Progression subtypes in Parkinson's disease identified by a data-driven multi cohort analysis

Area of Science:

• Neurology
• Data Science
• Clinical Trials

Background:

• Parkinson's disease (PD) progression varies significantly among patients, complicating treatment strategies and clinical trial design.
• Distinct PD subtypes may necessitate tailored therapeutic approaches.

Purpose of the Study:

• To identify and validate distinct Parkinson's disease progression subtypes using a data-driven approach.
• To investigate if observed heterogeneity in PD can be attributed to underlying progression subtypes.

Main Methods:

• Analysis of multimodal longitudinal data from three large PD cohorts with cross-cohort validation.
• Application of a latent time joint mixed-effects model (LTJMM) for timescale alignment.
• Identification of progression subtypes using variational deep embedding with recurrence (VaDER).

Main Results:

• Two stable PD progression subtypes (fast and slow) were identified across cohorts.
• Subtypes differed in motor/non-motor symptoms, survival, treatment response, imaging, gait, and Alzheimer's pathology.
• Predictive models achieved an ROC-AUC of 0.79 for individual patient subtype prediction.
• Simulations indicated a 43% reduction in clinical trial size by enriching fast-progressing patients.

Conclusions:

• PD heterogeneity can be explained by two distinct, stable progression subtypes.
• These subtypes may align with the brain-first vs. body-first concept, offering biological insights.
• Predictive models can optimize clinical trials by enabling enrichment of fast-progressing patient cohorts.