Sudden cardiac arrest (CA)—the complete cessation of cardiac mechanical activity in a pulseless, apneic (or with agonal breathing) and unresponsive individual—continues to claim millions of lives across the world and is a major cause of long-term disability.1,2 The precise estimation of the epidemiologic impact of CA worldwide is difficult, as incidences and survival vary widely according to factors pertaining to details of arrest (location, etiology, and non-perfusing rhythm); whether the event was assisted by Emergency Medical Services (EMS); and in what region of the world it occurred. Table 1 shows a brief comparison of reported incidences across different world areas. In the US, over half million CA occur every year, with approximately 2/3 taking place outside medical facilities (out-of-hospital cardiac arrest – OHCA); less than half of these receive bystander cardiopulmonary resuscitation (CPR).3 In the hospital, CPR is performed in 2.73 and 2.37 per 1,000 in elderly and non-elderly adult admissions, respectively.4 5 Approximately 10% of OHCA and 1 in 4 adults who experience an in-hospital cardiac arrest (IHCA) survive to hospital discharge in the US.3 The presence of comorbidities seems to have little overall influence on short-term survival following OHCA,6 including malignancy,7 but comorbidities do influence outcomes in non-shockable rhythms in both OHCA8 and IHCA.< a href="#_ENREF_9">9,10 Survival rates are uniformly higher in shockable rhythms across all patient populations, and conflicting evidence supports lower survival in women, older, and non-white patients; however, these findings may be substantially influenced by several other factors acting as confounders, such as CA rhythm, hospital site, timing of urgent coronary reperfusion, use of targeted temperature management (TTM), duration of resuscitation efforts, and many more. There is an independent and inverse relationship between duration of resuscitation and favorable survival—in particular, for total downtimes greater than 35 minutes regardless of underlying CA rhythm.11,12 Nonetheless, a significant proportion of patients with prolonged arrest times may survive and achieve a good neurologic outcome should prompt aggressive care be sustained.13 Similarly to patients who received standard cardiopulmonary resuscitation (CPR), having shorter downtime, shockable rhythms, higher arterial pH value, and lower serum lactic acid concentrations at admission are also associated with better outcomes in patients undergoing extracorporeal CPR.14
In addition to overall survival, outcomes in CA studies also take into consideration functional status, and the Glasgow-Pittsburgh Cerebral Performance Category (CPC) scale is the most commonly used tool (Table 2). Death, regardless of its main determinant (recurrent and/or refractory CA, multisystem organ failure, brain death, or death related to withdrawal of life-sustaining therapies) is classified as CPC 5. A CPC score of 1 – 3 represents a conscious state (awakening); however, as CPC 3 reflects severe disability with complete dependence on others for daily living activities, it is often considered a poor outcome. Traditionally, good outcome is reserved for CPC scores of 1 – 2 in the CA literature. Families and individuals may consider regaining of consciousness an acceptable outcome in specific scenarios; thus, it is important to ascertain what is considered meaningful neurologic recovery for each individual before sharing neuroprognostic impressions. Besides, all these outcome categories include substantial heterogeneity of deficits, and a more granular analysis of functional outcomes is indispensable. The extended version of CPC (CPC-E) was developed to address this gap and potentially guide further interventions or the need for services following hospital discharge of CA survivors. This comprehensive evaluation comprises 10 domains: alertness, logical thinking, attention, short-term memory, motor function, basic and complex activities of daily living, mood, fatigue, and return to work. The end score is obtained in staged evaluations (at discharge and on follow up) and has demonstrated good to excellent inter-rater reliability,22 which makes it a promising comprehensive tool for outcome assessment in this population.
The war against death fought by a CA patient includes 5 major battles that occur from resuscitation to the rehabilitation period. These steps are demonstrated in Figure 1 and constitute promising targets for interventions to improve outcomes in this deadly disease.
