Peripheral GFAP Predicts Incident Dementia in Parkinson's Disease

Importance: Dementia is a common and disabling complication of Parkinson's disease (PD). Blood-based biomarkers that identify individuals at higher risk of future dementia could improve prognostication and trial stratification. Objective: To determine whether blood-based proteins are associated with future risk of dementia in PD. Design, Setting, and Participants: Prospective longitudinal cohort study with discovery and replication analyses in the Oxford Parkinson's Disease Centre (OPDC) Discovery cohort and the Parkinson's Progression Markers Initiative (PPMI). A total of 1,335 participants with PD and 431 healthy controls, with serum, plasma, or cerebrospinal fluid proteomic data and follow-up of up to 12 years, were included. Main Outcomes and Measures: Incident dementia, defined using a composite of MoCA scores, MDS-UPDRS items, and clinician diagnosis. Associations between baseline protein levels and time to dementia were evaluated. Results: Among 1,335 participants with PD, 168 developed incident dementia across cohorts (OPDC Discovery [serum], n = 108; PPMI Project 293 [plasma], n = 23; PPMI Project 181 [CSF], n = 37). In the OPDC cohort, GFAP was the only protein (of 5,408 tested) significantly associated with incident dementia (HR = 2.43; 95% CI: 1.79-3.30; p-adjust =1.35 x 10-4). Higher GFAP tertiles were associated with greater cumulative dementia incidence. Findings were replicated in the PPMI plasma project (HR = 2.42; 95% CI: 1.12-5.22; p = 0.024) but not in the CSF project (HR = 0.96; 95% CI: 0.66-1.39; p = 0.82). Higher baseline GFAP was significantly associated with lower baseline cognitive performance and greater longitudinal cognitive decline but no significant association with motor progression. In both cohorts, GFAP levels increased over time but showed no group-level differences. Conclusions and Relevance: Circulating GFAP is a robust and reproducible predictor of future dementia in PD, detectable early in the disease course. While the lack of differential longitudinal trajectories between PD patients with and without dementia suggests that GFAP does not act as a dynamic marker of cognitive decline, its relative stability supports its role as an early indicator of underlying biological vulnerability or subclinical pathology. These findings support serum GFAP as a promising, accessible biomarker for early dementia risk stratification.