< back to all PPMI Presentations

POSET Based Cognitive Function Impairment (PCFI): A Novel Approach For Delineating Heterogeneity Of Cognitive Impairment In Parkinson's Disease

Gupta, D.K.; Goldman, J.G.; Jaeger, J; Tatsuoka, C
MDS: San Diego, CA
Download Presentation: Gupta-et-al.-MDS-2015.pdf
To apply partially ordered set (POSET) modeling for classifying impairment in cognitive functions in Parkinson's disease (PD).

Mild cognitive impairment in PD (PD?MCI) is common and heterogeneous in nature. Current paradigms of classifying cognitive impairment lack specificity with regards to particular cognitive functions as many neuropsychological measures tap into several different or overlapping cognitive domains. POSET models serve as a basis for novel methods to classify the performance of subjects with respect to specific cognitive functions.

De novo PD subjects (n=418) and healthy controls (HC), n=195) from the Parkinson's Progression Marker Initiative (PPMI) study were studied. PD subjects were classified as PD?MCI level Ia (Montreal Cognitive Assessment, MoCA) and PD?MCI level Ib (< 5 domains tested) as per the International Parkinson and Movement Disorders Society (MDS) Task Force PD?MCI criteria. Based on expert opinion, we mapped four specific cognitive functions: Attention (ATTN), Visuospatial Judgment (VSJ), Cognitive Flexibility (CogFlex) and Episodic Memory (EM), to different cognitive test scores and selected MoCA subscores (table 1). Cognitive function scores, representing posterior probabilities of a subject's highest level of functioning, were derived for each subject through POSET models and Bayesian analysis. Based on 10th percentile cutoff of cognitive function scores of HC, and reported cognitive decline (from MDS?UPDRS Part I), the impairment in each cognitive function was classified and termed as POSET based cognitive function impairment (pCFI). We used t?tests and chi?square statistics for testing significant (p?
Cognitive test scores were significantly worse in PD compared to HCs for all measures except two (table 1). Although cognitive function scores were not significantly different in PD from HC, all four pCFI types occurred more significantly in PD (table 2). EM was the most frequently impaired cognitive function. The number of subjects with pCFI in one, two, three and all four domains was also significantly higher in PD compared to controls.

We report for the first time a novel approach for distinguishing different cognitive phenotypes of PD?MCI. This may have important clinical and research implications (e.g., biomarkers discovery for PD?MCI).