Figure 1 – Saving the Survivors of Cardiac Arrest in Five Steps
The mechanisms involved in the development of systemic and neurologic hypoxic injury result from the global ischemia during circulation cessation and from the pathologic state of reperfusion with restoration of forward flow. The “two-hit” model24 includes complex cascades involving widespread molecular, cellular, and organ-specific responses in addition to the ongoing pathophysiologic process that led to the cardiac arrest; the interplay of these factors is summarized in Figure 2. The new systemic state following restoration of spontaneous circulation, first recognized in the early 1970s, is now called Post-Cardiac Arrest Syndrome (PCAS), and includes brain injury, myocardial dysfunction, and diffuse endotheliopathy with systemic ischemia/reperfusion response.25 Early extra-cerebral organ dysfunction is common in non-survivors, and is directly related to the extent of global hypoxic-ischemic injury; this relationship with survival does not seem to occur when organ dysfunction happens late during the intensive care unit stay.26 Severe renal dysfunction is an independent predictor of early mortality,26 and may be present in over 50% of resuscitated patients,27 more often in prolonged shock, in older individuals, and in those with prior chronic renal disease.
No-flow Period and Primary Injury
Lacking energy stores, the brain relies heavily on sustained circulation to maintain its function. In fact, 20 – 25% of the cardiac output is used by the brain; thus, within minutes of CA, depletion of adenosine triphosphate (ATP) occurs, triggering neural ischemia and cell death due to massive ionic dysregulation and anoxic depolarization.24,28 Intracellular acidosis subsequently ensues, following the rise in cerebral lactic acid concentrations from the metabolism shift towards anaerobic pathways as energy substrates decrease.24 This culminates in an influx of Na+ and consequent cytotoxic edema, and the rise in intracellular Ca+2 activates lytic enzymes, promotes release of excitatory neurotransmitters, and activates apoptosis. Aquaporin-4 channels are disrupted, contributing further to cerebral edema formation.29
The effects of these events impact all organ systems, the brain in particular. Areas of increased aerobic metabolism are more vulnerable, such as the cerebellum, striatum, thalamus, hippocampus, and primary sensory cortex.30 On the other hand, hypothalamic and brainstem neurons display increased resilience to ischemia due to their intrinsic ability to resist anoxic depolarization; hence, the preservation of basic vital functions and reflexes in severe HIE survivors in persistent vegetative state.31 Several studies have demonstrated that consciousness disruption along with diffuse slowing followed by attenuation of cerebral rhythms on electroencephalogram (EEG) occur within 10 – 30 seconds of circulation cessation, regardless of the underlying CA rhythm.24,32,33 To some extent, the clinical manifestations of hypoperfusion during no-flow state are reversible, particularly in the setting of immediate high-quality CPR.
Figure 2 – Pathophysiology of Post-Cardiac Arrest Syndrome
Quite unnervingly, a recent survey of CA survivors demonstrated that approximately 2% of respondents described awareness with explicit recall of events related to their CPR.34 In fact, signs of regaining of consciousness during CPR such as spontaneous eye opening, random body movements, and even spontaneous speech may occur in up to 1% of resuscitated patients. Their clinical significance is still being elucidated, but evidence points towards an increased likelihood of survival to hospital discharge if these occur in unwitnessed/bystander-witnessed CA events compared to EMS-witnessed CA, as the former are less likely to receive sedating medications should these signs occur.35
Re-flow, Failure of Microcirculation and “No Re-flow,” Excitotoxicity, and Secondary Injury
The pathologic state that ensues upon resumption of forward circulation following whole body ischemia is probably the main determinant of injury burden and outcomes. The reperfusion injury is complex and involves a cascade of dysregulated inflammatory and prothrombotic events, as well as ongoing bioenergetic failure from the imbalance between metabolic demand and availability of energy substrates. The microcirculation was thought to be initially preserved in the early phase following return of spontaneous circulation (ROSC),36 but a recent pre-cliniccal study demonstrated theoccurrence of red blood cell flow arrest, or “no re-flow” phenomenon, as early as 5 minutes following ROSC.37 The subsequent ischemia, likely due to a combination of impaired autoregulation and factors related to the endotheliopathy that arises following reperfusion. This is the hallmark of secondary brain injury.24,38 Despite clear evidence of fibrinogenesis,39,40 microthrombi formation, and the benefits of systemic thrombolysis on the cerebral vasculature and overall perfusion in preclinical studies,41 the TROICA study (Thrombolysis during Resuscitation for OHCA) was interrupted early after enrolling 1050 patients demonstrating no differences in survival, ROSC rates, or neurologic outcomes between tenecteplase and placebo groups.42 Studies targeting endotheliopathy in PCAS are underway and may change the course of multisystem organ failure in this setting.43 In addition to endothelial damage and disruption of interstitial media, accumulation of intracellular Ca+2 coupled with decreased clearance of free radicals result in further neuronal dysfunction and cell death, thus constituting the final pathway in secondary brain injury.
