15.1 EEG, Standard and Processed (02.B)

EEG that changes management in the neuro ICU

Quick Answer: Order continuous electroencephalography (cEEG) for unexplained coma, status epilepticus (SE) that is not clearly over, possible nonconvulsive status epilepticus (NCSE), and titration of anesthetic burst suppression. Record 24–48 hours in high-risk patients because a routine 30–60 minute tracing misses many nonconvulsive seizures. Lateralized periodic discharges (LPDs), generalized periodic discharges (GPDs), and stimulus-induced rhythmic, periodic, or ictal-appearing discharges (SIRPIDs) live on the ictal-interictal continuum (IIC) in American Clinical Neurophysiology Society (ACNS) 2021 language — they are not automatic seizures. Bispectral index (BIS) and other processed frontal numbers are not ischemia or awareness monitors. After cardiac arrest, highly malignant patterns (suppressed background, burst-suppression, identical bursts) support poor prognosis only after rewarming and delayed multimodal assessment, typically after 72 hours.

Independent OpenExamPrep teaching in this section covers standard and processed EEG listed under Diagnostic studies in the ABPN Content Specifications. This is not an ACNS product and not a substitute for a neurophysiology read. The job on the examination is to know when to hook the patient up, how long to leave the wires on, what periodic patterns mean conceptually, why a forehead number is not a stroke monitor, and how not to withdraw care on an EEG recorded during hypothermia.

Why a spot EEG is not the whole test

A routine EEG is a 20–60 minute sample. It answers “what is the background right now?” and can catch frequent seizures. It does not answer “will this comatose patient have a five-minute electrographic seizure at 03:00?” Claassen and colleagues (2004) showed that among critically ill patients who eventually had electrographic seizures, only about half declared the first seizure in the first hour, about 88% by 24 hours, and about 93% by 48 hours. Comatose patients were more likely to declare late. ACNS consensus statements on critical-care cEEG therefore recommend recording at least 24 hours in most at-risk adults and 48 hours or more when the patient is comatose, already shows periodic discharges, or is deeply sedated.

Worked timing: you obtain a 40-minute EEG at 14:00 on a stuporous subarachnoid-hemorrhage patient. It is slow, reactive, without epileptiform discharges. That tracing has not ruled out NCSE overnight. If the question is salvage of nonconvulsive seizures, you convert to cEEG and you keep it on.

Indications you should not negotiate away

Salvage of NCSE means you treat electrographic SE you would have missed without wires. Duration of untreated nonconvulsive seizures is associated with worse outcome in several ICU cohorts. That is not a license to put every mildly confused hyponatremic patient on a 72-hour montage, but it is a license to treat unexplained coma as an EEG indication until proven otherwise.

Burst-suppression titration is a second, separate indication. When pentobarbital, midazolam, or propofol is used to treat refractory SE or to drop cerebral metabolic demand, the target is a specified burst-suppression pattern on raw EEG (commonly 1–2 bursts per 10 seconds — local protocols vary). You cannot titrate that from a Ramsay score or from a single BIS integer. Hypothermia, high-dose anesthetics, and severe anoxia all produce suppression; the tracing tells you whether you have drug effect or a dead cortex, but only if you look at the waveform and the clinical context together.

How long is 24–48 hours in practice?

Start cEEG as soon as NCSE is on the list, not after the next morning's technician shift if you can help it. If the first two hours show no epileptiform discharges in a non-comatose patient, seizure risk falls sharply (Westover and 2HELPS2B work). If the first hours show LPDs, GPDs, or fluctuating rhythmic delta, you have not “cleared” the patient at 24 hours — those patterns both raise seizure risk and may themselves sit on the IIC. Coma, periodic discharges, and anesthetic coma are the three reasons to plan 48 hours from the start rather than arguing with the lab at hour 23.

Neurocritical Care practice questionsPractice questions with detailed explanations

ACNS language without turning LPDs into pentobarbital

ACNS 2021 Standardized Critical Care EEG Terminology is the shared vocabulary. You do not need to grade every modifier like an epilepsy fellow. You do need the conceptual map.

Electrographic seizures and electrographic status epilepticus have explicit duration and evolution criteria (Salzburg-based ACNS definitions). Patterns that do not meet those criteria but still might be harming the brain are placed on the ictal-interictal continuum. ACNS treats IIC as a purely electrographic term, synonymous with “possible electrographic seizure,” not as a bedside diagnosis. Typical IIC ingredients include periodic discharges or spike-wave averaging >1.0 and <2.5 Hz, slower periodic patterns (>0.5 and <1.0 Hz) that carry a plus modifier or fluctuation, and lateralized rhythmic delta activity faster than 1 Hz with plus or fluctuation. Generalized rhythmic delta without other features is not on that IIC list.

LPDs (older name PLEDs) are unilateral or bilateral-asymmetric periodic discharges. They are common after stroke, hemorrhage, infection, and tumor. They raise the odds of electrographic seizures. They are not, by themselves, a mandate to induce burst suppression. GPDs are bilateral synchronous periodic discharges; they appear in anoxic injury, metabolic encephalopathy, cefepime neurotoxicity, and some degenerative diseases. Frequency and plus features (+F superimposed fast, +R rhythmic, +S spike/sharp) push the pattern toward the ictal end of the continuum. A diagnostic trial of a parenteral antiseizure drug with simultaneous EEG and a clinical/electographic endpoint is often more honest than either ignoring 2 Hz LPDs or calling them “status” in the chart.

