The aim of this study was to characterize the population pharmacokinetic of vinorelbine in elderly patients and to propose a limited-sampling strategy to estimate individual pharmacokinetic parameters. Vinorelbine was administered by a 10-min continuous infusion at a dose of 20–30 mg/m2. The population parameters were computed, using a three-compartment model, from an initial group of 27 patients. Twelve additional courses were used for model validation and evaluation of eight different limited-sampling strategies. The inter-individual variability of CL was explained by a linear dependency with age. The population average parameters and the interindividual variabilities (CV%) were: CL=47.1 l/h (31.7%),V=16.6 l (64%),k21=0.776 h−1(20%),k31=0.0346 h−1(15.2%),&agr;=0.431 h−1(6.84%) and&bgr;=0.0167 h−1(25%). Bayesian estimation with three measured levels (end of infusion, and 6 and 48 h) can be selected, because it allows adequate estimation of CL, elimination half-life and vinorelbine concentrations with a non-significant bias. Moreover, the choice of these three sampling times presents practicality advantages for the patient's comfort. Vinorelbine clearance decreasing with age and AUC being a good predictor of several toxicity end points during vinorelbine treatment, the limited-sampling strategy developed in this paper may be clinically relevant.