CONTENTS
- Rapid Reference 🚀
- Introduction
- Early brain death is (still) death
- Prognostic value of historical information
- Confounding factors & hypothermia
- Timing & serial evaluation
- Prognostic tests
- Neurological examination
- EEG
- Myoclonus
- SSEPs (Somatosensory evoked potentials)
- CT scan
- MRI
- Neuron-specific enolase
- Multimodal prognostication
- Podcast
- Questions & discussion
- Pitfalls
- PDF of this chapter (or create customized PDF)
timing of tests
neuroprognostication approach
Patients often sustain severe neurologic injury during cardiac arrest. In such cases, the decision to provide ongoing support often hinges on whether the patient might regain meaningful consciousness. Prognostication is extraordinarily important in these cases. Excessively pessimistic prognostication could lead to a premature withdrawal of care in a patient with the potential to recover. However, overly optimistic prognostication may lead to ongoing support for days or weeks in a hopeless situation.
The TTM2 trial recently demonstrated that hypothermia at 33C following cardiac arrest is unhelpful. The study also showed that maintaining patients at a target temperature of 37.5 degrees is safe. Maintaining patients in a normothermic temperature range (e.g., 36-37.5 C) while minimizing paralytics and sedatives facilitates more accurate neuroprognostication. Much of the literature obtained over the past 15 years has been obfuscated by the use of hypothermia, paralytics, and high doses of sedatives. As hypothermia falls out of favor, we may notice that diagnostic tests start working better than previously.
- Rarely, patients may rapidly develop cerebral edema that progresses to herniation and brain death. If the formal criteria for brain death are satisfied, then the patient is dead. There is no need to wait further (e.g., 72 hours) if the patient meets all criteria for brain death.
- The declaration of brain death should be extremely cautious in patients among whom the cause of cardiac arrest is unclear. For example, if drug overdose is a possibility, then residual intoxication could confound the diagnosis of brain death. As a rough rule, early brain death declaration should be pursued only if one of the following is present:
- (1) Unequivocal CT scan imaging of catastrophic structural brain injury (e.g., tonsillar herniation with diffuse cerebral edema).
- (2) Flow scan showing cerebral circulatory arrest.
- See the chapter on brain death for further information.
- Details regarding the cardiac arrest (e.g., down time) are often unreliable. Furthermore, it is often unclear what the rhythm and blood pressure truly were (e.g., the “pulseless” patient could be truly asystolic or they could have a non-palpable pulse).
- One possible exception is an arrest which was clearly asphyxial in mechanism (e.g., choking, airway loss during intubation, or asthma/COPD exacerbation which progressed to the point of cardiac arrest). Respiratory arrest will first cause hypoxemia and only later on progress to cardiac arrest. By the time the pulse is lost, the patient has been hypoxemic for a while. Thus, an asphyxial mechanism of cardiac arrest is a poor sign.
primary offender: sedatives & analgesics
- Perhaps the most important aspect of prognostication is avoidance of any sedating medications which can accumulate. Midazolam infusion is the worst offender, but even drugs like lorazepam and fentanyl may accumulate (especially when provided as a continuous infusion; more on fentanyl infusions here). In a patient with severe neurologic injury, it might not take much medication to delay awakening.
- Even propofol has a half-life which is longer than usually recognized (40 minutes initially, with some increase over time when provided as a continuous infusion). Thus, propofol should ideally be held for a considerable duration before neurological evaluation.
- 💡 For comatose or stuporous patients who are unresponsive to pain, the goal of sedation is to prevent dangerous ventilator dyssynchrony or shivering. The goal is not to prevent all movement nor to allow the patient to appear outwardly peaceful. The practice of providing high-dose opioid/sedative infusions to comatose patients is dubious, as this may make patients look better but is unlikely to truly help them (i.e., we're treating our discomfort, not the patient's).
normothermia facilitates early awakening & more accurate neuroprognostication
- TTM at 36-37.5C may generally be performed without the use of opioids or benzodiazepines. Shivering can be managed by a multimodal strategy as described here. Sedatives and analgesics should ideally be limited to agents with a shorter half-lives (e.g., propofol, dexmedetomidine, and pain-dose ketamine infusions). Liberalizing the target temperature to 37.5 may further limit shivering.
- If TTM can be performed at 36-37.5C with strict avoidance of opioids or benzodiazepines, then patients may awake earlier and neuroprognostication may be accurate at earlier time points (which would be similar to clinical practice before hypothermia became popular). The same general strategy for neuroprognostication should be used, but many tests will perform better in the absence of confounding medications.
additional confounding factors
- Prognostication can be confounded by any cause of altered mental status. For example, common confounders are:
- Untreated electrolyte abnormalities (e.g., hypernatremia, hypoglycemia).
