Objective: To examine longitudinal change in excessive daytime sleepiness (EDS) and to determine clinical and biological variables that predict new onset EDS in the cohort of at baseline de novo Parkinson’s disease (PD) patients.

Background: There is a substantial body of literature on the prevalence of EDS in PD. However, there are limited data on the baseline predictors of EDS in subjects with at baseline de novo PD.

Methods: Parkinson’s Progression Markers Initiative (PPMI) is a longitudinal case-control study of de novo, untreated PD participants at enrolment. Participants contribute a wide range of motor and non-motor measures, including biofluids and imaging biomarkers. EDS was defined as Epworth sleepiness scale (ESS) score ≥ 10. The machine learning method of random survival forests was used to examine the ability of baseline variables to predict years to diagnosis of EDS since baseline.

Results: Of the 423 PD subjects enrolled, 353 did not have baseline EDS and were included in the analysis. During the observation period, 160 patients developed an EDS diagnosis (38%). We used 33 baseline variables of demographics, clinical and biologic disease characteristics. Based on in-sample concordance and cross-validated prediction accuracy the best model had seven variables (SCOPA, STAI-Trait, MDS-UPDRS Part I, STAI-State, MDS-UPDRS Total score, MDS-UPDRS Part II, and spinal fluids (CSF) ptau/ t-tau), and modestly prediced time to EDS (C=0.65, pseudo-R^2=0.03). Prediction using the seven variables was very similar to using the entire set of 33. Hazard ratios (HR) indicated that a higher relative risk of EDS was associated with higher scores on the seven predictors. The strongest effect was for p-tau/t-tau (an estimated HR of 2 (95% CI = [0.9, 4.41]). A prognostic index (PI) for risk of EDS was computed based on the seven-predictor Cox model estimates, and four risk groups of unequal size were formed. Survival curves varied by PI risk group, with higher PI scores indicating greater relative risk of EDS.

Conclusions: Our analysis using a novel machine learning method establishes seven variables that predict EDS in early PD. Interestingly, CSF ptau/ t-tau had the strongest effect. Our prognostic index constructed based on the group-level survival curve can provide an indication of the risk of EDS for early PD patients based on functions of the seven top predictors.

« Back to 2016 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/clinical-and-biological-predictors-of-excessive-daytime-sleepiness-in-early-parkinsons-disease/