Hypoxic-Ischemic Brain Injury: After the Arrest, Before the Prognosis

A scan can show that the brain was injured and still not say when. The contest is the first three days after resuscitation, and the note that ended them.

14 minTobias B. Kulik, MD, FAAN

Most deaths after cardiac arrest that are attributed to brain injury do not occur at the moment of the injury. They follow a decision to withdraw life-sustaining treatment (stopping the ventilator and the drugs holding the circulation up), taken based on a predicted unfavorable neurological outcome, and the American Heart Association's 2025 post-arrest guideline opens its section on prognostication (the formal estimate of whether the patient will recover) by stating so.1 Whether a withdrawal decision breaks a causal chain is a legal question that the clinical facts do not settle. What they change is where the analytical work sits in a case of hypoxic-ischemic brain injury (HIBI, the damage the brain takes when its supply of oxygen or blood is interrupted). Imaging and examination usually establish that the brain was injured. What they do not establish is when the injury became irreversible, or whether the prediction that followed rested on the tests the guidelines ask for. Nor do they show whether the confounders (sedation, drug-induced paralysis, cooling, and metabolic upset, which alter an examination for reasons other than the brain injury) had cleared before anyone relied on the examination of a comatose patient (one whose eyes do not open and who follows no commands). The chart from the first days after return of spontaneous circulation (ROSC, the moment a pulse and a blood pressure come back) is where each of those questions was answered or left open.

Global Ischemia Does Not Image on the Stroke Clock

A cardiac arrest injures the whole brain at once, and that is what separates its imaging from stroke imaging. Loss of flow starves neurons of the energy that keeps water out of them, resulting in cytotoxic edema (swelling inside the cells themselves). On computed tomography (CT, the X-ray scan obtained first in almost every arrest), it shows as lost contrast between gray matter, where the nerve cell bodies sit, and white matter, the wiring between them, measured as the gray-white ratio (GWR). Many times the initial head CT will be read as "normal." On magnetic resonance imaging (MRI, the slower scan that resolves the injury itself), the edema shows as restricted diffusion on diffusion-weighted imaging (the sequence that finds water molecules which have stopped moving freely).1,2

None of that runs on the stroke clock, and assuming it does is the most consequential imaging error in this area. In a focal stroke the diffusion signal changes within minutes to hours,3 fast enough that an MRI pattern can stand in for a known time of onset. Nothing of the kind exists after cardiac arrest. The paper supplying that stroke figure reports that after global injury reduced diffusion may not be apparent in the initial hours: among its patients with a poor outcome, the two earliest scans, at 2 and 7 hours, were read as showing a normal cortex (the brain's outer layer of gray matter).3

CT runs on a different clock again. The 2025 European post-arrest guideline records that in most studies a reduced GWR was 100% specific for a poor outcome from as early as one hour after ROSC, and that the scan's sensitivity rises with time. Specificity is the share of patients who recovered in whom the test was correctly negative; sensitivity, the share who did badly in whom it was positive. The guideline's instruction follows: repeat the brain CT if the patient is unconscious at the time of prognostication, 72 to 96 hours after ROSC, and the first CT does not show signs of HIBI.2 A chart holding one early negative CT and no repeat has not excluded the injury. It stopped looking at the hour the test performs worst.

MRI moves later. Both 2025 guidelines fix the window for extensive restricted diffusion at 2 to 7 days after the arrest.1,2 An MRI in the first hours that shows none excludes nothing. The diffusion paper holds only three poor-outcome scans from the first two days, a limitation its authors name, so an opinion asserting what a 30-hour MRI means is asserting past the data.3 What a scan at a given hour can support in a given patient is work for an expert. What the record has to show is the hour.

The Temperature Evidence Does Not Reach Every Arrest

Targeted temperature management (TTM) is the active holding of core temperature (the body's internal temperature) at a set point with a cooling device, and the trial governing the modern conversation about it is TTM2, published in 2021. It randomized (assigned by chance) 1,900 adults comatose after an out-of-hospital cardiac arrest to hypothermia at 33°C (91.4°F) or to normothermia (a controlled normal temperature) with treatment of fever at or above 37.8°C (100.0°F), and at six months it did not detect a difference between the arms in death (50% against 48%) or in moderately severe disability or worse.4

