Once status epilepticus persists or recurs at least 24 hours after continuous anesthetic therapy begins, including recurrence during anesthetic reduction or withdrawal, the syndrome has crossed into super-refractory status epilepticus. The definition is based on treatment behavior rather than a discrete biologic transition, but it should change the management frame. The problem is no longer simply how to suppress seizures for another 24 hours. Seizure control, identification and treatment of the underlying disease, construction of a durable nonanesthetic antiseizure regimen, preservation of systemic physiology, and maintenance of future recoverability must now proceed simultaneously. Repeatedly producing a quiet EEG with deeper anesthesia while the causal process remains untreated is not definitive management. (Shorvon & Ferlisi, 2011) (PubMed)

Super-refractory status epilepticus, NORSE, and FIRES must be kept conceptually separate. Super-refractory status describes treatment duration and response. New-onset refractory status epilepticus, or NORSE, describes a clinical presentation: refractory status developing in a patient without active epilepsy or another relevant preexisting neurologic disorder when no acute or active structural, toxic, or metabolic cause is immediately apparent. FIRES is a subtype of NORSE in which a febrile infection began between two weeks and 24 hours before refractory status; fever need not still be present when seizures begin. NORSE should therefore be recognized at the refractory stage, potentially well before the patient qualifies as super-refractory. If HSV encephalitis, autoimmune encephalitis, malignancy, or another etiology is subsequently established, that diagnosis explains the NORSE presentation. “Cryptogenic NORSE” should be reserved for disease that remains unexplained after an appropriately extensive investigation, not merely after a normal CT, bland routine CSF, and negative commercial antibody panel. (Hirsch et al., 2018)

The first action at the super-refractory checkpoint is verification. Review the continuous EEG alongside the entire electroclinical trajectory and ask whether the current treatment target is genuinely recurrent seizure or evolving electrographic status. After prolonged seizures, anesthesia, metabolic disturbance, and structural injury, the EEG commonly contains periodic discharges, rhythmic delta activity, stimulus-induced patterns, discontinuity, or abnormalities along the ictal-interictal continuum. Not every ominous waveform is status epilepticus. Frequency, evolution, spatial spread, clinical and physiologic correlates, seizure burden, background organization, and response to a rational treatment trial should determine whether deeper coma is justified. Otherwise the team can begin treating the appearance of an EEG rather than an injurious ictal process.

At the same time, audit pharmacologic delivery. Confirm loading doses, body weight used for dosing, infusion interruptions, line function, measured concentrations when informative, protein binding, hepatic and renal clearance, extracorporeal removal, drug interactions, and enteral absorption. Paralysis can eliminate the motor phenotype without altering cortical seizure activity. When breakthrough seizures occur during an anesthetic infusion, an appropriate rebolus may restore effective concentrations more rapidly than slowly increasing the maintenance rate. Most importantly, continuous anesthesia cannot remain the only mechanism maintaining seizure control. A meaningful nonanesthetic regimen must already exist before another wean is attempted.

That regimen should be constructed by pharmacology rather than by counting medications. Levetiracetam or brivaracetam, valproate, lacosamide, fosphenytoin or phenytoin, phenobarbital, clobazam, perampanel, topiramate, and other agents may be combined depending on etiology and patient physiology, but no randomized evidence establishes an optimal SRSE cocktail. Mechanistic complementarity matters, as do cardiac conduction, platelet count, ammonia, hepatic and renal function, protein binding, drug interactions, bowel function, and whether enteral agents are actually being absorbed. Five nominal antiseizure medications do not necessarily create a stronger platform than three appropriately selected and adequately exposed agents.

The EEG endpoint under anesthesia likewise requires discipline. Burst suppression is a pharmacodynamic manifestation of profound cortical depression, not a validated surrogate for neurologic recovery. No high-quality evidence establishes a universal interburst interval or suppression ratio that improves outcome over reliable seizure cessation. In a 147-patient cohort of anesthetic-treated refractory status, induced burst suppression was not independently associated with persistent seizure termination, hospital survival, or return to premorbid neurologic function after adjustment. This observational result cannot prove that burst suppression is never useful, particularly when seizures cannot otherwise be controlled, but it argues strongly against treating a predetermined depth of suppression as intrinsically therapeutic. A 2024 systematic review similarly found that the choice of initial continuous intravenous anesthetic affected short-term treatment failure and adverse-event profiles, particularly hypotension, more clearly than mortality. The working target should therefore generally be durable electrographic seizure cessation, with deeper suppression used when necessary to achieve that target rather than pursued ritualistically. (Au et al., 2024; Fisch et al., 2023) (JAMA Network)

