Association Between Plasma GFAP and Medial Temporal Atrophy and Cognition in Amyloid-Negative Cerebrovascular Disease

Plasma glial fibrillary acidic protein (GFAP), a marker of astrocyte reactivity, is elevated across multiple neurodegenerative conditions, including Alzheimer disease. However, its role in neurodegeneration and cognitive decline driven by cerebrovascular pathology, independent of β-amyloid (Aβ) copathology, remains poorly characterized. We investigated whether plasma GFAP is associated with medial temporal atrophy and cognition across a spectrum of cerebrovascular burden in Aβ-negative cognitively impaired individuals.

In this cross-sectional multicenter study, Aβ PET-negative cognitively impaired participants were recruited from South Korean memory clinics. Plasma GFAP was measured using ultrasensitive Simoa assays. White matter hyperintensity burden was graded using the Fazekas scale and stratified into low (LVP: Fazekas 1) and high (HVP: Fazekas 2-3) cerebrovascular burden groups. Medial temporal gray matter density was assessed using voxel-based morphometry, and hippocampal and amygdalar volumes were derived from T1-weighted MRI adjusted for intracranial volume. Linear regression, interaction, and bootstrap mediation models were used to assess associations among GFAP, brain structure, and cognition.

A total of 324 participants were included (LVP n = 203; HVP n = 121; median age 73 years [interquartile range 66-78]; 67.9% female). Compared with LVP, HVP participants were older (75 vs 71 years; p < 0.0001), had lower Mini-Mental State Examination (MMSE) scores (22.7 vs 24.5; p = 0.005), and higher plasma GFAP (136.7 vs 112.1 pg/mL; p = 0.001). Higher GFAP was associated with lower medial temporal gray matter density in HVP (β = -0.311; p = 0.001) but not LVP (β = -0.012; p = 0.858), with a significant GFAP-vascular burden interaction (β = -0.309; p = 0.008). In HVP, higher GFAP was associated with smaller hippocampal (β = -0.179; p = 0.044) and amygdalar volumes (β = -0.169; p = 0.049) and lower MMSE (β = -0.194; p = 0.039). Medial temporal atrophy statistically explained the GFAP-MMSE association (indirect β = -0.071, 95% CI -0.140 to -0.010; p = 0.016). Vascular comorbidities (diabetes, dyslipidemia, hypertension) did not modify the GFAP-cognition association.

In Aβ-negative cognitively impaired individuals with high cerebrovascular burden, elevated plasma GFAP is associated with medial temporal atrophy and cognitive decline, suggesting GFAP may capture astrocyte-reactivity relevant to vascular cognitive impairment beyond amyloid pathology. These cross-sectional findings require confirmation in longitudinal and ethnically diverse cohorts.