- Untreated hepatic encephalopathy.
- Uremic encephalopathy.
- Receipt of sedative or analgesic medications.
- Multiorgan failure, especially due to septic shock.
- Unrecognized seizures (leading to postictal state and inability to wake up).
- Coexisting traumatic brain injury.
- Confounding factors should be aggressively sought out and corrected.
- In the presence of confounders which cannot be immediately corrected, a greater amount of caution is required with neuroprognostication.
progression of neurological function over time
- In the absence of confounding factors, neurologic function tends to track out along various curves as shown above. As time progresses, it becomes increasingly clear where the patient will end up neurologically.
- Prognostication depends on integrating information about neurologic function over time.
- More rapid improvement carries a better prognosis (e.g., regaining corneal reflexes within a few hours is a favorable sign).
- Sluggish improvement carries a worse prognosis (e.g., regaining corneal reflexes after 80 hours is a poor sign).
- 💡 Any piece of information about neurologic function (e.g., EEG, clinical exam) is meaningful only when placed within the context of time.
critical time windows for different investigations
- Prognostic tests often have specific time windows in which they reveal the most information. For example:
- Neurologic examination at 72 hours is especially important.
- EEG between 24-48 hours may be most informative.
- CT scan is optimally obtained within 72 hours.
- It is important to make efforts to obtain accurate information within these time windows. If no organized effort is made to neuroprognosticate the patient over the first several days of their ICU stay, this may lead to a murky situation later on.
time course & goals of care
- Often, 72 hours will be needed for definitive prognostication. However, malignant EEG patterns may alert the clinician after 24 hours that the patient is unlikely to recover.
- It is useful to have sequential meetings with the family throughout the prognostication process (e.g., an initial meeting after 24 hours, then another meeting after 72 hours). This may help the family process information over time.
- Aside from neurodiagnostic testing, other critical pieces of information are the patient's baseline level of function and the level of disability the patient might be satisfied living with.
- The key question to be addressed in neuroprognostication is whether the patient will be able to improve to a level of function which they would find satisfactory. This involves an intersection of neuroprognostication with the patient/family's wishes. Other active medical problems may also factor into this assessment.
- Prognostic tests can be roughly grouped in terms of:
- Structural tests – radiologic tests looking at brain structure.
- Functional tests – measurements of brain function.
- Functional tests are generally superior, although they may be more susceptible to confounding (e.g., due to medications).
- Structural tests do have some advantages:
- They are not confounded by medications or metabolic abnormalities.
- They can be objectively assessed by independent observers (neuroradiologists).
- They have the capacity to reveal alternative diagnoses which may not have been previously suspected (e.g., intracranial hemorrhage, other primary neurologic disorders).
- A common mistake made during neuroprognostication is overemphasis on imaging tests, with inadequate attention to EEG and neurological examination (which are probably the two most powerful prognostic tests). MRI images are pretty, but they are often less informative than serial evaluation of the neurologic examination and EEG over time.
motor response to pain
- Absence of motor response to pain is worrisome between 24-72 hours, but nonspecific.
- After 72 hours, inability to localize or withdraw from pain is worrisome for poor neurological outcome. However, recent data has cast some shade on the use of motor responses – specifically, a poor motor response at 72 hours may incorrectly suggest a poor outcome in patients who may in fact do well.(32915254)
pupillary reflexes
- Pupillary reflexes may be particularly useful, as they seem to be fairly resistant to the effects of sedatives or paralytics (since the pupils contain smooth muscle, they are not affected by paralytics).
- Between 0-24 hours, lack of a pupillary response is nonspecific. However, emergence of pupillary responses may be an optimistic sign, especially if this occurs rapidly following cardiac arrest.
- Between 24-72 hours, absence of a pupillary response is very worrisome yet nonspecific.
- After 72 hours, absence of any pupillary response is ~20% sensitive and ~99% specific for poor neurological outcome.
- After 96 hours, lack of any pupillary response bilaterally may approach 100% specificity for poor neurological outcome.(33765189)
- The specificity of absent pupillary reflexes may be reduced among patients with baseline small pupils, where even an intact light reflex may cause only tiny changes in pupillary size.