This distinction becomes most important during anesthetic withdrawal. Recurrence of unequivocal electrographic seizures means the wean has failed. Reappearance of periodic discharges does not necessarily mean the same thing. A rapidly evolving 2.5-Hz lateralized pattern with physiologic deterioration is fundamentally different from static 1-Hz periodic discharges over irreversibly injured cortex in a clinically improving patient. The danger of undertreating genuine status must be balanced against the harm of repeatedly returning a patient to profound coma because any increase in EEG activity is interpreted as relapse.

While electrical control continues, the etiologic search should intensify rather than simply continue at its original level. Repeat the history with family, emergency personnel, pharmacists, and relevant occupational or travel contacts. Revisit febrile prodromes, respiratory or gastrointestinal illness, behavioral change, insomnia, dysautonomia, recent medications, immune-checkpoint inhibitors, recreational drugs, alcohol or benzodiazepine withdrawal, supplements, toxic exposures, pregnancy or postpartum state, malignancy symptoms, animal or arthropod exposure, and family history suggesting mitochondrial, metabolic, or genetic disease. A disease that was clinically invisible during the first several hours may declare itself over the following days.

MRI should include diffusion, FLAIR, susceptibility-sensitive imaging, and gadolinium-enhanced sequences, with vascular imaging when the phenotype warrants it. A normal initial MRI does not close the investigation. Conversely, an abnormal MRI does not automatically identify the etiology. Prolonged seizures themselves can produce cortical or hippocampal diffusion restriction, FLAIR hyperintensity, thalamic abnormalities, perfusion changes, and enhancement. Serial imaging should therefore be interpreted against the timing and topography of the EEG. New cortical restriction precisely beneath days of lateralized seizure activity may be a consequence of status rather than proof of encephalitis; an inflammatory or limbic pattern preceding maximal seizure burden carries a different implication.

CSF investigation should likewise be intentionally broad. Routine cell count, differential, glucose, protein, cultures, and appropriately timed infectious testing are the foundation, with HSV and VZV receiving particular attention and additional pathogens selected according to geography, season, exposures, and immune status. Empiric antimicrobial therapy should continue when the clinical probability of a dangerous treatable infection remains meaningful. Paired serum and CSF autoimmune and paraneoplastic testing should generally be obtained because diagnostic sensitivity varies by compartment. Oligoclonal bands and IgG index can add context. Pretreatment serum and CSF should be stored when feasible because later IVIG, plasma exchange, or other immunotherapy may complicate interpretation, and expanded antibody testing, metagenomic sequencing, or research-level inflammatory studies may later become useful. Brain biopsy should not be a random fishing expedition; it becomes more defensible when imaging provides a target or when vasculitis, lymphoma, granulomatous disease, unusual infection, or another tissue diagnosis would materially alter management. The international NORSE consensus specifically emphasizes broad early investigation rather than postponing etiologic work until conventional seizure therapy has failed repeatedly. (Wickström et al., 2022) (PubMed Central (PMC))

In an unexplained adult, malignancy evaluation should be driven by the phenotype but may require cross-sectional chest, abdominal, and pelvic imaging, sex-specific pelvic or gonadal evaluation, and additional metabolic imaging when suspicion persists despite initially unrevealing studies. A neurologic syndrome may precede radiographic detection of the responsible malignancy, so selected patients require repeat screening. Age and phenotype similarly govern the metabolic and genetic branch. Younger patients raise greater concern for mitochondrial disease, channelopathies, and inborn metabolic disorders, but ammonia, porphyria, endocrine or electrolyte abnormalities, toxins, and medication withdrawal remain relevant at any age. Valproate deserves particular caution when POLG-related disease is biologically plausible.

Normal routine CSF and MRI do not exclude immune-mediated disease. Autoimmune encephalitis may present with little or no pleocytosis and initially normal imaging. Conversely, absence of a recognized neuronal antibody does not establish cryptogenic NORSE as autoimmune. The correct approach is probabilistic: progressively integrate history, serial imaging, EEG, CSF, systemic findings, antibody results, infectious studies, and treatment response rather than expecting one diagnostic test to classify the syndrome.

