Neuroprognostication After Pediatric Cardiac Arrest and ROSC: Prediction of Poor Neurological Outcome

Post-ROSC Clinical Examination

Last Full Review: ILCOR 2025

The brain is prone to hypoxic injury during cardiac arrest and following the return of spontaneous circulation (ROSC). Therefore, some patients will develop global cerebral edema followed by herniation and brain death within 24 hours, while others will remain comatose and at risk for severe neurological injury. Prognostication has been used to help families and physicians make decisions to limit or withdraw life support when unfavorable neurological outcomes are expected. A 2023 International Liaison Committee on Resuscitation (ILCOR) systematic review (Berg et al. 2023, e187) considered the use of functional and structural modalities to aid clinicians in predicting a good neurological outcome for infants and children with the return of circulation (ROC), including ROSC or mechanical circulation, after resuscitation from in-hospital or out-of-hospital cardiac arrest from any cause. For 2025, ILCOR assessed individual prognostic tests to predict poor neurological outcome. For evaluation of each intervention/test, the same population, timing of the test, comparator, outcome, study design and time frame for the literature search were used. Prediction of poor neurological outcome was considered imprecise with a false positive rate greater than 1%; evidence was considered reliable if the false positive rate was less than 1% with upper 95% confidence intervals less than 10%, and evidence was considered moderately reliable if the false positive rate was less than 1% but without a restriction on the width of the 95% confidence interval.

This review relates to the use of clinical examination for prediction of poor neurological outcome. Of note, all evaluated tests were used in combination with other tests by clinicians in the included studies.

 

Evidence Summary

A 2025 ILCOR systematic review and Consensus on Science with Treatment Recommendations (CoSTR) (Scholefield et al. 2024c; Scholefield et al. 2025, S116) studied the use of bedside clinical neurological examination for the prognosis of survival with poor neurological outcome in children under 18 years of age who achieve a return of spontaneous or mechanical circulation after resuscitation from in-hospital and out-of-hospital cardiac arrest from any cause. Components of the neurological examinations included pupillary response (manual light reflex or automated pupillometry), level of coma (Glasgow Coma Scale [GCS] or Full Outline of UnResponsiveness [FOUR] score), and brainstem reflexes.

For pupil reactivity, three studies (Abend et al. 2012, 32; Fink et al. 2014, 664; Oualha et al. 2013, 1306) with 139 total patients showed that absence of a pupillary light reflex at 48 and 72 hours after ROC had a false positive rate of less than 1%, but a confidence interval of 0 to 40% (Scholefield et al. 2024c; Scholefield et al. 2025, S116). A total GCS score of less than 7 (Nishisaki et al. 2007, 10) and a GCS motor score of less than 4 (Lin et al. 2020, 2151; Lin et al. 2014, 732) was assessed at less than 1 hour and at 4 to 6 hours, with a false positive rate of 69% and 50% to 83%, respectively, and thus this was not predictive of poor neurological outcomes (Scholefield et al. 2024c; Scholefield et al. 2025, S116).

Very limited evidence was identified for the presence of evoked brainstem reflexes (pain, cough or gag response), and ILCOR concluded that there is insufficient evidence to make a recommendation for or against the use of other brainstem or motor response tests to predict poor neurological outcome in children after cardiac arrest.

ILCOR produced a strong treatment recommendation that no single clinical examination test be used in isolation to predict poor neurological outcome in children after cardiac arrest. A good practice statement recommends that clinicians should use multiple tests in combination for poor neurological outcome prediction (Scholefield et al. 2024c; Scholefield et al. 2025, S116).

Additional ILCOR recommendations include:

• The absence of pupil reactivity to light at 48 and 72 hours after ROC may be considered as part of multimodal testing to predict poor neurological outcome in children after cardiac arrest. (Good practice statement)
• We suggest against using absence of pupil reactivity to light within 24 hours after ROC to predict poor neurological outcome in children after cardiac arrest (weak recommendation, low-certainty evidence).
• We suggest against using GCS within 24 hours after ROC to predict poor neurological outcome in children after cardiac arrest (weak recommendation, low-certainty evidence).

