Is this event actually a seizure?
Most misdiagnosed "epilepsy" is something else entirely. Before you ever classify a seizure, decide whether the paroxysmal event was epileptic at all. The two big pretenders are syncope and psychogenic non-epileptic events.
The single most useful tool here is a careful witnessed history — a phone video of the event is worth more than any test you can order in clinic. Anchor the interview on what happened before, during, and after.
| Feature | Epileptic seizure | Convulsive syncope | Psychogenic (PNES) |
|---|---|---|---|
| Trigger / prodrome | Sometimes aura (rising epigastric, déjà vu, focal sensory) | Upright posture, pain, heat, standing; greyout, tunnel vision, warmth, nausea | Emotional context; often prolonged, fluctuating build-up |
| Onset | Abrupt | Gradual over seconds | Variable, may be gradual |
| Motor pattern | Rhythmic, synchronous, decrementing; tonic then clonic | Brief (<15 s) arrhythmic multifocal jerks after collapse | Asynchronous, side-to-side head, pelvic thrust, waxing/waning, stop-start |
| Eyes | Open, may deviate | Open, rolled up | Often forcibly closed, resist opening |
| Duration | 1–3 min typical | Seconds | Often >2 min, can be very long |
| Tongue bite | Lateral tongue (specific) | Rare; tip if any | Tip of tongue, lips |
| Recovery | Post-ictal confusion, gradual over minutes | Rapid, oriented within seconds | Variable; may have rapid reorientation or prolonged unresponsiveness |
| Incontinence | Can occur | Can occur | Can occur — not discriminating |
Convulsive syncope is the most common false-positive. Brief jerks and even incontinence happen with any cause of cerebral hypoperfusion — they do not make the event a seizure. Ask what came first: collapse then jerks (syncope) or jerks then fall (seizure). PNES and epilepsy co-exist in a meaningful minority of patients; the presence of one does not exclude the other, and definitive diagnosis usually needs video-EEG.
Does this patient have epilepsy?
A single seizure is not epilepsy. The ILAE gives a practical, operational definition — and, importantly, an updated one that lets you diagnose epilepsy after a first unprovoked seizure when recurrence risk is high.
The ILAE 2014 operational definition
Epilepsy is a disease of the brain defined by any one of the following three conditions:
- At least two unprovoked (or reflex) seizures occurring more than 24 hours apart.
- One unprovoked seizure and a probability of further seizures similar to the general recurrence risk after two unprovoked seizures (at least ~60%) over the next 10 years.
- Diagnosis of an epilepsy syndrome (e.g., juvenile myoclonic epilepsy on the strength of history + EEG, even after one clinical event).
Provoked vs. unprovoked — this changes everything
An acute symptomatic (provoked) seizure occurs in close temporal relationship to an acute insult and does not by itself indicate epilepsy. Treat the cause, not the "epilepsy." Common provokers:
- Metabolic: hyponatremia (<115), hypoglycemia, hypocalcemia, uremia, hepatic failure.
- Toxic / withdrawal: alcohol withdrawal (classically 6–48 h), benzodiazepine withdrawal, cocaine, bupropion, tramadol.
- Acute brain insult within ~7 days: stroke, trauma, CNS infection, hypoxic-ischemic injury.
- Fever in a young child meeting febrile seizure criteria.
Definition #2 is where primary-care neurology earns its keep. After a first unprovoked seizure, features that push recurrence risk toward or above 60% — and therefore justify a diagnosis and treatment — include an epileptiform EEG, a symptomatic lesion on MRI, a remote symptomatic cause (prior stroke, trauma, infection), and a nocturnal seizure. A first seizure with a normal EEG, normal imaging, and no clear cause carries roughly a 30–40% two-year recurrence risk — often watched rather than treated after shared decision-making.
