Why we do What We do Critical Care Edition: ED Prediction of Neurological Outcome After Cardiac Arrest

Home / Journal Club / Why we do What We do Critical Care Edition: ED Prediction of Neurological Outcome After Cardiac Arrest


Andrew Williams, MD, PGY-4


Neuroprognostication in the emergency department is a thorny problem. Emergency departments in the United States treat approximately 175,000 victims of sudden cardiac arrest per year3. Of these, many will experience poor neurologic recovery while a select few will go on to have good outcomes. There is often pressure on emergency medicine physicians to offer a prognosis to family and staff with limited information in a field in which many misconceptions exist. Early and accurate neuro-prognostication would allow limited intensive care resources to be concentrated where they are needed most, help limit healthcare costs, and spare families extended periods of uncertainty.

The American Academy of Neurology released guidelines based on a systematic review of this topic in 20067. The physical exam remains the most important and frequently utilized tool. As patients following cardiac arrest are typically intubated and eye opening may occur early without implying a favorable diagnosis, the motor score potion of GCS and the pupillary and corneal reflexes are most useful. At the time of this review, data from 10 studies showed false positives for negative prognosis with a GCS motor score of 1 or 2 at 24 and 48 hours, but none at 72 hours. Similarly there were no false positives for absence of pupillary and corneal reflex at 72 hours. It is therefore a level A recommendation that prognosis is poor if there are absent corneal or pupillary reflexes or a GCS motor score of 2 or less 3 days after cardiac arrest.

Two studies found the presence of myoclonic status epilepticus to have a 0% false positive rate; it is a level A recommendation that the presence of myoclonic status epilepticus within the first 24 hours confers a poor prognosis. Interestingly, anoxia time, duration of CPR, and cause of cardiac arrest were found to have false positive rates for poor neurologic outcome between 20-27%. Therefore, while anoxia, duration of CPR and cause of cardiac arrest are associated with poor outcome, these variables can not accurately differentiate between those with good and poor outcomes and so it is a level B recommendation that prognosis not be made based on the history of CPR. The overall proposed prognostic algorithm is shown below.

neuroprog algorithm

From Hijdicks et al.

Of note for emergency medicine physicians the first step is excluding major confounding factors, which may include drugs (exogenous or iatrogenic), shock, acute kidney or liver failure/dysfunction, profound acidosis and other toxic metabolic factors. This can be a laborious process the completion of which is not always practical in the emergency department. According to the review there was inadequate evidence to precisely delineate which lesions on CT and MRI could conclusively predict poor outcome. More recent studies have shown that CT may have a role in predicting poor outcome using a method of the ratio of Hounsfield units in grey matter to those in white matter at the level of the basal ganglia and cortex4. However ideal cutoff values and timing of CT is yet to be established. Particularly MRI, using quantitative methods of diffusion-weighted imaging (DWI) or fluid-attenuated inversion recovery (FLAIR) may be useful in predicting long term cognitive and functional outcomes10. Other modalities discussed in the guidelines include somatosensory evoked potentials (SSEP), which are useful 24 hours after cardiac arrest, and biomarkers for hypoxic neurologic damage such as neuron specific enolase, which is not commonly available from the emergency department. Based on these recommendations, little can usually be said in the emergency department about the neurologic prognosis of a post cardiac arrest patient. History and physical exam findings are nonspecific except in the case of myoclonic status epilepticus and when a true diagnosis of brain death can be made.

The now widespread use of therapeutic hypothermia (TH) has added further questions2. A large prospective study of 111 adults treated with TH showed a poor prognosis false positive rate as high as 24% for GCS motor 1-2 at day 38. Likewise the false positive rate for even negative pupillary and corneal reflexes at 3 days in the same group is 4%. Therapeutic hypothermia also affects serum levels of neuron specific enolase6. There is now consensus that TH lengthens the period of neurologic recovery, but the overall effect on pattern of delay is not well characterized. It is understood that there is a need for new guidelines that account for the physiologic changes induced by TH as well as the finding of an increased number of patients awakening from post-anoxic coma.

