Abstract
Shandilya MCV, Addo-Osafo K, Ranasinghe KG, Shamas M, Staba R, Nagarajan SS, Vossel K. Brain Commun. 2025;7(1):fcaf041. doi:10.1093/braincomms/fcaf041 Alzheimer's disease increases the risk of developing epilepsy together with cognitive decline. Early diagnosis or prediction of parameters associated with epileptic activity can greatly help in managing disease outcomes. Network hyperexcitability is a candidate of interest as a neurophysiological biomarker of Alzheimer's disease. High-frequency oscillations are increasingly recognized as potential biomarkers of hyperexcitability and epileptic activity. However, they have not yet been identified in Alzheimer's disease. In this study, we measured high-frequency oscillations via magnetoencephalography recordings in Alzheimer's disease patients with and without epileptic activity, as part of a Phase 2a randomized, double blind clinical trial of the efficacy of levetiracetam to improve cognitive functions in Alzheimer's disease. To measure the high-frequency oscillations, we used 10-min magnetoencephalography recordings (275-channel and sampling rate 1200-4000 Hz) during awake resting periods in participants with Alzheimer's disease and healthy controls. Recordings from 14 Alzheimer's disease participants, with six having non-epileptic Alzheimer's disease (median age: 60.8, 2 M/4 F), eight having sub-clinical epileptic activity (median age: 54.9, 5 M/3 F) and eight as control (median age: 71, 5 M/3 F), were analysed using two software scripts: Delphos and a custom-made script, for detecting high-frequency oscillations. Levetiracetam 125 mg twice-a-day or placebo was administered for 4 weeks in between two magnetoencephalography recordings, and 4 weeks of washout before switching levetiracetam/placebo phases for each participant. High-frequency oscillations were categorized into ripples (80 to 250 Hz) and fast ripples (250 to 500 Hz). At baseline, Alzheimer's disease participants, both epileptic and non-epileptic had higher rate of ripples and fast ripples than controls in several left/right hemispheric sensor regions (P < 0.05). Additionally, compared to epileptic, non-epileptic had higher rate of ripples in left-frontal, left-temporal and cerebral fissure regions and higher rate of fast ripples in left-frontal regions (P < 0.05). In epileptic type, levetiracetam decreased ripples in bilateral-frontal, bilateral-occipital regions and cerebral fissure, whereas in non-epileptic type, levetiracetam increased both ripples and fast ripples in right central and left parietal regions, and ripples in the right parietal region (P < 0.05). Additionally, we found hemisphere asymmetry in epileptic type, with right temporal/occipital having more high-frequency oscillations than their counterpart region. Overall, Alzheimer's disease had a high level of high-frequency oscillations, with higher numbers observed in non-epileptic type. Levetiracetam decreased high-frequency oscillations in epileptic but increased high-frequency oscillations in non-epileptic. Thus, high-frequency oscillations can function as a biomarker of hyperexcitability in Alzheimer's disease and may be more pathological when asymmetric and coinciding with presence of epileptic activity. Levetiracetam has the potential for treating hyperactivity in patients with epileptic Alzheimer's disease.
“Doctor, why does my wife with ‘just memory problems’ now have seizures?”
This anxious question is gradually becoming increasingly common as we recognize that around 5% to 15 % of people with Alzheimer's disease (AD) will develop overt epilepsy, and many more harbor subclinical epileptiform discharges
The authors analyzed 10-minute resting-state magnetoencephalography (MEG) segments from 14 participants with mild AD (8 with subclinical epileptiform activity, “EAD,” and 6 without, “NEAD”) and 8 cognitively normal controls. MEG recordings were acquired during a double-blind cross-over trial of low-dose levetiracetam (LEV, 125 mg bid) previously shown to improve spatial memory in EAD. 4 Two automated detectors (Delphos and a custom MATLAB script) quantified ripples and fast ripples across 11 regional sensor groups. Each AD participant served as their own control for drug effects, with four MEG sessions (baseline, post-LEV, wash-out, post-placebo) spaced 4 weeks apart.