SIRPIDs are the umbrella for stimulus-induced (or stimulus-exacerbated) rhythmic, periodic, or ictal-appearing discharges. ACNS prefers that you name the specific SI-pattern (SI-LPDs, SI-seizures) rather than stopping at the umbrella. Stimulus-induced patterns can be epileptic, can be nonepileptic irritability, and can coexist with true seizures. Suctioning the endotracheal tube is a classic trigger. Do not treat every stimulus-induced run as NCSE, and do not dismiss an evolving, unreactive, high-frequency stimulus-induced pattern as “just stimulation.”

The examination wants this humility: IIC is a risk category and a treatment-trial category, not a binary. Overclaiming (“all GPDs are seizures”) and underclaiming (“periodic discharges never harm”) are both wrong.

Processed EEG and BIS: what the number cannot do

Processed EEG takes a limited montage, usually frontal, and compresses it into a 0–100 index, a suppression ratio, and a density spectral array. BIS is the proprietary example every intensivist has seen on an anesthesia monitor. Designed for hypnotic depth in the operating room, it is sometimes borrowed in the ICU to “watch sedation” during neuromuscular blockade.

Limitations that matter on this examination:

• Ischemia. Frontal electrodes do not see a posterior cerebral artery infarct, many MCA insular injuries, or delayed cerebral ischemia in a watershed that is not under the strip. A sudden BIS drop can accompany global hypoperfusion, but a stable BIS does not exclude regional ischemia. Carotid and thrombectomy series are mixed: some show a percentage drop with clamping, with poor positive predictive value; others show no reliable change at reperfusion. Processed EEG is not a substitute for the neurologic exam, transcranial Doppler, perfusion imaging, or cEEG when ischemia is the question.
• Awareness. A BIS of 40 does not prove the paralyzed patient is unaware. A BIS of 70 does not prove recall. Electromyographic activity inflates the index; neuromuscular blockade deflates it. Seizures, hypothermia, hypoglycemia, and metabolic suppression all scramble the number. The large anesthesia literature did not establish processed EEG as a foolproof awareness monitor even in the operating room; it certainly is not one in a structurally injured brain.
• Seizures. BIS cannot localize or reliably detect focal nonconvulsive seizures. A suppression ratio near 100 is compatible with burst suppression or with a disconnected lead. You still need the raw waveform.

Use processed EEG, if you use it at all, as a sedation adjunct while you already know the differential. When the question is NCSE, ischemia, or prognosis, order full-montage cEEG with video.

Post-anoxic EEG: highly malignant patterns and the 72-hour clock

Westhall and colleagues, reading TTM-trial EEGs after rewarming with ACNS terminology, grouped highly malignant patterns as (1) suppressed background (amplitude <10 µV throughout) without discharges, (2) suppressed background with continuous periodic discharges, and (3) burst-suppression (suppression >50% of the recording) with or without discharges. In that cohort, a highly malignant EEG after rewarming predicted poor outcome with very high specificity. Later ERC–ESICM guidance allows highly malignant suppression or burst-suppression after 24 hours only as one poor predictor among others, not as a stand-alone withdrawal trigger.

Burst-suppression with identical bursts (Hofmeijer) is a distinct anoxic signature: burst shapes are stereotyped, typically bilateral and synchronous, often appearing between about 12 and 36 hours after arrest. Identical bursts are not the usual appearance of anesthetic burst-suppression. Seeing them early is ominous. It is still not a 12-hour self-fulfilling prophecy. Sedation, temperature, shock, and metabolic failure must be stripped away. Independent teaching for this examination follows delayed, multimodal prognostication: wait until after rewarming, wait until major sedatives have had time to clear, and do not make a withdrawal decision on EEG alone before about 72 hours after ROSC (longer if midazolam, renal failure, or lingering hypothermia still confound the exam).

Malignant-but-not-highly-malignant features (unreactive background, some periodic patterns) are not specific enough in isolation. A recovering continuous, reactive background is the hopeful tracing — and it can appear after a grim first night. Record serial EEGs rather than tattooing the first one onto the family meeting.

Artifact versus seizure

If it does not evolve in frequency, amplitude, or field, be slow to call it a seizure. Common mimics:

True electrographic seizures recruit adjacent electrodes, evolve, and often have a post-ictal slowing. When paralyzed, you lose the motor clue — which is why video and a full montage still matter. If you are unsure, treat a high-risk evolving pattern and ask electrophysiology to review, rather than starting pentobarbital for chewing artifact or ignoring a 3 Hz left temporal march because “the BIS was 55.”

Exam traps

Stopping cEEG at 4 hours in a still-comatose ICH patient because “the first epoch was slow.” Calling 1 Hz LPDs “status” and inducing coma without a treatment trial or a clinical question. Using BIS to “rule out” delayed cerebral ischemia. Withdrawing on a suppressed EEG at hour 18 during 33 °C and a midazolam infusion. Treating electrode pop with fosphenytoin.