In every other operation the anaesthetic is a background condition. Here it is a variable in the measurement the surgeon is about to cut on. Your drugs can hide the focus, invent a focus that is not there, or start a seizure in a patient whose head is in pins — and the middle one is the one nobody watches for.
An anaesthetic for epilepsy surgery can fail in three directions, and they are not variations of one problem. It can hide the focus — propofol obscures spike-wave activity for up to 20 minutes after the infusion stops. It can invent a focus — methohexital activates spikes in 50 to 85% of patients but showed up to 43% inappropriate activation in one study, and a resection guided by a drug-made spike removes the wrong brain. Or it can start a seizure in a patient in head pins, where the first move is not a drug at all: cold saline irrigation of the exposed cortical surface.
Neuro 2 day 04 was the awake craniotomy, where the patient’s job was to talk while a tumour came out. This is the operation next door, and it looks similar enough to be dangerous: same theatre, same drapes, often the same surgeon, and a completely different relationship between your drugs and the surgery.
There, the anaesthetic had to let the patient perform. Here the anaesthetic is inside the measurement.
Epilepsy is present in 0.5 to 2.2% of the population, and 30 to 40% do not respond adequately to medication. Of those with medically refractory seizures, only 10 to 30% are appropriate candidates for surgery, and only about 1% eventually have it.
So the patient in front of you has passed through three filters, and it is worth knowing what surgery buys them. In a series of 375 patients, full employment rose from 39.5% to 42.8% — a modest change — while part-time employment nearly doubled from 6.9% to 12.4%. Quality of life improves, including reduced depression. Most patients continue anticonvulsant therapy afterwards.
That is a real benefit stated honestly, and it is worth carrying into the consent conversation rather than the version where surgery cures epilepsy.
The classification changed and the old words persist, so it is worth being current. Seizures are focal onset, generalized onset, or unknown onset, with psychogenic non-epileptiform seizures as a separate entity — seizure-like manifestations with no corresponding EEG activity, considered conversion reactions.
Focal seizures arise in one area of one hemisphere. Without impairment of awareness they last under a minute and consciousness is preserved; with impairment of awareness there is loss of consciousness and spread beyond the focus. They are further described as motor or non-motor. A focal seizure that generalises is now called “focal to bilateral tonic-clonic” — not “secondarily generalised”.
Generalized seizures have no demonstrated focal onset, affect both hemispheres, and involve loss of consciousness.
Everything that makes a good neuroanaesthetic makes a bad recording.
Propofol is the commonest agent for these cases and the most awkward. It depresses electrocorticographic recordings and decreases the frequency of epileptogenic spike activity, quieting existing foci particularly in the lateral and mesial temporal areas, while having minimal effect on spontaneous interictal discharges. It can obscure spike-wave activity for up to 20 minutes after the infusion is stopped, and it produces high-frequency beta activity for as long as 30 minutes afterwards — hence the recommendation to stop it 20 to 30 minutes before recording, though some investigators report the beta activity does not prevent interpretation.
Confusingly, low-dose propofol has also been reported to activate spikes, and isolated tonic-clonic seizures have occurred with it, particularly on emergence. Myoclonus unrelated to excitatory EEG activity is more common than either.
Benzodiazepine premedication is avoided, because raising the seizure threshold makes epileptogenic activity harder to record. Isoflurane may decrease the frequency and spatial distribution of spikes, though whether that persists at low concentration is unclear. Nitrous oxide and other inhalational agents are generally advised against during recording.
The general principle for this whole class: most sedative-hypnotics are neuroexcitatory at low dose and depressant at high dose, which is why the group has the most confusing profile in the chapter and why a dose written down without a purpose is meaningless here.
This is today’s pitfall, so here only the table of who does what.
| Agent | Behaviour on the recording |
|---|---|
| Methohexital 25–50 mg | Activates in 50–85%; up to 43% inappropriate activation |
| Alfentanil 20 µg/kg | The most active: 83% of patients |
| Etomidate 0.2 mg/kg | High activation rate; use in intraoperative recording has not been studied |
| Ketamine | Non-specific activation, especially limbic; clinical seizures mostly above 4 mg/kg |
| Enflurane | The classic activator, dose-dependent, enhanced by hypocapnia — obsolete |
| Sevoflurane | Convulsions and spike waves reported; increases with dose and with hyperventilation |
| Desflurane | Does not share sevoflurane’s convulsant reports |
Two of these deserve a sentence on their own. Etomidate at higher doses produces burst suppression and can break status epilepticus — the same drug at both ends of the spectrum. And enflurane is the one agent the source explicitly says should be avoided in epilepsy unless the desired effect is to trigger seizures during recording, which is a fair summary of the whole chapter’s tension in one sentence.
Cortical stimulation, light anaesthesia and brain manipulation all provoke seizures, and during awake craniotomy the reported incidence is 4.9%. An untreated seizure in a patient whose head is in pins is described as potentially catastrophic — scalp and soft tissue injury, brain swelling from straining, and a patient who must be controlled quickly.
