Hypoxic-Ischemic Encephalopathy

Hypoxic-ischemic encephalopathy after cardiac arrest is the setting where EEG carries its heaviest consequences, because its findings feed directly into decisions about continuation or withdrawal of life-sustaining therapy. Global cerebral anoxia injures the most metabolically demanding neurons first - pyramidal cells of cortical layers 3 and 5, the hippocampal CA1 sector, cerebellar Purkinje cells, and selected thalamic and basal ganglia populations - and the EEG mirrors the depth and reversibility of that injury as a trajectory over hours to days rather than as a fixed snapshot. The interpreter must therefore think less about a single recording and more about where on the recovery curve a given pattern falls, while holding firmly to the epistemic humility that the EEG is one input into a multimodal estimate and never, by itself, a verdict. No contemporary guideline endorses withdrawal of care on the basis of any EEG pattern alone, and the reasons for that restraint are as much about the structure of the evidence as about ethics.

The vocabulary of a failing background

Precise ACNS definitions matter here because prognostic categories are built on them and small definitional slips change the conclusion. A suppressed background is activity below 10 µV for essentially the entire record (the continuity convention is that more than 99 percent attenuation or suppression is scored as suppressed, and the highly malignant prognostic category requires suppression below 10 µV throughout). A low-voltage background is activity below 20 µV for most of the record but not meeting the suppression threshold. Attenuation, by contrast, denotes periods of lower voltage that are at least 10 µV but less than half the amplitude of the surrounding higher-voltage background; the distinction between attenuation and true suppression is not pedantic, because the malignant categories specify suppression below 10 µV. Burst-suppression (or burst-attenuation) is a background in which 50 to 99 percent of the record consists of suppression or attenuation alternating with bursts of higher-amplitude activity, where a burst is defined as a period lasting at least 0.5 second and containing at least four phases. These are not merely descriptive labels: a persistently suppressed background beyond the first day after arrest, in the absence of sedation and hypothermia, is among the most ominous findings in all of clinical neurophysiology.

Within burst-suppression, two modifiers carry decisive prognostic weight. Identical bursts are bursts whose first approximately 0.5 second is highly stereotyped and nearly superimposable from one burst to the next across a derivation; their stereotypy reflects a profoundly deafferented, pacemaker-driven cortex that has lost the moment-to-moment variability of a living network, and they are considered a highly malignant feature after anoxia. Burst-suppression with highly epileptiform bursts describes bursts laden with spikes, polyspikes, or rhythmic discharges; these sit on the ictal-interictal continuum and raise the question of post-anoxic status epilepticus. The mechanistic intuition is that anoxic cortex, stripped of normal inhibitory tone and thalamic modulation, oscillates between near-silence and hypersynchronous discharge, and that the more stereotyped and epileptiform that oscillation, the worse the underlying substrate.

The neurophysiology of burst-suppression

Burst-suppression is not unique to anoxia; it is a deep-coma signature shared by high-dose anesthesia, hypothermia, and several encephalopathies, which is precisely why understanding its mechanism is the key to interpreting it. Two complementary hypotheses dominate. The cortical hyperexcitability view, grounded in older work showing that the suppressed cortex remains paradoxically responsive and that bursts can be evoked by stimulation, frames the bursts as the discharge of a disinhibited network poised between silence and synchrony. The metabolic view, formalized in computational models, proposes that when cerebral metabolic rate falls far enough, intermittent depletion of adenosine triphosphate gates ATP-sensitive potassium channels open, hyperpolarizing neurons into suppression until metabolic recovery permits another burst; in this model the alternation between burst and suppression is a relaxation oscillation between metabolic supply and demand. These views are not mutually exclusive, and the modern synthesis treats burst-suppression as the behavior of a network in which both depressed metabolism and altered excitability conspire to produce the on-off dynamic.

This mechanistic understanding directly explains the prognostic rules. If burst-suppression can be produced reversibly by anesthetic depth or by hypothermia, then burst-suppression in those contexts says nothing about the integrity of the underlying cortex and cannot be read prognostically. If, however, the same pattern persists at normothermia after sedation has cleared, it indicates that the cortex cannot sustain continuous activity on its own metabolic and synaptic resources, which is a far graver statement. Identical bursts add a further layer: the loss of burst-to-burst variability implies that the cortex has been reduced to a stereotyped pacemaker, the antithesis of the variable, reactive activity that signals a recoverable brain. The neurophysiology thus converts an abstract prognostic category into a comprehensible claim about what the tissue can and cannot do.

