By the end of the established-status sequence, the patient should have received an adequately dosed benzodiazepine and a full loading dose of an appropriate nonbenzodiazepine antiseizure medication. If clinical or electrographic seizure activity continues despite both, the syndrome is refractory status epilepticus. No additional duration threshold is required. Before declaring pharmacologic resistance, verify that the benzodiazepine was actually weight appropriate, intravenous access was functional, the second-line load was completed, and persistent movement is truly ictal. That verification should take seconds, not become an excuse for another prolonged observation period. Ongoing generalized convulsive activity after two adequate treatment classes is not a cue to cycle leisurely through several additional maintenance drugs. It is a cue to build an ICU treatment platform capable of stopping the seizure while preserving systemic and cerebral physiology. The definition is straightforward; the decision to induce anesthetic coma is more phenotype dependent. (Brophy et al., 2012; Vossler et al., 2020)

The biologic rationale for rapid escalation is more sophisticated than “the first drugs did not work.” Experimental status epilepticus produces internalization of synaptic γ-aminobutyric acid type A receptors, reducing the receptor population through which benzodiazepines exert much of their antiseizure effect. Prolonged activity also disrupts chloride homeostasis and increases excitatory glutamatergic signaling, including recruitment of N-methyl-D-aspartate receptors to the neuronal surface. These observations come largely from experimental models. They do not prove that every human seizure follows an identical molecular clock, establish a precise bedside threshold, or demonstrate that one anesthetic strategy improves functional outcome. They do explain why repeatedly administering small benzodiazepine doses becomes progressively less rational and why mechanistically complementary therapy becomes attractive as status persists. (Naylor et al., 2005; Naylor et al., 2013)

The first attending-level distinction is whether the seizure phenotype makes continued activity more dangerous than the treatment required to suppress it. Persistent generalized convulsive status, subtle electrographic continuation after convulsions disappear, and nonconvulsive status in a comatose patient—particularly after preceding convulsive status—generally warrant rapid intubation, continuous EEG, and a continuous intravenous anesthetic. By contrast, focal aware motor status, epilepsia partialis continua, absence status, aphasic status, and some other focal nonconvulsive phenotypes may justify additional rapidly administered nonanesthetic medication before committing a hemodynamically stable patient to mechanical ventilation and coma. This is not therapeutic hesitation. It recognizes that the risk–benefit ratio of anesthesia changes with seizure type, level of consciousness, underlying lesion, comorbidity, and physiologic reserve. Current evidence most strongly supports continuous anesthetic treatment for convulsive refractory status and nonconvulsive status with coma; evidence is limited and the potential for treatment-related harm is greater in focal motor status or nonconvulsive status with preserved consciousness. (Fisch et al., 2026)

When anesthesia is indicated, intubation is not itself seizure treatment. It permits administration of doses that would otherwise abolish airway reflexes and ventilation. A neuromuscular blocker likewise stops neither cortical seizure activity nor seizure-mediated metabolic demand; it merely removes the motor display. An anesthetic antiseizure dose must precede paralysis, and continuous EEG should be connected as early as operationally possible. Otherwise, the team may see a quiet body while the brain remains in electrographic status.

Preparation for induction should anticipate the physiology that treatment will create. Obtain reliable access, establish continuous EEG, place an arterial catheter when feasible, prepare norepinephrine if pressure is marginal, and define an individualized blood-pressure target before escalation. Correct hypoxemia, hypoglycemia, clinically important sodium, calcium, or magnesium abnormalities, dangerous hyperthermia, and shock concurrently. In acute brain injury, the relevant harm is not only systemic hypotension but reduced cerebral perfusion across tissue with impaired autoregulation. A drug that suppresses seizures while collapsing arterial pressure may exchange excitotoxic injury for ischemic injury. Vasopressor support is often an accepted component of effective therapy, but a rising requirement should repeatedly prompt reassessment of volume status, myocardial function, sepsis, anesthetic choice, and whether the EEG target is deeper than necessary.

At the same time, refractory status must remain a phenotype rather than become the diagnosis. Repeat the etiologic inventory while suppression proceeds: acute ischemia or hemorrhage, CNS infection, medication withdrawal or nonadherence, toxic exposure, hypoglycemia, electrolyte disturbance, hypoxic–ischemic injury, tumor, trauma, eclampsia, autoimmune encephalitis, and less common metabolic or genetic disorders. Obtain antiseizure medication concentrations when interpretable, but do not wait for them before treating active status. Image the brain, sample cerebrospinal fluid when indicated and safe, and begin empiric antimicrobial treatment when encephalitis or meningitis remains plausible. The anesthetic purchases physiologic time; it does not treat HSV encephalitis, correct hyponatremia, remove cefepime, reverse isoniazid toxicity, deliver a fetus, or control an expanding hemorrhage. New-onset refractory status without a clear acute structural, toxic, or metabolic cause should trigger a NORSE-oriented evaluation early rather than after several days of anesthetic cycling.

