Refractory status epilepticus

1. Three patient vignettes#

Vignette A. Noah, 5 years, the canonical case#

Noah, 5 years old, 18 kg, healthy until six hours ago. Sudden generalised tonic-clonic at home; ambulance called after 5 minutes of seizing. Pre-hospital intramuscular midazolam 5 mg (0.28 mg/kg, an under-dose). On ED arrival he is still intermittently seizing. Glucose 4.4 mmol/L, electrolytes normal, ABG shows a metabolic acidosis from the convulsive activity itself. He is intubated for airway protection and transferred to the PICU. cEEG is hooked up at the 70-minute mark and shows continuous electrographic seizures despite the paralysis. Question: he has failed first-line benzodiazepine; what second-line drug, at what dose, and where is the cEEG endpoint that says we can stop? 1 2

Vignette B. Mira, 6 weeks, neonatal RSE#

Mira, 6 weeks old, 4.2 kg, presents from the postnatal ward at 36 hours of life with rhythmic right-arm jerking and intermittent apnoea. Continuous aEEG shows a saw-tooth rhythmic ictal envelope with bursts of suppression in between. She is loaded with phenobarbital 20 mg/kg IV (the neonatal first-line per the NEMO trial pathway), then a second 20 mg/kg if breakthrough seizures continue. cEEG confirms an electroclinical dissociation: clinical jerking resolves but rhythmic right hemispheric discharges continue at 1 Hz. Question: neonatal SE has different drug ordering, different cEEG patterns, and different prognostic implications. Where does Mira fit and what do we add next? 3 4

Vignette C. Asha, 14 years, NORSE/FIRES presentation#

Asha, 14 years old, 52 kg, well two weeks ago. Five days of low-grade fever, three days of behavioural change, then sudden GTC seizures escalating to status. She has failed midazolam, levetiracetam, valproate, and a midazolam infusion. cEEG shows multifocal, evolving discharges in a bilateral frontal and temporal distribution. CSF: mild lymphocytic pleocytosis, normal protein, no organisms. MRI day 3 shows symmetric mesial temporal T2 hyperintensities and cortical ribboning. Question: this looks like new-onset refractory SE (NORSE) with the febrile-prodrome subgroup (FIRES); what is the parallel workup and how does management differ from ordinary RSE? 5

2. The clinical question#

For each of these children: how do we recognise SE that is not stopping, how do we escalate without losing time, and how do we use bedside multimodal monitoring (cEEG, aEEG, NIRS, pupillometry) to know that we have actually achieved cessation versus chemically paralysed a child whose brain is still seizing?

3. Pathophysiology refresher#

Status epilepticus is the clinical and electrographic expression of a failure of the brain's inhibitory machinery. Two transitions matter mechanically.

The first transition: from a self-limiting seizure to SE. GABA-A receptors are trafficked off the post-synaptic membrane after sustained seizure activity. Within 10 to 30 minutes, the same dose of benzodiazepine that would have aborted a single seizure has progressively less effect; the receptors it targets are no longer there. NMDA receptors are trafficked on, increasing excitatory drive. This is the mechanistic reason that early treatment is qualitatively, not just quantitatively, more effective: the receptor pharmacology changes underneath you while you wait. 5

The second transition: from RSE to SRSE. After 24 hours of continuous seizing, neuronal injury compounds. Hippocampal pyramidal cells (CA1, CA3) and cerebellar Purkinje cells are most vulnerable. Mitochondrial dysfunction, calcium overload, and excitotoxic apoptosis produce both the radiological signature (T2 hyperintensities, restricted diffusion) and the clinical consequence (post-SE cognitive and motor deficits, chronic epilepsy). The longer the seizure goes, the harder it becomes to stop, the worse the substrate of the recovering brain.