 

Insights and Implications

Limitations of this review included the small number of studies and patients, a high risk of bias, lack of blinding, and variability in test assessment, performance, and outcome measurement. Only one study each was identified for the total GCS, GCS motor score, and brainstem tests, which prohibited making a formal suggestion or recommendation for these tests. Lastly, confounding from the effect of sedatives or hypothermic temperature control was not reported in the included studies. The new Red Cross guidelines are informed by this ILCOR review.

 

Biomarkers

Last Full Review: ILCOR 2025

The brain is prone to hypoxic injury during cardiac arrest and following the return of spontaneous circulation (ROSC). Therefore, some patients will develop global cerebral edema followed by herniation and brain death within 24 hours, while others will remain comatose and at risk for severe neurological injury. Prognostication has been used to help families and physicians make decisions to limit or withdraw life support when unfavorable neurological outcomes are expected. A 2023 International Liaison Committee on Resuscitation (ILCOR) systematic review (Berg et al. 2023, 91) considered the use of functional and structural modalities to aid clinicians in predicting a good neurological outcome for infants and children with the return of circulation (ROC), including ROSC or mechanical circulation, after resuscitation from in-hospital or out-of-hospital cardiac arrest from any cause. For 2025, ILCOR assessed individual prognostic tests to predict poor neurological outcome. For evaluation of each intervention/test, the same population, timing of the test, comparator, outcome, study design and time frame for the literature search were used. Prediction of poor neurological outcome was considered imprecise with a false positive rate greater than 1%; evidence was considered reliable if the false positive rate was less than 1% with upper 95% confidence intervals less than 10%, and evidence was considered moderately reliable if the false positive rate was 1% but without a restriction on the width of the 95% confidence interval.

A 2025 review relates to the use of serum brain injury biomarkers or markers of inflammation and poor perfusion for prediction of poor neurological outcome.

 

Evidence Summary

A 2025 ILCOR systematic review and Consensus on Science with Treatment Recommendations (CoSTR) (Scholefield et al. 2024a; Scholefield et al.2025, S116) evaluated evidence for the use of blood biomarkers for the prognosis of survival with poor neurological outcome in children under 18 years of age who achieve a return of spontaneous or mechanical circulation after resuscitation from in-hospital and out-of-hospital cardiac arrest from any cause. Biomarkers of interest included serum biomarkers either specific to neuronal damage (e.g., neuron-specific enolase [NSE], S100 calcium-binding protein B [S100B], glial fibrillary acidic protein, neurofilament light chain) or blood markers of inflammation or systemic ischemic reperfusion (e.g., blood pH or lactate).

For studies of lactate, two out of six included studies identified a false positive rate of less than 1% for poor neurological outcome prediction. Kramer et al. (Kramer et al. 2018, 113) used a lactate threshold greater than 28.8 milliosmoles (mOsm) per liter (L) at under 1 hour, while De La Llana (de la Llana et al. 2021, 202) used failure of lactate clearance to less than 2 mOsm/L by 48 hours. Other tests with a lactate greater than 2 mOsm/L at 6 to 12, 24 and 48 hours had false positive rates up to 84%, and the review concluded that lactate was not a reliable prognostic test. Various pH thresholds at resuscitation and time points up to 24 hours post-ROC had false positive rates of 3% to 4% but low sensitivity for predicting poor neurological outcome (Scholefield et al.et al. 2025; Scholefield et al.2025 In Press).

For testing of neuronal biomarkers, NSE and S100B values were reported by three studies (Bangshøj et al. 2022, 1659; Fink et al. 2014, 664; Kramer et al. 2018, 113) with 156 children post-ROC. Cut off values were calculated at different time intervals up to 72 hours. At 24 hours, S100B and NSE both predicted poor neurological outcome with a false positive rate of 0% and with a low sensitivity of 29% to 38% and 19% to 26%, respectively. In a single study (Fink et al. 2014, 664), myelin basic protein (MBP) also predicted poor neurological outcome at 24 and 48 hours with a false positive rate of 0% (95% CI, 0%–20%). Overall, sensitivity of these biomarkers was low, and cutoff thresholds in the individual studies varied widely, limiting clinical utility (Scholefield et al. 2024a; Scholefield et al. 2025, S116.