How far that result travels is fixed by who was enrolled. An arrest inside a hospital was ineligible to enter the trial; the cause had to be presumed cardiac or unknown, and roughly three-quarters of the patients analyzed had a shockable rhythm (one a defibrillator can reset).4,5 Normothermia meant active fever treatment with a device, not the absence of temperature management, and both arms were held in mandatory deep sedation for the 40 hours of the intervention, so the temperature protocol itself delays the first examination that can be trusted.4,5

The guidelines reach further than the trial, and they have kept changing since it was reported. In January 2022, the European societies replaced the temperature section of their guideline with a document whose scope is adults comatose after in-hospital or out-of-hospital cardiac arrest, regardless of rhythm. It recommends continuous core-temperature monitoring and active prevention of fever (above 37.7°C) for at least 72 hours, and on the target it records "insufficient evidence to recommend for or against temperature control at 32-36 °C or early cooling after cardiac arrest."6 In October 2025, both the American and European post-arrest guidelines were replaced. The American document recommends, at its strongest grade, maintaining a temperature between 32°C and 37.5°C in patients unresponsive after ROSC, regardless of the arrest location or rhythm. The European document suggests preventing fever by targeting 37.5°C or below for 36 to 72 hours, and notes that whether any subpopulation benefits from 32°C to 34°C remains uncertain.1,2

The asymmetry inside the 2022 document is what a chart from January 2022 to October 2025 is read against, though an American clinician was never bound by a European document. Choosing 33°C breached no recommendation in it, because it declined to recommend for or against any target in that range. Failing to monitor core temperature, or letting a fever run inside 72 hours, breached a recommendation it does make. An opinion treating 33°C as impeachable in that window and an opinion treating it as required are both asserting past that document.6

Both 2025 documents are adult guidelines confined to cardiac arrest, and none of the evidence cited here reports separately how any imaging window, temperature target or prognostic threshold performs after airway loss under anesthesia, respiratory failure, submersion or overdose.1,2 Which document reaches a particular patient on a particular date, and whether its evidence included a patient of that kind, is where an expert's work begins.

Prognostication Runs on Two Clocks, From Two Different Zero Points

None of what follows concerns brain death, which is a separate determination that the patient has died rather than a prediction. The distinction has to be checked in every file, because a family, a complaint and a deposition will all use the words brain dead to mean a very poor prognosis.

The two current documents do not start the clock in the same place. The American guideline calls it reasonable to interpret the prognostic assessment no earlier than 72 hours after the patient is back to normal temperature and off sedatives.1 The European guideline places prognostication at 72 hours or later after ROSC.2 Normothermia and sedative clearance can sit a day or more past ROSC in a patient who was cooled and rewarmed, so a chart can satisfy one document and fail the other, and an opinion stating the rule with a single anchor has stated half of it. Both date from October 2025; what their predecessors required of an earlier chart is for an expert.

Neither document is law. Each is evidence of careful practice in its year, and where the two agree the agreement carries the weight. What both recommend most strongly is a rule no single test can satisfy: prognostication must be multimodal and must not rest on any one finding.1,2

Beyond the imaging, the battery of tests has four parts, and each carries its own window and its own weakness.

  • Pupillary and corneal reflexes. Whether the pupils constrict to light, and whether touching the cornea produces a blink. Absence of either reaches a 0% false-positive rate (the share of patients who recovered in whom the test was wrongly positive) only late, the pupillary reflex after 96 hours and the corneal after 72.2 Sedation can suppress both and a paralytic (a drug that paralyzes the muscles and erases every motor response the examination scores) abolishes the corneal reflex, so either finding needs the medication record beside it. Because some tests mature after the 72-hour mark, an indeterminate assessment at 72 hours is a normal result rather than a failure.
  • Somatosensory evoked potentials (SSEP). An electrical stimulus at the wrist recorded over the sensory cortex. Bilateral absence of the cortical response is among the most specific single signs available, 99.8% at 48 to 96 hours (roughly one wrong call in five hundred), and is found in fewer than half of the patients who do badly.2
  • Electroencephalography (EEG). Continuous scalp recording of the brain's electrical activity. The European guideline suggests using the highly malignant patterns (a nearly flat tracing, or one broken only by brief bursts) after 24 hours. Read locally at each center in a substudy of TTM2, those patterns were 93% specific: of every hundred patients who recovered, roughly seven were read as having a malignant pattern.2 Both 2025 documents have withdrawn one finding: absence of EEG reactivity (whether the tracing changes when the patient is stimulated) is no longer to be used to support a poor prognosis.1,2
  • Neuron-specific enolase (NSE). A blood test for an enzyme released by injured neurons. The European threshold is above 60 µg/L at 48 hours, at 72 hours, or at both, with serial samples suggested to show the trend and to catch hemolysis (rupture of red cells in the sample, which releases the same enzyme).2 A single value with no trend and no hemolysis check is a weaker datum than its number looks.

