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

Commentary

In 1870, in A Study of Convulsions, John Hughlings Jackson argued that a convulsion was not a disease, but a symptom of disorderly cortical discharge, an idea later refined in his famous definition of epilepsy as a localized discharge of gray matter. ¹ It was a remarkable act of inference: Jackson placed epilepsy in the cortex decades before clinicians could examine the living brain in meaningful detail. A century and a half later, despite exquisitely advanced imaging, nearly half of adults with late-life seizures have no explanatory magnetic resonance imaging (MRI) lesion, an entity referred to as late-onset unexplained epilepsy (LOUE). ² However, Kek-Laflamme et al ³ challenge the premise that “unexplained” means structurally normal, reframing LOUE as a structural-cognitive syndrome rather than a diagnosis of absence. The gray matter Jackson placed at the center of epilepsy, it seems, may have been altered all along.

Using whole-brain surface-based morphometry, the authors compared 59 individuals with LOUE to 53 matched healthy older adults. ³ LOUE demonstrated reduced cortical thickness in bilateral sensorimotor and visual cortices and the left mesiotemporal lobe, along with reduced volumes in the pallidum and putamen. ³ The first advance is anatomical: LOUE extends beyond the mesiotemporal atrophy described in earlier, smaller studies. ⁴ Thus, LOUE may be less anatomically silent than previously recognized.

Equally compelling are the cognitive correlations of structural alterations. Individuals with LOUE scored lower than controls on delayed verbal memory, letter fluency, category fluency, and Preclinical Alzheimer's Cognitive Composite 5(PACC-5), a measure of early amyloid-related cognitive change. ³ Importantly, structural reductions correlated specifically with lower category fluency and global cognition.

This is the conceptual heart of the article. The relationship between late-onset epilepsy and cognitive decline has resisted simple chronology: Is epilepsy the driver, the consequence, or merely a fellow traveler? The authors suggest that the more relevant question may be one of the shared substrate: epilepsy and cognitive decline may be 2 visible expressions of a single process. ³ Category fluency depends on semantic access, executive retrieval, temporal-limbic integrity, and network efficiency. ⁵ Its association with structural reductions in LOUE points toward a process at the intersection of epilepsy and cognitive aging. The regions showing thinning, including the mesiotemporal cortex and networks supporting semantic retrieval and global cognition, are also among the territories most vulnerable in Alzheimer's disease (AD). ⁶ This overlap does not make LOUE prodromal AD. Rather, it reframes LOUE as a clinical waypoint where aging, epileptogenesis, neurodegeneration, vascular risk, inflammation, and diminished reserve may converge to lower seizure threshold while corroding cognitive resilience. ⁷

Age cautiously reinforces this possibility. Older age was associated with widespread cortical and subcortical reductions in LOUE but not in controls. ³ While the cross-sectional design forbids any claim about trajectory and the age-by-group interaction was not significant, the finding raises an important hypothesis: in LOUE, aging may interact with disease-specific vulnerability rather than simply mirror ordinary senescence.

The most puzzling finding is not loss, but apparent preservation or enlargement. LOUE showed increased thickness in the left inferior frontal gyrus and increased bilateral thalamic volume, with marked leftward asymmetry that did not correspond to seizure lateralization. ³ These increases were more pronounced in those with focal seizures with retained consciousness. The inferior frontal gyrus contains Broca's area, while the thalamus serves as a hub for cortical-subcortical integration. A compensatory interpretation is tempting: these regions may form scaffolding against seizure propagation. Alternatively, increased thickness or volume may reflect inflammation, remodeling, maladaptive plasticity, hyperexcitability, or relative preservation within an atrophying brain. ³ The absence of correlation with cognition further argues against a simple compensatory mechanism. Until longitudinal data show whether these regions persist, progress, or recede, these findings should be viewed with caution.

The asymmetry analyses further reveal the structural signature of LOUE. LOUE demonstrated greater leftward asymmetry in frontal regions, pallidum, insula, thalamus, and hippocampus. This is concordant with prior observations that LOUE may more commonly lateralize to the left hemisphere. ⁴ Whether this reflects the biological vulnerability of the dominant hemisphere or greater clinical detectability of left-sided seizures remains uncertain. The language-dominant hemisphere may be both more symptomatic when disrupted and more vulnerable to the interacting pressures of aging, seizure susceptibility, and cognitive decline. Notably, asymmetry did not simply map onto seizure focus; it may instead reflect a broader disease architecture.

Methodologically, the study has several strengths. The cohort is large relative to prior LOUE studies, prospectively recruited, and carefully phenotyped. A well-characterized aging cohort provides a strong control framework, and sensitivity analyses support the robustness of the principal findings. Long defined by absence, LOUE here begins to show what it contains.

The limitations temper, but do not diminish, the central insight. Because the study is cross-sectional, the proposed epilepsy-neurodegeneration continuum is inferred rather than observed. Whether structural alterations precede seizures, follow them, reflect medication exposure, or arise from shared pathology remains unresolved. Independent cohorts with distinct scanning protocols also introduce potential differences in recruitment and unmeasured characteristics. Controls were drawn from a healthy aging study with stringent cognitive criteria, whereas individuals with LOUE were clinically recruited, potentially amplifying group differences. Amyloid biomarker data were available for only 7 patients, leaving the neurodegenerative hypothesis untested. Finally, both cohorts were predominantly White, highly educated, and non-Hispanic, limiting generalizability, and the sample size was modest for subgroup analyses.

These caveats define the work ahead. LOUE research needs longitudinal imaging, multimodal biomarkers, vascular and inflammatory measures, electroencephalogram metrics, and repeated cognitive assessment. We need to know who remains stable, who progresses to dementia, and which imaging patterns predict seizure recurrence, treatment response, or decline. The field must also move beyond “lesional” and “nonlesional.”

For clinicians, LOUE should no longer function as a diagnosis of exclusion alone. It should prompt assessment of cognitive health, vascular burden, sleep, mood, and neurodegenerative risk. ⁸ If gray matter reductions track with PACC-5 and category fluency, baseline cognitive assessment is a tool for identifying individuals whose epilepsy is the leading edge of a broader brain disorder.^9,10 The label “unexplained” should therefore be held lightly, as a statement about the limits of current detection rather than underlying biology.

Jackson understood the seizure as a clue rather than a conclusion: a symptom of an underlying disease. ¹ Kek-Laflamme et al, ³ armed with tools he could not have imagined, extend that insight into the living gray matter itself. Their study moves LOUE out of the category of idiopathic late-life epilepsy and into a space where epileptogenesis and neurodegeneration intertwine. The cognitive correlations hint at networks under quiet siege before dementia is clinically apparent. The next task is longitudinal: to follow these brains over time. For now, the “unremarkable” scan has become remarkable. The gray matter is speaking; we are only beginning to learn how to hear it.