With current limitations, there is a demand for novel methods of neuro-prognostication after cardiac arrest. One technique that holds significant promise is noninvasive cerebral oxygenation monitoring. This utilizes near infrared spectrometry to measure regional cerebral tissue oxygenation (rSO2), typically near the frontal lobe. This tool already has applications in cardiac surgery5. Previous studies have shown rSO2 to significantly correlate with jugular bulb venous saturation7. The Japan-Prediction of neurological Outcomes in patients Post-cardiac arrest (J-POP) Registry, a recent multi-center prospective study in Japan, investigated the utility of using rSO2 to predict neurologic outcomes in cardiac arrest survivors1.  rSO2 was measured in out of hospital cardiac arrest (OHCA) patients immediately upon hospital arrival in a non-blinded fashion and measured for at least one minute. The primary outcome was neurologic status at 90 days as defined by Glasgow-Pittsburgh Cerebral Performance Categories. CPC scores of 1 (good outcome) or 2 (moderate disability) were deemed good neurologic outcome. The lowest rSO2 found was used.

672 of 1017 OHCA patients were enrolled. Inclusion criteria were unresponsive patients during and after resuscitation at hospital arrival after OHCA. Exclusion criteria included trauma, accidental hypothermia, age under 18, previous DNR form and GCS >8 at hospital arrival. 72 were excluded because the rSO2 was not measured, mostly for inadequate staffing.

190 of the 672 survived to admission, an additional 152 died after admission. 38 (5.7%) survived to 90 days. Of these 29 had a good neurologic outcome. A higher rSO2 was found on hospital arrival in those who achieved ROSC in the field vs. those who did not, those who had ROSC in the ED vs. those who did not and those who survived to hospital admission vs. those who did not. Most importantly, rSO2 on arrival correlated with 90 day neurologic outcomes. Setting an rSO2 of >42%, good neurologic outcome could be predicted with a sensitivity of 79% (23/29) and specificity of 95% (610/643).

The study has limitations. It was not blinded, which may have influenced staff’s efforts at resuscitation. rSO2 was measured essentially only briefly and the lowest measurement recorded. Continuous monitoring beginning with initiation of CPR would be preferable. Further, EMS is not permitted to terminate resuscitation efforts in Japan, leading to perhaps more patients with poor prognosis and possibly limiting the generalizability of the study.

The results are impressive but insufficient in themselves to guide resource utilization or withdrawal of lifesaving interventions. A significant number of patients with potentially good neurological outcomes are missed with the cutoff recommended by the study. It is conceivable that clinicians could now with a combination of physical exam findings and testing available in the ED make a determination on their own with a similar sensitivity and specificity to this test. Although the measurement does not represent definitive information, it may be clinically significant data and in my opinion represents a small step forward. It is possible that improving this technology, applying data from continuous monitoring, or perhaps using this information in combination with other clinical data in a decision rule could ultimately allow accurate neuro-prognostication after cardiac arrest in the emergency department.


Take home points: 

  • There is usually no reliable way to prognosticate neurologic recovery after cardiac arrest in the emergency department. 
  • A diagnosis of brain death can be pursued only after a thorough search for reversible causes of neurologic dysfunction. 
  • Therapeutic hypothermia lengthens neurologic recovery time in ways that are incompletely understood. 
  • Noninvasive cerebral oxygen monitoring provides a number that correlates with, but does not by itself yet predict, neurologic prognosis after cardiac arrest. 




1. Ito et al. Noninvasive regional cerebral oxygen saturation for neurological prognostication of patients with out-of-hospital cardiac arrest: A prospective multicenter observational study. Resuscitation (2014).

2. Greer and Rosenthal. Nat Rev Neurol. 2014

3. Johnson et al. Resuscitation. 2013. 84(3)292-7.

4. Metter et al. Resuscitation. 2011. 82. 1180-1185

5. Murkin and Arango. British Journal of Anaesthesiology. 2009. Dec;103 Suppl 1.

6. Oddo and Rossetti. Current Opinion in Critical Care. 2011. 17: 254-259.

7. Rosenthal. Journal of Neurosurgery. 2014 Jan 31.

8. Rossetti et al. Annals of Neurology. 2010. 67. 301-307

9. Wijdicks et al. Neurology. 2006. 67. 203-210.

10. Wijman et al. Annals of Neurology. 2009. 65: 394-402.



Leave a Reply