Antiseizure Medications in Pediatric Epilepsy

A practical framework for choosing, avoiding, and combining ASMs

Maksim Parfyonov, MD FRCPC

October 5, 2026

Objectives

By the end of this session, you should be able to:

  • Explain the major mechanistic classes and clinical spectra of ASMs
  • Distinguish broad-spectrum from narrow-spectrum treatment
  • Recognize common drug–syndrome mismatches and paradoxical seizure aggravation
  • Select reasonable first-line therapy for common pediatric epilepsy syndromes
  • Incorporate age, etiology, adverse effects, reproductive considerations, and interactions
  • Approach monotherapy failure and rational polytherapy
  • Recognize when medication is not the most important next treatment

How to think about ASM choice

The right antiseizure medication is determined by the epilepsy syndrome you think the child has.

Three questions to guide you:

  1. Seizure or not seizure? If so, what is the seizure type?
  2. What epilepsy type or syndrome is this? - ELECTRO-CLINICAL diagnosis
  3. What is the likely etiology?

The same medication can be an excellent choice in one child and a poor choice in another.

How do ASMs work?

Most ASMs reduce the probability that a neuronal network will generate or propagate a seizure. They generally do this by:

  • reducing voltage-gated sodium or calcium currents
  • increasing inhibitory GABAergic transmission
  • reducing excitatory glutamatergic transmission
  • modifying vesicle release / synaptic transmission
  • acting through less well-defined or multiple mechanisms

So… Keppra for everybody?

Mechanism ≠ spectrum ≠ syndrome-specific efficacy

A mechanistically “broad” drug is not automatically the best drug for every generalized epilepsy.

A practical pharmacology map

Mechanism / target Examples Clinical shorthand
Na⁺ channel inhibition carbamazepine, oxcarbazepine, phenytoin, lacosamide, lamotrigine Strong focal utility; some can worsen generalized epilepsies (lamotrigine is the exception — used as broad-spectrum despite Na⁺ action)
Na⁺ (persistent current) + GABA-A cenobamate Drug-resistant focal seizures (FDA-approved ≥18 y); slow titration for DRESS risk
SV2A levetiracetam, brivaracetam Broad clinical use; few interactions
T-type Ca²⁺ channels ethosuximide Absence seizures
GABA-A / GABAergic benzodiazepines, phenobarbital, clobazam, vigabatrin Useful, but sedation/tolerance can limit chronic use
Multiple mechanisms valproate, topiramate, zonisamide Broad-spectrum; trade-offs matter
AMPA antagonism perampanel Broad clinical utility; behavioral adverse effects
Mixed / distinctive cannabidiol, stiripentol, fenfluramine Syndrome-specific roles, especially developmental epileptic encephalopathies

Where the drugs act

Broad-spectrum vs narrow-spectrum

Narrower-spectrum

Ethosuximide
Excellent for absence; poor choice if also has other seizure types

Carbamazepine / oxcarbazepine
Excellent focal-seizure drugs; problematic in several generalized epilepsies and Dravet syndrome.

Broad-spectrum

Valproate, levetiracetam, lamotrigine, topiramate, zonisamide

Can cover multiple seizure types — but broad does not mean benign.

The practical question is:

Which drug gives this child the best seizure control for the lowest overall burden?

Before starting an ASM: the five checks

1. Diagnosis
Is this actually epilepsy? Sometimes the best treatment is not starting an ASM yet.

2. Seizure / syndrome
Focal? Generalized? Developmental and epileptic encephalopathy?

3. Etiology
Structural? Genetic? Metabolic? Immune? Unknown?

4. Patient factors
Age, cognition, behavior, sleep, comorbidities, reproductive potential.

5. Drug factors
Spectrum, interactions, titration burden, monitoring, formulation, rescue plan.

When should we treat?

Once epilepsy is established, treatment is generally indicated.