One sign carries a spread wide enough to decide a dispute by itself. Myoclonus (brief involuntary muscle jerks) is common after cardiac arrest, and the American guideline gives its strongest grade to prompt EEG in any patient who has it, because up to half prove to have seizure activity the bedside examination cannot detect.1 In a systematic review, recovery to independence was reported in 5.7% of patients with myoclonic status epilepticus (jerking with continuous or repeated seizure activity) and in 27 of the 33 patients with a recorded outcome (81.8%) who had early Lance-Adams syndrome (jerking triggered by movement, in a patient who is waking up).7

Not every case ends in a withdrawal. A large share involve a patient who lived, and the outcome scales in this literature, which sort patients into independent and not, say almost nothing about that patient.1 What follows global hypoxia-ischemia is a pattern of deficits (of memory, vision, attention, executive function and movement) rather than a number, and the chart describes it in the discharge summary, the rehabilitation medicine evaluation and the neuropsychological testing, usually performed months later.

The Decisions the Post-Arrest Chart Has to Document

The first week after resuscitation asks the chart to document eight things. What the chart omits from these eight is itself a finding.

  • The arrest interval, in minutes. Whether the arrest was witnessed and monitored, and the times to first chest compression, first shock and ROSC, which the code record, the defibrillator download and the telemetry strip carry. Time to defibrillation is measured in the in-hospital registry literature against a 2-minute mark drawn from expert guidelines.8 A witnessed anesthetic event is timed to the minute instead by the anesthesia record and the capnography trace (the continuous readout of exhaled carbon dioxide).
  • The temperature record, hour by hour. Target chosen, device used, whether core temperature was monitored continuously, and whether fever was prevented for the period the era's guideline required: at least 72 hours under the 2022 document, at least 36 under both 2025 documents.1,2,6 The temperature flowsheet (the running chart of recorded temperatures) is where that lives.
  • Sedation and paralysis, with a clearance time. The agents, the doses and the last administration, from the medication administration record, and the recorded judgment that they had cleared, in the physician progress note of the day the examination was relied upon.
  • The cause of the arrest, and whether it was treated. A brain-injury analysis that never establishes why the heart stopped has skipped the variable deciding whether the insult was single or ongoing. The answer sits in the cardiology and catheterization record rather than in the neurology note.
  • The prognostication timeline and its zero point. Which clock was run, from what moment, and what data existed at that moment. Where the assessment was indeterminate, the date of its repetition belongs in the same record.
  • Which tests were performed, when, and by whom read. Imaging measurement is not standardized and the European guideline directs centers to derive their own threshold values, so a radiology report naming a ratio without the local threshold has not documented the test.2 Where the first CT was negative and the patient still unconscious, the repeat either happened or did not.
  • The assessment for signs of potential recovery. Both 2025 guidelines now treat the prediction of a favorable outcome as a separate exercise,1,2 and the European document's signs include a motor response that withdraws from or localizes pain, a normal NSE and a continuous EEG background without discharges.2 In a chart written before October 2025 the same findings may sit in the record unlabeled.
  • The surrogate conversation, and what the patient had already said. The American guideline gives its strongest grade to early, regular and transparent discussion with surrogates (those who speak for a patient who cannot).1 The European guideline instructs that the discussion of withdrawal be kept separate from the assessment of prognosis.2 The documents recording the patient's own voice are separate from the family-meeting notes: an advance directive, a health care proxy designation or a POLST form (a portable medical order for life-sustaining treatment). A withdrawal consistent with a written directive is a different record from one resting on a surrogate's inference.

How These Charts Fail, and How Claims Overreach

Post-arrest charts fail in a short list of recurring ways.