Why does cEEG matter so much? Two empirical facts. First, convulsive activity stops before electrographic activity stops in roughly half of patients with CSE; the muscular contractions damp first but cortical discharges continue. Without cEEG you will believe the seizure ended when it has not. Second, in the paralysed, ventilated patient on midazolam infusion, clinical exam is uninformative; cEEG is the only window into whether you have actually achieved cessation. The ACNS pediatric cEEG indications cite SE and RSE as Tier 1 (strongly indicated) precisely because of these two facts. 6 7 8

NIRS and pupillometry come in as sentinels for secondary injury. Sustained convulsive seizing roughly doubles regional CMRO2, outstripping autoregulatory flow reserve in the actively epileptogenic cortex. NIRS rSO2 asymmetry (the affected hemisphere drops first) and a falling NPi (often unilateral with focal injury) flag tissue that is being damaged in real time, even as cEEG documents the discharges. 9 10

Pediatric pharmacokinetics differ. Children clear midazolam, propofol, and pentobarbital faster than adults; weight-banded dosing matters; the propofol-infusion syndrome ceiling (5 mg/kg/h for more than 48 h) is more restrictive in younger children. 11

4. The multimodal picture#

5. Decision tree#

flowchart TD
  Start[Seizing > 5 min] --> Benzo[Benzodiazepine load]
  Benzo --> Stop1{Stopped?}
  Stop1 -->|Yes, GCS recovering| Watch[Watch and observe]
  Stop1 -->|No after 5 min| Second[Second-line AED]
  Second --> Stop2{Stopped at 20 min?}
  Stop2 -->|Yes| cEEG6[cEEG for 6 to 24 h]
  Stop2 -->|No| RSE[RSE, anaesthetic infusion]
  RSE --> Midaz[Midazolam infusion]
  Midaz --> EndPoint{cEEG endpoint?}
  EndPoint -->|Seizure free 24 h| Wean[Slow wean]
  EndPoint -->|Breakthrough| Esc[Escalate dose or add drug]
  Esc --> Pento[Pentobarbital or ketamine]
  Pento --> SRSE{Still seizing after 24 h?}
  SRSE -->|Yes| NORSE[NORSE/FIRES workup]
  SRSE -->|No| Wean
  NORSE --> Immuno[Immunotherapy + KD]

6. Step-by-step bedside actions#

For Noah (5 years, 18 kg), the canonical case. Times are from seizure onset.

1. 0 to 5 min: airway, IV, glucose, benzodiazepine. Lateral position, suction, high-flow oxygen. IV or IO access if not present. Glucose check (target greater than 3.3 mmol/L). Midazolam 0.2 mg/kg IM (max 10 mg) or 0.1 mg/kg IV if access is in. Noah's 18 kg target is 3.6 mg IM or 1.8 mg IV. Pre-hospital under-dosing is the rule, not the exception; document what was actually given. 2
2. 5 to 10 min: second benzodiazepine if needed. Repeat IV midazolam 0.1 mg/kg if still seizing. Two doses maximum before moving to second-line. Recheck airway.
3. 10 to 20 min: second-line AED, ESETT-validated. Three options have equivalent efficacy in the ESETT trial: levetiracetam 60 mg/kg IV (max 4500 mg) over 10 min; fosphenytoin 20 mg PE/kg IV (max 1500 mg) at 150 mg PE/min; valproate 40 mg/kg IV over 10 min. Levetiracetam is forgiving (no cardiac monitoring required, no infusion-rate hazards, no IV infiltration injury); fosphenytoin needs cardiac monitoring; valproate is contraindicated in suspected mitochondrial disease. Pick one and dose adequately. 1
4. 20 to 40 min: get cEEG on and look at the trace. While the second-line is loading, get the cEEG techs in. If full montage is delayed, start a reduced-channel aEEG immediately. Treat persistent rhythmic ictal envelope as ongoing SE even if convulsive activity has stopped. 7
5. 40 to 60 min: established RSE, anaesthetic infusion. Persistent electrographic SE after benzodiazepine plus one second-line agent is RSE. Start midazolam infusion 0.2 mg/kg bolus then 0.1 to 2 mg/kg/h titrated to cEEG endpoint. Intubate (most are intubated already at this point for airway). Add an arterial line for continuous BP.
6. 60 to 90 min: define and confirm the cEEG endpoint. Two acceptable endpoints: (a) seizure cessation with continuous background (the preferred endpoint when achievable); (b) burst-suppression with BSR 50 to 90% for refractory cases where (a) is unobtainable. Pure isoelectric EEG is not the goal; burst-suppression preserves some background.
7. 2 to 6 h: NIRS and pupillometry sentinels. Bilateral NIRS and 4-hourly pupillometry. NIRS asymmetry greater than 10% in the suspected focus or a falling NPi (less than 3, or asymmetric) flags secondary injury and should prompt imaging.
8. 6 to 24 h: maintain endpoint, watch for breakthrough. Continuous cEEG review every 4 to 6 hours. Breakthrough discharges trigger escalation: midazolam dose up; add ketamine 1 to 3 mg/kg bolus then 1 to 5 mg/kg/h; if pentobarbital is the chosen escalation, load 5 to 10 mg/kg then 1 to 5 mg/kg/h.
9. 24 to 48 h: declare SRSE if continuing, broaden workup. If still seizing after 24 h of anaesthetic infusion: send autoimmune encephalitis panel (anti-NMDA, anti-LGI1, anti-GAD65, paraneoplastic), viral panel, metabolic workup; consider methylprednisolone 30 mg/kg/day; consider IVIG 2 g/kg over 5 days; consider plasmapheresis; consider ketogenic diet (start as early as day 3 in SRSE).
10. 48 to 96 h: cautious wean. When the cEEG has been seizure-free for 24 to 48 hours, wean midazolam by 20 to 30% every 6 to 12 hours under continuous cEEG, looking for breakthrough discharges. If breakthrough, return to the prior step and try again 24 h later. Some children need a 5 to 14 day taper; some wean cleanly in 48 hours.