The art of neuroprognostication in the modern era is a complicated and dynamic process. Despite this, significant headway has been made, with a trend toward using a multi-modality approach (rather than the results of a single test) incorporating complementary data achieved via different means to give a more cohesive picture. Also, more attention has been gained for the concept of the self-fulfilling prophecy of premature termination of life-sustaining therapy, as many important reports have surfaced that have drawn into question the pre-existing dogmas and fatalistic approaches. It is clear that the coming years will bring a calibration of our approach to neuroprognostication in this population, one that will likely be more comprehensive and appropriately patient. It seems prudent to account for comorbidities when neuroprognosticating;212 however, comorbidities have not been associatetd with short-term6 or 1-year survival rates.213
Prior to the widespread use of therapeutic hypothermia post-CA, there was some agreement as to the most useful clinical examination findings and ancillary test results, as well as the best time frames in which to utilize them for prognostication. However, the use of TTM has muddied the waters, as the impact on both the clinical examination and ancillary testing can be significant.214 There are conflicting reports on the impact of the exposure to TTM on awakening, and reports of delayed regaining of consciousness in patients exposed to deeper hypothermia (32-33°C) compared with those with higher temperature targets (>34°C) suggest that careful delayed prognostic assessments may be warranted in this setting.215,216 This is for two reasons: 1) neurons that are hypothermic may take longer to recover; and 2) medications used during TTM to keep the patient comfortable, such as sedatives, are likely to remain in the system longer than usual in the setting of hypothermia. Furthermore, sedatives will also take longer to clear in the setting of hepatic and/or renal dysfunction, both of which are not uncommon in this population, either pre-morbidly or post-CA. Finally, the target temperature for post-CA patients is now a moving target, as many centers are now using 36°C based on the TTM trial,101 while some are still using 33°C. The amount of sedative medications may vary quite a bit between these temperatures, as patients cooled to 33°C may not require paralysis to keep them from shivering, thus minimizing the amount type of sedative medications necessary. Finally, the upcoming TTM2 study will provide further data, as it will be studying 33°C vs. 37°C, although patients in this range will likely still need some medications to keep them comfortable, depending on the temperature control method. Caregivers concerned with prognosis will need to take all of these factors into account when prognosticating, and likely many more factors as well. Unfortunately, a significant proportion of ICUs still does not use a standard operating procedure for neurologic assessment and prognostication when using TTM.120 Several risk scores have been proposed for the early prediction of poor outcome in OHCA over the years and, most recently, the TTM score has showed promise with an area under the receiver operating curve of 0.842 (0.818 corrected for optimism), whereas the CAHP and OHCA risk scores were similarly inferior to TTM at 0.746.217 The factors most reliably predicting poor prognosis were older age, CA occurring at home, non-shockable initial rhythm, longer duration of no-flow and low-flow times, early administration of epinephrine in shockable rhythms, pupillary and corneal areflexia, no motor response to noxious stimulation, lower pH, delayed lactate clearance and hypocarbia.217
Recent studies have looked at outcome prediction scores, including attempting to predict good outcome rather than poor outcome, such as the GO-FAR score, although these lack external validation.218
The clinical examination remains central to assessing prognosis; in fact, its importance should not be overlooked or underemphasized. Of all of the findings that have been assessed for accuracy of prediction of prognosis, none have rivaled the clinical examination, which has stood the test of time. However, like all testing, staying true to proper technique of examination and correct interpretation of examination findings are integral to the process. Furthermore, the timing of the clinical examination is crucial, and, in the setting of therapeutic hypothermia, this should be delayed until at least 5 days post-arrest, if not longer.219
