Hi, I am Dr. Sterling Shepard, a neurocritical care attending at Piedmont Atlanta Hospital in Atlanta, Georgia, and I am going to speak to you today on neuroprognostication after cardiac arrest. I have no relevant disclosures. I would like to acknowledge Dr. Sharma, one of my colleagues, who graciously granted some of his slides. I am going to first discuss the epidemiology of cardiac arrest, discuss the need and history of neuroprognostication, and then go over the new 2023 Neurocritical Care Society guidelines for neuroprognostication in comatose adult survivors of cardiac arrest. So in the U.S., about 89 out of every 100,000 patients experience an out-of-hospital cardiac arrest, according to a cardiac arrest registry from 2020. Only 24% of out-of-hospital cardiac arrest patients survive to hospital admission, and 9% survive to hospital discharge. Of those who survive to discharge, 79% actually will have a good neurological outcome. How about patients who have a cardiac arrest in the hospital? The American Heart Association reported that about 10 out of every 1,000 hospital admissions, you know, those patients have a cardiac arrest, and 27% of in-hospital cardiac arrest patients survive to discharge, with 80% having a good neurological outcome. So from both groups, in-hospital or out-of-hospital, if you survive until discharge, you will more than likely have a pretty good neurological outcome. So how does a cardiac arrest affect the brain? You can think of a cardiac arrest as a lack of blood flow to the entire brain all at once, which we call a stroke. So cardiac arrest is basically a big, bad stroke to the entire brain. Different parts of the brain have different susceptibilities to hypoxia, and the brainstem is relatively tolerant to ischemia compared to other parts of the brain. On the right, you can see that the duration of global brain ischemia is very important. The longer the time that goes by where, you know, your brain is not receiving enough oxygen, the more severe deficits a patient will have, leading all the way potentially to brain death. So analytic brain injury can be seen on CT scans and MRI. Our top image is a CT scan of the brain. On the left, you can see a relatively normal CT scan, while on the right, you can see that this brain is swollen. There's loss of gray-white differentiation, and there's also suckle effacement, which is evidence of global cerebral edema as a result of hypoxic ischemic injury. On the left, we have an MRI showing diffusion-weighted imaging and its corresponding ADC component. And you can see on diffusion-weighted imaging that the entire brain is brighter than it should be. It is hyper-intense, and we call this sometimes cortical ribboning. And on ADC, you can see there's corresponding hypo-intensity. So this question I'll pose to us all. In the United States, after a cardiac arrest, what is the most common cause of death in this patient population? And the most common cause of death after a cardiac arrest is actually withdrawing care. It is withdrawal of life-sustaining therapy. It occurs in about 40% to 80% of comatose survivors, according to numerous studies. So as clinicians, we should be aware of the self-fulfilling prophecy of withdrawal of life-sustaining therapy. When we're giving advice to families, to loved ones of those who've suffered a cardiac arrest and are presently comatose, we should approach it with facts and data in terms of, you know, according to that particular patient's imaging, EEG, et cetera, do we think that there's an overall good prognosis, including, of course, their clinical and neurological examination. So the concept of neuroprognostication was not even a thing before the 1970s, right? In the 1970s, ventilators came into vogue. So before that, these patients didn't have ventilators. They would either live or die. There was no sustaining of their life. In the 1970s, after the advent of ventilators, patient autonomy actually became an important concept. You know, should we continue to keep, you know, the heart going, organs going, breathing going in patients who seem to have a poor overall prognosis and are unlikely to ever wake up again? Plum and Posner, also in the 1960s, published their textbook, Diagnosis of Stupor and Coma, which really brought to the forefront, for the first time, a good description and understanding of coma itself. EEG was also gaining credibility as a bedside neurophysiologic tool around this time. And in the United States, there was a legal case that was pretty well known in the media at that time. It involved a young woman, Karen Ann Quinlan, and her parents. She was comatose on a ventilator. Her parents were trying to take her off of a ventilator, which at that time was illegal. And it led to the right-to-die law, where, you know, families and surrogate decision makers could choose to, you know, remove life-sustaining therapies if the overall prognosis was not aligned with their goals of care. So why is neuroprognostication important? It does establish realistic expectations for families and surrogate decision makers. It determines resource mobilization, and it helps with risk stratification and decisions for therapy and clinical trials. It's very important also for clinical intervention trials, you know, because withdrawal of life-sustaining therapy would preclude appreciation of specific treatment effects. If we withdraw care before a patient will wake up, we will always get the same sort of outcome, which is death. It's even more important in our younger patient population, but I will not go over that in this talk. And shared decision-making is important in the world that we live in, where we believe in patient autonomy. There's often a divergence in terms of what us as clinicians believe is a good outcome versus what surrogate decision makers may believe is a good outcome. The first paper looking at predicting overall outcome from hypothesis ischemic coma was published in JAMA in 1985, based primarily on the clinical examination itself. In 2006, the American Academy of Neurology published a process parameter to help assist us all in terms of neuroprognostication after cardiac arrest. In this guideline, you know, if a patient is comatose, we first would exclude any major confounders. If there are no brainstem reflexes at any time, then, you know, you would proceed with brain death testing. They recommended that if on the very first day after cardiac arrest, the patient went into myoclonic status epilepticus, then this suggested a poor outcome. They looked at neurophysiologic testing, such as SSCP. If there were absent N20 responses, then this suggested a poor outcome. If serum neuron-specific enolase was greater than the specific value, then this suggested a poor outcome. And if by, you know, 72 hours after the cardiac arrest, there were still absent pupillary or corneal reflexes, or if you had extensor or absent mobile responses, this was also suggested a poor outcome. And the only other outcome that we could recommend as neurologists was an indeterminate outcome. There were several criticisms of the 2006 AAN review, especially as medicine has advanced and treatment of post-cardiac arrest patients has advanced. It was actually based on studies conducted before the advent of therapeutic hypothermia or normothermia for post-resuscitation care. You know, therapeutic hypothermia itself could also interfere with the clinical examination and prognostication indices. And the predictive value of these indices really needed to be re-evaluated in an era where a lot of patients are either normothermic or slightly hypothermic after cardiac arrest. And then also these thresholds for biomarkers as a serum neuron-specific enolase, there was no specific value found in studies even before the review. So a lot of neurologists and other scientists were criticizing this specific value of 33 in terms of the serum neuron-specific enolase. Criticisms for prognostic tools such as EEG and imaging studies were also limited at the time of this review and needed some re-evaluation. And the AAN 2006 review did not adequately address limitations of prognostication studies, including the risk of self-fulfilling prophecy and withdrawal of life-sustaining therapies. In Europe, in 2014, due to this overall criticism, the European Society of Intensive Care Medicine also created their own algorithm for neuroprognostication that was being used in Europe since 2014, where they based their overall neuroprognostication on the clinical examination, also on neuron-specific enolase levels if the patient went into stasis myoclonus, but also included EEG findings and CT and MRI findings for the first time. This led to the Neurocritical Care Society, which actually is a global society with members from countries all around the world, leading to a work group to create a new guideline for neuroprognostication in comatose adult survivors of cardiac arrest that could be used around the world, especially, you know, considering EEG, CT, MRI findings, and then also addressing different biomarkers. So in this guideline, they first started with establishing the variables that they wanted to look at. They decided on these 11 variables. Age, you know, does a patient's age affect overall prognosis? You know, will a 21-year-old do better than a 90-year-old, assuming everything else is the same? Initial cardiac rhythm, does this make a difference? The time to ROSC itself, does that make a difference? You know, we believe it does, but has it been shown to make a difference in various studies? The motor response, you know, the patient's clinical examination, is there a pre-priority light response? Is there myoclonus? And then they incorporated for the first time our CT and MRI findings, EEG findings. Since neuron-specific enolase and N20 waves have been mentioned both in the American Academy of Neurology guidelines and the European Society of Critical Care Medicine guidelines, they also decided to look at those two variables. They also looked at these clinical prediction models, the out-of-hospital cardiac arrest model, the cardiac arrest hospital prognosis model, and good outcome following attempted resuscitation. And every variable was assessed in the form of a PCOS question, looking at population, the intervention, a comparator, outcome, timing, and setting. So when counseling surrogates of comatose adult survivors of cardiac arrest, should age, should the MRI findings, should the neuron-specific enolase value, et cetera, be considered a reliable predictor of whatever that outcome is, a good outcome, a poor outcome, an indeterminate outcome, within what time frame? And they decided to assess their outcomes at three months or later. They first started with over 3,500 records and studies and wrote it down to 72 studies that were eligible to support recommendations. This slide just goes over how we grade guidelines in general, which you can review. They also looked at cerebral performance categories, which range from 1 to 5, 1 being minor or no disability and 5 being death. Before even assessing, you know, prognosis overall, they decided that they would issue sort of good practice statements to all of us as clinicians. The first statement was that they actually recommend deferring any sort of neuroprognostication in comatose survivors of cardiac arrest for at least three days, following resuscitation of spontaneous circulation. In patients not treated with therapeutic hypothermia, so if you have a patient that comes in, they're not cooled, then you wait three days before you even give a prognosis. If the patients are cooled and then rewarmed, then you need to wait three days from that period of time, which, you know, will be about day four or five into the hospital stay. However, persistence of coma beyond this period of time should not be automatically equated with a poor neurological prognosis, based on several new studies. 10 to 20-ish percent of all cardiac arrest survivors in the ICU not treated with hypothermia actually will wake up more than 72 hours from ROSC, that's about 1 in 5, and also about 1 in 5 patients treated with hypothermia will wake up more than three days from rewarming. There was a study in Taiwan published in Critical Care Medicine in March 2022 that pretty much has changed our entire approach to this observation period for neurological recovery after cardiac arrest. In this study, they found that nearly 10% of cardiac arrest patients had recovery of consciousness with a favorable outcome more than a week out after ROSC. So if we went, you know, based on the original American Academy of Neurology guidelines, you know, the longest period of time in those guidelines pretty much would be three days, at which time we either say it's a poor prognosis or an indeterminate prognosis. So this study actually changed the length of time that it's recommended that we observe before giving families a prognostication value. Of course, you want to make sure that, you know, in your prognostication, you sort of removed any other confounders, such as sedation, you know, if the patient's in liver failure and renal failure, you want to make sure that you give those patients sufficient time for these substances to be cleared. They also recommended that we don't only consider the brain. You know, if you have a cardiac arrest patient that also has end-stage renal disease, also has liver failure and needs a liver transplant, also has cardiac failure and needs a heart transplant, et cetera, that these things do impact overall prognosis and should be considered, including, you know, baseline level of functioning. Is the patient already bed-bound before this has occurred? Are there any pre-existing illnesses? Is there terminal stage four cancer with mass to the brain? Is there multi-organ failure? All of these should be considered so that there's a holistic picture in the approach and what we tell families. In today's world, you know, it is 2024. They recommend that with the different modalities that are now present, that were not present in the 1980s or 90s or the early 2000s, universally in a lot of hospitals, at least in the developed world, that, you know, this assessment be multi-modal and that we do not depend on only one clinical scenario and one variable in terms of, you know, what the patient is going to have to go through, you know, in order to be able to make a decision about whether or not they're going to have a heart transplant, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, et cetera, recovery can occur, may occur, but may take several days to several months to maybe never. And based on particularly this study in Taiwan, they also recommended that an extended period of observation, if the patient has an indeterminate prognosis, be recommended if that is aligned with overall goals of care. Meaning, you know, you can say, I don't know if they will wake up. However, we'll need a tracheostomy, we'll need a feeding tube, we'll need to go to a long-term acute care hospital. Are these things aligned with your overall goals of care and what your loved one would want? So let's look at the first variable. So in these guidelines, age actually did not come out as a reliable predictor of poor functional outcome assessed at three months on its own. You know, we do have this overall feeling and that a younger patient will do better than an older patient, but that by itself did not predict a poor functional outcome. The initial cardiac rhythm also from these studies, they recommended be not considered a reliable predictor of poor functional outcome as well. How about time to ROSC? As stated, we do know that the longer the time period that the brain is not receiving oxygen, the higher the likelihood of severe disability and even brain death. However, in these studies, they found that time of return of spontaneous circulation alone should not be considered a reliable predictor of poor functional outcome either. This sort of makes sense because every patient's different. You know, some people have really good sort of collateral brain compensation. You know, we don't know actually in this time to ROSC the quality of the CPR. So they found that looking at time alone actually did not help us either in establishing what is a poor outcome. How about our examination more than three days out if the patient does not have a pupillary reflex? This was stated in the European trial and the European guidelines and also the American Academy of Neurology guidelines and also in the Neurocritical Care Society guidelines. This was found to be a reliable predictor of poor functional outcome. So if, you know, day three from ROSC patients do not have any pupillary response, they have blown pupils, then this is a poor outcome. How about the lack of corneal reflexes more than or equal to 72 hours out from ROSC? This was also found to be a reliable predictor of poor functional outcome. So, so far, you know, your pupillary light response if it's absent and the corneal reflexes if they're absent are both predictors of poor functional outcome. How about our MOHA response? You know, the American Academy of Neurology guidelines did suggest that if patients were extending then this was an overall poor prognosis. However, they found that this was not the case in more recent studies and recommended that absent or extensor MOHA responses alone not be considered a reliable predictor of poor functional outcome. I remember they started out by saying they want you to take a multi-modal approach and not just establish a poor outcome based on one factor itself. How about patients who have myoclonus less than or equal to 48 hours from ROSC? This also was not found to be a reliable predictor of poor functional outcome. So, as you can see, these new guidelines in 2023 have pretty much disrupted what even I was taught in medical school and residency in terms of what is a poor outcome after cardiac arrest. So, looking at CT brain findings for the first time, you know, since the U.S. guidelines did not really include that, they did find that if we waited more than 72 hours from ROSC and got a CT scan at day three, so often in practice I'll see, you know, there's a scan done at day one, it looks perfectly fine, or the next day after cardiac arrest it looks fine, and no one really thinks of repeating the CT scan if there is no improvement in the coma exam, but we should repeat a CT scan more than three days from ROSC, so that's pretty important, and this was considered a moderately reliable predictor of poor functional outcome. MRIs, so an MRI can be done between days two to seven from ROSC, and if we see, you know, a diffuse pattern across vascular distributions in the bilateral anterior and posterior circulation, there's involvement of the cerebral cortex, there's loss, there's restricted diffusion throughout, then this is considered a moderately reliable predictor of poor functional outcome. How about our EEG findings? You know, if the patient is not sedated, there are no confounding factors, and we do an EEG more than or equal to 72 hours from ROSC, this also, and find that there is a suppressed or birth suppressed background with or without periodic discharges, then this is a moderately reliable predictor of poor functional outcome, and remember they're looking at all of these variables alone, not together as a whole, but just alone. How about our SSEPs? So these are electrophysiologic tests, this is somatosensory evoked potentials, we look at absence of an N20 wave, so if there's absence of the N20 wave, and it's performed, this testing is performed at least 48 hours from ROSC, this is considered a reliable predictor of poor functional outcome. However, as we know, I mean, I've worked in several kind of large academic neurological and neurosurgical programs, and I have never done an SSEP in the patient population, they're not that easy to do, especially in our ICUs, but if you can get them done, then if the N20 wave is absent, then this is a reliable predictor of poor functional outcome. This just shows what is done during an SSEP N20 measurement, and this is the N20 wave, if this is absent, then it suggests a poor functional outcome. How about serum neuron-specific enolase, which was mentioned in the American guidelines, and also the European guidelines, they found that there is no established threshold, you know, we don't know the baseline serum level of neuron-specific enolase in anyone, so how can we say that a value greater than 33 is a poor functional outcome? They established that we cannot, so neuron-specific enolase is no longer recommended in suggesting that there is a poor functional outcome. They also looked at different prediction models, as I mentioned, the out-of-hospital cardiac arrest model, the cardiac arrest hospital prognosis model, and the good outcome following attempted resuscitation prediction model, and basically said that the only way to one that seemed to, actually, they said that, you know, none of these models seem to really help us. Okay, so this is just a table that summarizes some of these findings, which I have already discussed on the previous slides, like what is important in terms of suggesting what is a poor functional outcome. This led to an overall algorithm that has changed significantly from our previous American Academy of Neurology guidelines in 2006. It incorporates targeted temperature management. So now, you know, we have a patient that comes into the hospital, or they're already in the hospital, they have return of spontaneous circulation. On day one, you know, we would initiate targeted temperature management. We would not really prognosticate on day one. They recommended performing continuous EEG monitoring on day one, if available. If unavailable, we should at least perform a routine EEG to evaluate for any seizure activity. But you could go ahead and consult your friendly neurologist to follow, to ultimately give sort of neuroprognostication. On day two, you would begin rewarming if therapeutic hypothermia was used. And then on day three, make sure that you lean completely off sedation. And at this time, you could obtain a CT scan to see, you know, is there a diffuse rheumatoid edema, which just suggests a poor overall functional outcome, or an MRI. Or if you are, you know, extremely lucky, you could also order somatosensory evoked potentials at that time, if available at your hospital. If therapeutic hypothermia was not used, you're kind of on day four, or day three, day four, then you could do these CT findings. Of course, your clinical examination is still very important. Is there a pupillary light response? Are there any coronary refluxes? And you can consider EEG findings as well. And of course, if you have a pupillometer, which is a device that takes away the subjective component of the pupillary light response, as we know, if I look at, if I shine a light into the pupils, I could say, you know, pupils are three milliliters, minimally reactive. The nurse may say they're five milliliters and not reactive. And the pupillometer is a device that will tell you the exact size, and if the pupils are reactive or not, and take away that human subjective component. This, again, is the same sort of algorithm. However, for the first time, they decided to include what is a good overall prognosis. So return of spontaneous circulation, is the patient following commands at any time? If that's a yes, then it's a good prognosis to awaken from coma. If the patient meets all neurological criteria for brain death, and you know, there are new brain death or death by neurological criteria guidelines that has pretty much changed how we approach death by neurological criteria. If we follow those guidelines and check all of our boxes, and patient meets the criteria, then we can declare the patient dead at that time. If the patient does not meet all neurological criteria for brain death, then are there any predictors of good outcome? Is the patient withdrawing or localizing? That actually was found to be a predictor of good outcome. If we do an MRI, are there no diffusion-weighted imaging lesions, no restricted diffusion, or is it only small or isolated? And then that could actually be a good outcome as well. Do you have a reactive background in your EEG in a comatose patient? Then that is a good outcome. And if you're able to do SSCPs, and the amplitude is above a certain value, then that also is a good outcome. How about predictors of poor outcome? If these things are absent, so you know, our patient is not withdrawing or localizing, the MRI findings show the few restricted diffusion, or EEG suppressed, then, you know, that does lead us to the column that says poor outcome is very likely, but some uncertainty exists. And what is called a moderately reliable predictor of poor outcome, are CT scan findings, or MRI findings, or EEG findings. You know, if they show diffuse gray-white differentiation, loss of gray-white differentiation, cell co-defacement, you know, our brain lights up everywhere, the EEG is very suppressed, then the poor outcome is likely, but significant uncertainty exists. Indeterminate outcome is still also a suggestion, just following this algorithm. So these are actually the first guidelines to also focus on the good. And as I already mentioned, these are the things that actually can predict a good outcome. So for the first time in our notes, we can say that we think there will be a good outcome, or there will be a poor outcome, or there will be an indeterminate outcome. Okay, so I'm going to go through some cases. A 65-year-old man with a history of hypertension, diabetes, was admitted to the ICU following a witness out-of-hospital ventricular fibrillation cardiac arrest. CPR was initiated within three minutes by bystanders. Total downtime was 15 minutes. ROSC was achieved. Post-resuscitation care included targeted temperature management. And then on day two, post-ROSC, patient is still sedated, not being actively cooled. The patient remains comatose with a GCS score of 3T. And you are working with a neurologist on call, and you're consulted by the cardiology team to give prognosis. According to the guidelines, how long should clinicians wait before attempting neuroprognostication in a comatose patient in a comatose patient post-ROSC when not using hypothermia? Should we wait 24 hours, 72 hours, 48 hours, immediately after there's return of spontaneous circulation? And the answer, of course, is 72 hours. So you would tell your cardiology colleagues, sure, I can come see the patient. I can write a prelim note. However, I cannot prognosticate today. So then you return to examine the patient on day three, and the patient's GCS is 3T. Which of the following is considered a reliable predictor for poor functional outcome when assessed correctly? Is it the patient's motor response? Is it bilateral absence of the pupillary light response? Is it an EEG that shows generalized slowing or age alone? And the answer, of course, is the bilateral absence of pupillary light response. We'll go to our second case. There's a 70-year-old woman. She has a history of chronic kidney disease and atrial fibrillation. She has an in-hospital PEA or cardiac arrest, and there's ROSC after 20 minutes of CPR. She underwent targeted temperature management for 24 hours, and now it's day four post-ROSC. She still is comatose, and her GCS is 4. She's currently on renal replacement therapy, CRT, and has received sedatives. Which of the following is crucial before performing neuroprognostication in this patient? A, waiting for sedation effects to wear off. B, ensuring renal function is stabilized. C, both A and B. D, immediate neuroimaging regardless of confounders. And the answer, of course, is C. We should wait for sedation effects to wear off, and we should ensure renal function is stabilized. We should not just jump to a CT scan or an MRI and base our overall neuroprognostication on that alone when we don't even have a clinical exam. Remember, the first step is always our neurological examination. In the presence of confounding factors, which approach enhances the accuracy of neuroprognostication? Is it still reliance on clinical examination? Utilizing a multimodal approach, including EEG and MRI? Immediate withdrawal of life-sustaining therapy based on our initial impressions or our initial examination? Reliance solely on patient age and pre-existing conditions? And the answer is that we should use a multimodal approach, including EEG and MRI. If, you know, we wait, we do a clinical examination, but we're still concerned that there may be confounding factors, we can, you know, use our exam plus our imaging findings plus our EEG to come up with an overall neuroprognostication. So, just some take-home practical applications. You know, we should involve family and surrogate decision makers early in discussion regarding goals of care. Goals of care, you actually want to sort of prep your families. So, even on, you know, the first day or the second day of the hospital admission, you can sort of prep them in terms of what will come in terms of overall prognostication and not being sure yet how to predict, but things that you can do and will do to try to help guide an overall decision making. You know, after neuroguidelines, we actually institute more normothermia these days and hypothermia. Also, if patients seem to progress to brain death, families should also be prepped about brain death and what it is before we even do that, according to the new brain death guidelines published earlier this year. We should always aim to provide continuous EEG monitoring in all patients after cardiac arrest or comatose. If we don't have that available, we should do routine EEG because these patients have a high likelihood of developing seizures and non-convulsive seizures. And there should also be a multidisciplinary and multispecialty approach to coherent decision making. You know, we often get these consults from the medical ICUs, from the cardiac ICUs, about overall neurological prognosis. While the patient is having liver failure, renal failure, you know, we have a poor neurological examination, but we can look at these things collectively as a whole and not the brain alone itself. And of course, we should involve our case managers and care coordination teams early. All right. Thank you, everyone, and have a great day.

neuroprognostication

cardiac arrest

2023 guidelines

multimodal assessment

prognostication timing

neuron-specific enolase

holistic care