  • Fever left unwatched once normothermia is declared. Active fever prevention is a positive recommendation in the 2022 document and in both 2025 guidelines, and the temperature flowsheet either shows it or does not.1,2,6
  • An examination relied upon before sedation and paralysis clear. This is the confounder both 2025 guidelines put first.1,2 It contaminates the bedside examination and the EEG; the SSEP, the blood markers and the imaging are the three the European guideline names as not materially affected, so an unclear sedation record does not leave a chart with no usable prognostic data.2
  • A prognosis declared before 72 hours. By the European guideline's account, the frequency of withdrawal for neurological reasons before 72 hours varied between 20% and 60% across nine United States hospitals.2 A range that wide is unlikely to be case mix alone: the timing of a withdrawal decision is not fixed by the patient's brain injury alone.
  • A single test treated as dispositive. A note resting on one absent reflex breaks the one rule both documents make at full strength.1,2
  • Myoclonus called a poor-prognosis sign without an EEG. One word covering a 5.7% entity and an 81.8% entity.7
  • A temperature standard borrowed from the wrong year. A 2019 case judged against the 2021 trial and a 2026 case judged against the superseded 2022 document are the same error with opposite signs.
  • A retired practice parameter cited as current authority. The American Academy of Neurology practice parameter on this question recommended single findings as reliable predictors and set NSE at above 33 µg/L. The Academy retired it on May 22, 2021, and the tag [RETIRED] now sits inside its indexed title.9 Retirement is not retraction, and the paper remains a legitimate source for what the standard was. Cited unmarked as current authority, it marks an opinion that stopped tracking the literature.
  • Withdrawal recommended to a surrogate never told the prediction was provisional. The uncertainty is documented in the guidelines themselves; whether it reached the family is documented, or not, in the chart.

These failures are not independent. They compound. Fever that goes unwatched produces a worse examination; a worse examination is read before sedation has cleared; an examination read early produces a prognosis declared early; a prognosis declared early produces a withdrawal recommendation; and the death that follows confirms the prediction and enters the literature as a correct call. That chain is the self-fulfilling prophecy, and it is a measured property of the evidence base: the European guideline reports that 64 of 73 studies (88%) were at risk of it in a 2013 systematic review.2

The error runs in both directions, and the European guideline says so outright: a prognosis that is inaccurately pessimistic can end the life of a patient who would otherwise have reached a good functional outcome, while one that is too conservative can leave a patient in a severely disabled state neither the patient nor the family would have chosen.2 A theory of the case that names only the first error has read half the document.

The chart supplies answers of its own, and a claim that ignores them overreaches. A prognosis made on guideline timing, on concordant tests (independent tests pointing the same way), is very hard to impeach: among 337 patients still comatose 72 hours after out-of-hospital arrest and followed prospectively, the European algorithm predicted a poor outcome in 175 without a false positive.10 An incomplete battery is a documentation failure, not evidence of a good outcome, and an early negative scan proves little for either side.2,3

How Far This Evidence Travels

Two further limits govern what this literature will carry.

The battery predicts poor outcome at something close to perfect specificity, and in only part of the population: the same cohort left the other 162 of its 337 patients, roughly half, indeterminate.10 Neither guideline converts an unresolved assessment into a poor prognosis, and what the chart has to show for that patient is a date and a repetition rather than a conclusion.1,2 An opinion that engages the indeterminate half survives a cross-examination that an opinion eliding it does not.

The second limit is the distance between a population and a patient. Across 6,789 patients with an in-hospital arrest from a ventricular arrhythmia (an abnormal rhythm arising in the heart's pumping chambers), survival to discharge was 39.3% when defibrillation was delivered inside 2 minutes and 22.2% when it was not, with a graded relationship across each further minute of delay.8 Both figures sit well under half: a patient defibrillated inside 2 minutes was still more likely to die than to live, so a but-for argument built on these figures asks a court to accept that a specific death would probably have been avoided when the counterfactual survival the registry supplies is 39.3%. That is the problem the loss-of-chance doctrine exists to address, and how a given state treats a lost chance is a question of law that nothing in the medicine decides. What the medicine can say is which number is which: the two rates are population figures, the difference between them is an association, and the per-minute gradient is what makes a specific interval arguable at all.

Reading the First Week After an Arrest

A defensible review of a post-arrest brain-injury case, plaintiff or defense, works through a short list of questions addressed to named documents.

