Wrist accelerometry and machine learning sensitively capture disease progression in prodromal Parkinson’s disease

Sensitive motor measures are needed to support trials in Parkinson's disease (PD). Wrist sensor data was collected continuously at home from 269 individuals with PD (106 with prodromal PD). Submovements were smaller, slower, and less variable in PD and prodromal PD. A machine-learned composite measure captured disease progression in prodromal PD more sensitively than the MDS-UPDRS Part III motor score. Wearable sensor-based measures may be useful in upcoming clinical trials.