7. Management ladder and endpoints#

Success looks like: seizure cessation on cEEG with preserved background or burst-suppression at BSR 50 to 90%; haemodynamics stable on titrated infusion; NPi recovering toward 4 to 5; NIRS symmetry restored; daily AED levels in target range.

Failure looks like: breakthrough electrographic seizures within 6 hours of dose; refractory hypotension on the infusion (consider noradrenaline support; consider switching agent); evolving NIRS asymmetry; falling NPi; new neurological deficit; oedema on imaging.

When to escalate:

• Breakthrough at 24 h on midazolam alone, add ketamine or switch to pentobarbital.
• Breakthrough at 48 to 72 h after two infusions, declare SRSE, broaden workup, immunotherapy, consider ketogenic diet.
• Breakthrough at 5 to 7 d, consider compassionate-use agents (perampanel, lacosamide, brivaracetam), specialist centre transfer.

When to de-escalate:

• 24 h seizure-free on cEEG with stable BSR target met.
• Hyponatraemia, fever, hypoglycaemia all corrected.
• No active autoimmune or infectious driver requiring escalation.
• AED levels in maintenance range.
• Family conversation about goals of care is current.

8. Variant subsections#

8.1 Convulsive versus non-convulsive SE#

Convulsive SE (CSE) has visible motor activity and is recognised within minutes. Non-convulsive SE (NCSE) is electrographic seizing without sustained motor signs; it accounts for roughly half of all post-convulsive comas and is missed entirely without cEEG. NCSE in the obtunded post-cardiac-arrest patient, the post-TBI patient, or the post-meningitis patient is one of the strongest indications for cEEG in any PICU. 6 12 7

8.2 Super-refractory SE (SRSE)#

Defined as SE that continues 24 hours or recurs after attempted wean from a continuous anaesthetic infusion. Mortality is 30 to 50% even in well-resourced centres. Mainstays of management: (i) broaden the differential (autoimmune, paraneoplastic, metabolic, mitochondrial); (ii) consider second and third anaesthetic agents in combination (midazolam plus ketamine; pentobarbital coma); (iii) ketogenic diet, ideally by day 3 to 5 of SRSE; (iv) immunotherapy stack (steroids, IVIG, plasmapheresis, rituximab); (v) transfer to a centre with experience in NORSE/FIRES.