No continuous intravenous anesthetic has been proved superior across refractory status epilepticus. The American Epilepsy Society review found predominantly insufficient comparative evidence for the commonly used anesthetic and nonanesthetic third-line agents. A 2024 systematic review included 66 heterogeneous studies and 1,637 patients. Initial anesthetic choice was associated with differences in short-term treatment failure, hypotension, and switching agents, but not with a convincing mortality difference; etiology was much more strongly associated with outcome. Barbiturates appeared to produce fewer acute treatment failures than midazolam or propofol but substantially more hypotension. The included studies were largely observational, used inconsistent definitions and EEG targets, and were highly vulnerable to confounding by indication. The bedside conclusion is not that the agents are interchangeable. It is that selection should be driven by seizure phenotype, cardiovascular reserve, organ function, concomitant brain injury, anticipated duration, and the pharmacology required at that moment. (Au et al., 2024; Vossler et al., 2020)

Whichever agent is selected, a recurrent operational error is starting an infusion without an adequate loading bolus. Changing an infusion rate alters the concentration slowly; an active seizure requires rapid attainment of an effective brain concentration. If definite seizures continue or recur during titration, re-bolus and then increase the infusion rather than repeatedly nudging the pump while waiting for equilibration. Doses should be titrated to the EEG and systemic response, not to the rates used for routine ventilator sedation. The commonly quoted ranges that follow derive chiefly from protocols, observational practice, and the 2012 Neurocritical Care Society guideline, not comparative dose-finding trials. (Brophy et al., 2012)

Midazolam is frequently chosen first because it acts quickly, is familiar, and is generally easier to support hemodynamically than propofol or a barbiturate. A commonly used adult regimen begins with approximately 0.2 milligrams per kilogram intravenously, followed by additional boluses for breakthrough seizures and an infusion titrated across a broad range of approximately 0.05 to 2 milligrams per kilogram per hour. Tachyphylaxis may develop, and prolonged high-dose exposure produces accumulation, especially with obesity, hepatic dysfunction, renal failure, or accumulation of the active glucuronidated metabolite. Breakthrough seizures during treatment and recurrence during tapering are common enough that apparent clinical quiet should never prompt premature EEG discontinuation. Delayed awakening after the infusion stops may reflect pharmacology rather than irreversible neurologic injury.

Propofol offers rapid onset, precise titratability, and potentially faster reassessment after a short exposure. Guideline-derived regimens use a loading dose of approximately 1 to 2 milligrams per kilogram followed by roughly 30 to 200 micrograms per kilogram per minute, although patient-specific dosing and local protocols should govern practice. Its reduction of cerebral metabolism can be useful when status coexists with intracranial hypertension, but the costs are vasodilation, myocardial depression, hypertriglyceridemia, pancreatitis risk, and substantial lipid calories. Propofol infusion syndrome is the major prolonged-exposure hazard. No dose-duration boundary is perfectly safe, although the older guideline specifically cautions about rates above 80 micrograms per kilogram per minute continued beyond 48 hours. New high-anion-gap or lactic acidosis, rising creatine kinase, rhabdomyolysis, hyperkalemia, acute kidney injury, bradyarrhythmia, or cardiovascular collapse should trigger immediate reassessment. With substantial exposure, follow acid–base status, lactate, creatine kinase, potassium, renal function, triglycerides, ECG and hemodynamic findings, and total caloric delivery rather than waiting for the complete syndrome to declare itself.

Ketamine supplies a mechanistically complementary option through NMDA-receptor antagonism rather than primary GABAergic potentiation. Adult protocols vary, but loading doses around 1 to 2 milligrams per kilogram followed by infusions commonly within approximately 1 to 5 milligrams per kilogram per hour illustrate contemporary practice rather than a validated optimum. A 2026 systematic review and meta-analysis pooled 14 observational studies comprising 388 adults and estimated seizure cessation in approximately 64%. That number must not be interpreted as a 64% independent drug effect: nearly all patients received concurrent anesthetics and antiseizure medications, definitions and timing varied, and confounding by disease severity was substantial. Earlier introduction is physiologically attractive when GABAergic escalation is causing hypotension or late status may be increasingly NMDA dependent, but prospective evidence remains inadequate. Ketamine often preserves pressure better than propofol or pentobarbital, yet a catecholamine-depleted patient can still develop myocardial depression. Ketamine also does not reliably create conventional burst suppression, so failure to produce that pattern is not treatment failure if electrographic seizures have stopped. (Lele et al., 2026; Fisch et al., 2026)

Pentobarbital remains a potent rescue option when seizures persist despite better-tolerated regimens or when deeper suppression is deliberately required. Guideline-derived regimens generally load approximately 5 to 15 milligrams per kilogram, with additional loading if necessary, followed by approximately 0.5 to 5 milligrams per kilogram per hour. Its strength is reliable cortical suppression. Its liabilities are vasoplegia, myocardial depression, frequent vasopressor dependence, respiratory suppression, paralytic ileus, propylene-glycol exposure, withdrawal seizures, and a long, unpredictable emergence. Escalating pentobarbital without a hemodynamic plan, bowel surveillance, infection vigilance, and explicit recognition of its prolonged effect is incomplete therapy.

