The purpose of the study was to describe the population pharmacokinetics of levodopa in patients with Parkinson's disease studied in 5 trials (10 occasions) over 4 years. Twenty previously untreated Parkinsonian patients were investigated. Each trial consisted of a 2-hr IV infusion of levodopa (1 mg/kg/h) with concomitant oral carbidopa given on two occasions separated by 72 hr with no levodopa in between. This trial design was repeated at 6, 12, 24 and 48 months. A two-compartment pharmacokinetic model with central volume (V1), peripheral volume (V2), clearance (CL) and inter-compartmental clearance (CL(ic)) was used to fit plasma levodopa concentrations. The model accounted for levodopa dosing prior to each trial and endogenous levodopa synthesis. Population parameter estimates (geometric mean) and population parameter variability (PPV; SD of normal distribution) were V1 11.4 l/70 kg (0.44), CL 30.9 l/h/70 kg (0.25), V2 27.3 l/70 kg (0.27), and CL(ic) 34.6 l/h/70 kg (0.48). PPV was partitioned into between subject variability (BSV) which was 0.12 V1, 0.13 CL, 0.15 V(2), 0.28 CL(ic), within trial variability (WTV) which was 0.16 V1, 0.13 CL, 0.08 V2, 0.18 CL(ic) and between trial variability (BTV) which was 0.40 V1, 0.17 CL, 0.21 V2, 0.34 CL(ic.) Neither structural nor random levodopa pharmacokinetic parameters were associated with the time course of development of fluctuation in motor response. Variability in levodopa pharmacokinetic parameters (particularly V1) may result in variability in plasma levodopa concentrations that could contribute to fluctuations in motor response.