Predicting the potential for neurologic recovery—or lack thereof—after hypoxic-ischemic brain injury is a cornerstone of postarrest care. Recommendations guiding approaches to neuroprognostication for unconscious cardiac arrest survivors provided by guidelines (1,2) are supported by a large body of data (3,4). In fact, of all types of severe acute brain injury, hypoxic-ischemic encephalopathy is the one for which the most high-quality, regularly updated guidance for neuroprognostication exists.

The 2021 European Resuscitation Council/European Society of Intensive Care Medicine (ERC/ESICM) guidelines (1) include one of the most clinically useful algorithms—it is a practical, evidence-based, stepwise multimodal algorithm for poor outcome prediction. Easy to implement in clinical practice, this algorithm applies to patients who remain unresponsive to verbal commands greater than or equal to 72 hours postarrest with a motor examination that demonstrates abnormal limb flexion, extension, or no response to pain (i.e., Glasgow Coma Scale motor subscore 1–3; GCS-M) in the absence of confounders that might reversibly prevent wakefulness. Because the ERC/ESICM algorithm combines multiple elements that individually predict poor outcomes with high specificity (Table 1), it is most useful in excluding the potential for recovery. In practice, the resulting pessimistic prognostic impression—that is, a prediction of inability to recover to a functional state—often leads to limitations of care or withdrawal of life-sustaining therapies (WLST); these patients rarely survive hospital discharge. Hence, and appropriately so, the algorithm is conservative and geared toward mitigating the risk of false predictions of poor outcomes at the expense of sensitivity (5,6). Prior work has aimed at improving the sensitivity without sacrificing specificity, but the 2021 ERC/ESICM guidelines remain conservative for their algorithm entry point: a poor motor examination is required with absent or reflex-only responses to noxious stimuli (i.e., GCS-M 1–3) and only after confounding factors (e.g., sedatives) have been excluded (1). This leaves clinicians in the dark and without guidance on how to prognosticate in many clinical scenarios, including for patients who are unresponsive to verbal commands but withdraw from or localize to pain (i.e., GCS-M 4 or 5) and those for whom the examination findings are potentially affected by residual confounders.

TABLE 1. - European Resuscitation Council/European Society of Intensive Care Medicine Additional Unfavorable Criteria in Outcome Prediction Following Cardiac Arrest Clinical examination Absent ocular reflexes after day 3 postarrest
Continuous generalized myoclonus persisting for at least 30 min (status myoclonus) before day 3 postarrest Chemical biomarkers of injury Neuron-specific enolase > 60 µg/L at day 2 or 3 postarrest Electroencephalogram Suppressed background without periodic discharges Suppressed background with periodic discharges Burst suppression Somatosensory-evoked potentials N20 peak bilaterally absent (primary sensory cortex latencies)

In this issue of Critical Care Medicine, Arctaedius et al (7) reported the result from a large, retrospective study of a mixed cohort of out-of-hospital and in-hospital cardiac arrest patients aiming at reappraising the entry point for the ERC/ESICM algorithm. The cohort included patients admitted between 2014 and 2018, a timeframe during which the ERC/ESICM 2015 algorithm (8)—and thus, a stricter entry point, allowing only extensor posturing or absent motor response (i.e., GCS-M 1 or 2)—was routinely used in clinical practice at the four Swedish study sites. The authors assessed the prediction performance for poor outcome prediction (defined as death or functional dependence) when the updated 2021 ERC/ESICM algorithm was applied to progressively broader patient populations.