But pediatric epilepsy contains important exceptions and nuances:

  • Some self-limited epilepsies may reasonably be observed
  • A first seizure does not automatically equal lifelong medication
  • Seizure burden, developmental risk, syndrome, EEG, imaging, and family preferences all matter
  • A medication should not be continued indefinitely simply because it was started

Treatment is a longitudinal decision, not a one-time prescription.

Case 1

A 6-year-old has multiple daily few second staring episodes +/- eyelid fluttering.

School has noticed them for months.

Case 1: the EEG

Which ASM would you consider? What is the mechanism of action? What other drugs would be reasonable alternatives?

Absence epilepsy - ethosuximide

Ethosuximide is the classic first-line drug for pure absence epilepsy.

Why?

  • Excellent efficacy against typical absence seizures
  • Relatively narrow target: T-type calcium currents
  • Avoids unnecessary exposure to a broad-spectrum drug when no other seizure type is present

The landmark randomized pediatric trial found ethosuximide and valproate superior to lamotrigine for seizure freedom; ethosuximide had the most favorable attentional profile.

However, important to ask: Are these really the only seizures?

If generalized tonic-clonic or myoclonic seizures are present or likely, the treatment calculation changes.

The evidence: the childhood absence trial

Glauser et al., NEJM 2010. Freedom from treatment failure over 20 weeks: ethosuximide and valproate were comparable and both superior to lamotrigine; ethosuximide had fewer attentional adverse effects than valproate.

Take-home: for pure absence, ethosuximide matches valproate for control with the best attentional profile — and avoids valproate’s systemic and reproductive burden.

Absence: the avoid list

Aggravate absence — strongest evidence

  • Carbamazepine
  • Oxcarbazepine
  • Phenytoin
  • Tiagabine, vigabatrin (GABAergic → enhance thalamocortical oscillation)

Weaker evidence: phenobarbital

Aggravate myoclonic seizures

  • Gabapentin
  • Pregabalin
  • Vigabatrin
  • Lamotrigine (in some patients with JME)

Mechanism: carbamazepine acts in the thalamic reticular nucleus to promote absence-generating rhythms (Jang et al., PNAS 2025).

Case 2

An 8-year-old presented to outside hospital about a year ago with 2 nocturnal episodes of drooling, facial clonic activity with retained awareness. Carbamazepine initially started. Over the past year, he has continued to decline with ongoing seizures and worsening school performance, attention, and behavioral decline.

What is the diagnosis? What happened afterwards?

Case 2: the EEG

The initial diagnosis: SeLECTS

Self-limited epilepsy with centrotemporal spikes (SeLECTS) — formerly “benign rolandic epilepsy.”

Mandatory features all present here (Specchio et al., Epilepsia, 2022):

  • age 3–13 at onset
  • nocturnal, brief focal seizures
  • orofacial motor signs, sialorrhea, speech arrest with retained awareness
  • centrotemporal spikes, activated by sleep

But this is NOT a benign course

The year of seizure worsening + cognitive/behavioral/academic regression is the alarm.

This picture is most consistent with evolution to epileptic encephalopathy with spike-and-wave activation in sleep (EE-SWAS / ESES) — occurs in ~1.3–4.6% of SeLECTS (He et al., CNS Neurosci Ther, 2025; Tovia et al., Epilepsia, 2011).

Warning

Regression in a child with “benign rolandic epilepsy” demands a sleep EEG to assess spike-wave activation.

What happened next

  • Repeat EEG in sleep → markedly activated / near-continuous spike-wave (SWI of 90%)
  • Carbamazepine implicated as a likely aggravating factor
  • Switch away from sodium-channel blockers; may require other treatments for EE-SWAS (e.g. steroids, nocturnal diazepam) and reassessment

SeLECTS: treatment is not always necessary

SeLECTS seizures are often:

  • infrequent
  • nocturnal
  • developmentally self-limited (most remit by adolescence)

The decision to treat depends on:

  • seizure burden
  • generalized tonic–clonic or daytime seizures
  • cognitive / behavioral concerns and school performance
  • EEG burden, particularly during sleep

SeLECTS: where the controversy lives

Carbamazepine and oxcarbazepine may rarely exacerbate seizures or precipitate atypical evolution (e.g., atypical absences, negative myoclonus, continuous spike-wave in sleep) in susceptible children (Corda et al., Epilepsia, 2001).

