Anti-Seizure Medications — Proven
Sodium Channel Blockers PROVEN
Carbamazepine, lamotrigine, phenytoin, lacosamide. Block voltage-gated Na channels in high-frequency repetitive firing mode. First-line for focal epilepsy. Lamotrigine safe in pregnancy.
Multiple RCTs. Marson et al. 2007 Lancet (SANAD trial — carbamazepine vs others)
Valproate PROVEN
GABA enhancement + Na/Ca channel effects. Broad-spectrum (focal + generalized). MOST EFFECTIVE for generalized epilepsy including JME. Contraindicated in pregnancy (teratogen).
Marson et al. 2007 Lancet (SANAD) — valproate superior for generalized
Levetiracetam (Keppra) PROVEN
SV2A (synaptic vesicle) modulator. Unique mechanism. Broad-spectrum. Favorable safety/interaction profile. First-line in many centers. Psychiatric side effects (irritability) ~10%.
Ben-Menachem et al. 2000 (pivotal trial)
Surgery — Focal Epilepsy PROVEN
Temporal lobe resection: 60–70% seizure-free at 1yr (vs 8% medical). Anterior temporal lobectomy = most evidence. Requires presurgical evaluation (MRI, EEG, neuropsych, fMRI language).
Wiebe et al. 2001 NEJM (first RCT — surgery vs medical, n=80) · Engel et al. 2012 NEJM (early surgery)
Responsive Neurostimulation (RNS) STRONG
Closed-loop device: detects ictal onset EEG → delivers cortical stimulation to abort seizure. 50% responder rate in drug-resistant focal epilepsy. FDA approved 2013.
Nair et al. 2020 Epilepsia (long-term outcomes)
EEG Atlas — Canonical Patterns PROVEN
| 3Hz Spike-Wave | Absence seizure signature. Generalized, symmetric, bisynchronous. Onset/offset abrupt. Correlates with thalamo-cortical loop 3Hz resonance. Hyperventilation provokes. Penfield & Jasper 1954 (original description) |
| Polyspike + Wave | Myoclonic epilepsy signature (JME, PME). Multiple spikes before slow wave. Photoparoxysmal response in JME. Valproate-responsive. |
| Focal sharp waves | Temporal, frontal, or parietal depending on seizure focus. Interictal epileptiform discharges (IEDs) — guide surgical planning. Spike negativity = cortical hyperexcitability at electrode. |
| HFOs (80–500Hz) | High-frequency oscillations: ripples (80–250Hz) + fast ripples (250–500Hz). Ripples = physiological (memory) OR pathological. Fast ripples = pathological only. HFO resection predicts good surgical outcome. Staba et al. 2002 Science |
| Ictal onset | Low-voltage fast activity (LVFA) or theta/delta run at seizure onset. Evolves → slows → terminates → post-ictal suppression. Most diagnostic signal in SEEG for surgical planning. |
| Post-ictal suppress. | Diffuse voltage attenuation after GTC seizure. Caused by adenosine release + synaptic depression. Duration correlates with seizure severity. SUDEP risk marker: prolonged suppression. |
Thinkbeat Bridge — Most Literal
Epilepsy IS the 4-state classifier in its most literal form:
Interictal = SLOW_MONOTONE (baseline, IEDs present but controlled)
Pre-ictal = OSCILLATING (building synchrony, detectable by RNS)
Ictal = CHAOTIC (hypersynchronous discharge, the seizure itself)
Post-ictal = SLOW_MONOTONE recovery (suppression then return to baseline)
RNS closed-loop device = implemented δ control: detects OSCILLATING → intervenes before CHAOTIC.
Novel Patterns (8)
PROVENNP-EPI-001 · Epilepsy is the most literal EEG disease — every seizure type has a canonical waveform · EEG is both diagnostic AND mechanistic · Fisher 2014 Epilepsia
PROVENNP-EPI-002 · PV+ fast-spiking interneurons are the seizure brake · their selective vulnerability in multiple epilepsy syndromes (Dravet, TLE, post-status) · Cossart 2001 Science
PROVENNP-EPI-003 · SCN1A paradox: Na channel LOSS (not gain) → seizures via selective interneuron failure · Na blockers contraindicated in Dravet · Claes 2001
PROVENNP-EPI-004 · Surgery superiority: 60–70% seizure-free vs 8% medical therapy for drug-resistant TLE · Wiebe 2001 NEJM RCT
PROVENNP-EPI-005 · HFOs (fast ripples 250–500Hz) = most specific biomarker of epileptogenic zone · HFO resection = best surgical outcome predictor · Staba 2002 Science
OPENNP-EPI-006 · RNS closed-loop stimulation = implemented 4-state grammar control: OSCILLATING detected → stimulation before CHAOTIC · 50% responder rate · mechanism unclear
SORRYNP-EPI-007 · 30% drug-resistant — mechanism of pharmacoresistance unknown · multi-drug transporter hypothesis (MDR1/P-gp) vs target hypothesis · no validated biomarker predicts resistance
SORRYNP-EPI-008 · SUDEP (sudden unexpected death) mechanism — post-ictal cardiorespiratory suppression? · incidence 1:1000/yr in controlled, 1:150/yr in drug-resistant · no validated prevention strategy
Open Sorries (5)
SORRY-1: 30% drug-resistant — mechanism unknown. MDR1/P-glycoprotein overexpression vs target hypothesis (altered Na channels). No biomarker predicts who will become drug-resistant.
SORRY-2: SUDEP mechanism — post-ictal cardiac arrhythmia? Central apnea? Spreading depression? Incidence 1:1000–1:150/yr. No validated prevention.
SORRY-3: Surgical outcome prediction — MRI-negative epilepsy surgery success rate drops to 30–40%. Best predictor (HFO resection) still imperfect.
SORRY-4: Status epilepticus second-line treatment — after benzodiazepines fail: valproate vs levetiracetam vs lacosamide. Three head-to-head trials (ESETT) showed equivalent (all bad) efficacy for refractory SE.
SORRY-5: Ketogenic diet mechanism — reduces seizures in 50% of drug-resistant pediatric epilepsy. Why? GABA enhancement? Direct neuronal metabolic shift? Multiple proposed mechanisms, none definitive.