Data-driven characterization of distinct cognitive subtypes in Parkinson's disease dementia

Individual cognitive profiles of patients with Parkinson's disease dementia (PDD) are highly heterogeneous, suggesting possible biological subtypes. We studied 75 PD patients who developed dementia in the course of the Parkinson's Progression Markers Initiative study to investigate data-driven evidence for the existence of distinct cognitive subtypes of PDD. Using Ward's hierarchical clustering on neuropsychological test data, we identified two distinct cognitive subtypes. Despite similar dementia severity (MoCA: 20.6 vs 20.0), cluster-A exhibited more pronounced memory deficits (n = 50), whereas cluster-B showed greater visuospatial impairments (n = 25). The subtypes did not differ in demographic, motor, or MRI-based neurodegeneration measures. However, the visuospatial-predominant cluster-B had a higher prevalence of GBA mutations (p = 0.003) and hallucinations (p = 0.009). No differences were found in APOE-ε4 prevalence or cerebrospinal fluid biomarkers of Alzheimer's pathology. These findings reveal distinct memory-predominant and visuospatial-predominant PDD subtypes, which associate with different clinical and genetic features but are independent of comorbid Alzheimer's pathology.