The difficulty is that the treatment destroys the recording. Anything that stops the seizure suppresses the activity the operation is being planned around. So the sequence is deliberately unusual:
That last clause is the useful one. Once the recording is over, the constraint is gone. Knowing exactly when that moment arrives is part of the job.
Antiepileptics change your anaesthetic before you give any of it.
Neuromuscular blockade. Long-term phenytoin or carbamazepine produces resistance to nondepolarizing blockers — pancuronium, vecuronium, metocurine, cisatracurium, rocuronium; less so atracurium. The mechanism is both pharmacodynamic, through upregulation of acetylcholine receptors, and pharmacokinetic, through increased hepatic metabolism.
Opioid requirement. A direct relationship has been reported between the number of anticonvulsants a patient takes and the fentanyl dose needed for maintenance. Not the class, the count.
Systemic effects worth checking before the case:
And one premedication rule that is easy to break by accident: antihistamines can activate seizure foci and should be avoided.
Several conditions associated with epilepsy carry anaesthetic problems of their own, and the reason to know them is that the epilepsy is what got the patient onto your list.
Neurofibromatosis — variable expression, from asymptomatic tumours to acoustic neuromas and significant intracranial lesions. Airway and respiratory tract tumours can cause chronic aspiration, pulmonary fibrosing alveolitis, pulmonary hypertension and cor pulmonale.
Tuberous sclerosis — benign tumours in brain, heart, lung, kidney, skin and eye. Obstructive hydrocephalus, dysrhythmias, cerebrovascular embolisation, renal dysfunction, arterial aneurysms. Intracardiac rhabdomyomas are found in 32.8 to 48%, which is why these patients need a full preoperative cardiac evaluation.
Down, Angelman and Sturge–Weber syndromes are also associated with epileptiform activity.
Subdural grid placement. A craniotomy under general anaesthesia with no intraoperative testing, so agents may be chosen freely. Hyperventilation for brain relaxation is weighed against precipitating a seizure — and may be less effective anyway, because patients with focal seizures can have lower CO₂ reactivity of cerebral blood flow.
Stereo EEG depth electrodes, which have largely replaced grids in the United States as the phase II localisation procedure. This anaesthetic inverts the usual habits:
Resection, with or without mapping. Without mapping it is an ordinary craniotomy: arterial line, access adequate for rapid blood loss, brain relaxation, immobility, and rapid emergence for neurological examination — remifentanil helps here, with a longer-acting opioid added for postoperative pain. With mapping, everything above applies.
If cortical motor stimulation is needed, neuromuscular blockade must be minimal, with neostigmine or sugammadex to fully reverse any residual.
Bradycardia during amygdalohippocampectomy. Severe bradycardia has been reported during amygdalohippocampectomy that is not seen during routine anterior temporal lobe resection, thought to result from surgical limbic stimulation enhancing vagal activity. If it happens, the operation is the cause.
Hemispherectomy is a different order of case. Performed mostly in children for diffuse foci, it requires a very large craniectomy and carries massive blood loss, electrolyte and metabolic disturbance, coagulopathy, cerebral haemorrhage, dural sinus tearing and reported air embolism. Comparing techniques, lateral hemispherectomy had the lowest blood loss, shortest intensive care stay and lowest complication rate; functional had the highest reoperation rate; anatomical had the longest stays, most need for CSF diversion and most postoperative fever. Patients with cortical dysplasia bled most.
Two details from that literature travel well. In children, urine output is a poor indicator of volume status because of frequent massive glycosuria. And patients undergoing major brain resection become hypercoagulable as early as dural closure, so thrombotic complications should be anticipated.
Extubation after a case with large-volume resuscitation and a high potential for complications is described as unwise; recovery belongs in intensive care.
A non-cranial neurostimulator for refractory epilepsy, reducing seizure frequency and severity by a mechanism that is not well understood — proposed to involve the limbic system, locus coeruleus and amygdala.
The one fact to keep: it goes on the left vagus nerve, to avoid the fibres supplying the sinoatrial node that travel with the right vagus, and therefore to avoid clinically significant bradycardia. The patient is supine with the head turned right.
Everywhere else in this curriculum, the anaesthetic supports the operation. Here it participates in the diagnosis, and it can corrupt the diagnosis in three distinguishable ways.
Hiding the focus wastes an operation. Igniting a seizure is an emergency with a protocol, and the protocol starts with cold water rather than a drug.
Inventing a focus is the one that leaves no trace — and the defence against it is not pharmacological. It is saying out loud, before you give the activating agent, what its false-positive record is, so that the decision belongs to the three people who should be making it together.
No randomized trial addresses this.