Absolute risk is low (~1 aggravation in 40 CBZ-exposed in one series), but the consequence can be severe. Comparative data favor sulthiame, levetiracetam, and clobazam over carbamazepine/oxcarbazepine/topiramate for seizure freedom (Gerstl et al., Clin Neuropharmacol, 2021). Overall evidence that any drug improves long-term seizure or cognitive outcome remains limited (Cochrane, 2014). Current NICE guidance continues to evolve, with a dedicated SeLECTS update anticipated.

Case 3

An 18-month-old develops:

  • atonic seizures
  • myoclonic seizures
  • developmental regression

EEG: generalized slow spike-and-wave. MRI obtained.

The team proposes valproate — broad-spectrum, covers atonic + myoclonic seizures.

Case 3: the MRI

MRI in this child.

What do you think?

Valproate under age 2–3: the etiology problem

In a young child with new-onset epilepsy, the etiology is often unknown — and an undiagnosed POLG-related mitochondrial disorder cannot be excluded.

  • Highest risk of VPA-induced acute liver failure is in the first 2 years of life, in children with developmental delay and refractory epilepsy — the exact phenotype of undiagnosed POLG disease (Dulac, Lancet Neurol 2014).
  • POLG disease often presents with seizures before the diagnosis is confirmed, so children may already be on VPA when the metabolic disorder declares itself.
  • In reported series, POLG patients developed acute liver failure 2–3 months after starting VPA for intractable seizures (Saneto, Neurology 2011).

Warning

VPA can trigger fatal acute liver failure in POLG disease before the diagnosis is ever made.

What is POLG?

  • POLG encodes mitochondrial DNA polymerase γ — the enzyme that replicates and repairs mitochondrial DNA (mtDNA) (Hynynen, Epilepsia 2018).
  • Pathogenic variants cause mtDNA depletion / deletions → failure of cellular energy production.
  • The spectrum includes Alpers–Huttenlocher syndrome: refractory epilepsy + hepatopathy + regression (Stewart, Hepatology 2010).
  • Most common variants p.A467T and p.W748S (~2/3 of autosomal-recessive disease); carrier frequency ~0.5–0.8% in Northern European populations.

The earliest-onset POLG phenotypes are the hepatopathic ones — precisely the infants in whom VPA is most dangerous.

Why VPA is lethal in POLG: the mechanism

VPA hepatotoxicity is multifactorial mitochondrial injury. In a POLG-deficient hepatocyte the mitochondrion has no reserve, so injury crosses the threshold to cell death:

  • inhibits β-oxidation → microvesicular steatosis; depletes carnitine and CoA
  • inhibits oxidative phosphorylation (complex I/IV, cytochrome c oxidase) → ↓ ATP
  • further mtDNA depletion (HDAC inhibition), compounding the underlying POLG defect
  • lowers the threshold for mitochondrial permeability transition pore opening → apoptosis (Li, Hepatology 2015)

Heterozygous POLG variation was strongly associated with VPA hepatotoxicity: OR 23.6 (95% CI 8.4–65.8) in the DILIN cohort (Stewart, Hepatology 2010).

POLG: red flags and the practical rule

FDA contraindications: a known POLG mitochondrial disorder — and any child < 2 years suspected of a POLG-related disorder.

Red flags to suspect POLG before prescribing VPA: unexplained encephalopathy · refractory focal/myoclonic seizures · status epilepticus at onset · psychomotor regression · neuropathy · ataxia · ophthalmoplegia.

