Objective: To use latent mixed modeling (LMM) for the identification of subgroups of people living with Parkinson’s disease (PD) who have statistically distinct trajectories of motor symptom scores.

Background: Clinical progression of motor symptoms in PD is characterized by substantial heterogeneity. Identification of PD progression subgroups with distinct trajectories of motor symptoms would provide a more refined characterization of disease progression. LMM is a statistical approach that is useful in identifying homogenous trajectory subgroups within a larger heterogeneous population.

Method: Longitudinal data came from the Parkinson’s Progression Markers Initiative early de novo PD cohort. LMM was used to identify PD progression subgroups demonstrating distinct patterns of change in motor severity over four years. Separate models were constructed for MDS-UPDRS part II, part III (untreated and OFF scores), and ambulatory capacity (questions 2.12, 2.13, 3.10, 3.11, 3.12) scores. To choose the optimal number of latent classes (i.e., subgroups), we estimated models with one to six classes for each outcome measure, and selected the models that provided the best fit to data according to minimization of the Bayesian Information Criterion. For each optimal model, we report the number of classes (with annualized change of score from baseline for each class; and percentage of individuals in each class).

Results: 413 participants with early PD were included. Four classes of MDS-UPDRS part II changes were identified: rapid (5 points/year; 4%), moderate (2 points/year; 29%), mild (0.4 points/year; 59%), and improved (-1 point/year; 8%). Five classes of MDS-UPDRS part III changes were identified: rapid (10 points/year; 5%), moderate (6 points/year; 19%), mild (3 points/year; 50%), minimum (0.3 points/year; 23%), and improved (-4 points/year; 3%). Five classes of ambulatory capacity changes were identified: rapid (3 points/year; 5%), moderate (1 point/year; 6%), mild (0.3 points/year; 45%), minimum (0.02 points/year; 42%), and improved (-0.6 points/year; 3%).

Conclusion: Distinct subgroups of changes in motor severity scores over 4 years were identified among an early PD cohort. Next steps include summary and comparison of patient characteristics within each subgroup, and validation of subgroups on external data.

« Back to MDS Virtual Congress 2020

MDS Abstracts - https://www.mdsabstracts.org/abstract/estimation-of-latent-class-mixed-models-towards-discovery-of-subgroup-trajectories-of-mds-updrs-part-ii-part-iii-and-ambulatory-capacity-in-parkinsons-disease/