Accelerometry-Derived REM Sleep Behavior Disorder Predicts Future Parkinson's Disease in the UK Biobank

Estimating Parkinson's disease (PD) risk years before diagnosis remains an unmet need. We applied a validated machine learning classifier for REM sleep behavior disorder (RBD) detection to 7-day wrist accelerometry data in 87,975 UK Biobank participants followed for 10 years. Participants in the highest RBD risk stratum (>99th percentile) had an approximately fivefold increased hazard of incident PD compared with the lowest-risk group (0-90th percentile), with a dose-dependent relationship across the full score distribution. Among non-converters, higher RBD risk was associated with baseline cognitive deficits and longitudinal enrichment of autonomic and psychiatric prodromal features. The association with incident PD remained independent of PD polygenic risk score, while RBD score and genetic risk were synergistic. The combined high-risk group achieved a positive likelihood ratio of 7.91, approximately threefold higher than questionnaire-based RBD screening. These findings support wrist accelerometry as a scalable approach for prodromal PD risk enrichment in population screening.