Replication and reliability of Parkinson's disease clinical subtypes

• Published in: Parkinsonism & related disorders Vol. 124; p. 107016
• Main Authors: Cash, Therese V., Lessov-Schlaggar, Christina N., Foster, Erin R., Myers, Peter S., Jackson, Joshua J., Maiti, Baijayanta, Kotzbauer, Paul T., Perlmutter, Joel S., Campbell, Meghan C.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.07.2024
• Subjects: Classification; Latent class analysis; Parkinson disease; Psychometrics; Parkinson disease; Psychometrics; Latent class analysis; Classification
• ISSN: 1353-8020; 1873-5126; 1873-5126
• DOI: 10.1016/j.parkreldis.2024.107016

We recently identified three distinct Parkinson's disease subtypes: “motor only” (predominant motor deficits with intact cognition and psychiatric function); “psychiatric & motor” (prominent psychiatric symptoms and moderate motor deficits); “cognitive & motor” (cognitive and motor deficits). We used an independent cohort to replicate and assess reliability of these Parkinson's disease subtypes. We tested our original subtype classification with an independent cohort (N = 100) of Parkinson's disease participants without dementia and the same comprehensive evaluations assessing motor, cognitive, and psychiatric function. Next, we combined the original (N = 162) and replication (N = 100) datasets to test the classification model with the full combined dataset (N = 262). We also generated 10 random split-half samples of the combined dataset to establish the reliability of the subtype classifications. Latent class analyses were applied to the replication, combined, and split-half samples to determine subtype classification. First, LCA supported the three-class solution – Motor Only, Psychiatric & Motor, and Cognitive & Motor– in the replication sample. Next, using the larger, combined sample, LCA again supported the three subtype groups, with the emergence of a potential fourth group defined by more severe motor deficits. Finally, split-half analyses showed that the three-class model also had the best fit in 13/20 (65%) split-half samples; two-class and four-class solutions provided the best model fit in five (25%) and two (10%) split-half replications, respectively. These results support the reproducibility and reliability of the Parkinson's disease behavioral subtypes of motor only, psychiatric & motor, and cognitive & motor groups. •Clinical subtypes—“Motor Only”, “Psychiatric & Motor”, and “Cognitive & Motor”—were reproduced using latent class analyses.•Reliability analyses supported the robustness of the three clinical subtypes driven by psychiatric and cognitive features.•With a larger, combined sample, a “Severe Motor” group emerged, while the original three clinical subtypes were preserved.