Highly malignant, malignant, and benign patterns

Contemporary post-arrest prognostication, shaped by the targeted-temperature-management neurophysiology substudies and codified in the European Resuscitation Council and European Society of Intensive Care Medicine guidelines (the 2021 framework, reaffirmed and refined in the 2025 update), collapses the EEG into three tiers. Highly malignant patterns - a suppressed background with or without continuous periodic discharges, and burst-suppression (with or without identical bursts and with or without superimposed discharges) - are defined to carry very high specificity for poor outcome, approaching but not reaching 100 percent in the best series, when recorded at or beyond 24 hours off sedation. Even so, current ERC and ESICM guidance does not endorse a highly malignant EEG as a stand-alone trigger for a poor-prognosis conclusion; it should be combined with at least one other concordant predictor, so that no fewer than two independent findings agree. Malignant patterns form an intermediate category that includes abundant periodic or rhythmic discharges, electrographic seizures, a discontinuous but not suppressed background, low-voltage activity, a reversed anterior-posterior gradient, and - crucially - the absence of background reactivity. Benign patterns - a continuous, reactive background that re-emerges early - strongly predict good neurologic recovery and are, in several series, more reliable for predicting a good outcome than the malignant patterns are for predicting a poor one. That asymmetry is clinically useful: a reassuring EEG can be trusted more confidently than an alarming one.

Timing is therefore inseparable from interpretation. In the first hours after return of spontaneous circulation, and during targeted temperature management, suppression and discontinuity are expected and minimally informative, because hypothermia and sedatives both suppress the cortex through exactly the mechanisms described above. The prognostically meaningful recordings are those obtained after rewarming and after sedation has been allowed to clear, conventionally at or beyond 24 to 72 hours and with the most confident statements reserved for 72 hours or later. A burst-suppression pattern at 6 hours on a midazolam infusion at 33 degrees Celsius is nearly uninterpretable for outcome; the identical pattern at 72 hours, normothermic and sedation-free, is gravely meaningful. A common and dangerous error is to treat an early malignant-appearing record as if its timing did not matter.

Post-anoxic status epilepticus and rhythmic patterns

Rhythmic and periodic patterns after arrest occupy an uncomfortable middle ground. Myoclonic status epilepticus, with clinically obvious jerks time-locked to generalized bursts, and electrographic status epilepticus arising from a burst-suppression background were historically equated with futility. That blanket pessimism has softened in the modern literature: a minority of patients with post-anoxic status epilepticus, particularly those with a continuous and reactive background between discharges, preserved brainstem reflexes, and an early electrophysiologic substrate, can recover with aggressive and sustained treatment. The pattern's prognosis is thus modified by the background it sits upon - status on a suppressed, unreactive, identical-burst background remains dismal, whereas status on a continuous, reactive background warrants a genuine treatment trial. This is precisely the ictal-interictal-continuum reasoning of ACNS terminology applied to its highest-stakes setting, and it is the strongest argument against reflexive withdrawal when seizures appear after arrest.

The defensible clinical posture is multimodality with explicit time. A robust poor-prognosis conclusion requires the convergence of independent modalities, measured no earlier than 72 hours after arrest in a normothermic, sedation-free patient. Those modalities are a highly malignant EEG; bilaterally absent cortical N20 somatosensory evoked potentials, with the emerging refinement that a low N20 amplitude carries graded information and a clearly preserved N20 amplitude favors recovery; a markedly elevated serum neuron-specific enolase, with values rising over 24 to 72 hours and high absolute values at 48 to 72 hours, around or above the often-cited 60 microgram-per-liter range, carrying more weight than any single value; diffuse anoxic injury on CT or MRI; and an examination lacking pupillary and corneal reflexes, including absent or unfavorable quantitative pupillometry. Each test has imperfect specificity in isolation; their agreement is what supports a confident statement. Where modalities disagree, or where confounders such as residual sedation, hypothermia, or organ failure persist, the correct action is continued observation, not a forced verdict.

Two technical caveats protect against error. First, reactivity testing must be performed and documented using a standardized graded stimulus protocol interpreted by an experienced reader, because absent reactivity is one of the malignant criteria and is easily missed if it is never tested; an untested record cannot claim that reactivity is absent. Second, the EEG-EMG distinction matters in the paralyzed-then-awakening patient, where shivering, fasciculation, post-rewarming tremor, and the chewing artifact of returning tone generate myogenic activity that can be mistaken for fast cortical activity or seizure; recognizing that the rhythm exceeds physiologic cortical frequencies, that it is maximal over muscle rather than following a cortical field, and that it disappears with neuromuscular blockade prevents both false reassurance and false alarm. A related trap is mistaking the regular jerks of post-anoxic myoclonus, with their EMG correlate, for an evolving electrographic seizure when the cortical correlate is in fact a stereotyped burst.

It is worth stating plainly how the 2026 consensus differs from the pessimism of two decades ago. The field has moved from a search for a single infallible test toward an explicitly probabilistic, multimodal, time-anchored process that treats every marker as a likelihood ratio rather than a switch. It has internalized that benign EEG findings are powerful predictors of good recovery and that they justify continued aggressive care. It has named the self-fulfilling prophecy as a structural threat to the evidence base and built guideline safeguards against it. And it has accepted that some post-anoxic status epilepticus is survivable when the background substrate is favorable. The practical consequence for the interpreter is a posture of calibrated restraint: describe the pattern precisely, anchor it to its timing and confounders, place it in the multimodal picture, and resist the pull toward a premature verdict that the structure of the evidence cannot support.