L.Nguyen, P. Variol, B. Tranchand* and C. Puozzo.
Institut de recherche Pierre Fabre [Castres/France] * Centre Leon Berard [Lyon/France]
Navelbine (vinorelbine tartrate) is a vinca-alkaloid antimitotic agent which has been marketed since 1989 by Pierre Fabre Laboratory as iv form and used against non-small-cell lung and metastatic breast cancer. In order to support an oral form development, a pharmacokinetic population study using NONMEM system (version IV level 1.1) was performed on both iv and oral phase I rich data selected from 139 patients (64 iv / 75 p.o.).
IV and oral data were analysed separately and respectively using three-compartement (+ zero order absorption) and two-compartment model (+ first order absorption), with first order method and an exponential error model to describe inter and residual variabilities.
Screening of covariates was conducted through a multiple linear stepwise regression approach and tested individually on NONMEM to confirm significant influence. Data were randomly splited in two parts : model building was developped on a training dataset (2/3 of the whole data) and the remaining dataset was assigned to validate a sparse sampling strategy.
The final covariate models on clearance parameter were as follows :
CLiv (1.h-1)= 27.6 . Bsa . (1 – 0.0012 Plat) . (1 – 0.0126 Bili) + 16.9 Wt/Ser
and
CL/Fp.o. (1.h-1) = 3.78 . Wt . (1 – 0.0056 Age) . (1 – 4.55 10-4 GGT)
where Bsa, Wt, Plat, Bili and GGT were respectively body surface area (m2), weight (Kg), platelet count (nb/mm3), bilirubinemia (pM) and Gamma Glutamyl Transferase (UI/ml).
IV and oral distribution volume were correlated with gender showing a decrease of Vd in female. Inter-individual variabilites on CLiv and CL/F p.o. decreased respectively from 35% and 76% in basic model to 22% and 56% in final model. Residual variability was estimated to 22% and 24% in iv and oral route respectively.
Based on iv and oral population model, a limited sampling strategy was undertaken to explore parmacokinetic and PK/PD relationships in ongoing phase II studies. The best sampling designs tested on validation dataset (approximatively 30 patients) needed three samples to accurately and precisely determine clearance parameter.: around 0.33 (end of perfusion); 3 and 24h after iv administration and around 1.5, 3 and 24h after oral intake. Low bias were obtained with both oral and iv sampling : median predicted error between actual (model independant) and bayesian estimated clearance were respectively -2.8% and + 4.5%.
Median predicted absolute error were 10.2% [range : 0.1 – 27.1%] in patients treated by iv route and 9.1% (range : 1.0 – 22.4%] in patients following oral administration of Navelbine. According to the sparse sampling strategy, data from Phase II trials will be included in the analysis to improve performance predictivity of the models, to test inter-occasion variability and to gain information on PK/PD relationships.
Reference: PAGE 7 (1998) Abstr 689 [www.page-meeting.org/?abstract=689]
Poster: poster