- Automated pupillometry has superior performance, but is not yet widely available. This technology may allow for the pupillary reflex to have a high predictive value early in the patient's course.
corneal reflexes
- After 72 hours, absence of any corneal reflex is ~30% sensitive and ~95% specific for poor neurologic outcome.
- After 96 hours, absence of corneal reflexes may approach 100% specificity for a poor neurologic outcome.
- Some technical details regarding the corneal reflex:
- (1) It is essential to stimulate the cornea itself. One study showed that a quarter of physicians test this reflex by stimulating the temporal conjunctiva, rather than the actual cornea.(32915254)
- (2) For the purposes of prognostication, it's probably desirable to physically touch the cornea with gauze (rather than dropping saline flushes on the cornea).
absence of gag or cough reflex
- Absence of a gag or cough reflex (e.g., following endotracheal tube suctioning) after >48 hours was 100% specific for poor neurologic outcome in a few studies. However, further replication is required to determine the precise specificity of this finding.(32915254)
optimal timing of EEG
- EEG patterns evolve over time following anoxic brain injury, with patients often transitioning from more malignant-appearing EEGs to more benign-appearing EEGs.
- The EEG pattern may be best interpreted within the context of when it occurs, for example:
- Benign patterns: The sooner the patient is able to achieve a favorable EEG pattern, the more likely they are to recover. Patients with poor prognosis may eventually develop benign patterns several days after arrest, so a benign pattern is less meaningful at later time points.
- Malignant patterns: The longer the patient remains stuck in a malignant pattern, the worse their prognosis is. Malignant patterns in the first 12 hours may be nonspecific. However, remaining in a malignant pattern for >24 hours carries a very poor prognosis.
- Therefore, the timing of various EEG patterns is critical. The best single time point to evaluate EEG seems to be ~24 hours after arrest. Beyond 24 hours, EEGs will often improve over time, thereby compromising their sensitivity to detect patients with poor neurological prognosis.(32915254)
obtaining high-quality EEG data
- Medications (e.g., propofol) can affect the EEG. These should be held for several half-lives, to obtain an optimal recording.
- Motor artefact (e.g., myoclonus, shivering, respiration) can obscure the EEG. For patients who are clinically comatose, chemical paralysis (typically one dose of 10 mg vecuronium) is often required to obtain a high-quality EEG.
- 🛑 Before paralyzing the patient, a neurologic examination should be performed. Paralysis without sedation is acceptable only in a comatose patient.
- Paralysis is similarly helpful to obtain accurate somatosensory evoked potentials.(SSEPs, more on this below).
The EEG may be understood as a combination of the background pattern plus any superimposed activity. The background pattern is the single most important aspect of the EEG, so this will be discussed first.
EEG background pattern
- This is the single most important aspect for EEG prognostication. There are four background patterns, listed in order from benign to malignant:
- (#1) Continuous background activity
- If a continuous background can be achieved within <12 hours, this is a particularly optimistic sign. (29525975)
- Among patients with good neurological outcome, 80% will achieve a continuous baseline by 30 hours.(32562686) Nearly all patients with a favorable outcome will achieve a continuous baseline within <60 hours.(32915254)
- (#2) Discontinuous background
- This refers to periods of suppression (flat EEG) that occur between 10-40% of the time.
- This pattern is nonspecific prognostically.
- (#3) Burst suppression
- This is defined as bursts of EEG activity present on a flat background, with no EEG activity for the majority of the tracing (i.e., periods of suppression occur for >50% of the time).
- A burst suppression pattern occurring >24 hours after arrest indicated a poor prognosis in 99% of patients within most studies.(32915254)
- Burst suppression with identical bursts is highly malignant. This is defined as the first 0.5 second of each burst being identical, or the presence of stereotyped clusters of bursts appearing repeatedly.
- (#4) Suppression (>99% activity has a <10 uV amplitude)
- A suppressed EEG occurring >24 hours after ROSC is one of the worst possible EEG findings. It is a reliable sign of poor prognosis, in the absence of confounders (with nearly all studies showing 100% specificity for a poor neurological outcome).
- A “low voltage” pattern is similar to suppression, albeit with an amplitude <20 uV. This pattern also carries a poor prognosis, but ~5% of patients with this pattern may have a good outcome.(32915254)
periodic discharges & generalized periodic discharges (GPDs)
- A periodic discharge is a waveform that occurs repeatedly, with a quantifiable interval between waveforms.
