Objective To determine the time course of hazard for motor and nonmotor milestones of Parkinson disease (PD) in the long term and to investigate whether risk scales nonlinearly with time is instrumental in identifying changes in pathological processes and evaluating disease-modifying therapies in PD. Methods Outpatients with PD at the Lyon University Movement Disorders Center were evaluated for 7 clinical milestones in this retrospective cohort study, encompassing 4 domains of PD progression: (1) motor (motor fluctuations, dyskinesias); (2) axial (postural instability and falls, freezing of gait); (3) neuropsychiatric (impulse control disorders, hallucinations); and (4) cognitive (dementia) complications. For each complication, we estimated the outcome-specific hazard using parsimonious smooth parametric Poisson regression models allowing for nonlinear scaling over disease duration, age at diagnosis, current age, and their interaction. Results A total of 1,232 patients with PD experienced 1,527 disease-related complications in up to 12 years of follow-up. Specific to each complication, hazard rates increased dramatically starting from diagnosis and were highest for motor fluctuations and lowest for dementia up to 6 years after diagnosis in patients aged 65 years at diagnosis. Nonlinear patterns indicated dramatic changes in the course of PD after 5 years and predicted more severe axial prognosis after 70 years and for motor fluctuations, dyskinesias, and impulse control disorders before 60 years at diagnosis. Conclusion Time course of motor and nonmotor milestones in PD is determined by disease duration and age at diagnosis in nonlinear patterns and their interaction. This indicates disease- and age-specific thresholds across the multiple neurodegenerative processes accumulating in PD at different paces.