Dubruc C, *Mesnil F, Thenot JP, *Mentré F.
Synthélabo Recherche, Chilly-Mazarin, *INSERM U436, Pitié Salpêtrière, Paris, France
The pharmacokinetic profile of mizolastine, a new antihistamine agent, was assessed firstly in 18 healthy young volunteers after a single oral doses of two 10 mg formulations (capsule and tablet) (19 – 20 blood samples per patient id 711 plasma concentrations) and secondly with sparse sampling (1 – 4 blood sample per patient id 960 plasma concentrations at steady state) in 696 allergic patients included in 10 Phase III trials after repeated once a day administrations of different doses ( 7.5, 10 or 15 mg) of the tablet or capsule.
In healthy young volunteers, the data could be fitted with an open two-compartment model with a zero order absorption phase. Five parameters were estimated in the model: Tabs (time to peak concentration), V/F (oral volume of the central compartment), CL/F (oral total clearance), alpha (rate constant of the rapid elimination phase) and beta (rate constant of the slow elimination phase). A multiplicative model for random effects was used. The error variance model was heteroscedastic. The population mean values were Tabs = 0.509 h, V/F = 20.1 L, CL/F = 8.32 L/h, alpha = 0.608 h-1 and beta = 0.0857 h-1 with respective coefficients of variation of 31.2%, 53.9%, 34.5%, 61.0% and 35.4%. The estimated coefficient of variation of the error was 14%.
In patients, it was impossible to fit the data with the same PK model as in healthy young subjects. When taking initial values of estimates equal to those obtained in healthy young subjects with the same model, no satisfactory convergence could be achieved even with constraints on parameters and/or not allowing variability on some of them. As 70% of patients had only one concentrations measurement, two possibilities were considered: either try an open one compartment model with a zero order absorption phase or add the concentrations measured in healthy young subjects to the data set of patients since the distributions of ages and weights in patients were similar to those in healthy young subjects. In patients, there was approximately the same number of males and females whereas healthy young subjects were only males. The first possibility was tested. A multiplicative model for random effects was again used and the same model as in healthy subjects for the error variance was chosen. Three parameters were estimated: V/F, CL/F and Tabs and seven covariates were selected (galenic form, sex, age, weight, serum creatinine, ASAT and ALAT). No variability on Tabs was allowed. The population mean values were Tabs = 0.501 h, V/F 143 L and CL/F = 4.31 L/h with respective coefficients of variation for V/F and CL/F of 239% and 56.2%. The coefficient of variation of the error was 51.4%. No influence of the covariates on parameters could be shown. This analysis will be further refined by adding concentrations measured in healthy volunteers to the data set obtained in patients.
Reference: PAGE 6 (1997) Abstr 654 [www.page-meeting.org/?abstract=654]
Poster: poster