The first steps in the evaluation of an unresponsive patient after cardiac arrest are observation and stimulation. Observation should include an evaluation of the ventilator settings, such as whether the patient is over-breathing the set rate on the ventilator. The breathing pattern of the patient should be assessed, as this can give an indication of the degree of brain injury and possible brainstem dysfunction. The patient’s limbs should be uncovered during the examination (always maintaining decency) so that responsiveness to stimulation at one location (e.g., pressure on the supra-orbital notch) can be observed in a remote location (e.g., foot movement). The patient should be observed closely for any spontaneous movements, such as myoclonic or seizure-like movements, or spontaneous eye opening. Next, the patient should be stimulated, and this should include auditory, visual, and tactile stimulation. We recommend working from mild to more noxious stimulation in order to assess what degree of stimulation is necessary to establish consciousness. A common mistake of junior clinicians is to not be loud or noxious enough. Auditory stimulation should reach the point of yelling the patient’s name and clapping loudly. Visual stimulation is accomplished by opening the eyes of the patient (assuming they are not already open spontaneously) and testing for a blink to visual threat by approaching the patient with the edge of the hand or tips of fingers (so as to not create an air current and thus test a corneal reflex inadvertently), first coming from the lateral fields, then down the center of vision. Tactile stimulation should reach the point of deep noxious stimulation, both on the extremities and head (the head must be stimulated, as the patient may have an unknown spinal cord injury or severe peripheral neuropathy that could preclude responsiveness to noxious peripheral stimulation). Cranial stimulation should take place first with a “nasal tickle” using a Q-tip inserted into the nares, followed by deep pressure on the supra-orbital notch as well as the temporo-mandibular joint. Pressure on the extremities should be on the nail bed first, and then at more proximal locations to determine if a given response is stereotyped or not. The goal of this stimulation is to try to establish whether the patient is comatose or not. The strict definition of coma is that the patient shows no evidence of responsiveness to any stimulation; the only responses that are seen are reflexive, including brainstem and spinal.
Next, brainstem testing is performed, with particular attention to pupillary and corneal reflexes, which have shown the highest reliability in prognostication. Again, technique is paramount. Pupils should be tested for reactivity in a dimly lit room, and optimally, a pupillometer should be used, especially in cases of questionable reactivity. In the absence of a pupillometer, a magnifying glass should be used. The reliability of the “swinging flashlight” test is likely quite poor in this setting, as the halo from conventional flashlights likely initiates pupillary constriction prior to the pupil coming into focus, reducing the likelihood that the clinician will be able to appreciate reactivity in subtle cases. The light, which should be LED to maximize power and focus, should be shined off and on repeatedly, looking for pupillary constriction as well as the speed of constriction, if present. The pupillary light reflex may be affected by drug overdose, states of low cardiac output, and/or resuscitation drugs; thus, early non-reactivity may still reverse following the acute period of resuscitation.220 Additionally, despite taking every measure to improve sensitivity of testing, it still remains a subjective test with moderate inter-rater reliability based on the skill of the examiner.221 The use of automated pupillometry shows promise in increasing the reliability of this finding by means of an objective assessment.222,223 The corneal reflex should be tested by touching the cornea where it has significant innervation (which is over the iris and aperture of the pupil, and not on the conjunctiva, where innervation is less robust); we recommend testing at the border of the iris with the conjunctiva. Intensivists surprisingly often miss the target when testing the corneal, per a recent large international survey (Figure 5). Furthermore, to say that a corneal reflex is absent requires sufficient stimulation; squirting the eye with sterile saline/water or lightly touching the eye with a wisp of cotton is not sufficient to determine that a corneal reflex is absent. In the setting of neuroprognostication, the cornea should be pressed with a sterile Q-tip. Other brainstem reflexes should also be tested, including oculocephalic, oculovestibular, cough, and gag reflexes, although their current impact on neuroprognostication is uncertain and needs to be studied.
Traditionally, the motor examination findings were felt to carry prognostic weight, specifically extensor or no motor response to noxious stimulation. However, especially in the era of therapeutic hypothermia, poor motor response has been associated with unacceptably high false positive rates in prediction of poor outcome, up to 24%, making it a suspect prognostic finding at best. Again, attention to technique is likely important, as outlined above.
