Objective: To identify subgroups of people living with Parkinson’s disease (PD) who have statistically distinct trajectories of Montreal Cognitive Assessment (MoCA) scores over four years, and to explore baseline features predictive of MoCA progression trajectories.

Background: Cognitive decline is one of the most common and detrimental non-motor symptoms in PD. Predicting progression of cognitive changes in PD remains an active area of research.

Method: Longitudinal data came from the Parkinson’s Progression Markers Initiative early de novo PD cohort. Latent class mixed modeling (LCMM) was used to identify PD progression subgroups demonstrating distinct patterns of MoCA scores over four years. We considered models with one to six latent classes (i.e., subgroups) and selected the model that provided the best fit to the data according to minimization of the Bayesian Information Criterion. We then trained a Random Forest (RF) classifier to predict subgroup classifications assigned by the LCMM. The data features were ranked by predictive power based on RF feature importance, and reduced RF models were then trained iteratively using the top one to 162 features.

Results: 413 participants with early PD were included. A three-class model was chosen to describe MoCA scores over 4 years. The three classes were categorized as rapid (mean rate of progression: -4 points/year), moderate (-1 point/year) and stable (-0.05 points/year) progression trajectories, and included 1%, 13%, and 86% of the cohort, respectively. Due to the low number of individuals in the rapid trajectory class, RF models were trained to predict either progressive (rapid or moderate) or stable trajectories. Model accuracy reached 93.7% using 162 different features, and 92.2% using the top 10 features. Compared to the stable trajectory group, those in the progressive group were older, reported more constipation problems, and performed worse on different measures of cognitive domains including memory, executive, and processing speed at baseline.

Conclusion: We identified distinct subgroups of cognition changes over 4 years and their predictors. Next steps include validation of subgroups and predictors on external data.

« Back to MDS Virtual Congress 2020

MDS Abstracts - https://www.mdsabstracts.org/abstract/predictors-of-different-progressive-trajectories-of-cognition-in-parkinsons-disease/