In cryptogenic NORSE and FIRES, inflammation is increasingly understood as a plausible component of the self-sustaining ictal state. Prolonged seizures disrupt the blood-brain barrier and activate innate immune pathways, while inflammatory signaling can further lower seizure threshold and amplify excitotoxicity. Hanin and colleagues demonstrated abnormal serum and CSF cytokine and chemokine profiles in NORSE, with prominent innate immune activation and associations between higher inflammatory signals and worse outcome. That observation supports biologic plausibility but does not resolve causality. Inflammation may initiate seizures, result from seizure burden, or participate in a bidirectional feedback loop. Cytokine concentrations are therefore not currently validated diagnostic tests for NORSE and cannot yet select anakinra versus tocilizumab with precision. (Hanin et al., 2023) (PubMed)

That uncertainty should not become therapeutic paralysis. The international NORSE consensus recommends initiating first-line immunotherapy within the first 72 hours of seizure onset when NORSE/FIRES remains unexplained and common infectious etiologies have been reasonably addressed, with many panelists favoring initiation as early as 48 hours. Corticosteroids, IVIG, and therapeutic plasma exchange are the principal first-line strategies. A commonly recommended corticosteroid regimen is IV methylprednisolone 20–30 mg/kg/day, maximum 1 g/day, for three to five days; IVIG may be administered at a total dose of 2 g/kg over two to five days. Plasma exchange may also be appropriate but should not delay subsequent therapy. These are consensus recommendations supported by limited observational evidence, not randomized proof that empiric immunotherapy improves outcome in every patient with cryptogenic NORSE. The rationale is time sensitivity: if the process is immune mediated, waiting days to weeks for antibody results may sacrifice a modifiable phase of disease. (Wickström et al., 2022) (Wiley Online Library)

The converse is equally important. Cryptogenic does not mean autoimmune, and apparent steroid responsiveness does not establish autoimmune encephalitis. Infectious treatment and diagnostic reassessment must continue when warranted. If a specific neuronal-surface antibody or another defined immune syndrome is identified, therapy should move toward that disease-specific pathway rather than automatically toward cytokine blockade.

When noninfectious NORSE remains uncontrolled despite first-line therapy, consensus favors moving during the first week to both ketogenic therapy and second-line immunotherapy rather than spending several weeks cycling anesthetics first. In a conventional antibody-mediated phenotype, rituximab may be mechanistically appropriate. In cryptogenic NORSE and FIRES, interest has focused on cytokine-directed therapy, particularly the IL-1 receptor antagonist anakinra and the IL-6 receptor blocker tocilizumab. Neither has proven superiority. The supporting literature consists principally of case series, retrospective cohorts, and registries, and a 2024 review concluded that available evidence cannot establish which second-line immunotherapy is best or identify a validated responder phenotype. (Hanin et al., 2024) (PubMed)

The foundational anakinra experience is particularly pediatric, whereas the initial tocilizumab NORSE literature consisted of very small uncontrolled series. These reports are important signals but remain vulnerable to selection bias, publication bias, simultaneous therapies, and spontaneous late improvement. A large PCORI-funded study is now intended to compare anakinra with tocilizumab directly in cryptogenic NORSE. As of February 18, 2026, PCORI lists the project as an in-progress randomized controlled trial with an observational arm and as not yet recruiting; it therefore has no outcome data available to guide current treatment. Cytokine blockade remains biologically rational and consensus compatible, but it should not be described as biomarker-proven precision therapy. (PCORI)

Ketamine addresses a different component of the prolonged ictal state. Experimental status is associated with altered inhibitory receptor trafficking, including reduced synaptic GABA-A receptor availability, while NMDA-mediated excitation becomes increasingly relevant. NMDA antagonism therefore provides a mechanistically distinct strategy when repeated GABAergic therapy is failing. Ketamine also often produces less vasodilation and myocardial depression than propofol or barbiturates, although its hemodynamic behavior becomes less predictable in profoundly catecholamine-depleted shock.