There were several key findings. Both AD subgroups exhibited higher ripple and fast-ripple rates than controls across bilateral temporal, parietal, and occipital sensor regions, with the greatest burden in right temporal and occipital regions . NEAD showed even higher ripple rates in left frontal and temporal regions and more fast ripples in left frontal regions than EAD. EAD alone demonstrated striking right-hemisphere asymmetry of HFOs, echoing prior MEG reports of lateralized epileptiform discharges in AD. 5 LEV suppressed ripples in EAD (bilateral frontal and occipital regions cerebral fissure) yet paradoxically increased both ripple and fast-ripple rates in NEAD (right-central and left parietal) and additionally ripples in right parietal regions. Controls were older than AD participants, yet still had fewer HFOs, arguing against age alone driving the signal.
HFOs have been promoted as electrophysiological “fingerprints” of epileptogenic tissue in drug-resistant focal epilepsy, with fast ripples especially enriched at seizure-onset zones and predictive of surgical outcome.6,7 Their detection outside traditional spike analysis could expose otherwise invisible hyperexcitable circuits. Shandilya et al extend that concept to a neurodegenerative context, demonstrating that HFOs are detectable non-invasively in mild AD using brief, awake MEG recordings. Higher rates and hemispheric asymmetry track with subclinical epileptiform activity suggesting that not all ripples are created equal. Rapid modulation by LEV (a synaptic-vesicle protein 2A ligand) implies that at least a subset of AD-related HFOs reflect dynamic synaptic disinhibition rather than static neurodegeneration.
Strengths of the study include the first-ever demonstration of HFOs in AD, and innovative repurposing of an existing cross-over trial yields within-subject drug/placebo comparisons without additional participant burden. Prior work from the same cohort reported LEV-related improvements in executive and spatial navigation performance only in EAD and the current electrophysiological data offer a plausible mechanistic correlate. 4
Limitations of the study include small, demographically skewed cohort where controls were ≈12 years older and recordings were limited to a single center, tempering generalizability. Ten-minute wake recordings risk under-sampling sleep-related ripples, which often dominate HFO spectra. Longer, multi-state acquisitions and simultaneous scalp EEG will be needed for clinical translation.
LEV 125 mg bid is a fraction of typical anti-seizure doses; the heterogeneous regional response hints that higher or individualized dosing may be required to uniformly dampen hyperexcitability. The LEV-induced rise in HFOs where epileptiform activity is absent is provocative. Differential SV2A expression, network compensation, or detector misclassification of physiological ripples are all on the table and mandate caution before blanket LEV use in all AD.
For the practicing neurologist, these data nudge us toward incorporating advanced electrophysiology when faced with puzzling cognitive fluctuations or behavioral spells in AD. While MEG remains confined to specialized centers, scalp EEG HFO detection is improving and could soon complement overnight EEG for risk stratification. 7 Importantly, the bidirectional drug effect underscores that “one size fits all” therapy is unlikely; biomarkers that distinguish harmful from compensatory hyperactivity will be essential before widespread prophylactic anti-seizure treatment.
Future directions include validation in larger, more diverse cohorts, including Black persons who shoulder disproportionate AD burdens. Longitudinal tracking to determine whether fast-ripple predominance in NEAD forecasts later seizure emergence is a fascinating possibility. Multimodal coupling of HFO metrics with tau PET, SV2A PET, and intracranial EEG could unravel whether ripples accelerate or merely report tau spread. 8 Therapeutic trials that titrate LEV (or alternative agents) to electrophysiological targets rather than clinical seizures alone might finally test whether dampening hyperexcitability slows cognitive decline.
Shandilya et al shine a spotlight on millisecond-scale biomarkers lurking beneath the slow swell of amyloid and tau. Their findings suggest that HFOs, when asymmetric, abundant, and suppressible, may mark the very circuits that tip mild Alzheimer disease toward epilepsy and faster dementia. The work is an invitation to listen more closely to the brain's high-frequency whispers; what we hear could guide a new era of precision neuromodulation in neurodegenerative disease.
Footnotes
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