8.3 Anti-NMDA receptor encephalitis and autoimmune-related RSE#

Subacute behavioural change, then movement disorder, then orofacial dyskinesias, then RSE. Often young women and adolescent girls; sometimes preceded by HSV encephalitis or ovarian teratoma. CSF anti-NMDA antibody is diagnostic. Treatment: tumour removal if applicable, steroids, IVIG, plasmapheresis, rituximab, cyclophosphamide. Recovery may be slow (weeks to months) but is often substantial.

8.4 FIRES (febrile infection-related epilepsy syndrome)#

Previously well child, febrile prodrome of 2 to 14 days, then explosive onset RSE that becomes SRSE. MRI is often normal early; later shows mesial temporal T2 hyperintensities. No specific autoantibody usually found; CSF and serum often unremarkable; cytokine storm physiology suspected. Ketogenic diet has reasonable evidence; anakinra (IL-1 receptor antagonist) and tocilizumab (IL-6 inhibitor) are being used in specialist centres. Mortality and long-term morbidity are high. 5

8.5 Neonatal status epilepticus#

Different pharmacology and different first-line. Phenobarbital 20 mg/kg IV is first-line (NEMO trial); levetiracetam is reasonable second-line; midazolam infusion is the bridge to pyridoxine trial and metabolic workup. cEEG patterns differ (more multifocal, more clinical-electrical dissociation). Pyridoxine-dependent epilepsy needs a single 100 mg dose to exclude; pyridoxal-5-phosphate-responsive forms require 30 mg/kg/day. Long-term prognosis is worse than in older children, particularly when SE follows HIE or cortical malformations. 3 4

8.6 Post-cardiac-arrest non-convulsive SE#

The post-arrest paralysed and cooled patient is the textbook NCSE-misses-it scenario. Up to 30% of comatose post-arrest patients have electrographic SE on cEEG. Treatment improves seizure control; whether it improves outcome in the dead-but-not-yet-declared brain is the subject of TELSTAR-style ongoing trials. Reasonable practice: cEEG on every cooled or recently cooled post-arrest patient; treat electrographic SE if present; cap aggression on the third escalation if the background is severely suppressed. 11 13

9. Multimodal integration matrix#

10. Worked alternative scenarios#

10.1 What if it is not really status?#

A 6-year-old with prolonged shivering, eyes deviated, post-anaesthetic emergence. cEEG is normal. Reassess: this is post-anaesthetic delirium, not SE. Do not start a second-line AED; reverse the precipitant. The discriminator is cEEG before escalation.

10.2 What if the second-line drug is contraindicated?#

A 15-month-old with global developmental delay, hypotonia, episodic lactic acidosis. Suspected mitochondrial disease. Valproate is contraindicated (POLG1-related hepatotoxicity risk). Use levetiracetam first; if breakthrough, phenobarbital rather than fosphenytoin (phenytoin is also pro-mitochondrial). Send a metabolic panel; pyridoxine trial 100 mg IV. 14

10.3 What if the cEEG endpoint is unobtainable?#

A 12-year-old septic patient, post-meningitis RSE, midazolam at 1.5 mg/kg/h with hypotension at MAP 50 needing noradrenaline. Cannot escalate sedation without further haemodynamic compromise. Accept a higher seizure-burden endpoint (intermittent breakthrough discharges acceptable in this physiology), prioritise sepsis source control, broaden cover, manage haemodynamics, and reassess for further sedation escalation once stabilised. The goal is the best achievable, not the textbook ideal.