- Periodic discharges are a sign of poor prognosis, but this is less important than the background:
- Generalized Periodic Discharges (GPDs) on a suppressed background are a highly malignant pattern, which invariably predicts poor prognosis (figure below).(32915254)
- Periodic discharges on a continuous background don't necessarily indicate a poor outcome.
sporadic epileptiform discharges
- This describes sharp waves or spikes resembling those seen in patients with epilepsy, but lacking the regularity of a periodic pattern.
- Sporadic epileptiform discharges may be rare (<1 per hour) or abundant (>1 every ten seconds).
- This is a concerning sign, but it lacks specificity.
seizures
- Unequivocal seizures are defined as generalized rhythmic spike-and-wave discharges with a frequency >3 Hz or clearly evolving discharges of any type >4 Hz. Unequivocal seizures during the first 72 hours after ROSC have a low sensitivity, but high specificity for poor outcome (>>95%).
- “Electrographic status epilepticus” has various definitions, limiting its use in prognostication. As with periodic discharges, the background may be critical to interpretation:(32915254)
- Status epilepticus occurring within the context of a discontinuous or burst suppression background predicts a poor outcome.
- Status epilepticus occurring within the context of a benign background (continuous or nearly continuous) can be consistent with a favorable outcome.
alpha coma pattern
- This refers to a pattern of frontal predominant, monotonous, unreactive activity in the low alpha range.
- An alpha coma pattern is strongly linked to a poor outcome.(32562686)
EEG reactivity
- EEG reactivity is limited by poor standardization and low interobserver agreement.
generalized status myoclonus
- Generalized status myoclonus may be defined as follows:
- Onset within 24-48 hours of arrest (usually <24 hours).
- Generalized myoclonus involves four limbs and the face. This should be synchronized, repetitive, symmetric, and unequivocal (not just a little jerking of the hands or face).
- Persists for >30 minutes.
- Associated with clinical coma.
- EEG shows either seizures (i.e., status epilepticus) or a malignant pattern (e.g., burst suppression pattern). In order to obtain a good EEG, paralysis will be needed to remove muscle artefact.
- 🛑 If the EEG shows a continuous or reactive background, or the presence of spike-wave discharges synchronized with myoclonic jerks, that indicates a potential for good neurologic outcome. These EEG patterns are consistent with Lance-Adams syndrome, as described below.(33765189)
- When all of the above criteria are met, this is almost invariably associated with poor neurological outcome (specificity of 99-100%).(33765189, 27017468)
Lance-Adams syndrome
- This syndrome is defined by the following features:
- Myoclonus emerges later after the occurrence of cardiac arrest, often after several days.
- Patients may not be comatose.
- Myoclonus occurs while the patient is attempting to move (intention myoclonus).
- Lance-Adams syndrome does not carry the same negative prognostic implications as status myoclonus. About half of patients with Lance-Adams syndrome may evolve towards a slow neurological recovery.(32915254)
- SSEPs involve stimulating the hand with electricity and tracing the nerve impulse from the periphery to the sensory cortex of the brain. Failure of the nerve impulse to reach the cortex indicates cortical damage (absent N20 wave).
- Bilateral absence of a cortical signal more than 24 hours after arrest strongly predicts a poor outcome (e.g., with a sensitivity of ~45% and a specificity of >99%).(32562686)
- Intact SSEPs are nonspecific, as such patients often still fare poorly. In this situation, the amplitude of the N20 waves may be used to provide additional information. If the amplitude of the N20 waves is >2.5 uV, this more strongly implies a favorable outcome. Alternatively, if the amplitude is <0.62 uV, this suggests an unfavorable outcome – even despite the presence of an N20 wave.(32562686)
- An absent N20 wave on one side combined with a low-voltage N20 wave on the contralateral side appears to predict a poor neurological outcome.(32915254)
- SSEPs can be depressed by a barbiturate coma, but are preserved with other sedatives.(33765189) SSEPs are not affected by hypothermia.(30739213)
- The specificity of absent N20 waves approaches 100% with attention to technical details.(32562686)
- 🛑 Noise from muscle activity may be problematic, but this can be eliminated using paralytics. Thus, paralysis should be considered whenever possible.(33765189)
- SSEPs probably lack utility in patients with a benign EEG pattern. Such patients appear to reliably have normal SSEPs.(31394155)
signs of anoxic brain injury on CT scan
- CT findings in anoxic brain injury may include:
- Reduction of the grey matter/white matter interface due to cytotoxic edema, which reduces the density of the grey matter. Similarly, this may cause hypoattenuation of the basal ganglia.(28865528)
- Diffuse edema, revealed by the presence of effacement of cortical sulci, effacement of the cisterns, and reduced ventricle size.
