Clustering Absorption Profiles of Rivaroxaban Using Between Subject Model Mixture
E Ollier (1,2), I Gouin-Thibault (3), MA Loriot (4), P Mismetti (1), S Laporte (1), X Delavenne (1).
(1) Groupe de Recherche sur la Thrombose, EA3065, Université Jean Monnet - Saint Etienne (France), (2) U.M.P.A., Ecole Normale Supérieure de Lyon, CNRS UMR 5669 & INRIA, Project-team NUMED - Lyon (France) (3) Université Paris Descartes, INSERM U765 - Paris (France), (4) UF Pharmacogénétique et Oncologie Moléculaire, Hôpital Européen Georges Pompidou, Inserm UMR-S 775, Université Paris Descartes - Paris (France).
Objectives: Rivaroxaban is a direct oral anticoagulant that is characterised by a very low solubility independent from the pH conditions. Moreover for high dosing regimen, the bioavailability is incomplete and absorption could be delayed. This characteristic induces also high variability in the early rivaroxaban absorption phase (monophasic, multiphasic). The objective of this work was to characterize absorption profile of rivaroxaban and to classify it in different groups of absorption.
Methods: This work consists in the analysis of 60 rich pharmacokinetic profiles of rivaroxaban (40 mg) from the DRIVING study (ClinicalTrials.gov NCT01627665). The absorption rate of rivaroxaban was broken down into a sum of inverse Gaussian functions [1]. To regularize the identifiably issue due to the estimation of bioavailability only from oral data, we introduce prior probability distribution on bioavailability, volume of distribution and clearance parameters. These priors were constructed using previously published data following intravenous administration of rivaroxaban [2]. Finally, a Between Subjects Model Mixture (BSMM) [3] was built in order to cluster absorption profiles (early monophasic vs. early biphasic absorption). Data analysis was performed using a non-linear mixed effect model with MONOLIX® software.
Results: A one-compartment model with absorption rate broken down into a sum of three inverse Gaussian functions (IG1, IG2, IG3) best described the data. These 3 inverse Gaussian functions have an increased mean absorption time (49.5 min, 3h14 min and 21 h for IG1, IG2 and IG3 respectively). The BSMM consists of two models corresponding to two different absorption profiles. The corresponding absorption rates are AR1 and AR2 respectively:
i) Early monophasic absorption and late absorption
AR1 = IG1 + IG3
ii) Early biphasic absorption and late absorption
AR2 = IG1 + IG2 + IG3
The proportion of subjects who are adequately described by an early biphasic absorption was estimated to 30%.
Conclusions: BSMM effectively classified rivaroxaban absorption into two separate early absorption profiles. Moreover, rivaroxaban 40 mg appears to undergo an unexpected late absorption. This last result will allow us to propose optimal strategy for anticoagulant effect reversion.
References:
[1] Csajka, C., Drover, D., & Verotta, D. (2005). The use of a sum of inverse Gaussian functions to describe the absorption profile of drugs exhibiting complex absorption. Pharmaceutical Research, 22(8), 1227–1235.
[2] Stampfuss, J., Kubitza, D., Becka, M., & Mueck, W. (2013). The effect of food on the absorption and pharmacokinetics of rivaroxaban. International journal of clinical pharmacology and therapeutics, 51(7), 549-561.
[3] Lavielle, M., & Mbogning, C. (2014). An improved SAEM algorithm for maximum likelihood estimation in mixtures of non linear mixed effects models. Statistics and Computing, 24(5), 693–707.