The continuous anesthetic must be accompanied by a durable antiseizure scaffold. It is a temporary bridge, not the regimen the patient will retain after awakening. Confirm that the initial nonbenzodiazepine medication reached an adequate load, then establish scheduled therapy with one or more complementary agents selected for the seizure type, etiology, organ function, conduction system, hematologic profile, interactions, pregnancy status, and anticipated route of administration. Before anesthetic withdrawal, verify that every intended maintenance agent has actually been loaded and administered, enteral drugs are being absorbed, renal or hepatic changes have been incorporated, and serum concentrations are defensible where measurement is clinically meaningful. Failure to establish maintenance treatment before withdrawal and excessively rapid tapering are recognized causes of recurrence. (Mullhi et al., 2025)

Continuous EEG is the control surface for this treatment. The first defensible target for most refractory status epilepticus is complete cessation of electrographic seizures. Under the 2021 American Clinical Neurophysiology Society terminology, an electrographic seizure consists of epileptiform discharges averaging more than 2.5 hertz for at least 10 seconds, or any pattern with definite evolution lasting at least 10 seconds. Electrographic status is present when seizure activity continues for at least 10 minutes or occupies more than 20% of a 60-minute period. These definitions standardize communication; they do not make every rhythmic or periodic pattern a seizure. (Hirsch et al., 2021)

The ictal–interictal continuum contains patterns whose significance varies with frequency, evolution, burden, plus modifiers, clinical correlation, metabolic stress, and the vulnerability of the underlying brain. A medication trial can help, but it must be interpreted carefully. EEG improvement accompanied by meaningful clinical improvement supports an electroclinical seizure diagnosis. EEG improvement alone after a sedating benzodiazepine is weaker evidence because nonspecific periodic patterns may attenuate while the examination worsens from medication. Conversely, lack of immediate awakening does not disprove an ictal process when anesthetic accumulation, postictal dysfunction, structural injury, or organ failure obscures the examination. Quantitative EEG trends can identify changes in burden and direct attention to an event, but raw EEG review remains necessary. Reduced-montage EEG may provide a useful bridge or monitoring option when full-montage acquisition is unavailable, but its lower spatial coverage must be remembered during prolonged anesthetic titration.

Traditional guidance permits titration either to seizure cessation or to burst suppression. Burst suppression is a possible deeper target, not a universally proven destination. It is spatially heterogeneous, difficult to maintain at a fixed depth, dependent on both drug and brain state, and may contain highly epileptiform bursts. A suppressed scalp background does not establish that every cortical region is electrically silent, and an arbitrary number of bursts per minute or prescribed interburst interval has not been validated as an outcome-improving dose.

In a 2023 cohort of 147 anesthetic-treated adults, 102 had nonanoxic refractory status. Complete burst suppression was achieved in 21% of that nonanoxic subgroup for a median of 51 hours and was associated with more hypotension requiring vasopressors, but not with persistent seizure termination, in-hospital survival, or return to premorbid neurologic function after adjustment. The 2024 systematic review identified an apparent mortality association favoring burst suppression over seizure suppression, but that analysis rested on only 22 seizure-suppression patients from three publications and 98 burst-suppression patients from 12 publications, with major risks of selection bias and confounding. It is not practice-changing evidence. The current synthesis therefore supports electrographic seizure cessation as the initial target for most nonanoxic refractory status. Deeper suppression remains reasonable when seizures repeatedly break through at lighter levels, when ictal activity is embedded within highly epileptiform bursts, or in selected super-refractory cases, but it should be a conscious escalation purchased with additional hypotension and delayed awakening rather than an automatic endpoint. Postanoxic refractory status is biologically and prognostically distinct and should not be used to define treatment targets for other etiologies. (Fisch et al., 2023; Au et al., 2024; Fisch et al., 2026)

Duration is another area in which convention has outrun evidence. The 2012 Neurocritical Care Society guideline recommends maintaining electrographic control for approximately 24 to 48 hours before slow withdrawal, but explicitly grades that recommendation as weak and based on very-low-quality evidence. The 2025 UK multidisciplinary guidance recommends a minimum of 24 hours before tapering. No randomized trial establishes 24, 36, or 48 hours as a biological threshold that prevents recurrence. Longer exposure may prevent early breakthrough in some patients but also increases hypotension, infection, cardiac depression, ileus, mechanical ventilation, drug accumulation, delayed extubation, and delirium. A rapidly corrected toxic or metabolic trigger with a stable EEG background may not require the same treatment window as active encephalitis with highly epileptiform bursts. The control interval should be individualized and reconsidered daily rather than automatically renewed. (Brophy et al., 2012; Mullhi et al., 2025; Fisch et al., 2026)