Of 794 patients in the cohort, 218 (27.5%) were unconscious on day 3; this subset of patients was further explored as three distinct subgroups, defined by their GCS-M subscore at 72 hours and presence or absence of sedating medications: GCS-M 1–3 without sedation (n = 163); GCS-M 1–3 with sedation (n = 23) and GCS-M 4 or 5 with (n = 6) or without (n = 23) sedation. Survival with independence, a good outcome, was observed in all subgroups (1/163 with GCS-M 1–3 off sedation; 1/23 with GCS-M 1–3 on sedation; 11/29 in the GCS-M 4 or 5 group). Sensitivities for the 2021 ERC/ESICM algorithm were 71.0% (95% CI, 63.6–77.4%) and 69.6% (95% CI, 62.6–75.8%) for the GCS-M 1–3 groups with and without sedation, respectively, and 62.9% (95% CI, 56.1–69.2%) when including all 218 unconscious patients. Across all subgroups, no patients with greater than or equal to two prognostic test results pointing toward a poor chance for recovery (any two from Table 1) achieved a good outcome. As such, in each expanded entry-criteria group, the modified 2021 ERC/ESICM algorithm had a false-positive rate (FPR) of 0%. Importantly, given the small group sizes, there was considerable uncertainty around the FPR point estimates reflected in broad 95% CIs (e.g., spanning 0–79.4% for the GCS-M 1–3 plus sedation group). Hence, additional data are needed to gain more confidence in accuracy, and therefore safety, of this approach.

The authors are commended for their efforts in addressing such an impactful clinical question. Exploring the assessment of patients while on sedation and with higher GCS motor subscores in this study was driven by practicality: it is often difficult to completely hold sedation in this setting due to shivering, myoclonus, coughing, or agitation while mechanically ventilated. Adherence to the algorithm with few missing data, consistent observation beyond 72 hours before prognostication in the analyzed cohort, and a detailed review of reasons for WLST are strengths of this study.

Nonetheless, it remains unclear to what extent the poor outcomes observed in the cohort resulted from a self-fulfilling prophecy. Of the 218 patients who were unconscious on day 3, 168 (77%) died after the decision for WLST at a median of 4.5 days. During the study period, the previous but similar version of the ERC/ESICM algorithm (8) was used clinically and would have recommended WLST for many of these patients who were interpreted to have a predicted poor outcome with the new algorithm. Among the subgroups of patients with expanded criteria explored in this study, WLST was also common: 17 of 23 (74%) patients with GCS M 1–3 on sedation (i.e., absent or reflex-only motor responses) succumbed following WLST. The timing of these deaths was no later in the postarrest course than the deaths of patients whose examination findings did not have the confounding effect of sedatives, unveiling a major risk of self-fulfilling prophecy. Among patients with GCS-M 4 or 5, 30% (n = 9/30) were subjected to WLST—although, in this subgroup, reassuringly, the wait time before withdrawal was considerably longer, spanning a median of 13 days for better motor function in presence of sedation.

Separate from testing ERC/ESICM algorithmic performance, this cohort reaffirms several important epidemiological facts. First, many patients hospitalized after resuscitation from out-of-hospital cardiac arrest enjoy excellent outcomes. Indeed, over one-third (n = 273/794) of patients in the entire cohort were alive and functionally independent at the 6-month follow-up. This finding underscores that therapeutic nihilism during early postarrest care is misguided and not supported by evidence. Second, when maintaining the currently suggested minimum period of at least 72 hours postarrest to allow for early awakening, many patients will indeed wake up, and only a fraction of patients hospitalized after cardiac arrest require neurologic prognostication. In this cohort, less than one-third (n = 218/794) were alive yet unconscious at the 72-hour postarrest. Most patients treated with a bundle of care aiming at brain resuscitation will declare themselves within several days and deserve at least this duration of aggressive treatment and observation for recovery potential. While survival with poor functional recovery is a feared clinical outcome that may prompt early limitations in aggressive care, it is exceedingly uncommon, as reflected by the single-digit frequencies of CPC 4 (persistent vegetative state) in most clinical trials of cardiac arrest. Largely, this is because postarrest patients who remain unresponsive at this time are at very high risk of WLST, as evidenced by the 77% prevalence of WLST in this cohort (n = 168/218). Third, as suggested by this Swedish cohort, guideline-concordant neuroprognostication timed no earlier than 72 hours postarrest with normothermia and strictly adhering to multiple prognostic modalities is feasible in routine clinical practice. This is in stark contrast to data from large cohorts of postarrest patients treated in the United States and the United Kingdom, where early WLST remains the norm (9,10) and multimodal testing infrequent (11). Although efforts to increase the usefulness of evidence-based prognostication algorithms are important, it is imperative that clinicians apply existing guidelines in their practice at a minimum. When nearly one-third of patients recover consciousness after 48 hours postarrest, with a median time of 93 hours after discontinuation of sedatives (12), and close to 10% of patients who achieve independence recovering consciousness after 7 days postarrest (13), 72 hours of supportive care is not too much to ask.

