Clinicians recognize levodopa has a short-duration response (measured in hr) and a long-duration response (measured in days) in Parkinson's disease. In addition there is a diurnal pattern of motor function with better function in the morning. Previous pharmacokinetic-pharmacodynamic modeling has quantified only the short-duration response. We have developed a pharmacokinetic-pharmacodynamic model for the short- and long-duration responses to exogenous levodopa and the effects of residual endogenous levodopa synthesis in patients with Parkinson's disease. Thirteen previously untreated (de novo) patients with Parkinson's disease and twelve patients who had received levodopa orally for 9.7+/-4.0 years (chronic) were investigated. A 2 hr IV infusion of levodopa with concomitant oral carbidopa was given on two occasions separated by 3 days with no levodopa in between. A two compartment pharmacokinetic model was used to fit plasma levodopa concentrations. A sigmoid Emax model was used to relate concentrations from endogenous and exogenous sources to tapping rate (a measure of motor response). A model incorporating three effect compartments (fast equilibration (half life, Teqf). slow equilibration (Teqs) and dopa synthesis (Teqd)), yielded the most descriptive model for levodopa pharmacokinetics and pharmacodynamics. Baseline tapping rate reflected endogenous levodopa synthesis and the long-duration response. Partial loss of the long-duration response during the 3 days without levodopa in the chronic group lowered baseline tapping (36+/-7%, mean+/-SEM) and increased maximum levodopa induced response above baseline (112+/-31%). The maximum levodopa induced response after the drug holiday is a result of lowered baseline tapping due to the loss of long-duration response and not due to a change in levodopa pharmacokinetics or pharmacodynamics.