AT(N) Framework in Older Adults with Epilepsy: Plasma Biomarkers and Associations with Demographic, Clinical, and Cognitive Features

Background and Objectives: Older adults with epilepsy have a 2- to 4-fold increased risk of dementia, including Alzheimer's disease (AD), yet underlying mechanisms remain poorly defined. The NIA-AA classifies AD using amyloid (A), tau (T), and neurodegeneration [(N)] biomarkers. We applied this framework to characterize AT(N) profiles and clinical correlates in epilepsy. Methods: Eighty-four older adults with focal epilepsy (mean age=66.3 years) from the Brain Aging and Cognition in Epilepsy (BrACE) study were classified as A+, T+, and/or (N)+ using plasma {beta}-amyloid (A{beta}) 42/40 ratio, phosphorylated tau 181 (p-tau181), and neurofilament light chain (NfL) levels, and grouped into normal, AD-continuum, and non-AD pathologic change. Demographic, clinical, and cognitive characteristics were compared. Cognition was assessed using the International Classification of Cognitive Disorders in Epilepsy (IC-CoDE) and the Montreal Cognitive Assessment (MoCA). Memory was examined using IC-CoDE memory domain classification, with word-list delayed recall analyzed separately. Associations with cognition were modeled using logistic and linear regression. Secondary analyses examined biomarkers continuously, including p-tau217, and substituted hippocampal volume for NfL. Results: Only 32% of participants had normal biomarkers, while 37% were on the AD-continuum and 31% showed non-AD pathologic change. Participants with normal biomarkers were younger with shorter epilepsy duration, whereas APOE-{epsilon}4 carriers were enriched in the AD-continuum group. Early-onset compared to late-onset epilepsy (cutoff: [≥]55 years) showed higher odds of biomarker abnormality (aOR=8.84, 95% CI [2.35, 41.89], P=0.003), driven by elevated p-tau217, NfL, and greater amyloid burden. While categorical AT(N) profiles were not associated with cognition, higher p-tau181 levels were independently associated with lower word-list delayed recall (95% CI [-10.31, -0.86], P=0.021). Substituting hippocampal volume for NfL shifted more participants to normal profiles (48% vs. 32%) and fewer to non-AD pathologic change (15% vs. 31%). Discussion: AT(N) biomarker profiles showed substantial heterogeneity, with higher abnormality rates than in aging populations, particularly among those with early-onset epilepsy. Continuous p-tau181 was associated with memory performance while categorical AT(N) profiles were not, and NfL and hippocampal volume showed discordant classifications, highlighting divergence across neurodegeneration markers. These findings underscore the complexity of applying AD-centric frameworks to epilepsy and support multimodal, epilepsy-adapted biomarker approaches to characterize neurodegenerative risk.