CICM examiners are rather fond of asking about the factors that determine neurological outcome after cardiac arrest, as well as the clinical signs and tests used to fashion a prognosis out of the uncertainty that follows ROSC. Extensive tracts have been written by experts on this topic, and this short summary is written mainly to digest that wisdom into an easily digestable slurry. For an indication of what is expected in the CICM Part II, the reader is invited to explore past paper questions on this topic.
They mainly come in the form of "list 'em" questions, where the candidate is expected to produce a table of advantages and disadvantages for all the possible prognostic findings in the cardiac arrest survivor.
The literature which covers this topic is vast, and publications rapidly become outdated. For example, a review of the evidence was been published in Neurology (2006) by the American Academy of Neurology, which seems perfect for these questions because it outlined the main factors and prognostic features which influence neurological outcome after cardiac arrest. However, most of these studies were conducted in the pre-hypothermia era. One Seattle study which is often quoted aggregates data from 1977 to 2001! Who knows what those EMS hippies were doing in the 70s. This 2006 statement was then superseded by the ERC/ESICM statement (Sandroni et al, 2014), which incorporated a lot of hypothermia era data. Sandroni et al (2018) was a modernised revision of the same material, and ERC/ESICM issued their take on prognostication as a part of their 2021 guidelines on post-resuscitation care. Unfortunately, incorporating hypothermia-era data is no longer an attractive characteristic of society guidelines, as following TTM and TTM2 professional opinion regarding hypothermia had cooled somewhat, effectively ending the hypothermia ice age and returning us to basically the same position that we were in in the early 2000s.
In order to simplify revision, a tabulated summary is offered here:
|Predictive sign or investigation||Predictive utility||Confounding factors|
|Absent pupillary reflex||
0% false positive rate at 72 hours, irrespective of cooling
|Absent corneal reflex||0-15% false positive rate at 72 hours|
|Extensor motor response, or worse||May be associated with poor outcomes||
|Myoclonic status epilepticus||Persisting myoclonic status epilepticus has a 0% false positive rate within the first 24 hours||
|Somatosensory evoked potentials:
absence of the N20 component
|Absence of N20 predicts poor outcome with a 0% false positive rate.
Presence of N20 does not rule out a poor outcome.
N20 responses may disappear on repeat testing.
N20 responses may reappear, but this does not suggest a good prognosis.
|Burst suppression on EEG||May be associated with poor outcome|| Poor predicitive value;
cannot be used for prognostication.
|Absence of EEG reactivity, or "malignant" EEG pattern||Low false positive rate (0-5%)||Confounded by sedation|
|Neuron-specific enolase||NSE over 33μg/L at 1-3 days post CPR predicts poor outcome with a 0% false positive rate||
NSE may be elevated for reasons other than brain injury; for instance, it may be secreted by neuroendocrine tumours
|CT brain||On CT, an inversed gray/white matter ratio in Hounsfield units was found in patients who failed to awaken after cardiac resuscitation. However, the predictive value of CT findings is not known||
If performed too early, the CT may not demonstrate any findings.
Before launching into a detailed discussion of each predictive index, it is interesting to compare them together in a group. The numbers in the table below were derived from a relatively recent meta-analysis (Golan et al, 2014) which seems to be widely quoted (and which seems to form the basis of prognostic recommendations made by various society guidelines, for instance the new Canadian Guidelines).
|Predictive Test||False Positive Rate in % (and 95% CI)|
|Bilateral absence of pupillary reflexes more than 24 hours after ROSC||2% (1-6%)|
|Bilateral absence of somatosensory-evoked potentials between days 1 and 7 after ROSC||3% (1-7%)|
|Bilateral absence of corneal reflexes more than 24 hours after ROSC||4% (1-9%)|
|Myoclonic status epilepticus||5% (2-11%)|
|Unfavourable EEG patterns||7% (4-12%)|
|Motor score showing extensor posturing or worse||9% (6-13%)|
|Neuron-specific enolase >33mcg/L||12% (6-23%)|
The following recommendations are based on the 2021 ESC/ESICM statement:
Historically, the recommendations offered by the college in their answer to Question 20 from the first paper of 2015 also asked for the prognisticator to wait for a specific time period to pass in order to conclusively exclude the confounders (i.e 72 hours following the return of normothermia). These recommendations seem to be based on the multimodal prognostication algorithm from the then-recent 2014 statement. The situation is confused by the fact that many testing modalities have their own ideal time window. Neuron-specific enolase can be tested at 24 hours following ROSC, EEGs and SSEPs can be performed during hypothermia (i.e. 24-48 hours post ROSC) and most of the studies of clinical examination findings waited until 72 hours after ROSC. Given that the 2014 consensus statement repetitively recommends the simultaneous use of multiple markers, one is almost obliged to wait three days. The fundamental reason for the delay remains the same, even if you are not intending to cool the patient: it gives society the assurance that all mind-altering and paralysing substances have been cleared, and the patient's true neurology is being assessed.
