Emilie HENIN, Catherine GUITTET, Christian LAVEILLE
1. Calvagone, Chazay d’Azergues, France; 2. Meletios Therapeutics, Paris, France
Objectives:
Meletios Therapeutics is currently developing MLT103 (compound initially evaluated in the 1990s) with various pharmacological properties: it is a modulator of the sterol metabolic pathway, and is an anti-inflammatory and immunomodulatory agent that has been initially tested in clinical settings in rheumatoid arthritis as a disease modifying anti-rheumatic drug (DMARD), in multiple myeloma, in psoriasis and in prostate cancer. Recently, it has been tested in preclinical settings in viral infectious diseases (Coronavirus diseases (Covid), Middle East respiratory syndrome (MERS), Influenza).
The objective of this study was to develop a population PK model of MLT103, in healthy subjects and patients with psoriasis and rheumatoid arthritis, accounting for the large heterogeneity in PK profiles due to CYP2D6 polymorphism and fasting/fed conditions. The purpose of the PK model is, in a first intent to better characterise the PK profile of MLT103 and identify key covariates; and in a second intent, to evaluate the exposure-safety relationships, based on data from other studies where fewer PK samples are available in each subject and CYP2D6 status has not been reported.
Data & Methods:
Data from 5 studies, performed between 1991 and 1997, were considered for model development. A total of 1667 concentration measurements in 227 subjects were available. In those studies, MLT103 was administered either as single dose, daily for 7 days, daily for 22 days or daily for 12 weeks, with dose ranging from 25 to 1050 mg. CYP2D6 phenotype was known in 184 subjects, classified as 177 extensive metabolisers and 7 poor metabolisers. Regarding food conditions, 42 subjects were in fasting conditions, 175 subjects had a meal at the time of drug intake, and 10 subjects had a meal 30 minutes after drug intake.
Population approach allows the quantification of inter-individual variability. MLT103 pharmacokinetics was known to be influenced by CYP2D6 polymorphism. When a strong covariate, such as poor or extensive metaboliser, is not available, mixture modelling allows estimating separate parameter sets for each subpopulation while predicting to which subpopulation each subject belongs, by maximising conditional likelihood.
Model development and parameter estimation were performed using NONMEM 7.5. Simulations were performed using e-Campsis (https://ecapsis.com).
Results:
MLT103 disposition was well described by a 2-compartment model. Absorption model consisted in a 4-transit compartment chain, with a mean absorption time of 1.4 hours, in a typical fasting subject. No indication of dose non-linearity was identified. Food intake at the time of drug intake was found to slow absorption by a 30% and increase relative bioavailability by a factor 2.3. Too few subjects had their meal 30 minutes after drug intake to allow the implementation of a dynamic meal effect; however, those subjects showed an absorption rate similar to fasting subjects and a relative bioavailability similar to fed subjects.
In a first approach, CYP2D6 status was ignored, and two subpopulations were discriminated by mixture modelling, estimating a different clearance, mean absorption time and relative bioavailability. 180 out of 184 subjects were well assigned to a given population, when CYP2D6 status information is available (175/177 extensive metabolisers and 7/7 poor metabolisers); 2 subjects reported as extensive metabolisers were assigned to the mixture with lower clearance. Of note, CYP2D6 polymorphism nowadays include 2 additional categories, so-called “intermediate“ and “ultra-rapid metabolisers”, while only “extensive“ and “poor” metabolisers were considered at the time of these studies.
In a second approach, subjects were assigned to the sub-population corresponding to their observed status when available, and assigned to the most likely subpopulation when not available. Parameter estimates were comparable to ones obtained in the first approach (<5% difference).
In the future, the here-proposed PK model will be used to predict individual exposure from other studies evaluating exposure-efficacy and exposure-safety relationships.
Acknowledgements:
The authors would like to extend their regards to Sanofi for sharing data.
Reference: PAGE 32 (2024) Abstr 10915 [www.page-meeting.org/?abstract=10915]
Poster: Methodology - Covariate/Variability Models