Objective: To evaluate feasibility and performance of multi-night at-home recordings using a smartwatch to (1) detect REM sleep behavior disorder (RBD) compared with polysomnography (PSG) and (2) assess performance relative to a screening questionnaire (REM Sleep Behavior Disorder Screening Questionnaire; RBDSQ).

Background: RBD is a prodromal manifestation of Parkinson’s disease (PD) and other synucleinopathies. Identification RBD relies on PSG, which is typically limited to one or two nights despite substantial night-to-night variability of symptoms. Accessible and reliable approaches for longitudinal screening are needed. This study is part of AI-PROGNOSIS, a research project funded by the European Union (Grant Agreement No. 101080581).

Method: The digital biomarkers development study (dBM-DEV) recruited 81 participants across three European sites (Germany, Spain, France). The development cohort included 29 participants with PSG-confirmed RBD (27 idiopathic RBD [iRBD], 2 PD with RBD) and 24 controls. The confirmation cohort included 28 PD participants with unknown RBD status. Participants wore a smartwatch nightly for up to three months. PD participants underwent PSG. RBD-associated features were extracted, and machine-learning algorithms were developed for subject-level RBD detection. In PD participants, smartwatch-derived digital biomarker (dBM) results were compared with RBDSQ (cut-off ≥6) and one-night PSG.

Results: Controls and iRBD participants in the development cohort were classified with high accuracy using seven nights of data (88%, XGBoost). In the confirmation cohort, 22 of 23 PD participants with sufficient smartwatch data were classified as dBM-positive. PSG was performed in 24 PD participants. Of 16 conclusive PSGs, 13 were RBD-positive and 3 RBD-negative, while 8 were inconclusive. The observed PSG-based RBD prevalence in PD was 81%. The pre-specified endpoint (higher specificity of dBMs vs RBDSQ) was inconclusive due to too few PSG-negative cases. Sensitivity was not statistically different (dBM 91% vs RBDSQ 100%).

Conclusion: Multi-night smartwatch recordings are feasible and show high sensitivity for RBD detection in real-world settings. The high prevalence of RBD in PD participants and frequent inconclusive single-night PSG findings support longitudinal dBM assessments, which may aid identification of individuals with RBD for clinical trials. Further studies are needed to test the specificity of dBMs.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/smartwach-actigraphy-for-detection-of-rem-sleep-behavior-disorder-the-dbm-dev-study/