Before the first wean, make the prerequisites explicit. The provoking cause has been treated as far as possible or remains under an active time-sensitive evaluation. Glucose, sodium, calcium, magnesium, temperature, oxygenation, perfusion, and acid–base disturbances have been corrected. The maintenance antiseizure regimen is fully loaded, scheduled, delivered, and adjusted for organ function and interactions. The EEG has remained free of definite seizures for the chosen interval, and the hemodynamic and ventilatory consequences of emergence are manageable. Anesthesia should never be withdrawn into a pharmacologic vacuum.

The infusion should then be reduced gradually under uninterrupted EEG surveillance, generally one continuous anesthetic at a time. There is no evidence-based universal decrement or interval; the rate should reflect the drug’s half-life, duration of exposure, seizure severity, EEG background, and maintenance-drug coverage. Routine daily sedation interruption is inappropriate while the sedative is functioning as active status treatment. Reassessment instead means confirming the indication, electrographic target, effective dose, complications, and readiness for a deliberate monitored taper.

If a definite electrographic seizure recurs, re-bolus the active agent and promptly return to the last effective infusion rate rather than waiting for a slow pump change to take effect. The recurrence should trigger a diagnostic and pharmacologic review: Was the maintenance load completed? Was a scheduled dose missed during transport or a procedure? Has renal clearance or enteral absorption changed? Is a new fever, electrolyte disturbance, infection, withdrawal state, or interacting medication present? Does the patient need a mechanistically complementary drug rather than another identical day of coma? If only isolated periodic discharges or a brief non-evolving rhythmic pattern emerges during awakening, pause and reassess before recreating deep anesthesia. The question is whether status has returned, not whether the EEG has stopped looking normal.

Continuous EEG should remain in place throughout anesthetic treatment and withdrawal. The ACNS consensus recommends documenting at least 24 hours of seizure control, monitoring during the entire period of continuous intravenous anesthetic therapy, and commonly continuing for at least 24 hours after withdrawal because recurrence is frequent. The exact duration should increase when the background remains highly epileptiform, seizures previously recurred during weaning, or the underlying process remains active. (Herman et al., 2015)

The surrounding ICU care is part of seizure treatment. Avoid hypoxemia, hypotension, fever, and extremes of glucose; use lung-protective ventilation; anticipate aspiration, atelectasis, and ventilator-associated infection; account for propofol calories; and provide venous-thromboembolism prevention, bowel and bladder management, pressure-injury prevention, eye care, nutrition, and line stewardship. These are not peripheral details. Once seizures are controlled, treatment complications may become the dominant determinant of outcome.

Prognosis should be anchored primarily to etiology, premorbid reserve, structural injury, and systemic complications rather than the number of failed drugs or the dramatic appearance of multiple infusion pumps. Epilepsy-related refractory status has substantially lower reported mortality than status caused by many acute destructive etiologies. Duration matters, but it is entangled with cause, treatment delay, and physiologic complications. Delayed awakening after midazolam, pentobarbital, or multiorgan failure should not be equated prematurely with irreversible brain injury. Cortical diffusion restriction or FLAIR hyperintensity on MRI may be peri-ictal and reversible; its distribution, vascular conformity, associated perfusion, and evolution determine its meaning. Neither prolonged anesthetic exposure nor a failed first wean is, by itself, a test of futility.

Several bedside failures recur across this phase: mistaking paralysis for seizure cessation, using routine sedation doses for an anesthetic-level emergency, increasing an infusion without re-bolusing while seizures continue, pursuing burst suppression despite escalating vasopressor requirements, beginning withdrawal before maintenance medications are established, treating every ictal–interictal-continuum pattern as definite status, and interpreting drug accumulation as devastating neurologic injury. Avoiding these failures requires the same discipline as selecting the drug: name the target, measure whether it has been reached, and continually compare the neurologic benefit with the systemic price.

If seizures continue or recur despite an adequate continuous anesthetic for more than 24 hours, including recurrence during anesthetic withdrawal, the syndrome has crossed into super-refractory status epilepticus. At that point, repeatedly deepening the same GABAergic coma is no longer a complete strategy. The next installment will address super-refractory status and NORSE/FIRES: reopening the etiologic search, deciding when infectious and autoimmune investigation or empiric immunotherapy becomes urgent, and evaluating ketamine, ketogenic therapy, inhaled anesthetics, immunomodulation, surgery, and neuromodulation without overstating the quality of their evidence.

References