Between Epilepsy and Neurodegeneration: The Hidden Architecture of Late-Onset Unexplained Epilepsy

Brain Structural Changes and Cognitive-Clinical Profiles in Late-Onset Unexplained Epilepsy Kek-Laflamme A, Schaper FLWVJ, Whittingstall K, Pennell PB, Young GS, Marshall GA, Larivière S, Sarkis RA. Neurology. 2026;106(3): e214575. doi: 10.1212/WNL.0000000000214575 . Epub 2026 Jan 9. PMID: 41512209; PMCID: PMC12805585 Background and objectives: Epilepsy incidence peaks in childhood and again after age 55. Up to half of individuals with late-onset epilepsy lack an identifiable cause, but previous imaging studies suggest subtle mesiotemporal atrophy. In this study, we aimed to quantify whole-brain cortical and deep gray matter alterations in late-onset unexplained epilepsy (LOUE) and to examine their associations with cognitive and clinical measures. Methods: We prospectively recruited patients with LOUE through Brigham and Women's Hospital and its affiliated sites and compared them with sociodemographically matched healthy older adults from the Harvard Aging Brain Study. Inclusion criteria for LOUE were at least 1 unexplained seizure after age 55, onset within the past 5 years, and no identifiable cortical lesion on magnetic resonance imaging (MRI). All participants underwent 3T structural MRI and neuropsychological testing. Cortical thickness and deep gray matter volumes were extracted, and linear models were used to compare patients with LOUE and controls. Correlations were computed between structural alterations and demographic, cognitive, and clinical measures. Results: We included 59 patients with LOUE (mean age 71.2 ± 7.0 years, 49% female) and 53 controls (mean age 70.7 ± 5.2 years, 53% female). Patients with LOUE showed reduced cortical thickness in sensory and mesiotemporal cortices ( d = −0.75, 95% confidence interval [CI] [−1.14 to −0.37], FDR-corrected P values [ P FDR ] < .05) and reduced deep gray matter volumes in the pallidum and putamen ( d = −0.55, 95% CI [−0.93 to −0.17], P FDR <.05). These structural reductions correlated with lower performance on a category fluency task ( r = 0.31, P FDR = .016) and the extended Preclinical Alzheimer's Cognitive Composite ( r = 0.37, P FDR = .0041). Conversely, patients showed increased thickness in the left inferior frontal gyrus ( d = 0.82, 95% CI [0.43 to 1.20], P FDR < .05) and increased thalamic volume ( d = 1.13, 95% CI [0.73 to 1.53], P FDR < .05), which were more pronounced in those with focal seizures sparing consciousness ( d = −0.62, P uncorr = .021). Discussion: Structural brain changes in LOUE are more extensive than previously recognized and are associated with cognitive vulnerability. Although the cross-sectional design and use of independent cohorts limit conclusions about disease progression, the findings suggest that LOUE may fall along a continuum between epilepsy and neurodegenerative disease.