Background: The MDS-Unified Parkinson’s disease (PD) Rating Scale (MDS-UPDRS) is the most used scale in clinical trials. Little is known about the predictive potential of its single items. Objective: To systematically dissect MDS-UPDRS to predict PD progression. Methods: 574 de novo PD patients and 305 healthy controls were investigated at baseline (BL) in the single-center DeNoPa (6-year follow-up) and multi-center PPMI (8-year follow-up) cohorts. We calculated cumulative link mixed models of single MDS-UPDRS items for odds ratios (OR) for class change within the scale. Models were adjusted for age, sex, time, and levodopa equivalent daily dose. Annual change and progression of the square roots of the MDS-UDPRS subscores and Total Score were estimated by linear mixed modeling. Results: Baseline demographics revealed more common tremor dominant subtype in DeNoPa and postural instability and gait disorders-subtype and multiethnicity in PPMI. Subscore progression estimates were higher in PPMI but showed similar slopes and progression in both cohorts. Increased ORs for faster progression were found from BL subscores I and II (activities of daily living; ADL) most marked for subscore III (rigidity of neck/lower extremities, agility of the legs, gait, hands, and global spontaneity of movements). Tremor items showed low ORs/negative values. Conclusion: Higher scores at baseline for ADL, freezing, and rigidity were predictors of faster deterioration in both cohorts. Precision and predictability of the MDS-UPDRS were higher in the single-center setting, indicating the need for rigorous training and/or video documentation to improve its use in multi-center cohorts, for example, clinical trials.