The agent-by-agent effects on the electrocorticogram are well described and most of the day's practical content comes from them: propofol suppresses and lingers, methohexital and etomidate and alfentanil activate, dexmedetomidine does neither, enflurane was the classic activator and is obsolete, desflurane is quiet where sevoflurane is not. These are consistent observations across many small studies, not contested claims.
The interaction with antiepileptic drugs is equally settled. Long-term phenytoin or carbamazepine produces resistance to nondepolarizing neuromuscular blockade — through both receptor upregulation and increased hepatic metabolism — and there is a direct relationship between the number of anticonvulsants a patient takes and the fentanyl dose required.
What is genuinely unsettled is whether to activate at all. The practice of giving a proconvulsant to raise a focus has been discouraged because of the risk of non-specific activation, with the possible exception of remifentanil, and the source states plainly that controversy exists over whether pharmacologically elicited spikes correspond to the patient's native epileptogenic foci. That is a live disagreement about a decision you may be asked to make in the room.
And the outcome data are more modest than the surgery's reputation. In a series of 375 patients, full employment rose only from 39.5% to 42.8% after epilepsy surgery, though part-time employment nearly doubled from 6.9% to 12.4%. Quality of life improves, including reduced depression. Most patients continue anticonvulsant therapy.
Nothing here rests on a randomized trial, and the source says so about the thing residents most want a rule for: prospective studies have not been sufficiently powered to determine the impact of anaesthetic technique on neurologic and functional outcome after craniotomy.
The recording is quiet. The neurophysiologist cannot find the spikes. The surgeon looks at you and asks whether you can bring them out.
You can. The doses are known: methohexital 25 to 50 mg, alfentanil 20 µg/kg, etomidate 0.2 mg/kg. Alfentanil is the most active, provoking abnormal spike activity in 83% of patients against methohexital's 50%. You give it, spikes appear, the team is relieved, and the resection is planned around them.
Here is what nobody in the room can see.
Methohexital's activation rate of 50 to 85% comes with up to 43% inappropriate activation in one study. The source states the problem directly: controversy exists over the correlation of pharmacologically elicited seizure spikes with the patient's native epileptogenic foci. And the practice as a whole has been discouraged because of the risk of non-specific activation, with the possible exception of remifentanil.
A drug-induced spike looks exactly like a real one. There is no feature that distinguishes them intraoperatively, no confirmatory test available on the table, and no feedback loop — the patient will not be back next week to tell you the resection missed. A false focus produces an operation that goes smoothly, finishes on time, and takes out brain that was never the problem.
Two more findings sharpen the same point, and both concern things done for unrelated reasons.
Sevoflurane's excitation is not localising. Widespread neuroexcitatory activity associated with sevoflurane did not facilitate seizure focus localization in patients with temporal lobe epilepsy. Spikes, yes. Information, no.
Hyperventilation degrades the map. Hypocapnia increases sevoflurane's spike-wave frequency, and it decreases the prediction specificity of leads with ictal spikes. So a manoeuvre you would do reflexively for brain relaxation — and you may still need mild-to-moderate hypocapnia at 30 to 35 mm Hg for exposure — makes the recording less trustworthy at the same time. If you hyperventilate during sevoflurane anaesthesia, say so, so the people reading the traces know.
So the rule is not "never activate." It is:
Activation is a request that should be answered out loud, with its uncertainty attached. If you are asked to bring out spikes, say what the agent does and what its false-positive record is, and let the electroencephalographer and the surgeon decide whether the information is worth having. The source frames this decision as one made in consultation among all three of you, and that is the safeguard — not a better drug.
And notice which way the asymmetry runs. If you hide the focus, the recording is non-diagnostic and everyone knows it; the operation stops or waits. If you ignite a seizure, it is dramatic and there is a protocol. If you invent a focus, everything looks like success. The failure mode with no alarm attached is the one worth being afraid of.
Full text · Preston goes through LibKey to whatever Preston holds — the PDF where there is one — and works on a phone. The doi link always reaches the publisher's record, whether or not we hold a subscription. If you are reading on a computer, LibKey Nomad adds the same one-click PDF button to PubMed and publisher pages everywhere else; it is a desktop browser extension, so it does nothing on a phone or tablet.
During electrocorticography no spikes are seen, and the surgeon asks you to give methohexital to bring out the focus. What is the principal objection?
A patient is undergoing temporal lobe resection with intraoperative electrocorticography. Which plan best preserves the recording?
A patient presents for stereo EEG depth electrode placement. Which statement about the anaesthetic is correct?
That is the whole orientation rule. Fat is bright on T1 and intermediate on T2, cortical bone gives no signal, and flowing blood produces a signal void. The practical summary is that anatomy is better seen on T1 and pathology on T2 — and post-contrast sequences are T1 weighted for exactly the reason given.
Ferromagnetic aneurysm clips are a contraindication; non-ferromagnetic ones are not. The risk from a ferromagnetic object is magnetically induced movement, with dislodgement and injury, and heating — image artefact is the secondary concern. The information you need is in the old operation note, not on any current imaging.