P005 Comparison of auto-detected REM sleep without Atonia before- and after-editing

Introduction This study evaluates the impact of editing on auto-detection of REM sleep without atonia (RSWA) in three cohorts with different expected frequencies of abnormal RSWA. Methods Participants underwent in-home Sleep Profiler recordings with chin and arm EMG. Twenty-seven were diagnosed with REM sleep behavior disorder (RBD: women = 26%, age = 67+/-9.7), 45 with mild cognitive impartment (MCI = 39) or Alzheimer’s disease (AD = 6) (MCI/AD: women = 56%, age = 72+/-8.5) and 27 controls (CG: women = 70%, age = 64+/-11.9). Recordings were auto-scored for stage REM and phasic- or tonic-RSWA (“Auto”) using previously described procedures and then visually edited (“Edited”). Abnormal RSWA densities were assigned based on previously determined auto-detection thresholds for the Chin (>15%), Arm (>12.5%) and chin and/or arm (“Any” > 25%). Results In the RBD group, 18/22 (82%) participants were identified with abnormal Auto Chin-RSWA vs. Edited = 21/26 (81%), while 20/24(83%) had abnormal Auto Arm-RSWA vs. Edited = 24/27 (89%), and 18/24 (75%) had abnormal Any-RSWA vs. Edited = 23/27 (85%). For the MCI/AD cohort, 9/45 (20%) had abnormal Auto Chin-RSWA vs. edited = 7/45 (16%), while 4/45 had abnormal auto Arm-RSWA (9%) vs. Edited = 2/45 (4%), and 6/45 (13%) had Auto and Edited abnormal Any-RSWA. For the control group, Auto and Edited Chin-RSWA densities were abnormal in 3/27 (11%) and 4/27 (15%), while 4/27 (15%) had abnormal Auto Arm-RSWA vs. Edited = 3/27 (11%), and abnormal Auto Any-RSWA was observed in 2/27 (7%) vs. Edited = 3/27 (11%). Conclusions The accuracy of the auto-scoring suggests a foundational base that could help limit between-site variability during large-scale RBD screening in future multicenter clinical trials and population-based research studies.