Practical rules:

  1. Do not use VPA first-line under age 2–3 when the etiology is unknown.
  2. Screen for red flags; send POLG testing if mitochondrial disease is suspected — increasingly proposed as standard of care before starting VPA.
  3. If mitochondrial disease is possible in a child > 2 y, use VPA only after other agents fail — prefer levetiracetam or sodium-channel agents.
  4. If VPA is unavoidable in a young child: sole agent, close LFT monitoring, stop immediately for hepatic dysfunction.

2026: a useful new infantile-epilepsy teaching point

The 2026 American Epilepsy Society infantile epilepsy guideline addresses children 1 month to <36 months.

For new-onset epilepsy:

  • Levetiracetam is suggested over phenobarbital
  • Valproate is suggested against when etiology is uncertain
  • For new-onset focal epilepsy, oxcarbazepine is suggested over levetiracetam, with important caveats
  • Etiologic evaluation and genetic testing are explicitly part of the treatment pathway

The evidence certainty is generally low or very low.

Case 4

A 7-month-old has recurrent prolonged hemiclonic seizures, initially associated with fever.

What is the diagnosis? Which ASM to choose?

Dravet: the sodium-channel paradox

In SCN1A-related Dravet syndrome, several sodium-channel blockers can worsen seizures:

  • carbamazepine
  • oxcarbazepine
  • phenytoin
  • lamotrigine

NICE also flags lacosamide among medications that may exacerbate seizures.

Diagnostic step: send SCN1A testing (epilepsy gene panel) — a pathogenic variant is found in ~85% of Dravet syndrome.

Preferred syndrome-directed therapies include

  • Valproate
  • Clobazam
  • Stiripentol
  • Fenfluramine
  • Cannabidiol

Dravet: the treatment hierarchy

Consensus treatment sequence for Dravet syndrome (Wirrell et al., Epilepsia 2022): valproate first-line, then fenfluramine / stiripentol / clobazam, then cannabidiol, with topiramate and ketogenic diet as later options. Carbamazepine, oxcarbazepine, lamotrigine and phenytoin are explicitly avoided.

Dravet / LGS: know the newer drugs by role

  • Cannabidiol
    • FDA-approved for Dravet syndrome, Lennox-Gastaut syndrome, and TSC-associated seizures.
    • MOA - distinct from THC; not primarily CB1/CB2. Proposed: GPR55 antagonism, TRPV1 modulation, adenosine signaling
    • ↑ N-desmethylclobazam with clobazam; transaminase elevations, especially with valproate
  • Stiripentol
    • Dravet-specific therapy, typically used with clobazam
    • GABA-A modulator
  • Fenfluramine
    • FDA-approved for Dravet and LGS
    • multiple - serotonin release + 5-HT1D, 5-HT2A and 5-HT2C agonism; sigma-1 receptor modulation
    • Important safety issue: echocardiogram before, every 6 months during, and 3–6 months after treatment (REMS) because of valvular heart disease and pulmonary arterial hypertension risk.

A practical adverse-effect framework

Don’t memorize 30 adverse-effect lists. Think in domains:

Domain Examples to remember
Sedation phenobarbital, benzodiazepines, clobazam
Behavioral levetiracetam, perampanel
Weight gain valproate
Weight loss topiramate, zonisamide
Kidney stones / acidosis topiramate, zonisamide
Rash / SJS lamotrigine, carbamazepine, oxcarbazepine
Hepatotoxicity valproate, felbamate
Hematologic toxicity felbamate
Hyponatremia oxcarbazepine, carbamazepine
Drug interactions enzyme inducers; valproate; clobazam/stiripentol interactions
Reproductive / fetal risk valproate, topiramate

Let the comorbidity choose the drug

Sometimes a “side effect” is what you want — treat two problems with one drug:

The child also has… Lean toward Avoid / caution
Migraine topiramate, valproate —
Wants weight loss topiramate, zonisamide valproate (weight gain)
Needs mood stabilization lamotrigine (favorable mood) levetiracetam, phenobarbital, topiramate
Behavior / ADHD / aggression / ASD sodium-channel agents, lamotrigine levetiracetam, perampanel, clobazam
Poor feeding / underweight valproate topiramate, zonisamide (appetite ↓)