The minimum first-seizure workup
| Test | What it answers | Note |
|---|---|---|
| History + witness / video | Was it a seizure? Focal features at onset? | Highest yield of anything here |
| EEG | Epileptiform activity → recurrence risk, classification | A normal routine EEG does not exclude epilepsy; sleep-deprived / repeat raises yield |
| MRI brain (epilepsy protocol) | Structural lesion → focal epilepsy, surgical target | Preferred over CT except in the acute/emergent setting |
| Labs | Provoked cause? (glucose, Na, Ca, Mg, renal, tox) | Rules provocation in or out |
| ECG | Syncope mimic, long-QT, arrhythmia | Cheap, catches a dangerous mimic |
Classify the seizure type
The ILAE 2017 framework classifies at three levels: seizure type → epilepsy type → syndrome. Everything starts with one question that colors the rest of the workup: where did the seizure begin?
The three onset categories
Focal onset
Begins in networks limited to one hemisphere. Sub-classify by awareness (aware vs. impaired) and by first prominent feature (motor vs. non-motor). May evolve to a focal-to-bilateral tonic-clonic seizure.
Generalized onset
Engages bilateral networks from the start. Awareness is not used to sub-classify (it's usually impaired). Split into motor (tonic-clonic, myoclonic, tonic, atonic) and non-motor / absence.
Unknown onset
Used when onset isn't witnessed or is unclear. Still describe observable features (e.g., "unknown-onset tonic-clonic"). Reclassify later as data arrive.
Interactive classifier
Work through the same questions you'd ask a witness. This walks the ILAE branch points and lands on a seizure-type label. It is a teaching aid, not a diagnosis.
Classify the epilepsy type & look for a syndrome
Once you know the seizure type(s), the epilepsy type usually follows. The 2017 framework adds a category most primary-care neurologists under-use: combined generalized and focal epilepsy.
| Epilepsy type | Typical evidence | Prescribing implication |
|---|---|---|
| Focal | Focal seizures; focal epileptiform EEG; structural MRI lesion | Broad list works; the "narrow-spectrum" sodium-channel agents are safe here |
| Generalized | Generalized seizure types; generalized spike-wave EEG; usually normal MRI | Choose a broad-spectrum drug. Narrow-spectrum agents can worsen absence/myoclonus |
| Combined focal & generalized | Both seizure types (e.g., Dravet, Lennox-Gastaut) | Broad-spectrum only |
| Unknown | Insufficient information / normal EEG with unclear semiology | Default to broad-spectrum to avoid aggravation |
When you are unsure whether an epilepsy is focal or generalized, prescribe as if it were generalized and pick a broad-spectrum agent. Narrow-spectrum sodium-channel blockers (carbamazepine, oxcarbazepine, phenytoin, and to a degree gabapentin/pregabalin) can aggravate absence and myoclonic seizures. The cost of guessing "generalized" wrong is small; the cost of guessing "focal" wrong can be worsened seizures.
Common syndromes worth recognizing on sight
A syndrome bundles seizure type(s), age of onset, EEG signature, and prognosis into a package that often dictates drug choice directly.
Juvenile myoclonic epilepsy
Early-morning myoclonic jerks (dropped coffee cup), generalized tonic-clonic seizures, sometimes absences. Photosensitive. Sleep deprivation & alcohol provoke.
Childhood absence epilepsy
Brief (5–10 s) staring spells, dozens/day, provoked by hyperventilation; 3 Hz generalized spike-wave. Normal development.
Self-limited epilepsy with centrotemporal spikes
Nocturnal focal seizures — hemifacial twitching, drooling, speech arrest — from sleep. Centrotemporal spikes on EEG. Benign, self-limited.
Mesial temporal lobe epilepsy
Focal impaired-awareness seizures with rising epigastric aura, déjà vu, automatisms. Hippocampal sclerosis on MRI. Frequently drug-resistant.
Who these patients are — traits & comorbidities by type
Epilepsy is a network disorder, and the same network that generates the seizures also shapes cognition, mood and behaviour. That is why the "traits" cluster by type. But group-level tendencies are not verdicts about a person — read this as clinical texture, not a personality horoscope.
