Search for a command to run...
Parkinson's disease varies widely in clinical manifestations, course of progression and biomarker profiles from person to person. Identification of distinct Parkinson's disease subtypes is of great priority to illuminate underlying pathophysiology, predict progression and develop more efficient personalized care approaches. There is currently no clear way to define and divide subtypes in Parkinson's disease. Using data from the Parkinson's Progression Markers Initiative, we aimed to identify distinct subgroups via cluster analysis of a comprehensive dataset at baseline (i.e. cross-sectionally) consisting of clinical characteristics, neuroimaging, biospecimen and genetic information, then to develop criteria to assign patients to a Parkinson's disease subtype. Four hundred and twenty-one individuals with de novo early Parkinson's disease were included from this prospective longitudinal multicentre cohort. Hierarchical cluster analysis was performed using data on demographic and genetic information, motor symptoms and signs, neuropsychological testing and other non-motor manifestations. The key classifiers in cluster analysis were a motor summary score and three non-motor features (cognitive impairment, rapid eye movement sleep behaviour disorder and dysautonomia). We then defined three distinct subtypes of Parkinson's disease patients: 223 patients were classified as 'mild motor-predominant' (defined as composite motor and all three non-motor scores below the 75th percentile), 52 as 'diffuse malignant' (composite motor score plus either ≥1/3 non-motor score >75th percentile, or all three non-motor scores >75th percentile) and 146 as 'intermediate'. On biomarkers, people with diffuse malignant Parkinson's disease had the lowest level of cerebrospinal fluid amyloid-β (329.0 ± 96.7 pg/ml, P = 0.006) and amyloid-β/total-tau ratio (8.2 ± 3.0, P = 0.032). Data from deformation-based magnetic resonance imaging morphometry demonstrated a Parkinson's disease-specific brain network had more atrophy in the diffuse malignant subtype, with the mild motor-predominant subtype having the least atrophy. Although disease duration at initial visit and follow-up time were similar between subtypes, patients with diffuse malignant Parkinson's disease progressed faster in overall prognosis (global composite outcome), with greater decline in cognition and in dopamine functional neuroimaging after an average of 2.7 years. In conclusion, we introduce new clinical criteria for subtyping Parkinson's disease based on a comprehensive list of clinical manifestations and biomarkers. This clinical subtyping can now be applied to individual patients for use in clinical practice using baseline clinical information. Even though all participants had a recent diagnosis of Parkinson's disease, patients with the diffuse malignant subtype already demonstrated a more profound dopaminergic deficit, increased atrophy in Parkinson's disease brain networks, a more Alzheimer's disease-like cerebrospinal fluid profile and faster progression of motor and cognitive deficits.