Treatment recommendations by ILCOR stemming from this review begin with the strong recommendation that no single blood-based biomarker be used in isolation to predict poor neurological outcome in children after cardiac arrest. A good practice statement recommends that clinicians use multiple tests in combination for poor neurological outcome prediction. There was insufficient evidence to make a recommendation for or against the use of other blood-based biomarkers after ROC for predicting poor neurological outcome in children after cardiac arrest at any time point. The International Liaison Committee for Resuscitation suggests against using lactate and pH after ROC for predicting poor neurological outcome in children after cardiac arrest at any time point (Scholefield et al. 2024a; Scholefield et al. 2025, S116.).

Insights and Implications

Limitations of this review included a lack of blinding and lack of assessment of the confounding influence of medication (e.g., sedatives). The biomarkers evaluated are not commonly available for clinical use, and the assessment of their applicability for neuroprognostication was based on studies not designed to test prognosis of blood biomarkers.

 

Electrophysiology Tests

Last Full Review: ILCOR 2025

The brain is prone to hypoxic injury during cardiac arrest and following the return of spontaneous circulation (ROSC). Therefore, some patients will develop global cerebral edema followed by herniation and brain death within 24 hours, while others will remain comatose and at risk for severe neurological injury. Prognostication has been used to help families and physicians make decisions to limit or withdraw life support when unfavorable neurological outcomes are expected. A 2023 International Liaison Committee on Resuscitation (ILCOR) systematic review (Berg et al. 2023, 91) considered the use of functional and structural modalities to aid clinicians in predicting a good neurological outcome for infants and children with the return of circulation (ROC), including ROSC or mechanical circulation, after resuscitation from in-hospital or out-of-hospital cardiac arrest from any cause. For 2025, ILCOR assessed individual prognostic tests to predict poor neurological outcome. For evaluation of each intervention/test, the same population, timing of the test, comparator, outcome, study design and time frame for the literature search were used. Prediction of poor neurological outcome was considered imprecise with a false positive rate greater than 1%; evidence was considered reliable if the false positive rate was less than 1% with upper 95% confidence intervals less than 10%, and the evidence was considered moderately reliable if the false positive rate was less than 1% but without a restriction on the width of the 95% confidence interval.

A 2025 review relates to the use of electrophysiology testing for prediction of poor neurological outcome.

 

Evidence Summary

A 2025 ILCOR systematic review and Consensus on Science with Treatment Recommendations (CoSTR) (Scholefield et al. 2024d; Scholefield et al. 2025, S116) evaluated evidence for the use of electrophysiology testing for the prognosis of survival with poor neurological outcome in children under 18 years of age who achieve a return of spontaneous or mechanical circulation after resuscitation from in-hospital and out-of-hospital cardiac arrest from any cause. Testing included surface bioelectrical recordings from the central nervous system, such as electroencephalogram (EEG), and evoked potentials, such as brainstem auditory evoked potentials and short-latency somatosensory evoked potentials (SSEPs). The reader is encouraged to review the full CoSTR and summary table of electrophysiology tests, time scale and prediction accuracy online at:
https://costr.ilcor.org/document/electrophysiology-testing-for-the-prediction-of-survival-with-good-neurological-outcome-after-return-of-circulation-following-pediatric-cardiac-arrest-pls-tfsr

Multiple studies included in the systematic review reported the relationship between the presence of clinical and/or electrographic seizures in a total of 1,165 children post-cardiac arrest and poor neurological outcomes at intervals between intensive care until and 12 months post discharge. Based on their false positive rate and low sensitivity, these were considered unreliable as a prognostic test (Scholefield et al. 2024d; Scholefield et al. 2025, S116).

Studies reporting the presence of status epilepticus in a total of 299 children showed a low false positive rate (0% to 5%; upper limit of 95% CI range, 13%–41%) and sensitivity from 9% to 25% for the prediction of poor neurological outcome at intensive care unit/hospital discharge; this was considered moderately reliable as a prognostic test (Scholefield et al. 2024d; Scholefield et al. 2025, S116).