- “White cerebellum” sign – diffuse edema of the cerebral hemispheres with sparing of the posterior fossa structures results in apparent hyperdensity of the cerebellum and brainstem.(28865528)
- “Pseudo-subarachnoid hemorrhage” pattern may result from increased density in the basal cisterns due to dilation of superficial venous structures as a result of increased intracranial pressure (figure below).(28865528)
- In extreme cases, herniation may occur.
- CT scan can be used to estimate the intracranial pressure based on optic nerve sheath diameter (similar to ultrasonography). Some evidence suggests that an average optic nerve sheath diameter >6.5 mm would suggest a poor prognosis.(32915254)
use of CT scan for prognostication
- CT may be useful for neuroprognostication.(33765189) Edema may become more prominent over ~3-5 days following cardiac arrest.
- CT scan is not very sensitive for anoxic brain injury. However, when anoxic injury is unequivocally seen on CT scan, this is a grave sign.
- 🛑 Diagnosis of ischemia on CT scan can be difficult and subjective in some cases. Be very careful about overcalling this. An official CT interpretation by a board-certified attending neuroradiologist is preferred, as an unbiased evaluation of imaging data.
- In particular, young women may normally have relatively large brains. Combined with low lying cerebellar tonsils (“benign tonsillar ectopia”), this may easily be mistaken as a combination of cerebral edema plus tonsillar herniation.
findings due to anoxic brain injury on MRI
- Cytotoxic edema due to anoxia has a similar MRI pattern compared to other forms of cytotoxic edema (e.g., due to ischemic stroke):
- Hyperintensity on diffusion weighted imaging (DWI). This involves the cerebral cortex extensively and the basal ganglia.
- Hypointense on ADC mapping (a.k.a. restricted diffusion).
- The extent of hypointensity on ADC mapping is generally used for neuroprognostication.
- Several studies suggest that extensive ADC hypointensity can predict poor prognosis with good sensitivity (>50%) and high specificity. However, variable cutoffs and modes of analysis may limit the precision and generalizability of these results.
MRI use for prognostication
- MRI is useful only if the CT doesn't show severe anoxic injury.(32562686) If the CT scan shows unequivocal anoxic injury, the MRI will inevitably reveal the same thing (since the MRI is more sensitive than the CT scan).
- The optimal time window for obtaining an MRI might be 2-7 days after cardiac arrest.(33765189)
- Occasionally, an MRI may be seen which doesn't show any substantive anoxic injury. This is a red flag that the diagnosis is incorrect – the patient likely doesn't have severe anoxic brain injury. This should prompt an evaluation for the presence of other causes of altered consciousness. (e.g., see the delirium chapter here).
- Global mild MRI changes can be compatible with a good outcome.(32562686) However, severe changes are fairly specific for poor outcome (e.g., with specificity in the >95% range). Unfortunately, no objective grading scheme exists, rendering the division between mild and severe disease somewhat subjective.
- NSE is an enzyme released due to brain injury or hemolysis.
- Elevated or rising levels of NSE predict poor outcome.
- Limitations to the use of NSE include the following:
- Lack of availability at many centers.
- Disagreement regarding optimal cutoffs, due to an absence of calibration standards between laboratories.(32562686)
- False-positive elevation of NSE due to hemolysis or neuroendocrine tumors.
- Turnaround time is often long.
- It remains to be demonstrated whether NSE adds additional, independent prognostic information above and beyond what is available from other tests.
- If NSE is used, serial NSE levels at 24, 48, and 72 hours may help avoid confounding by a hemolyzed specimen. (33765189)
general principle of multimodality neuroprognostication
- Neuroprognostication should be based on combining multiple sources of information. For example, most patients will have neuroimaging (at least a CT scan), serial neurological examinations, and video EEG.
- When all sources of information are in agreement, this may increase confidence. For example, a combination of cerebral edema on CT scan, poor serial neurologic examinations, and a malignant EEG pattern could indicate a poor prognosis with high confidence.
- When sources of information disagree with one another, then additional testing and/or additional time is required to allow for prognostication. For example, a poor neurological examination combined with a normal MRI scan would imply the presence of an alternative diagnosis (e.g., metabolic encephalopathy), and potentially a favorable outcome.
general approach to neuroprognostication
- One approach to neuroprognostication is shown above, based on the European Society of Intensive Care Medicine 2021 guidelines.(33765189) Although this provides a general approach, it shouldn't be followed blindly. Rather, the specifics of each patient must be considered carefully (e.g., the exact results of EEG and imaging tests).
