This model-based analysis quantifies the population pharmacokinetic-pharmacodynamic efficacy and safety/tolerability relationships of orally administered istradefylline, a selective adenosine A(2A) receptor antagonist, in healthy participants and patients with Parkinson disease. Data from 6 phase 2/3 clinical trials comprised the population database, with 1760 and 1798 patients contributing to the efficacy and safety/tolerability analyses, respectively. The relationship between istradefylline area under the curve at steady state and percentage OFF time was described by a nonlinear model (Emax) based on time for the disease progression/placebo response component and an Emax model for the effect of istradefylline. The typical maximum decrease in percentage OFF time due to istradefylline exposure would be 5.79% (95% confidence interval = 4.09%-7.49%) with one-half of the maximum effect reached at an exposure of 1690 ng × hr/mL (95% confidence interval = 199-3180 ng × hr/mL). The pharmacokinetic-pharmacodynamic relationships for dyskinesia and dizziness were described by an Emax model, and for nausea, a power model was used. The probabilities of dyskinesia and dizziness are expected to plateau at a dose of 40 mg/d, and the probability of nausea is expected to continually rise as the dose is increased. Collectively, these results support a starting istradefylline dose of 20 to 40 mg/d.