  • Was the arrest in-hospital or out-of-hospital, what was the initial rhythm, and how many minutes ran from collapse to first compression, to first shock and to ROSC? (Code record, defibrillator download and telemetry strip; for a perioperative event, the anesthesia record.)
  • Does the chart show a documented temperature-control strategy, continuous core-temperature monitoring, and any temperature above the threshold the era's guideline set? (Temperature flowsheet; intensive care order set.)
  • What were the sedative and paralytic agents and doses, when was the last one given, and is a clearance judgment recorded before the examination that was relied upon? (Medication administration record; physician progress note of that day.)
  • Which prognostic tests were performed, on what days counted from which zero point, by whom read, and against what locally derived threshold? (Radiology report; EEG and neurophysiology reports; laboratory record for serial NSE.)
  • Did the prognostic note rest on a single finding, was any assessment for signs of potential recovery recorded, and if the assessment was indeterminate, when was it repeated? (Prognostication or neurology consult note, and the serial notes after it.)
  • Does any note record whether the myoclonus was continuous or intermittent, whether the patient was comatose or waking, and whether an EEG showed matching seizure activity? (Neurology consult note; EEG report.)
  • Was the surrogate told brain death or a poor prognosis, when, against what data, and had the patient already spoken for themselves? (Family-meeting and goals-of-care notes; advance directive, health care proxy or POLST; where the patient survived, the discharge summary with the rehabilitation and neuropsychology evaluations; where the patient died, the autopsy and any later neuropathology addendum.)

Placed in order against the clock, those questions usually settle which way a case runs. Some that look strong on the complaint dissolve against a chart with a continuous temperature record, a documented sedation clearance, four tests performed after the later of the two zero points, and two conversations in which the family was told the prediction was provisional. Others that look strong on the defense collapse against a chart in which one early examination, performed while a paralytic was still infusing, produced a prognosis on day two and a withdrawal conversation on day three.

The scan is a picture of tissue. The prognostication note is a record of what was known, and when. The case is decided in the second.

References

Footnotes

  1. Hirsch KG, Amorim E, Coppler PJ, et al. Part 11: Post-Cardiac Arrest Care: 2025 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2025;152(16_suppl_2):S673–S718. doi:10.1161/CIR.0000000000001375 ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15 ↩16 ↩17

  2. Nolan JP, Sandroni C, Cariou A, et al. European Resuscitation Council and European Society of Intensive Care Medicine Guidelines 2025 Post-Resuscitation Care. Resuscitation. 2025;215 Suppl 1:110809. doi:10.1016/j.resuscitation.2025.110809 ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15 ↩16 ↩17 ↩18 ↩19 ↩20 ↩21 ↩22 ↩23 ↩24 ↩25 ↩26

  3. Mlynash M, Campbell DM, Leproust EM, et al. Temporal and spatial profile of brain diffusion-weighted MRI after cardiac arrest. Stroke. 2010;41(8):1665–1672. doi:10.1161/STROKEAHA.110.582452 ↩ ↩2 ↩3 ↩4

  4. Dankiewicz J, Cronberg T, Lilja G, et al. Hypothermia versus Normothermia after Out-of-Hospital Cardiac Arrest. N Engl J Med. 2021;384(24):2283–2294. doi:10.1056/NEJMoa2100591 ↩ ↩2 ↩3

  5. Ceric A, Dankiewicz J, Cronberg T, et al. Sedation and analgesia in post-cardiac arrest care: a post hoc analysis of the TTM2 trial. Crit Care. 2025;29(1):247. doi:10.1186/s13054-025-05461-0 ↩ ↩2

  6. Sandroni C, Nolan JP, Andersen LW, et al. ERC-ESICM guidelines on temperature control after cardiac arrest in adults. Intensive Care Med. 2022;48(3):261–269. doi:10.1007/s00134-022-06620-5 ↩ ↩2 ↩3 ↩4

  7. De Stefano P, Leitinger M, Misirocchi F, Quintard H, Degano G, Trinka E. Myoclonus After Cardiac Arrest: Need for Standardization-A Systematic Review and Research Proposal on Terminology. Crit Care Med. 2025;53(2):e410–e423. doi:10.1097/CCM.0000000000006521 ↩ ↩2

  8. Chan PS, Krumholz HM, Nichol G, Nallamothu BK. Delayed time to defibrillation after in-hospital cardiac arrest. N Engl J Med. 2008;358(1):9–17. doi:10.1056/NEJMoa0706467 ↩ ↩2

  9. Wijdicks EFM, Hijdra A, Young GB, Bassetti CL, Wiebe S. Practice parameter: prediction of outcome in comatose survivors after cardiopulmonary resuscitation (an evidence-based review) [RETIRED]: report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2006;67(2):203–210. doi:10.1212/01.wnl.0000227183.21314.cd ↩

  10. Bougouin W, Lascarrou JB, Chelly J, et al. Performance of the ERC/ESICM-recommendations for neuroprognostication after cardiac arrest: Insights from a prospective multicenter cohort. Resuscitation. 2024;202:110362. doi:10.1016/j.resuscitation.2024.110362 ↩ ↩2