Overall, this work advances our understanding of the real-world performance of a widely used, pragmatic prognostic approach and suggests potential opportunities to enhance its clinical utility. Furthermore, it is a noble attempt to shed light on a very important knowledge gap in cardiac arrest literature: the approach to outcome prediction for patients falling in the indeterminate category, for whom guidance is not addressed by guidelines. The focus of much research in prognostication has been centered on the prediction of a poor outcome, which intends to minimize the risk of survival with unbearable deficits and to guide resource allocation—all are extremely important. The question remains whether, with the increasing availability of excellent postarrest care, the fraction of survivors who could awaken, even in a delayed fashion, is also increasing—and are we missing them? Instead of turning a blind eye to the pervasive impact of self-fulfilling prophecy, we must shed light and embrace the challenge so that we can mitigate the daunting effect of this blind spot.

1. Nolan JP, Sandroni C, Böttiger BW, et al.: European Resuscitation Council and European Society of Intensive Care Medicine Guidelines 2021: Post-resuscitation care. Intensive Care Med 2021; 47:369–421 2. Berg KM, Soar J, Andersen LW, et al.; Adult Advanced Life Support Collaborators: Adult advanced life support: 2020 international consensus on cardiopulmonary resuscitation and emergency cardiovascular care science with treatment recommendations. Circulation 2020; 142(16_suppl_1):S92–S139 3. Sandroni C, D’Arrigo S, Cacciola S, et al.: Prediction of good neurological outcome in comatose survivors of cardiac arrest: A systematic review. Intensive Care Med 2022; 48:389–413 4. Sandroni C, D’Arrigo S, Cacciola S, et al.: Prediction of poor neurological outcome in comatose survivors of cardiac arrest: A systematic review. Intensive Care Med 2020; 46:1803–1851 5. Zhou SE, Maciel CB, Ormseth CH, et al.: Distinct predictive values of current neuroprognostic guidelines in post-cardiac arrest patients. Resuscitation 2019; 139:343–350 6. Moseby-Knappe M, Westhall E, Backman S, et al.: Performance of a guideline-recommended algorithm for prognostication of poor neurological outcome after cardiac arrest. Intensive Care Med 2020; 46:1852–1862 7. Arctaedius I, Levin H, Larsson M, et al.: 2021 European Resuscitation Council/European Society of Intensive Care Medicine Algorithm for Prognostication of Poor Neurological Outcome After Cardiac Arrest—Can Entry Criteria Be Broadened? Crit Care Med 2023; 52:531–541 8. Nolan JP, Soar J, Cariou A, et al.: European Resuscitation Council and European Society of Intensive Care Medicine 2015 guidelines for post-resuscitation care. Intensive Care Med 2015; 41:2039–2056 9. Vlachos S, Rubenfeld G, Menon D, et al.: Early and late withdrawal of life-sustaining treatment after out-of-hospital cardiac arrest in the United Kingdom: Institutional variation and association with hospital mortality. Resuscitation 2023 Sep1. [online ahead of print] 10. Elmer J, Torres C, Aufderheide TP, et al.; Resuscitation Outcomes Consortium: Association of early withdrawal of life-sustaining therapy for perceived neurological prognosis with mortality after cardiac arrest. Resuscitation 2016; 102:127–135 11. Elmer J, Steinberg A, Callaway CW: Paucity of neuroprognostic testing after cardiac arrest in the United States. Resuscitation 2023; 188:109762 12. Paul M, Bougouin W, Geri G, et al.: Delayed awakening after cardiac arrest: Prevalence and risk factors in the Parisian registry. Intensive Care Med 2016; 42:1128–1136 13. Tsai MS, Chen WJ, Chen WT, et al.: Should we prolong the observation period for neurological recovery after cardiac arrest? Crit Care Med 2021; 50.