OBJECTIVE: To develop a population pharmacokinetics of vinorelbine in a population of non-small-cell lung cancer (NSCLC) patients using a Bayesian estimation in order to calculate for any further patient, individual pharmacokinetic parameters from few blood samples. METHODS: Vinorelbine was given by a 15-min infusion (30 mg x m(-2)) to eight patients with NSCLC. Its serum concentration was determined by HPLC and its pharmacokinetics was described by a three-compartment open model with elimination from the central compartment. Volume of the central compartment (V1) and rate constants (k10, k12, k21, k13, k31) were selected as population pharmacokinetic parameters and computed by non-linear regression (two-step approach) from 14 to 18 concentration measurements per course. Subsequently, these parameters were used by the Bayesian estimator to calculate individual pharmacokinetics from only 2 or 3 measured concentrations. RESULTS: The population mean values (CV%) of V1, k10, k12, k21, k13, k31, CL, t1/2gamma were respectively 21 l (55%), 3.2 h(-1) (29%), 7.7 h(-1) (74%), 1.3 h(-1) (67%), 4.7 h(-1) (53%), 0.04 h(-1) (20%), 57 l x h(-1) (31%) and 43 h (36%). The comparison of results obtained from the Bayesian estimator and from the three-compartment model showed that CL and t1/2gamma were well predicted (relative deviation: +/- 12 to 22%) by the Bayesian method using only two blood samples. CONCLUSION: We demonstrated that Bayesian estimation allows, at minimal cost and minimal disturbance for the patient, the determination of several vinorelbine pharmacokinetic parameters and therefore dose adaptation from as few as two drug concentrations, measured at 6 h and 24 h after infusion.
OBJECTIVE: To develop a population pharmacokinetics of vinorelbine in a population of non-small-cell lung cancer (NSCLC) patients using a Bayesian estimation in order to calculate for any further patient, individual pharmacokinetic parameters from few blood samples. METHODS:Vinorelbine was given by a 15-min infusion (30 mg x m(-2)) to eight patients with NSCLC. Its serum concentration was determined by HPLC and its pharmacokinetics was described by a three-compartment open model with elimination from the central compartment. Volume of the central compartment (V1) and rate constants (k10, k12, k21, k13, k31) were selected as population pharmacokinetic parameters and computed by non-linear regression (two-step approach) from 14 to 18 concentration measurements per course. Subsequently, these parameters were used by the Bayesian estimator to calculate individual pharmacokinetics from only 2 or 3 measured concentrations. RESULTS: The population mean values (CV%) of V1, k10, k12, k21, k13, k31, CL, t1/2gamma were respectively 21 l (55%), 3.2 h(-1) (29%), 7.7 h(-1) (74%), 1.3 h(-1) (67%), 4.7 h(-1) (53%), 0.04 h(-1) (20%), 57 l x h(-1) (31%) and 43 h (36%). The comparison of results obtained from the Bayesian estimator and from the three-compartment model showed that CL and t1/2gamma were well predicted (relative deviation: +/- 12 to 22%) by the Bayesian method using only two blood samples. CONCLUSION: We demonstrated that Bayesian estimation allows, at minimal cost and minimal disturbance for the patient, the determination of several vinorelbine pharmacokinetic parameters and therefore dose adaptation from as few as two drug concentrations, measured at 6 h and 24 h after infusion.