were employed to describe the three-phase absorption delays. [2] The input rate in the depot compartment was modeled as the weighted sum of the transit model function and the first-order absorption.
Clustering: A preprocessing step of the raw PK data was performed before PPK modeling i.e., cluster analysis, in order to identify subpopulations. We applied the shape-respecting variation of k-means clustering method implemented in R package kmlShape. [3] This method, which was developed for the analysis of longitudinal data, takes into account the horizontal offset because modest variations on delays may be of limited importance, and yet account to large distances according to the classical k-means method. Normalization of the PK data allowed the identification of different patterns in release kinetics, without the influence of apparent clearance and total exposure. Cluster analysis results were handled as a categorical covariate and the inclusion in the final model was based on the likelihood ratio test and the overall performance of the PPK model to describe the data.
Results: The cluster analysis allowed the identification of two different patterns in PK data. 87% of the subjects exhibit the typical multi-phase pattern of the initial burst followed by 3 delayed peaks. A sub-population (13% of the subjects) was characterized by an early, extended phase of absorption, followed by a slow delayed release phase, which corresponds to a small part of the total exposure.
The base model consisting of a simple one-compartment PK model with linear elimination, coupled with the empirical absorption model, successfully described the complex and highly variable individual PK profiles. The PPK model we applied was found to be structurally identifiable. The inclusion of the cluster-defined subpopulations as covariate of f1, f2 and CL allowed a better characterization of the observed heterogeneity and variability of the study and resulted in a drop in the objective function value, ΔOFV= -115.158. The performance of the final model was evaluated based on goodness-of-fit plots, visual predictive checks (VPCs) and bootstrap. VPCs comparing model predicted BE metrics to observed values calculated by NCA, showed that the final population PK model described better Cmax, AUC(0-t) and partial AUCs.
Conclusions: The present model is the first to describe the multiple-peak absorption pattern observed after octreotide LAR administration and may be useful to provide insights and validate hypotheses regarding release from PLGA-based formulations. We proposed a workflow, showing that cluster analysis may be valuable in cases where sub-populations are present.
References:
[1] Park et al. J. Control. Release 2019, 304, 125–134.
[2] Savic et al. J. Pharmacokinet. Pharmacodyn. 2007, 34, 711–726.
[3] Genolini et al. PLoS One 2016, 11, e0150738.
Reference: PAGE 29 (2021) Abstr 9825 [www.page-meeting.org/?abstract=9825]
Poster: Methodology – AI/Machine Learning