Finally, the presence of myoclonic jerking should be noted, as the presence of myoclonic status epilepticus (MSE) has been associated with poor outcome in prior studies.224,225 However, recent pool analyses demonstrate that approximately 10% of patients with MSE may, in fact, achieve a good neurologic outcome.226 Additionally, the definition of MSE varies from study to study, and may or may not require an EEG. The most common definition is the presence of non-rhythmic but repetitive and unrelenting jerking of the face, eyes, eyelids, limbs, and axial musculature. We would caution that it is impossible to know whether or not a patient with overt clinical MSE is having electrographic seizures, and when MSE is seen, an EEG should be performed (and if seizures are seen, they should be treated with antiseizure drugs). The EEG itself may be difficult to interpret in the setting of MSE, as much EMG artifact is generated; in this case, a dose of short-acting paralytics may be administered. Emerging evidence from the use of continuous EEG in patients with cortical myoclonus points towards two distinct clinic-encephalographic phenotypes: high voltage polyspikes time-locked with myoclonus in a burst suppressed background with an ominous prognosis, versus midline spike-wave discharges time-locked with myoclonus in a continuous background, which are associated with delayed but potentially good neurologic recovery in some patients.164,173,227 MSE is often treated with antiseizure drugs and muscle relaxants, with marginal effect. However, the use of these medications likely leads to a significant sedative effect, thereby hindering awakening in these patients. There is likely a strong self-fulfilling prophecy bias, as MSE is difficult for families and ancillary staff to observe, and often leads to over-treatment with pharmacologic agents and early withdrawal of life-sustaining therapies (WLST). We would caution that MSE does not necessarily indicate a poor prognosis, and continued prudent care may often lead to good patient outcomes.
Electrophysiological evaluations in post-CA patients are crucial, lending keen insights to patterns of injury and recovery. The two techniques that are commonly used are evoked potentials (specifically, somatosensory evoked potentials SSEP) and EEG.
The data in support of neuroprognostication using SSEP typically has evaluated median nerve stimulation with measurement of the contralateral cortical (N20) responses. Bilaterally absent N20 potentials within the first several days after arrest has been associated with poor outcome in multiple studies. The N20 peak represents the thalamo-cortical pathway, and it is important to note that the integrity of the pathway from the median nerve through the brachial plexus and brainstem must be intact, and structural lesions (e.g., prior stroke or subdural hematoma) can influence the findings as well. N70 potentials, representing the cortico-cortical pathways, have also been evaluated in cardiac arrest, but are technically more challenging and not used in many centers. The largest study to evaluate SSEP in a prospective fashion was the PROPAC study,228 which served as the basis for the 2006 guidelines for prognostication from the American Academy of Neurology.229 PROPAC evaluated 407 post-CA patients from 32 Dutch ICUs, and found that no patients with absent N20 peaks bilaterally survived to a good outcome. However, this study has been criticized for a lack of blinding, as well as a clear self-fulfilling prophecy bias: 23% of patients has WLST at 24 hours post-arrest, an alarmingly early time for decision-making regarding neuroprognosis. Furthermore, false positives have been reported in patients with absent N20 potentials shortly after undergoing TH, who later recovered function (and N20 potentials). Environmental noise, commonly seen in ICU due to equipment in the room, and shivering artifacts frequently limit interpretation of SSEP studies in this patient population.230 Technical aspects of adequacy of SSEP studies for interpretation are important and basic requirements of testing are summarized in Table 12. Despite these limitations, SSEPs are felt to be a valuable ancillary test in the assessment of prognosis post-CA, when put into the proper context and employed at the right time points. (Figure 6) A recent small study investigated the use of brainstem auditory evoked potentials (BAEPs), but found them to be of limited use.231
EEG gives complementary electrical information, and is an emerging, useful tool in the evaluation of prognosis and in the diagnosis and treatment of post-anoxic seizures. EEG can be very helpful in the early setting, when a patient is displaying signs of myoclonus, to ascertain if these are cortical or subcortical in nature. Non-convulsive status epilepticus (NCSE) cannot be detected without EEG, especially in the setting of sedation (and paralysis) when undergoing TTM. As a neuroprognostic tool, several EEG findings have been reported as useful, but none have maintained 100% positive predictive value for poor outcome, including burst-suppression, diffuse attenuation, and lack of variability with identical bursts.236-239 Reactivity has also emerged as important finding;158 patients who lack EEG reactivity to external graduated stimulation appear to be destined for a poor outcome. Nonetheless, subjectivity of reactivity assessment and lack of standardization of stimulation continue to