Objective: To characterize the cognitive profile of Parkinson’s Disease (PD) patients that develop probable REM sleep behavior disorder (pRBD) within 8 years after PD diagnosis versus those who did not and those who had pRBD at baseline.

Background: PD patients with RBD may experience a faster decline across cognitive domains [1,2]. However, it is unknown if cognitive impairments precedes RBD in PD patients or what kind of impairments those may be. Identifying the existence of cognitive impairments before RBD onset and determining the profile of these impairments may shed light on its pathogenesis.

Method: The extreme gradient boosting supervised machine learning model was used to assess PPMI data for which longitudinal variations in cognitive assessments were the most important factors associated with subjects who developed pRBD (RBDSQ score >5) after baseline. Assessments that were highly correlated with each other (>75%) were removed before training. The model was then split into 80% training and 20% testing. Mean scores of the top exams selected by the model were plotted over time for those without pRBD, those who developed pRBD after PD diagnosis, and those with pRBD at baseline, during 8 years of observation post PD diagnosis. Post hoc analysis with a Cox mixed effects model was used to determine which assessments were associated with the risk of developing pRBD after PD diagnosis.

Results: The extreme gradient boosted model selected several cognitive tests, based on gain, cover, and frequency, as the tests most associated with development of pRBD (Table 2) (AUC = 0.80). Mean scores for those who had pRBD at baseline, and those with and without conversion to pRBD, were plotted over time (Figures 1-2).  Cox-hazard analysis demonstrated statistically significant but weak inverse relationships between the Benton Judgement of Line Orientation and MOCA Memory scores and risk of developing pRBD (HR= 0.96, p=0.028; HR=0.72, p<0.001, respectively).

Conclusion: Higher scaled scores on the Benton and higher total scores on the MOCA Memory section had a weak correlation with lower risk of pRBD. The association of memory impairment with pRBD is of particular interest given the prior observation of CSF Aβ1–42 level’s association with cognitive decline in early PD with RBD [3], highlighting the potential role of amyloid co-pathology which warrants further investigation.

Demographics Table

XGBoost Results

Line Plots for Executive Function and Memory

Line Plots for Language and Attention

Cox Mixed Effects Model Results

References: 1. Jozwiak, N., Postuma, R. B., Montplaisir, J., Latreille, V., Panisset, M., Chouinard, S., Bourgouin, P. A., & Gagnon, J. F. (2017). REM Sleep Behavior Disorder and Cognitive Impairment in Parkinson’s Disease. Sleep, 40(8), zsx101. https://doi.org/10.1093/sleep/zsx101
2. Postuma RB, Gagnon JF, Montplaisir J. Cognition in REM sleep behavior disorder – A window into preclinical dementia?. Sleep Med. 2008;9(4):341-342. doi:10.1016/j.sleep.2007.08.001
3. Ba M, Yu G, Kong M, Liang H, Yu L. CSF Aβ1-42 level is associated with cognitive decline in early Parkinson’s disease with rapid eye movement sleep behavior disorder. Transl Neurodegener. 2018 Oct 8;7:22. doi: 10.1186/s40035-018-0129-5. PMID: 30338062; PMCID: PMC6174574.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/longitudinal-analysis-of-cognitive-changes-in-parkinsons-disease-subjects-at-risk-of-developing-probable-rem-sleep-behavior-disorder-rbd/