Mentre F. (1), Comets E. (1), Pousset F. (2), Plaud B. (2), Diquet B. (2), Montalescot G. (3), Ankri A. (4), Mallet A. (1), Lechat Ph. (2)
(1) INSERM U436, (2) Service de Pharmacologie, (3) Service de Cardiologie, (4)Service dÃHemostase, CHU Pitie-Salpetriere, 91 Bd de lÃHopital, 75013 Paris, France.
Fluindione is an oral anticoagulant with a long half-life that inhibits the synthesis of vitamin K-dependent clotting factors. It is used in France in various cardiologic diseases for the prevention of thromboembolism. Although it is well known that the individual dose to achieve a target INR varies widely across patients, no study of the sources of variability in pharmacokinetics (PK) or pharmacodynamics (PD) of fluindione has been performed. The objective of the present study is to estimate the population characteristics of PK and PD parameters in patients.
Data of 49 patients who initiate an anticoagulant treatment were prospectively collected. No constraint on fluindione dosage was defined (usually 20 mg daily), the time of administration was randomly chosen between 8, 12 or 20 h. One blood sample was taken the day before treatment, and after 1, 3 and 5 doses at 8 h. In each blood sample were measured fluindione concentration by HPLC, INR, Prothombin Complex Activity (PCA) and four clotting factors (FII, FVII, FIX, FX). Therefore, three samples during treatment per patient are available 12, 18 and 24 hours after administration.
These data were analyzed using a population method given a parametric PK/PD model. The PK model is an open one-compartment model involving two parameters: oral clearance (CL/F) and oral volume (V/F). For each studied outcome, the PD model is similar to the one already derived for warfarin from îrich data” in healthy volunteers and is expressed as a differential equation.
A first-order elimination is assumed with rate constant Kd. The rate of synthesis is related to the current plasma concentration of fluindione through an Emax model: C50 is the concentration that reduces by 50% the synthesis rate and Tlag is the time delay on onset of effect of fluindione. This model was used to analyze PK and PD data for six different outcomes: the four clotting factors, PCA and INR. Assuming a constant coefficient of variation (CV) error model (10% and 20% for PK and PD data respectively), the population characteristics were estimated using the Non Parametric Maximum Likelihood estimation method. Several nested PD models were compared for the clotting factors.
The predictability of the best PD model was assessed on data of 24 similar further patients. Fluindione administration was performed at 20 h and blood samples were taken at 8 h. Data were obtained at Day 2,3 or 4, and 6 and only INR was assayed in each blood sample. The predictability of INR at D6 using the individual Bayesian estimate of the PK/PD parameters from INR measurements at D2 to D4 with the previous model was evaluated. This tool will be further evaluated for individualization of fluindione dosage in new patients.
Reference: PAGE 5 (1996) Abstr 552 [www.page-meeting.org/?abstract=552]
Poster: poster