challenge the prognostic yield of this EEG feature. It should be noted, however, that the presence of reactivity on EEG does not ensure a good outcome. The use of quantitative EEG analysis holds promise by offering increased objectivity.240 Seizures and epileptiform findings have been associated with poor outcomes with modest specificity in retrospective analysis, likely due to factors linked to how aggressive treatment was pursued and consequent self-fulfilling prophecy bias. Conversely, lack of epileptiform activity may be an optimistic sign in the absence of other ominous signs.241,242 Surprisingly, patients with epileptiform patterns may not demonstrate imaging evidence of hypoxic brain injury, according to a recent retrospective review of clinical, magnetic resonance imaging, and electroencephalography data of TTM treated CA survivors.243 (Figure 7)
Continuous EEG is preferable in the first several days after arrest, but is not feasible or available at all centers, and sequential EEGs may be a reasonable alternative.244 The EEG typically evolves during the course of monitoring,245 and may be significantly affected by temperature, sedating and non-sedating antiseizure drugs246 and anesthetics, renal/hepatic impairment and hypoperfused states; thus, it is important to not use a single snapshot of a patient’s EEG to make prognostic judgments. EEGs with initially poor or malignant appearing patterns may improve and the recording may display positive trends over time; hence, the need for repeated testing comprising different stages in the post-CA course. In fact, approximately one in 6 patients with initially non-reactive backgrounds to external stimuli during TTM regain reactivity during normothermia. Promising emerging neurophysiologic strategies in outcome prediction include longitudinal analysis of continuous quantitative EEG variables247 and progression of auditory discrimination based on EEG responses during unresponsiveness.248 It is prudent to note that, although widely used,249 neuroprognostic assessments based on EEG continue to be challenged by the lack of high levels of evidence, vulnerability of findings to the effect of confounders, lack of uniform classification in traditional studies based on standardized nomenclature,241 inconstant inter-rater variability,250 and the impact of self-fulfilling prophecy bias on existing evidence. Nonetheless, when used in a multimodal approach, EEG findings provide valuable data enriching prognostic models.156,251-254
Multiple chemical biomarkers have been evaluated as a method of detecting the degree of neuronal injury and breakdown, but all have their limitations in terms of accuracy and testing. The biomarker with the most data in this setting is neuron specific enolase (NSE), but it is subject to variations for multiple reasons: 1) multiple assays are available for testing, and laboratory variations can be significant;255 2) it is often a send-out lab, and delays on results can be lengthy; and 3) NSE is also found in red blood cells and platelets, and its release from these can result in elevated levels that are due to hemolysis or systemic factors and not necessarily reflective of neuronal injury. Thus, NSE can be a “dirty” biomarker, and relying on single values is likely a fallacy. It is probably more prudent to evaluate NSE on a serial basis over the first 1-5 days post-arrest, as this is a time when the patient is often improving from a medical/systematic standpoint, and thus NSE levels that are trending up during this time are likely indicative of ongoing neuronal injury and poor prognosis. This approach needs to be studied prospectively and in an unbiased population.
It should be noted that hypothermia attenuates NSE levels, but despite this, recent studies have found false positives for elevated NSE levels predicting prognosis. The primary data for NSE also came from the PROPAC study228 (see above in SSEP section), in which a cutoff value of >33 mcg/L was uniformly associated with a poor outcome. However, since that time there have been multiple patients with good outcome despite having NSE levels well above that. NSE has also been studied in patients on ECMO, and may be a useful marker of cerebral injury in this subset of patients.256
Post hoc analyses of the TTM trial demonstrated a promising role of light chain neurofilament, Tau and admission lactate (although not lactate clearance).257-259 The neutrophil-to-lymphocyte ratio (NLR) has also been the subject of limited studies, with some suggesting an NLR >6 at 72 hours post-ROSC is associated with worse outcomes.260
A quick word on what would make the optimal chemical biomarker: one of the challenges with the biomarkers studied to date is that they have primarily been studied in the blood, which is a contaminated environment. The optimal chemical biomarker would be most reflective of neuronal damage, or at least procured from a non-contaminated environment. The CSF is an obvious such environment, but studies of biomarkers obtained from the CSF are sparse. Clinicians may have pause in the concept of performing lumbar puncture in patients post-CA due to concerns of causing herniation. However, in a patient with radiographic signs of severe cerebral edema, a CSF examination is not necessary, as the prognosis is not in question (it is poor!). Lumbar puncture could be reserved for patients without these radiographic findings, which are the vast majority.
