The spectrum of cognitive disorders in Parkinson's disease: a data-driven approach

Abstract

The objective of this study was to identify different cognitive phenotypes in Parkinson's disease (PD) using a data-driven approach. A model-based cluster analysis was conducted on the neuropsychological test results of 558 PD patients from 2 European movement disorder centers (Lille, n = 403; Maastricht, n = 155). The number of clusters was determined according to their clinical relevance as well as on the basis of 3 statistical criteria: the cubic cluster criterion, the pseudo F statistic, and the total squared correlation ratio (R(2)). A factorial discriminant analysis was performed to assess the quality of the cluster's separation. Descriptive variables were used to further characterize the clusters. A 5-cluster model was considered the clinically most relevant. The 5 clusters comprised: (1) cognitively intact patients (19.39%); (2) patients without cognitive deficits but with slight mental slowing (41.29%); (3) patients with slightly impaired overall cognitive efficiency and deficits in all cognitive domains except recognition memory (12.93%); (4) patients with severe mental slowing, impaired overall cognitive efficiency, and severe cognitive impairment in all domains, including memory (23.88%); and (5) patients with very severe impairment in all cognitive domains (2.51%). Cognitively intact patients were significantly younger and had received more years of formal education. Patients in the last 3 clusters had more severe motor symptoms, longer disease duration, and more axial signs. In the last cluster, most patients were demented. Our results confirm the heterogeneity of cognitive presentations in PD, ranging from cognitively intact patients with rather high levels of performance in each cognitive domain to very severely impaired patients.