11. Outcome data#

• ESETT (Kapur 2019, Glauser 2016): in the second-line AED trial, levetiracetam, fosphenytoin, and valproate produced clinical seizure cessation in roughly 50% at 60 minutes (45% to 52% across arms; no significant difference). The trial settled the long debate about second-line drug choice and reframed the question as one of dose adequacy and early administration. 1 2
• Time-to-treatment effect (Trinka 2015 definition; multiple cohorts): every 10-minute delay in benzodiazepine reduces the probability of cessation by approximately 5%. Seizure control at 60 minutes drops from 80% (treated within 30 min) to 30 to 40% (treated after 60 min). The earlier the better, with no plateau. 5
• ACNS pediatric cEEG indications (Herman 2015; Hirsch 2021 nomenclature): cEEG is recommended (Tier 1) for SE / RSE / suspected NCSE. The 2021 ACNS standardised nomenclature has improved inter-rater reliability for ictal-interictal continuum patterns. 7 8
• Neonatal SE (Pressler 2017 NEMO trial): phenobarbital first-line achieves seizure cessation in roughly 50% of neonates; bumetanide as an adjunct did not improve outcomes in NEMO. 3
• Post-arrest NCSE (Topjian 2021 AHA, Naim 2023): electrographic seizures are present in 10 to 30% of comatose post-arrest children; their treatment improves short-term seizure burden, with uncertain effect on neurological outcome. 11 13
• Foreman 2022 cEEG review: modern pediatric cEEG yield (proportion of monitored patients with electrographic seizures) is 10 to 40% depending on indication; highest in clinical SE, intermediate in post-cardiac-arrest, lower in routine neuro-checks. 15

12. Pitfalls#

• Under-dosing the benzodiazepine. Both ESETT and pre-hospital studies (Eclipse-SE) show that the typical benzo dose given is 50 to 70% of the weight-based target. Document and re-dose if needed. 2
• Believing the seizure has stopped because the convulsion has stopped. Roughly half of CSE that "looks stopped" still has electrographic ictus on cEEG. Until cEEG is reviewed, treat as ongoing.
• Waiting for cEEG before treating. cEEG informs what to do next; do not delay second-line or anaesthetic infusion for the trace. Use clinical and aEEG findings to act.
• Hyperventilating to manage suspected ICP. SE is a CMRO2 problem first; hypocapnia worsens regional ischaemia in the active focus. Keep PaCO2 in the normal range.
• Choosing valproate in suspected mitochondrial disease. POLG1-related hepatotoxicity is fatal; default to levetiracetam first-line in any infant with lactic acidosis or developmental regression. 14
• Propofol infusion syndrome in children. PRIS risk rises sharply above 5 mg/kg/h or beyond 48 h of infusion in children; pentobarbital is the preferred prolonged anaesthetic for paediatric SE.
• Setting the cEEG endpoint at isoelectric EEG. Burst-suppression at BSR 50 to 90% is the target; isoelectric brain is not the goal and exposes the patient to unnecessary haemodynamic risk.
• Forgetting electrolyte and glucose monitoring. SIADH is common; hyponatraemia worsens seizures; correct slowly. Recheck glucose, sodium, calcium, magnesium every 60 minutes early.

13. Pediatric considerations#

14. Combine with#

• EEG / aEEG modality page: full-montage versus reduced-channel, ACNS nomenclature, ictal-interictal continuum.
• NIRS modality page: asymmetry as a localising tool during SE.
• Pupillometry / NPi: bedside neurochecks that survive paralysis.
• Integration: EEG-TCD non-convulsive seizure pair: the discordant aEEG-with-TCD-pulsations scenario.
• Integration: Family communication: conversations during prolonged RSE and SRSE.
• Foundations: cerebral metabolism: why SE doubles CMRO2 and what the consequences are.

15. Evidence summary#

16. Recent literature (2022 to 2025)#

• Foreman 2022 review updates the case for cEEG in the modern PICU; key change since the 2012 piece is the wider availability of reduced-montage continuous trace as a 24/7 bedside resource. 15
• Hirsch 2021 ACNS nomenclature has improved inter-rater reliability for ictal-interictal continuum patterns, particularly periodic discharges. 8
• Sansevere 2023 neonatal cEEG review describes the rising adoption of full-montage continuous EEG in level III NICUs and the residual barriers (technician availability overnight, neurology coverage). 4
• Topjian 2021 pediatric AHA consolidates post-arrest cEEG as standard of care for the cooled or recently cooled child. 11
• Naim 2023 PCCM brain injury after arrest quantifies the relationship between electrographic seizure burden and 12-month neurological outcome. 13

17. Self-check#