Absence of a benign continuous EEG background pattern in studies with a total of 563 patients was found to have widely variable false positive rates and sensitivity at all time points reported for predicting poor neurological outcome, while absence of an attenuated, isoelectric, or flat EEG in studies with up to 526 patients (possibly overlapping between studies) was found to have a false positive rate less than 10% (85% CI upper limit, 4%–52%) and sensitivities between 18 and 58%. At a false positive rate less than 1% cutoff, this was found to be an imprecise finding in over 50% of the studies included in the systematic review (Scholefield et al. 2024d; Scholefield et al. 2025, S116).

The presence of burst suppression, burst attenuation, or generalized periodic epileptiform discharges (GPEDs) on EEG background at various time points were reported in a total of 395 patients. In three of four studies (Brooks et al. 2018, 324; Oualha et al. 2013, 1306; Yang et al. 2019, 223), a false positive rate less than 1% (95% CI upper limit, 16%–54%) was reported at 24, 28 and 72 hours, with sensitivities ranging from 0% to 67%, which was considered moderately reliable for the prediction of poor neurological outcome from 24 to 72 hours (Scholefield et al.et al. 2025; Scholefield et al.2025 In Press).

Fewer studies reported on the absence of reactivity within an EEG, sleep II architecture, or sleep spindles and variability on EEG. These test findings were reported to have wide ranges of false positive rates and sensitivities for predicting poor neurological outcome and were considered inaccurate and unreliable (Scholefield et al. 2024d; Scholefield et al. 2025, S116).

A single study (McDevitt et al. 2021, 30) reported on SSEPs, evaluating bilateral absence of negative peak at 20 milliseconds (N20) waves for the prediction of poor neurological outcome. At 24 and 48 hours, the false positive rate was 0% (95% CI range, 0%–52%) with 100% sensitivity (95% CI range, 29%–100%). At 72 hours, the false positive rate rose to 17%. Although the test was rated as moderately reliable for the prediction of poor neurological outcome, because only one small study was identified, this was considered insufficient evidence to make a recommendation for or against this test (Scholefield et al. 2025; Scholefield et al.2025 In Press).

The ILCOR recommendations stemming from this comprehensive review include (Scholefield et al.et al. 2024d; Scholefield et al. 2025, S116):

• We recommend that no single electrophysiology test be used in isolation to predict poor neurological outcome in children after cardiac arrest at any time point (strong recommendation, very low-certainty evidence).
• Clinicians should consider using multiple tests in combination for poor neurological outcome prediction. (Good practice statement)
• The presence of status epilepticus between 24 and 72 hours after ROC and the presence of burst suppression, burst attenuation, or GPEDs between 24 and 72 hours after ROC all had moderate reliability and may be considered as part of multimodal testing to predict poor neurological outcome in children after cardiac arrest. (Good practice statement)
• We suggest against using the following EEG features for predicting poor neurological outcome (weak recommendation, very-low certainty evidence):
  • Presence of clinical or electrographic seizures
  • Absence of sleep spindle and sleep II architecture on EEG
  • Continuous or normal background EEG
  • EEG reactivity and EEG variability at any time point
• There was insufficient evidence to make a recommendation for or against the use of presence of attenuated, isoelectric, or flat EEG; absence of N20 response on SSEPs; presence of myoclonic status epilepticus; or quantitative EEG score to predict poor neurological outcome in children after cardiac arrest at any time point.

 

Insights and Implications

Like the other systematic reviews of testing for prognostication of poor neurological outcome, this review was limited by the high risk of bias across studies, the small number of patients, lack of blinding, variation in test assessment and performance, and variability in outcome measurement. This prohibited meta-analysis of findings. The concurrent use of sedatives and other treatments, such as active hypothermic temperature control, was not described in relation to the testing or time of test. The review notes that identification of seizures allows treatment, which may prevent secondary injury after a hypoxic-ischemic injury.

One study on the use of SSEPs in children was considered moderately reliable for the prediction of poor neurological outcome. Further studies are needed to confirm this finding.Further research is also needed on multimodal assessments—including clinical evaluations, biomarkers, electrophysiological studies and neuroimaging—to identify the most accurate strategies for predicting poor neurological outcomes, thereby improving clinical decision-making in critical care.