Neuroimaging has the distinct advantage of providing structural information about the brain, and both CT and MRI are valuable in the acute and subacute setting. Advanced MRI techniques are also able to tell us about brain perfusion and the integrity of white matter pathways and connectivity, which will all be important concepts moving forward. But even conventional CT imaging remains highly useful from a practical standpoint, and MRIs may not be able to be performed in all patients due to safety or compatibility issues. It should be noted that neither CTs nor MRIs have been validated for use in prognostication,261 but from a practical standpoint, both provide important information, and can and should be used in the context of the overall information gathered about the patient and their prognosis.
CT imaging is commonly performed on cardiac arrest patients at the time of admission, but this is not often a time point that is useful from a neuroprognostication standpoint. However, it is useful to rule out a cerebral cause for the arrest (such as in a younger patient or one without evidence of cardiac ischemia on ECG or enzyme analysis), or to evaluate for traumatic brain/spine injury in patients who are found down or have a witnessed fall with head/neck impact. From a prognostication standpoint, a CT scan performed on day 2-3 post-arrest is an appropriate time point, as ischemic changes that may have been delayed initially should have appeared by then, and it can give useful information to corroborate the clinical exam and other ancillary testing. Widespread hypodensities and/or diffuse cerebral edema with loss of grey-white differentiation likely portend a poor outcome, and specific brain regions, such as the Hounsfield unit ratio between the putamen (grey matter) and internal capsule (white matter) have correlated with outcome in several studies. Global Hounsfield unit reduction, especially in combination with the clinical examination findings on day 3, appears to strongly correlate with poor outcome with high specificity.262
Figure 9 – CT in cardiac arrest – A 20-year-old male suffered a 20-minute cardiac arrest. His CT is shown on day 4 post-arrest, with the pseudosubarachnoid hemorrhage sign caused by engorgement of the veins at the base of the brain.
However, the CT is commonly normal post-cardiac arrest, and MRI may be necessary, as it is more sensitive to demonstrating the degree of ischemic injury.263 Again, the time point for testing is important, as premature testing may give falsely optimistic (e.g., minimal changes) results.264 The optimal time point for procuring the MRI post-arrest is likely 3-5 days post-arrest, as that is when the changes on diffusion-weighted imaging (DWI) are most likely to be maximal. If imaging is performed at other time points, it is important to keep in mind where to expect the changes, and with what sequence (DWI vs. FLAIR/T2); early changes typically involve the cerebellum and basal ganglia, which then migrate to the cerebral cortices, and then finally to the subcortical white matter.265 Some specific imaging findings have shown promise in predicting poor outcome, such as the “bright hippocampus sign,”266 but these need to be validated in prospective and unbiased studies.
Figure 10 – 42-year-old woman with cardiac arrest. MRI on day 4 post-arrest shows diffuse cortical restricted diffusion on (a) diffusion-weighted imaging and (b) apparent diffusion coefficient maps.