There is no valid "epileptic personality." The 19th- and 20th-century idea of a global temperament shared by everyone with epilepsy is discredited, stigmatising, and wrong. What is real and evidence-based is two things: (1) psychiatric and cognitive comorbidity is far more common in epilepsy than in the general population, and (2) specific epilepsy types carry characteristic cognitive-behavioural profiles tied to their brain network. Both are reasons to screen and support — never to label.
Comorbidity is the rule, not the exception
Across all epilepsy types, the biggest driver of quality of life is often not the seizures but the company they keep. The relationship is bidirectional — depression, for instance, is both more common after epilepsy begins and an independent risk factor for developing it — which points to shared underlying biology rather than a simple reaction to diagnosis.
| Comorbidity | Pattern in epilepsy | Practical move |
|---|---|---|
| Depression | The most common; affects roughly a third; frequently under-recognised and under-treated; bidirectional risk | Screen routinely (a tool like the NDDI-E takes a minute); SSRIs are generally safe and do not meaningfully lower the seizure threshold at therapeutic dose |
| Anxiety | Very common; distinguish an ictal fear aura (seconds, stereotyped) from sustained interictal anxiety and anticipatory fear of seizures | Ask specifically; treat; ictal fear points toward a focal (often temporal) onset |
| ADHD / attention | Strongly over-represented, especially in childhood epilepsies; the inattentive presentation predominates | Screen children; stimulants are generally safe in controlled epilepsy |
| Psychosis | Interictal and postictal psychosis, most linked to temporal lobe epilepsy; "forced normalisation" — psychiatric symptoms emerging as the EEG normalises — is a recognised paradox | Ask about the postictal window; refer |
| Suicidality | Elevated relative to the general population — a sensitive but important part of the picture | Ask directly when screening for mood; have resources ready |
| Cognition | Attention, memory and processing speed, driven by the underlying cause, seizure burden, interictal discharges and medication | Separate what is fixed (etiology) from what is reversible (drug, seizure control) before attributing decline |
Type-specific profiles
These are the characteristic textures each classification tends to carry. Every card describes a tendency at the group level — many individuals show none of it.
Temporal lobe epilepsy
The richest subjective world of any epilepsy. Auras of déjà vu / jamais vu, a rising epigastric sensation, sudden fear, dreamy states, or odd smells and tastes. Memory is affected in a material-specific way — dominant (usually left) temporal networks hit verbal memory, non-dominant hit visuospatial. Carries the highest psychiatric burden of the epilepsies: interictal dysphoric disorder, depression, anxiety, and the strongest tie to interictal and postictal psychosis.
Frontal lobe epilepsy
Seizures are brief, often nocturnal, hyperkinetic and bizarre with fast recovery — and are frequently mistaken for psychogenic events. The interictal profile follows the lobe: executive dysfunction — planning, working memory, inhibition and sustained attention — plus a tendency toward impulsivity and blunted social cognition when the seizures are frequent or the cause is diffuse.
Juvenile myoclonic epilepsy
Intelligence is typically normal, but a frontal-executive behavioural signature has been described — impulsivity, mood lability, difficulty with planning and time management, sometimes framed (unkindly and debatably) as "immaturity". Whatever the label, the clinically decisive fact is that seizures here are driven by sleep deprivation, alcohol and non-adherence.
Childhood absence epilepsy
Development is called "normal", yet subtle attention deficits and higher rates of ADHD (inattentive type), plus academic and psychosocial difficulty, are common — the absences themselves fragment attention dozens of times a day, and discharges can be subclinical.
West · Lennox-Gastaut · Dravet
Here the traits are central, not incidental. The epilepsy and the developmental impairment usually share a cause, and the relentless seizures and discharges themselves worsen development — the "encephalopathy". Expect intellectual disability, behavioural dysregulation, and autism-spectrum features (prominent in Dravet). Controlling the electrical activity can protect development, which is why these are treat-today diagnoses.
Self-limited centrotemporal epilepsy
Usually cognitively normal and outgrown by mid-teens, but during the active period transient difficulties with language, reading and attention are recognised. A reminder that "benign" describes the seizures, not always the school year.