Pharmacokinetics

  • Enzyme inducers
    • Carbamazepine, phenytoin, phenobarbital can lower concentrations of many co-medications.
  • Valproate
    • highly protein bound
    • inhibits metabolism of several drugs
    • clinically important interaction with lamotrigine
    • broad spectrum, but substantial reproductive and metabolic considerations
  • Levetiracetam
    • minimal hepatic metabolism
    • few drug interactions
    • renal elimination
    • convenient in medically complex children
  • Clobazam + stiripentol
    • A classic example of a deliberate pharmacokinetic interaction that is used therapeutically but requires dose adjustment and monitoring.

Lamotrigine + valproate

This combination is often effective.

But valproate substantially increases lamotrigine exposure.

Therefore:

Lamotrigine must be titrated slowly when valproate is present.

Why?

  • Valproate inhibits glucuronidation (UGT1A4) — lamotrigine’s main elimination route
  • Lamotrigine half-life roughly doubles (~30 → ~60 h); clearance falls ~50%
  • Faster rise in concentration → higher risk of rash / SJS

In practice

  • Start at about half the usual dose and titrate more slowly (FDA label: adults 25 mg every other day for weeks 1–2; children 2–12 y 0.15 mg/kg/day)
  • Lower lamotrigine maintenance targets are usually needed
  • Reverse effect when VPA is stopped: lamotrigine levels fall (Patsalos & Perucca, Lancet Neurol 2003)

What counts as treatment failure?

A child still having seizures does not automatically need a second medication.

Ask:

  • Was the first diagnosis correct?
  • Was the drug appropriate for the syndrome?
  • Was the dose adequate?
  • Was adherence adequate?
  • Was the medication tolerated?
  • Are these actually epileptic events?
  • Has the epilepsy evolved?

Then decide:

Switch → add-on → dietary therapy → surgery → neuromodulation → syndrome-specific therapy

depending on the child.

Rational polytherapy

Useful principles:

  • Prefer different mechanisms when reasonable
  • Avoid duplicating the same toxicity
  • Consider pharmacokinetic interactions
  • Minimize total medication burden
  • Stop ineffective drugs rather than accumulating them

Polytherapy: what you need to ask

Before adding drug #2:

1. What did drug #1 actually accomplish?

Complete response? 30% reduction? No effect?

2. What toxicity did drug #1 create?

Sedation? Behavior? Weight? Cognition?

3. What mechanism is missing?

4. Will drug #2 make the first drug more toxic?

5. Is there a non-pharmacologic treatment that should happen now?

If two appropriate ASMs have failed, don’t simply keep adding medications indefinitely.

Don’t forget surgery — and think about it early!

The single most common surgical error in pediatric epilepsy is referring too late.

  • After two appropriate ASMs fail, each further drug has diminishing returns — seizure freedom becomes progressively less likely and is often not sustained.
    • Caveat: the often-quoted “~5%” is the share of the whole newly diagnosed cohort freed by the third drug — not the conditional chance for the child who has already failed two, which is higher but still modest (Blond & Hirsch, Epilepsia 2020; Perucca et al., Lancet Neurol 2023).
  • Resective surgery in a well-selected child can be curative — and the developing brain has the most to gain from early seizure control.
  • In infants, the 2026 AES guideline makes a STRONG recommendation for hemispherectomy/resection in lesional drug-resistant epilepsy — surgery is not a last resort.

Diminishing returns: the evidence

Cumulative probability of seizure freedom in newly diagnosed epilepsy by number of ASM regimens; dotted lines are 95% CIs (Brodie & Kwan, Neurology 2002).

Take-home: most of the benefit comes from the first two regimens; the curve flattens after that. That plateau is the time to refer.

Five prescribing habits I want you to leave with

1. Diagnose before prescribing

Seizure → epilepsy → syndrome → etiology → drug.

