In late-onset epilepsy of no clear cause, sleep spindles carry part of the cognitive decline

Design

Prospective observational study with cognitive testing, plasma biomarkers, 24-hour EEG with manual sleep scoring, and mediation analysis

Population

85 people with late-onset unexplained epilepsy (onset at 55 or older, no cortical lesion on MRI); mean age 71.3, 49% female

Primary outcome

Association of plasma p-tau217 with cognition (extended Preclinical Alzheimer Cognitive Composite) and with sleep microarchitecture

Effect

PACC5 beta -0.66 (95% CI -1.19 to -0.13, p=0.0017); refractory epilepsy interaction -1.49 (-2.94 to -0.05, p=0.04); 24% mediated by spindle-slow oscillation coupling (3-85%, p=0.032)

Late-onset unexplained epilepsy - first unprovoked seizures at 55 or over, with no cortical lesion on MRI - is associated with faster cognitive decline, and nobody has established why. This prospective study recruited 85 such patients (mean age 71.3, 49% female) for cognitive testing, plasma p-tau217 as a measure of Alzheimer pathology, and 24-hour EEG with manually scored sleep.

Higher plasma p-tau217 was associated with worse cognition on the extended Preclinical Alzheimer Cognitive Composite (beta -0.66 per log unit, 95% CI -1.19 to -0.13, p=0.0017), after adjusting for age, sex and education. Epilepsy severity modulated it: patients with both raised p-tau217 and medication-refractory epilepsy did worst (interaction beta -1.49, -2.94 to -0.05, p=0.04). Higher p-tau217 also went with reduced coupling between sleep spindles and slow oscillations, and mediation analysis attributed about 24% of the p-tau217-cognition association to that coupling (95% CI 3-85%, p=0.032).

The mediation estimate is the part to hold loosely. A confidence interval running from 3% to 85% tells you the analysis cannot distinguish a trivial mechanism from the dominant one, and mediation analysis in cross-sectional data assumes a causal ordering it cannot demonstrate - poor sleep architecture could as easily be a consequence of the pathology as the route through which it acts.

What survives is worth having. Plasma p-tau217 is measurable, was associated with cognition in this population, and identifies a group whose seizures are also harder to control. That argues for taking cognition seriously in late-onset epilepsy rather than attributing it to the drugs, and for asking about sleep - which is modifiable, unlike amyloid.

• Assess cognition formally in late-onset unexplained epilepsy rather than assuming it is drug-related
• Ask about sleep quality and screen for sleep apnoea, which is common and treatable in this group
• Refractory seizures plus cognitive complaints should raise the question of coexisting Alzheimer pathology
• Plasma p-tau217 is a research measure here - it is not yet a test to order in an epilepsy clinic
• Treat the 24% mediation figure as a hypothesis; the interval runs from 3% to 85%

The statistics, in plain English

A mediation estimate of 24% with a confidence interval of 3% to 85% is barely distinguishable from noise: it says some of the association may run through sleep architecture, and cannot say how much. The interaction term for refractory epilepsy has an upper bound of -0.05, only just excluding zero, so that subgroup finding needs replication. And with 85 participants, an association that is statistically significant can still be driven by a handful of individuals - which is why this reads as a direction for further work rather than a result to act on.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.