- This strategy is based on data derived from patients who often received hypothermia and benzodiazepine and/or opioid infusions. As such, it may not apply perfectly to patients maintained at a normothermic temperature with no opioids or long-acting sedatives.
- For example, imagine a patient with asphyxial cardiac arrest followed by a completely suppressed EEG for 72 hours. There are no confounding factors (e.g., no hypothermia, no sedatives/opioids). The patient's CT scan shows no visible abnormalities. After 72 hours, clinical examination shows no motor response to pain, no corneal reflexes, but intact pupillary reflexes. Based on the above algorithm, this patient would be deemed to have an unclear prognosis. However, a completely suppressed EEG and very poor examination after 72 hours actually give this patient close to nil likelihood of a meaningful recovery.
delayed awakening & patients with indeterminate prognosis (grey box above)
- Some patients may regain consciousness after three days (late awakening) – especially due to sedative medications, hypothermia, or any other causes of delirium (e.g., renal or hepatic dysfunction, sepsis).
- One before-after study comparing midazolam/fentanyl sedation versus propofol/remifentanil sedation found that short-acting sedation lowered the odds of delayed awakening (with an odds ratio of 0.08).(22527063)
- Avoidance of any long-acting sedation (including fentanyl infusions) may help patients awaken earlier, thereby limiting their ICU course and avoiding iatrogenic complications. A recent review article wrote that “conclusions about prognosis should be delayed at least 72 hours after arrest to allow for the clearance of sedative drugs” – but a better approach may be to never use long-acting sedation in the first place.(32562686)
- Don't give these patients long-acting sedatives (even fentanyl should be avoided). Long-acting medications may complicate the neurologic examination.
- If the patient spent a few days at an outside hospital before transfer, this must be accounted for. For example, if a patient is transferred to your hospital two days after arrest, the clock starts at the time of arrest (not the time of transfer).
- Failure to obtain a detailed and thorough neurologic examination at 72 hours is very problematic. The neurologic examination continues to evolve over time, rendering it less specific over time (e.g., if corneal reflexes appear after five days, this is less reassuring than corneal reflexes being present after three days).
- Overemphasis on fancy tests (e.g., MRI and SSEPs), while overlooking the neurologic exam and EEG.
references
- 27017468 Rossetti AO, Rabinstein AA, Oddo M. Neurological prognostication of outcome in patients in coma after cardiac arrest. Lancet Neurol. 2016 May;15(6):597-609. doi: 10.1016/S1474-4422(16)00015-6 [PubMed]
- 29525975 Muhlhofer W, Szaflarski JP. Prognostic Value of EEG in Patients after Cardiac Arrest-An Updated Review. Curr Neurol Neurosci Rep. 2018 Mar 10;18(4):16. doi: 10.1007/s11910-018-0826-6 [PubMed]
- 30739213 Hawkes MA, Rabinstein AA. Neurological Prognostication After Cardiac Arrest in the Era of Target Temperature Management. Curr Neurol Neurosci Rep. 2019 Feb 9;19(2):10. doi: 10.1007/s11910-019-0922-2 [PubMed]
- 32562686 Cronberg T, Greer DM, Lilja G, Moulaert V, Swindell P, Rossetti AO. Brain injury after cardiac arrest: from prognostication of comatose patients to rehabilitation. Lancet Neurol. 2020 Jul;19(7):611-622. doi: 10.1016/S1474-4422(20)30117-4 [PubMed]
- 32915254 Sandroni C, D'Arrigo S, Cacciola S, Hoedemaekers CWE, Kamps MJA, Oddo M, Taccone FS, Di Rocco A, Meijer FJA, Westhall E, Antonelli M, Soar J, Nolan JP, Cronberg T. Prediction of poor neurological outcome in comatose survivors of cardiac arrest: a systematic review. Intensive Care Med. 2020 Oct;46(10):1803-1851. doi: 10.1007/s00134-020-06198-w [PubMed]
- 33765189 Nolan JP, Sandroni C, Böttiger BW, Cariou A, Cronberg T, Friberg H, Genbrugge C, Haywood K, Lilja G, Moulaert VRM, Nikolaou N, Olasveengen TM, Skrifvars MB, Taccone F, Soar J. European Resuscitation Council and European Society of Intensive Care Medicine guidelines 2021: post-resuscitation care. Intensive Care Med. 2021 Apr;47(4):369-421. doi: 10.1007/s00134-021-06368-4 [PubMed]