2. Know the “avoid” list

Especially: - generalized epilepsies - Dravet - selected self-limited epilepsies

3. Treat the child, not the EEG

But know when EEG burden itself has clinical significance.

4. Reassess every failure

Persistent seizures should trigger diagnostic reconsideration, not automatic polypharmacy.

5. Think beyond medication

In children with drug-resistant epilepsy, the next treatment may be a diet, surgery, neuromodulation, or precision therapy.

One-slide board review

Absence → ETHX = VPA > LTG for control; ETHX preferred for attention/tolerability

Generalized epilepsy → think broad-spectrum (VPA, LTG, LEV, ZNS); beware Na-channel blockers (LTG excepted, but may worsen myoclonus)

Dravet / SCN1A → avoid Na-channel blockers; think valproate + syndrome-specific agents (e.g. FFN, Stiri)

SeLECTS → often self-limited; avoid Na-channel if possible (CLOB, LEV)

Infantile spasms → urgent syndrome-specific treatment (steroids, ACTH, VGB)

Infant <36 months → avoid VPA

Valproate + lamotrigine → slow lamotrigine titration

Two appropriate ASM failures → think drug-resistant epilepsy and referral, not endless medication stacking

Selected references (1/3)

  1. NICE. Epilepsies in children, young people and adults (NG217). Updated 2025; SeLECTS update in development in 2026.
  2. Freedman D, et al. American Epilepsy Society Clinical Practice Guideline: Infantile Epilepsy. Epilepsy Currents. 2026.
  3. Glauser TA, et al. Ethosuximide, valproic acid, and lamotrigine in childhood absence epilepsy. N Engl J Med. 2010.
  4. Marson AG, et al. The SANAD II study of antiseizure medication treatment in newly diagnosed epilepsy. Lancet. 2021.
  5. AAN/AES/SMFM. Practice guideline: Teratogenesis, perinatal, and neurodevelopmental outcomes after in utero exposure to antiseizure medication. Neurology. 2024.
  6. Wirrell EC, et al. International consensus on diagnosis and management of Dravet syndrome. Epilepsia. 2022.

Selected references (2/3)

  1. ILAE. Updated classification of epileptic seizures. Epilepsia. 2025.
  2. Rahman S, Copeland WC. POLG-related disorders and their neurological manifestations. Nat Rev Neurol. 2019.
  3. Stewart JD, et al. POLG determines the severity of valproate-induced hepatotoxicity. Hepatology. 2010.
  4. Li S, et al. Valproate triggers apoptosis in POLG-deficient (Alpers) hepatocytes. Hepatology. 2015.
  5. Saneto RP, et al. POLG DNA testing as an emerging standard of care before valproate. Neurology. 2011.
  6. US FDA. Valproic acid prescribing information — Contraindications and Warnings.

Selected references (3/3)

  1. Blond BN, Hirsch LJ, Mattson RH. Misperceptions on the chance of seizure freedom with antiseizure medications after two failed trials. Epilepsia. 2020.
  2. Perucca E, Perucca P, White HS, Wirrell EC. Drug resistance in epilepsy. Lancet Neurol. 2023.
  3. Patsalos PN, Perucca E. Clinically important drug interactions in epilepsy: interactions between antiepileptic drugs. Lancet Neurol. 2003.
  4. Chu H, et al. Antiseizure medications for idiopathic generalized epilepsies: a systematic review and network meta-analysis. J Neurol. 2023.
  5. Jang SS, Agranonik N, Huguenard JR. Carbamazepine in the thalamic reticular nucleus: a potential mechanism of absence aggravation. PNAS. 2025.
  6. Strzelczyk A, Schubert-Bast S. A practical guide to the treatment of Dravet syndrome with anti-seizure medication. CNS Drugs. 2022.
  7. Erdemir G, et al. Diagnosis and management guidelines for infantile epileptic spasms syndrome around the world. Epilepsia. 2026.
  8. Brodie MJ, Kwan P. Staged approach to epilepsy management. Neurology. 2002.