All of these prognostic indicators are subject to some bias which is inherent in any scenario where the physician has an irresponsible amount of unilateral power over the decision to withdraw life-sustaining treatment. One must consider the possibility that early prognostication (i.e. 72 hours post ROSC) may produce clinical signs that give an inappropriately grim impression of the patient's neurological performance, and generate a self-fulfilling prophesy where the intensivist and family decide to stop life-sustaining treatment, reinforcing the value of those clinical signs as negative prognostic indicators. From this, one might conclude that logically, we should perform prognostic tests much later, to capture the maximum number of neurologically intact survivors. Sure, one might think, some of those people just need a bit more time?
Unfortunately, it appears that waiting longer to form an opinion does not identify many such patients. In support of this statement, we can look at Cronberg et al (2020) who quoted some data from countries where withdrawal of life-sustaining treatment is uncommon for cultural reasons. For example Kim et al (2016), reporting on a case series from Seoul, presented data from 279 patients of median age 62 who survived their out-of-hospital cardiac arrest. The group were pretty dismal to begin with (only 14.9% had a shockable rhythm, only 35.8% had bystander CPR), but the entire cohort survived long enough to be admitted to hospital. Over the subsequent month, 69.9% of them had died. Of the survivors, 57.2% were graded CPC 3 or CPC 4 (conscious with severe cerebral disability, or persistently comatose). In the subsequent six months, none of the CPC 4 patients and only two (4.1%) of the CPC 3 patients had demonstrated any improvement (up to CPC 2, "conscious and alert with moderate cerebral performance"). Of the two CPC 3 patients who graduated to CPC 2, both were young, had VF as the initial rhythm, and had less than 4 minutes of prehospital CPR. Something very similar was found by Scarpino et al (2019). In short, cardiac arrest survivors have a very early plateau of neurological recovery, and are unlikely to have any major functional breakthroughs unless they were spectacularly robust to begin with.
Thus, from the existing data it seems that normal early neuroprognostication (at 72 hours) leads to the selection of neurologically robust survivors. For example, Dyson et al (2019) reviewed twelve international cardiac arrest registries and found that, even though the overall survival to discharge was only 10%, of the survivors 80% had good neurological performance (CPS 1). It is hard to say how many of those who died had an elective withdrawal of therapy; generalising from Dragancea et al (2013), it would have been something like 30%. In that (admittedly small) cohort, there were very few severely disabled survivors, none of whom were persistently unconscious. In short, it appears that the current model of prognostication at 72 hours interrupts a process that would otherwise produce a population of heartbreakingly disabled persistently comatose loved ones.
The initial rhythm is the strongest predictor of mortality and morbidity.
VF arrest: This has the best prognosis. 69% survive the ambulance ride, and 40% survive until discharge from hospital. With cooling, roughly 49% of the survivors were either neurologically intact or suffered minor deficits.
Pulseless electrical activity: This has a poor prognosis. 23% survive the ambulance ride, and 11% survive until discharge from hospital. Of those 11%, roughly 80% were either neurologically intact or suffered minor deficits.
Asystole: This has the poorest prognosis. 10% survive the ambulance ride, and 2% survive until discharge from hospital. Of those 2%, roughly half (55%) were either neurologically intact or suffered minor deficits. None of the survivors were over 70 years old.
According to the Seattle series, you have a much better chance of survival if your cardiac arrest is witnessed, and occurs in a public place. The commencement of bystander CPR is the key feature here. No matter how useless the bystanders, any CPR is probably better than no CPR.
Interestingly, being in a hospital is not especially protective. Of patients suffering in-hospital cardiac arrest (of all aetiologies) 49% had a restoration of spontaneous circulation and 15% survived until discharge; in another study 27% of suvivors had "good neurological function".
Duration of time before CPR is commenced could be important. In one study of in-hospital arrest a delay of over 5 minutes was strongly associated with mortality. Duration of CPR itself could also be important. In the same study, CPR lasting longer than 20 minutes was strongly associated with increased mortality.
However, the AAN recommended that CPR duration be left out of the predictions regarding neurological outcome, as these variables do not discriminate accurately between those patients who will be neurologically damaged and those who will not. In their own words, "prognosis cannot be based on the duration of CPR".
Age is a weaker predictor than the initial rhythm.
Still, the older you are, the less likely you are to survive until discharge. The large-scale Seattle series calculated that the odds of survival decrease for every year of age, with an odds ratio of 0.97 per year. Mortality in one series was ~ 94% for the over-80s, and ~68% for the under-45s. However, age did not seem to be predictive of neurological outcome - only mortality.
Comorbidities are a negative prognostic indicator. Conditions such as CCF, NIDDM, COPD, AF, hypertension et cetera- all of these adds a certain risk of poor outcome, with an odds ratio of 0.84 per condition.
Apart from the abovementioned 2014 society statement, the best reference for prognostically interesting clinical examination findings seems to be this 2013 paper by Greer et al.
Immediately after a cardiac arrest the pupils can be fixed and mid-size (or dilated) for a whole variety of reasosns, and this should not be used to prognosticate. However, at 72 hours following ROSC a bilateraly absent pupillary reflex has a 0% false positive rate in predicting poor neurological outcome, irrespective of whether the patient was cooled or not.