Clinical evidence remains observational. In a retrospective series of 68 patients with super-refractory status, all also receiving midazolam, seizure burden decreased by at least 50% within 24 hours in 81%, and seizures stopped in 63%. The average ketamine infusion was approximately 2.2 mg/kg/hour for a median of two days. Because the study provided Class IV evidence without an untreated comparator, seizure cessation cannot be attributed to ketamine alone. Vasopressor requirements decreased, and the small subgroup undergoing invasive intracranial monitoring did not demonstrate a harmful ICP signal, further weakening the old assertion that ketamine is categorically contraindicated in intracranial disease. This does not establish universal safety or an outcome benefit. Ketamine is best regarded as a mechanistically rational component of SRSE therapy rather than a proven rescue drug. (Alkhachroum et al., 2020) (PubMed)

Ketamine is not physiologically free. Tachyarrhythmia, hypersalivation, hepatotoxicity during prolonged exposure, and myocardial depression in catecholamine-depleted states remain relevant. Barbiturates can produce powerful seizure suppression but at the cost of myocardial depression, vasoplegia, ileus, infection risk, tissue accumulation, and prolonged clearance. Propofol introduces hypotension, hypertriglyceridemia, metabolic acidosis, rhabdomyolysis, and the potentially catastrophic propofol infusion syndrome. Midazolam accumulates during prolonged exposure, particularly with organ dysfunction, and can make neurologic recovery appear worse than it is. The question is therefore never merely which anesthetic can abolish the EEG abnormality. It is which strategy can achieve the necessary electrical target at a systemic cost the patient can continue to tolerate. The available comparative literature shows differences in treatment failure and adverse events between anesthetic strategies but no demonstrated mortality advantage for one initial continuous anesthetic. (Au et al., 2024) (JAMA Network)

Ketogenic therapy should be operationalized early enough to work rather than introduced after every pharmacologic option has failed. Its potential actions extend beyond ketone production alone and include alterations in substrate metabolism, mitochondrial energetics, neurotransmitter handling, ion-channel activity, adenosine signaling, oxidative stress, and inflammatory pathways. The NORSE consensus recommends beginning ketogenic therapy within the first week in cryptogenic NORSE/FIRES with inadequate response to initial therapy. (Wiley Online Library)

In the ICU, this is a metabolic treatment rather than simply a tube-feed selection. Pharmacy and nutrition teams must identify carbohydrate in liquid medications, diluents, supplements, and feeding products. Serum beta-hydroxybutyrate is more useful than urine ketones for monitoring the intended metabolic state. Glucose, bicarbonate, anion gap, electrolytes, triglycerides, hepatic function, ammonia, lipase, caloric delivery, and fluid balance require surveillance. Topiramate and zonisamide can compound metabolic acidosis and nephrolithiasis risk; valproate may add hepatic or carnitine-related metabolic stress. Major disorders of fatty-acid oxidation are contraindications. Prolonged high-dose propofol and ketogenic therapy should generally not be assumed to be metabolically neutral when combined because both place substantial demands on fatty-acid and mitochondrial metabolism.

The most recent comparative adult data illustrate why enthusiasm should remain calibrated. In the 2026 Feil cohort, 18 adults receiving ketogenic therapy were compared with 16 severity-matched controls. Diet initiation occurred a mean of 16.6 days after status onset, and measurable ketosis was achieved in only six of 18 treated patients. Unadjusted SRSE resolution was 61.1% in ketogenic patients and 87.5% in controls, although ketogenic patients had substantially longer and more treatment-intensive status before therapy. Time-dependent and multivariable modeling generated an association between ketogenic treatment, particularly earlier initiation, and faster SRSE resolution and better later functional outcome. The authors appropriately characterize the result as a therapeutic signal requiring prospective evaluation. With only 34 patients, nonrandomized treatment allocation, residual group imbalance, and unavoidable concerns regarding treatment-selection and immortal-time bias, the study does not establish efficacy. It strengthens the rationale for prospective study and for not waiting until the final days of SRSE to consider a consensus-supported therapy. (Feil et al., 2026) (link.springer.com)

Beyond these approaches, rescue therapy becomes increasingly phenotype dependent. Inhalational anesthetics can rapidly suppress epileptiform activity in centers capable of delivering them safely, but suppression during exposure does not necessarily create durable remission after withdrawal. A multicenter isoflurane series demonstrated substantial on-treatment EEG suppression while emphasizing that establishing sustained remission remained difficult and systemic adverse effects were common. Volatile anesthesia is therefore better conceptualized as another bridge than as etiologic therapy. (Stetefeld et al., 2021) (PubMed)

Routine systemic hypothermia is not supported. In the randomized HYBERNATUS trial, adding 24 hours of cooling to 32–34 °C to standard treatment for ventilated convulsive status did not significantly improve 90-day functional outcome. Hypothermia can suppress seizures physiologically, but electrophysiologic plausibility does not equal clinical benefit, particularly when infection, arrhythmia, electrolyte disturbances, coagulopathy, and altered drug disposition accompany treatment. (Legriel et al., 2016) (PubMed)

Surgical thinking should occur much earlier when the electroclinical syndrome has a fixed, potentially resectable generator. A focal structural lesion or consistently unilateral ictal onset is fundamentally different from diffuse, multifocal, or migrating cryptogenic NORSE. Resection or disconnection may terminate otherwise unmanageable status in highly selected cases, while vagus-nerve stimulation, deep-brain stimulation, responsive neurostimulation, electroconvulsive therapy, and transcranial magnetic stimulation remain supported primarily by small observational series. These interventions should never delay treatment of infection, autoimmunity, inflammation, or a clearly remediable structural lesion.

The systemic platform becomes increasingly inseparable from seizure treatment as the course lengthens. Maintain adequate MAP and cerebral perfusion, lung-protective ventilation, nutrition, temperature control, venous-thromboembolism prevention, bowel function, skin and ocular protection, and aggressive surveillance for line and ventilator infection. Trend creatine kinase, triglycerides, lactate, acid-base status, hepatic function, and renal clearance according to treatment exposure. Remove ineffective medications and unnecessary invasive devices. Consider tracheostomy when prolonged ventilation has become the expected trajectory. A medication that suppresses seizures while generating profound hypotension, recurrent infection, metabolic failure, or weeks of unnecessary coma may simply exchange one form of secondary brain injury for another.

Neuroprognostication during SRSE requires equivalent restraint. In the 2023 systematic review and meta-analysis by Cornwall and colleagues, in-hospital mortality was approximately 24%, while only a minority of patients had no or slight disability at discharge. Yet seizure termination remained possible after extremely prolonged courses. These observational data do not show that weeks of aggressive treatment improve outcome—patients must survive and continue to be selected for treatment in order to receive prolonged therapy—but they demonstrate that duration alone is not a validated futility threshold. (Cornwall et al., 2023) (PubMed)

Etiology should carry far greater prognostic weight than the label SRSE itself. Postanoxic status accompanying overwhelming global hypoxic-ischemic injury is biologically different from cryptogenic NORSE in a previously healthy patient, an actionable autoimmune encephalitis, a focal surgically remediable lesion, or withdrawal-related status in a structurally preserved brain. Premorbid function, age, serial MRI, EEG background evolution, etiology, systemic organ injury, brainstem function, treatment complications, and evidence of emerging interaction should all contribute. Prolonged anesthetic accumulation itself can make the bedside examination nearly uninterpretable.

The risk of self-fulfilling prognosis is particularly real in NORSE. In the 2024 Stretti cohort, most patients developed super-refractory status, yet meaningful functional improvement occurred among a substantial proportion of survivors after discharge. The cohort is small and selected and cannot provide an individual prognosis, but it reinforces an important principle: prolonged coma and severe disability at ICU or hospital discharge do not reliably define the ultimate neurologic ceiling. (Stretti et al., 2024) (PubMed)

The super-refractory checkpoint therefore expands the treatment plane rather than merely escalating the same algorithm. Continue to suppress genuine injurious electrical activity, but continually reassess whether the EEG target still represents seizure. Build a nonanesthetic regimen capable of carrying the cortical network through withdrawal of continuous anesthesia. In NORSE, reopen the etiologic search early and treat potentially modifiable immune or infectious disease while diagnostic work proceeds. Initiate first-line immunotherapy on the consensus timescale when the phenotype supports it, and consider ketogenic and second-line immune therapy during the first week rather than after a month of pharmacologic failure. Use ketamine, barbiturates, inhaled anesthetics, surgery, and other rescue strategies according to the physiology and phenotype rather than as an automatic ladder. Throughout the course, preserve the systemic organism that the recovering brain will eventually require.

The successful exit from super-refractory status epilepticus is therefore not merely another 24 hours of burst suppression. It is a patient whose seizures are controlled without continuous anesthesia, whose underlying disease is being treated as specifically as the evidence permits, whose maintenance regimen can tolerate return of normal cortical activity, and whose cerebral and systemic physiology have survived the treatment required to reach that point.

References