Latent Class Trajectory Modeling Identifies Cerebrospinal Fluid Protein Signatures Associated with Motor Progression in Parkinson's Disease: Evidence from the Parkinson's Progression Markers Initiative Cohort

BACKGROUND: Motor progression in Parkinson's disease (PD) is highly heterogeneous, limiting prognosis and the efficiency of clinical trials that require accurate risk stratification. Traditional modeling approaches assume a single average trajectory and fail to capture meaningful divergence, whereas the biological determinants of trajectory-defined subtypes remain unclear. Multivariate proteomic biomarkers may improve early identification of rapid progressors. OBJECTIVES: We aimed to (1) delineate distinct motor progression subtypes in early PD using longitudinal Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) III scores and (2) identify cerebrospinal fluid (CSF) proteomic signatures that discriminate rapid from slow progressors. METHODS: We analyzed 5-year longitudinal data from 396 Parkinson's Progression Markers Initiative (PPMI) participants with early PD. Motor progression trajectories were derived using latent class trajectory modeling (LCTM) of repeated MDS-UPDRS III scores. CSF proteomics from 78 Olink samples were used for biomarker discovery, with least absolute shrinkage and selection operator (LASSO)-based feature selection followed by multivariable logistic regression. Validation was performed in 184 SOMAscan-profiled samples. RESULTS: LCTM identified distinct motor progression subtypes in early PD. Differential CSF protein expression across these trajectories highlighted candidate markers of rapid decline. A seven-protein signature (CCL27, CXCL9, NEFL, SERPINE1, GH1, TYMP, KLK1) exhibited good discriminative performance for distinguishing rapid from slow progressors (area under the curve [AUC] = 0.908). In an independent cohort, the combined model showed consistent directional effects and achieved the highest accuracy (AUC = 0.796). CONCLUSIONS: Integrating trajectory modeling with CSF proteomics identified a concise seven-protein signature associated with early PD progression, which showed reproducible patterns across two PPMI cohorts. This approach may support improved risk stratification in PD.