We will now discuss three guidelines for neuroprognostication post-CA, including their notable differences: the American Academy of Neurology Practice Parameters (AANPP) from 2006,229 the European Society of Intensive Care Medicine (ESICM) guidelines from 2015,269 and the American Heart Association (AHA) guidelines from 2015.123
The AANPP, the first of its kind, were primarily based on data prior to the widespread use of TH, and thus quickly became of questionable value in the modern era. They touted strong evidence for aspects of the clinical examination in predicting poor outcome, all measured at 72 hours post-CA, including absent pupillary and corneal reflexes, as well as motor exam findings of extensor posturing or no response, which has since fallen out of favor (although pupillary and corneal reflexes are still felt to be valid). The presence of myoclonic status epilepticus was also felt to indicate a poor prognosis, but with a lower level recommendation. Absent N20 responses on SSEP performed 1-3 days post-CA correlated with poor prognosis, but a burst-suppression pattern or generalized epileptiform discharges were felt to have insufficient evidence to achieve a strong recommendation. Serum NSE levels >33 mcg/L measured 1-3 days post-arrest were given a fairly strong recommendation (level B), but again, this was based on somewhat biased data and has since been debunked. There was felt to be insufficient evidence to support or refute the use of neuroimaging at that time.
The ESICM guidelines prudently took into account the impact of hypothermia on prognostication, and advocated for providing more time in situations of uncertainty. There is an up-front emphasis on excluding confounders, particularly residual effects of sedation. The motor examination at ≥72 hours post-arrest with findings of extension or no response is used as a sensitive sign for potential poor prognosis (in the setting of an unconscious patient), followed by one or both of the following at 3-5 days post-arrest: absent pupillary AND corneal reflexes, or bilaterally absent N20 peaks on SSEP. If these were not present, they suggested an additional waiting period of at least 24 hours, and then two or more of the following to indicate a high likelihood of poor outcome: status myoclonus ≤48 hours after ROSC, “high” NSE levels (not defined), unreactive burst-suppression or status epilepticus on EEG, or “diffuse” anoxic brain injury on brain CT/MRI (not defined). In the absence of these, the outcome should be considered indeterminate, and continued observation and re-evaluation is encouraged. The authors advocated the “use [of] multimodal prognostication whenever possible,” which is laudable.
The AHA guidelines also take into account the effect of TH/TTM, and advocate that the earliest time for prognostication using the clinical examination in this setting should be 72 hours after the return to normothermia, and longer if the residual effects of sedation or paralysis might potentially confound the examination. Absent pupillary and corneal reflexes again carried a strong recommendation for poor prognosis, but the motor exam findings of extensor or no movement were felt to have an unacceptably high false positive rate of 10% (95% CI, 7-15%). Any myoclonus was felt to predict a poor outcome, but with a 5% false positive rate (95% CI, 3-8%). Absence of EEG reactivity, persistent burst-suppression pattern, or status epilepticus were felt to correlate with poor outcome, but with a lower level of evidence. Absent N20 responses on SSEP again were felt to indicate a poor outcome, but the authors cautioned that technique and expertise were paramount in ensuring their validity. Both CT and MRI were felt to be reasonable to consider, and “marked reduction” in the grey-white ratio, or “extensive restriction” of diffusion on MRI were felt to be reasonable indicators of poor prognosis, in combination with other established predictors. Noting the high false positive rates, the AHA stated blood biomarkers (NSE and S-100B) should not be used alone to predict poor neurological outcome, but it would be reasonable to consider “high” serum values of NSE 48-72 hours post-arrest to support a possible poor prognosis, especially if repeated testing revealed persistently high levels.
Neuroprognostication after cardiac arrest remains challenging, and several principles should help guide the clinician. First, the presence of confounding should always be taken into account, including the effects of age, temperature, systemic disease, metabolic abnormalities and medications. When in doubt, err on the side of caution, correcting these abnormalities as much as possible, and giving the patient more time to recover, particularly younger patients and/or in the setting of limited indicators of poor outcome. A multi-modality approach should always be used; the clinician should take into consideration the findings in the context of ancillary data, and if the data is not internally consistent (e.g., the patient is not awakening, yet the neuroimaging shows little or no damage), this must be explained and the patient given more time to recover. Recent studies have explored the utility of multimodal approaches using different aspects of the investigative work up.270-272 We would advocate, in the proper clinical setting, that the finding of electrographic seizures or status epilepticus could be treated aggressively, as it is one of the few treatable conditions in the subacute setting. Finally, the effect of hypothermia should always be considered,273 not only for the clinical examination but also for ancillary testing. When in doubt, give more time.
The following table gives an algorithmic approach to specific testing and the time points at which they should be utilized.