At 72 hours post arrest, absent corneal reflexes have a false positive rate of only 5% in identifying poor neurological outcome. However, having a normal corneal reflex does not predict a good outcome.
Myoclonus status epilepticus (the repetitive, "unrelenting" jerking of the face and body) has been reported previously to have no association with prognosis; however the AAN reported that within their meta-analysis this feature (when present on day 1 post arrest) was strongly associated with poor neurological outcome even in the presence of brainstem reflexes or some motor response. The false positive rate was also 0%, which means you can use this early marker as a warning of poor outcome. This recommendation is upheld by the most recent 2014 psotion statement. Status myoclonus must begin within 48 hrs following ROSC in order to be avalid marker, and must be status myoclonus rather than intermittent jerks. The prognostic utility of this findign is tarnished somewhat by several case reports of patients who have recovered well after having early myoclonus.
Back in the days of "warm" arrest, the AAN strongly supported absent motor responses at 72 hours as predictors of poor outcome. The motor component of the Glasgow Coma Scale does not yield false positives after 72 hours, they said. Absent motor response or extensor responses were associated with poorer neurological outcomes: only 1% of such patients can hope to be independent. In short, you want to obey commands, localise to pain, or at least withdraw from it (63% of such patients went on to have a good neurological recovery, with some degree of independence).
However, in the hypothermia era the 2014 society statement has downgraded their support for this finding. More recent evidence has demonstrated that motor score has a very high false positive rate (almost 50% in some studies!). For example, in a post-hoc analysis of the TTM trial (33°C vs 36°C) performed by Dragancea et al (2015), a motor score of 2 or worse had a 19.6% false positive rate. Perhaps it was better in the pre-cooling era? The consensus is to only use this finding together with other, "more robust" predictive markers.
The 2021 ESC/ESICM statement describes a series of "malignant" EEG findings associated with poor outcome. The following list features links which open an image of the mentioned EEG pattern from the excellent hypoxic brain injury article by Cronberg et al (2020).
Additionally, the 2014 AANA statement has identified the following EEG findings which were associated with a poor outcome and had a false positive rate of around 0-7%:
In general, ICU trainees are often seen to scoff at the EEG in these patients, because they usually return a report of nonspecific and unhelpful-seeming findings ("diffuse cerebral dysfunction" again, blergh). This attitude is consistent with the historical 2006 AANA guidelines which suggested that EEG in general had poor predictive value in cardiac arrest survivors, not to be used for prognostication. With the addition of newer data, the opinions of the intensive care community (including the CICM examiners) have subsequently changed in favour of EEG (especially early EEG). For example, the model answer to Question 20 instead reports that "generalised suppression, burst suppression or generalised periodic complexes strongly associated with poor outcome".
Interestingly, burst suppression is actually compatible with a satisfactory neurological outcome. Low voltage EEG tracing in patients who are not being cooled also has a low false-positive rate, but the voltage magnitude of the EEG can be influenced by hypothermia and is therefore not very useful. In general, criticism of EEG in this setting (and of its use in ICU in general) is that experienced interpreters have very high confidence in their EEG interpretations, but low inter- and intra-rater reliability.
In 2006, the AANA recommended this as a more useful tool than the EEG. So interesting is this modality, that CICM made it the topic of an entire 10-mark SAQ (Question 11 from the second paper of 2014). The absence of the N20 component with median nerve stimulation accurately predicts poor outcome, but its presence does not rule out poor outcome - many patients without meaningful recovery will have normal N20 responses.
SSEPs are not confused by therapeutic hypothermia; in cooled patients bilateral absence of N20 SSEP is able to accurately predict a poor outcome with a false positive rate which approaches 0%. There are isolated case reports of false positive prediction, which are tarnished by allegations of poor tracing quality and interpretation error.
Overall, the new 2014 recommendations are as follows:
NSE over 33μg/L at 1-3 days post CPR predicts poor outcome with a 0% false positive rate and this does not seem to be affected by temperature control (Stammet et al, 2015). It may be more useful later rather than earlier. There is disagreement as to the precise threshold, but generally people agree that serial NSE levels over 60mcg/L are "very rarely associted with a good outcome". The 2021 ESC/ESICM guidelines do not specifically note any cut-off value, mentioning only "increasing values between 24 and 48 h or 72 h in combination with high values at 48 and 72 h".
S100b is expressed primarily by astrocytes; its presence in the bloodstream indicates some sort of glial damage. Threshold values differ study to study, and every author seems to b using a different measurement technique. Though potentially promising, this biomarker needs experiments with better standards before anybody is able to recommend it.
Question 20 from the second paper of 2022 asked the candidates to list CT and MRI features of severe hypoxic ischaemic encephalopathy. This is a topic large enough to require its own chapter, and what follows is an abbreviated list of findings and their value:
CT features of hypoxic ischaemic encephalopathy:
MRI features of hypoxic ischaemic encephalopathy:
Value of these features in neuroprognostication:
The following recommendations were made by ESC/ESICM in 2014: