When can Continuous EEG Help Tackle the Uncertainty of Prognostication After Cardiac Arrest?

Abstract

Electrographic and Clinical Determinants of Good Outcome After Postanoxic Status Epilepticus.

Admiraal MM, Backman S, Annborn M, Borgquist O, Dankiewicz J, Düring J, et al. Neurology. 2025;104(5):e210304.

Background and objectives: Postanoxic electrographic status epilepticus (PSE) affects up to a third of all comatose patients after cardiac arrest (CA) and is associated with high mortality. Late PSE onset (>24 h), from a restored continuous background pattern, and absence of established indicators of poor outcome at multimodal prognostication are described in survivors. We aimed to determine the increase in probability of good long-term outcome after PSE in patients presenting with this favorable PSE profile compared with all patients with PSE. Methods: This is a prospective observational substudy of the International Targeted Hypothermia versus Targeted Normothermia After Out-of-Hospital Cardiac Arrest trial (TTM2-trial, 2017-2020) including adult comatose patients resuscitated from CA with continuous electroencephalogram (cEEG) monitoring. EEG background pattern and type of PSE were determined using standardized EEG terminology of the American Clinical Neurophysiology Society, blinded to clinical data. On day 4, multimodal prognostication was performed according to the European postresuscitation guidelines. Good outcome was defined as a modified Rankin Scale score of 0 to 3 at 6 months. Detailed follow up was performed at 6 and 24 months. Results: A total of 191 patients were monitored with cEEG, of whom 52 (27%) developed possible or definite PSE at a median of 42 h [interquartile range (IQR) 32–46] after CA. The median age was 70 (IQR 63-77) years, and 35% were female. Favorable PSE profile was present in 20 patients (38%), of whom 12 patients (60%) survived until 6 months and 8 (40%) had good outcome; thus, the probability of good outcome increased 2.7 times. All patients lacking a favorable PSE profile had poor outcome. All patients with good outcome obeyed commands within the first 7 days. At 24 months, all 12 survivors were still alive and 7 had good functional outcome. Detailed follow up at 24 months showed that most had only mild cognitive impairment and overall life satisfaction was similar to the general population. Discussion: PSE is compatible with good outcome when onset is late and from a continuous background and no established indicators of poor outcome are present. One-third of patients with PSE had favorable PSE profile, of whom well over a third eventually had good outcome and showed improved level of consciousness within the first week.

Commentary

Does the epileptologist play a crucial role in managing and prognosticating postanoxic brain injury? ¹ The standard of care in most countries includes continuous electroencephalogram (cEEG) recordings for a variable period of time after cardiac arrest, ² but the use of this data for the counseling of caregivers during prognostication discussions remains incompletely understood. While certain patterns (eg, burst suppression and unreactive suppressed background ³ ) are known to be associated with poor outcomes, the predictive value of cEEG has typically been analyzed independently of clinical, biomarker, and neuroimaging features. ⁴ So the question remains: What is the independent value of cEEG? And more specifically, what additional information can postanoxic status epilepticus (PSE) give us? The answer likely depends on how we scale a clinically meaningful effect in this patient population for which key decisions must be made quickly and with a high level of certainty.

The investigators of the TTM2 trial, an international open-label clinical trial comparing the impact of normothermia versus hypothermia on mortality after cardiac arrest, ⁵ performed a nested observational study examining the association between patterns possibly and definitely consistent with PSE and functional outcome. Typical standard of care practices using cEEG data for clinical management were undertaken and the site investigators were not aware of the nested study on the impact of PSE on outcomes. The authors hope that this relative blinding translated into a minimal “self-fulfilling prophecy” effect, which is problematic in the majority of the postanoxic brain injury literature. ⁶ Still, one must imagine that investigators used existing assumptions about the impact of PSE on outcome in management and counseling of families.

In this observational substudy, ¹ 191 patients monitored with cEEG were included. About one-third (n = 52) developed possible or definite PSE, which the investigators defined in line with ACNS guidelines ⁷ of interictal-ictal continuum (IIC) and electrographic PSE, respectively. One strength of the study was the uniformity of sedation use, which was per protocol until 40 hours after randomization, after which sedation was weaned to allow for prognostication and only used again for a clinical indication. The investigators designed a 2-step process to predict outcome among the patients with PSE. Step 1 was performed in days 1 to 3 after cardiac arrest and consisted of categorizing a favorable PSE profile as onset of PSE >24 h after cardiac arrest, return of nearly continuous normal voltage background activity prior to the onset of PSE, and the absence of burst suppression with highly epileptiform bursts prior to PSE. Step 2 was then performed at the more typical prognostication timeline of 4 days or more after cardiac arrest, and included signs of improved consciousness or absence of typical predictors of poor outcome (combination of clinical, biomarker, computed tomography imaging, and interictal cEEG features) to label a patient in a favorable group. Fifty-two patients were then divided into 3 categories: favorable steps 1 and 2; favorable step 2, unfavorable step 2; and unfavorable step 1 regardless of step 2. Each category was then correlated with functional outcome at 6 and 24 months. They found that all patients in the latter 2 categories (ie, had classic signs of poor prognostication at the 4-day mark or an unfavorable early PSE profile) had poor outcome. Those with favorable steps 1 and 2 had a mix of good and poor functional outcomes. The authors performed a statistical analysis comparing the positive predictive value of each step (PSE early stratification and composite later features) and concluded that combining steps 1 and 2 improved the positive predictive value by a factor of 1.2 compared to step 2 (clinical standard of care) alone. Therefore, the addition of favorable PSE profile identified a few additional patients who eventually had a good outcome.

These results beg the question—How much more information does one need to glean from an additional piece of data to make it clinically impactful? Translating biostatistics into a family meeting in the neuro-intensive care unit forces one to quantify how much uncertainty one is willing to accept in life and death decisions. An ideal negative predictive value would be 100%—meaning one would not want to advise withdrawal of life support if there is even a small chance of good recovery. What about a clinically impactful positive predictive value—what chance of a good outcome does one need to counsel families that continuing aggressive measures may pay off in the long run? And lastly, what is a clinically acceptable confidence interval—that is, how much uncertainty in our prognostication counseling are we willing to accept? These questions would be best answered not just by clinicians, but by collaborative work with caregivers and survivors. Survivors with a good outcome did experience some level of disability and cognitive dysfunction, but rated their life satisfaction similarly to the general population. Interestingly, some members of the small group with poor functional outcomes could achieve some level of life satisfaction on par with that seen in the good outcome group. Therefore, the “acceptable” statistic on certainty of a good or bad outcome probably varies between patients, and the use of prognostication must be integrated with individual patient values during goals of care discussions.

Whether we choose to conclude that including early PSE data in our prognostication discussions with families can be helpful, many questions still remain. We continue to struggle with management decisions when it comes to definite and particularly, possible PSE. In the TELSTAR trial, all rhythmic and periodic patterns were suppressed in the treatment arm, and no differences were seen between the treated and untreated groups in their 3-month neurological outcomes. That study was underpowered to determine whether indeterminate patterns should be treated similarly to definite ones. ⁸ Analyses of individual patterns suggest that certain features, including possible PSE or IIC, generalized periodic discharges (GPDs) with triphasic morphology, or non-GPD patterns are indicative of better prognosis. Meanwhile, the simple act of weaning sedation may lead to transient rhythmic or periodic patterns. ⁹ Future studies may choose to design interventions to target patterns that may be associated with better outcomes. Interventional studies with prespecified treatment targets would help understand whether these indeterminate patterns reflect an underlying anoxic injury with a higher potential for recovery, or can benefit from treatment, which is likely currently being given as standard of care in many institutions. Until then, the epileptologist continues to play a central role in tandem with the intensivist, both living with the uncertainty of the existing literature on the independent importance of cEEG on outcomes of cardiac arrest.