Before you call it a "trait" — check the drug
A large share of what looks like personality is iatrogenic and reversible. Anti-seizure medications have their own behavioural fingerprints; changing the agent can change the "trait".
| Agent | Characteristic behavioural / cognitive effect |
|---|---|
| Levetiracetam | Irritability, aggression, anxiety and low mood are common — the well-known "behavioural" downside; pyridoxine sometimes helps, or switch to brivaracetam |
| Perampanel | Irritability and aggression, dose-related (carries a behavioural warning) |
| Topiramate | Cognitive slowing, word-finding difficulty, paraesthesia, occasionally low mood — the classic "cognitive" ADR profile |
| Zonisamide | Similar cognitive slowing and mood effects to topiramate |
| Phenobarbital / barbiturates | Sedation and depression in adults; paradoxical hyperactivity in children |
| Lamotrigine | Generally favourable — often mood-brightening and cognitively "clean"; watch for rash on fast titration |
| Valproate | Broadly mood-neutral to stabilising; watch tremor, weight, and hyperammonaemic encephalopathy |
When a patient seems changed, sort the contribution before you attribute it: the underlying etiology / lesion (fixed), the network location (temporal vs frontal shapes which functions), the seizure and discharge burden (improves with control), the medication (often reversible), and psychosocial and stigma factors (driving loss, employment, autonomy). Name the specific, modifiable contributor — never the diagnosis as a whole.
- Screen every visit for depression and anxiety; they hurt quality of life more than seizure frequency does and they respond to treatment.
- Ask what the medication did to mood and thinking before concluding anything about the person.
- Match the profile to the network: temporal → memory + mood + psychosis risk; frontal → executive + impulsivity; generalized → attention + adherence.
- Refuse the stereotype. "Epilepsy" is not a personality; describe specifics — a network, a burden, a drug — that you can actually act on.
Reading the EEG signatures
The EEG doesn't diagnose epilepsy on its own, but a handful of interictal patterns are strongly tied to specific classifications. This viewer animates the waveform morphology behind the words in your report.
- Generalized 3 Hz spike-and-wave: typical absence; classic in childhood absence epilepsy.
- Generalized polyspike-wave, often photosensitive: juvenile myoclonic epilepsy and other generalized epilepsies.
- Focal sharp waves / spikes: focal epilepsy; localization hints at the seizure network (e.g., temporal, centrotemporal).
- Hypsarrhythmia: chaotic high-voltage disorganization — the signature of infantile epileptic spasms; a can't-miss, refer-today pattern.
- Normal posterior-dominant α rhythm: reassuring but never rules epilepsy out — sensitivity of a single routine EEG is only ~30–50%.
Choosing a first medication
Drug choice follows classification, then gets tuned to the person — age, sex and pregnancy potential, comorbidities, other medications. Pick a rational agent, start low, titrate slowly, aim for monotherapy at the lowest effective dose.
- Pregnancy potential: valproate is teratogenic and impairs neurodevelopment — avoid in anyone who could become pregnant unless no alternative exists and under specialist oversight. Lamotrigine and levetiracetam are the usual preferred broad-spectrum choices in this group.
- HLA-B*15:02: in patients of Southeast/East Asian ancestry, screen before carbamazepine (and consider before oxcarbazepine/phenytoin) for risk of Stevens-Johnson syndrome / toxic epidermal necrolysis.
Broad- vs. narrow-spectrum: the mental model
Cover both focal and generalized seizures. When onset is uncertain, these are the safe default: lamotrigine, levetiracetam, valproate, topiramate, zonisamide, lacosamide (adjunct), perampanel, brivaracetam.
Effective for focal seizures but can worsen absence and myoclonic seizures: carbamazepine, oxcarbazepine, phenytoin, gabapentin, pregabalin, eslicarbazepine, vigabatrin. Don't reach for these when the epilepsy might be generalized.
- Monotherapy first. ~50% are controlled on the first well-chosen drug; another ~15% on the second.
- Start low, go slow — especially lamotrigine, where fast titration drives serious rash. Slow titration is a safety measure, not a delay.
- Titrate to effect or to tolerability, not to a target level. Levels guide adherence and toxicity, not the goal.
- If the first two appropriate drugs fail at adequate doses → drug-resistant epilepsy. That's a referral trigger, not a third-drug trigger.
Status epilepticus — the time-critical branch
A convulsive seizure lasting ≥5 minutes, or repeated seizures without recovery of awareness between them, is status epilepticus. This is a neurological emergency where the algorithm is a clock, not a decision tree.
ABCs, clock, glucose
- Airway, oxygen, IV access, continuous monitoring. Note the time.
- Check fingerstick glucose; treat hypoglycemia (with thiamine first if malnourished/alcohol).
- Send labs: electrolytes, Ca/Mg, renal, LFTs, AED levels, toxicology, consider ABG.
Give an adequately-dosed benzodiazepine — under-dosing is the classic error
- IM midazolam (no IV), IV lorazepam, or IV diazepam at weight-appropriate doses.
- May repeat once. Do not stall on this step — a second full dose beats waiting.
Load a second-line agent if seizures continue
- Choices with comparable efficacy: IV levetiracetam, IV fosphenytoin, or IV valproate.
- Choose by comorbidities and interactions; any one of the three is a reasonable first pick.
ICU, continuous EEG, anesthetic infusion
- Refractory status → continuous infusion (midazolam, propofol, or pentobarbital) with intubation.
- Continuous EEG monitoring, because ongoing seizures can become non-convulsive.
When to escalate or refer
Primary-care neurology handles the common, well-classified epilepsies. These situations should route to epileptology, video-EEG monitoring, or a comprehensive epilepsy center.
| Trigger | Why |
|---|---|
| Drug-resistant epilepsy | Failure of two tolerated, appropriately chosen and dosed regimens. Surgery, neurostimulation, and dietary therapy become options with high value in the right patient. |
| Diagnostic uncertainty | Events that could be PNES, or seizures that won't classify — video-EEG resolves both. |
| Structural lesion | MRI shows a resectable focus (e.g., hippocampal sclerosis, focal cortical dysplasia). |
| Pregnancy / planning | Regimen optimization and folate before conception; specialist co-management. |
| Infantile spasms / hypsarrhythmia | Emergency — time-sensitive treatment protects development. |
| SUDEP risk counseling | Poorly-controlled convulsive seizures carry real mortality; the conversation belongs in the plan. |
A reproducible classifier, in code
The onset-first logic above expressed as a small GNU Octave function — a compact way to encode the branch points so the reasoning is explicit and testable rather than implicit.
% classify_seizure.m — teaching model of the ILAE 2017 onset-first branch % Inputs are strings; returns a seizure-type label. Educational only. function label = classify_seizure(onset, awareness, feature) onset = lower(onset); % 'focal' | 'generalized' | 'unknown' awareness = lower(awareness); % 'aware' | 'impaired' | 'na' feature = lower(feature); % 'motor' | 'nonmotor' | 'tonicclonic' | 'absence' | 'myoclonic' switch onset case 'focal' aw = 'aware'; if strcmp(awareness,'impaired'), aw = 'impaired awareness'; end label = sprintf('Focal %s, %s onset', aw, feature); if strcmp(feature,'tonicclonic') label = 'Focal to bilateral tonic-clonic'; end case 'generalized' switch feature case 'absence', label = 'Generalized non-motor (absence)'; case 'myoclonic', label = 'Generalized motor: myoclonic'; case 'tonicclonic', label = 'Generalized motor: tonic-clonic'; otherwise, label = 'Generalized motor (specify)'; end otherwise label = sprintf('Unknown-onset %s', feature); end endfunction % >> classify_seizure('focal','impaired','motor') % ans = Focal impaired awareness, motor onset % >> classify_seizure('generalized','na','absence') % ans = Generalized non-motor (absence)