 

Brain Imaging

Last Full Review: ILCOR 2025

The brain is prone to hypoxic injury during cardiac arrest and following the return of spontaneous circulation (ROSC). Therefore, some patients will develop global cerebral edema followed by herniation and brain death within 24 hours, while others will remain comatose and at risk for severe neurological injury. Prognostication has been used to help families and physicians make decisions to limit or withdraw life support when unfavorable neurological outcomes are expected. A 2023 International Liaison Committee on Resuscitation (ILCOR) systematic review (Berg et al. 2023, 91) considered the use of functional and structural modalities to aid clinicians in predicting a good neurological outcome for infants and children with the return of circulation (ROC), including ROSC or mechanical circulation, after resuscitation from in-hospital or out-of-hospital cardiac arrest from any cause. For 2025, ILCOR assessed individual prognostic tests to predict poor neurological outcome. For evaluation of each intervention/test, the same population, timing of the test, comparator, outcome, study design and time frame for the literature search were used. Prediction of poor neurological outcome was considered imprecise with a false positive rate greater than 1%; evidence was considered reliable if the false positive rate was less than 1% with upper 95% confidence intervals less than 10%, and evidence was considered moderately reliable if the false positive rate was less than 1% but without a restriction on the width of the 95% confidence interval.

A 2025 review relates to the use of brain imaging for prediction of poor neurological outcome.

 

Evidence Summary

A 2025 ILCOR systematic review and Consensus on Science with Treatment Recommendations (CoSTR) (Scholefield et al. 2024b; Scholefield et al. 2025, S116) evaluated evidence for the use of brain imaging for the prognosis of survival with poor neurological outcome in children less than 18 years of age who achieve a return of spontaneous or mechanical circulation resuscitation from in-hospital and out-of-hospital cardiac arrest from any cause. Imaging modalities included head computer tomography (CT) and brain magnetic resonance imaging (MRI).

Three studies (Fink et al. 2014, 664; Starling et al. 2015, 542; Yang et al. 2019, 223) reported on the relationship of head CT findings to poor neurological outcome in 173 patients. The absence of gray-white matter differentiation at 24 hours had a false positive rate of 0% to 36% with a 20% to 30% sensitivity, making it a moderately reliable test for prediction of poor neurological outcome. Other head CT findings (effacement, lesions, edema, intracranial hemorrhage) had higher false positive rates with variable sensitivities and were considered unreliable for prediction of poor neurological outcome.

Five studies (Bach et al. 2024, e209134; Fink et al. 2014, 664; Fink et al. 2020, 185; Kirschen et al. 2021, e719; Yacoub et al. 2019, 103) with 305 patients reported on MRI findings to predict poor neurological outcome. Poor neurological outcome was predicted in three studies (Bach et al. 2024, e209134; Kirschen et al. 2021, e719; Yacoub et al. 2019, 103) by an apparent diffusion coefficient threshold of <650×10-6 mm² per second in 10% or more of brain volume (i.e., high ischemic burden) at a median of 4 days after ROC with a false positive rate of 0% to 6% (95% CI, 1–21%) and 49% to 52% sensitivity. The study by Yacoub et al. (Yacoub et al. 2019, 103) using apparent diffusion coefficient thresholds to identify high ischemic burden met the low false positive rate of less than 1% for moderate reliability to predict poor neurological outcome. Regional abnormalities on restricted diffusion, diffusion-weighted MRI, and T1-weighted and T2-weighted imaging were not considered reliable for prediction of poor neurological outcome.

The ILCOR treatment recommendations stemming from this review include (Scholefield et al.et al. 2024b; Scholefield et al. 2025 S116):

• We recommend no single imaging test be used alone to predict poor neurological outcome in children after cardiac arrest at any time point. (Strong recommendation, very-low certainty evidence)
• Clinicians should consider using multiple tests in combination for poor neurological outcome prediction. (Good practice statement)
• An abnormal MRI showing high ischemic burden on apparent diffusion coefficient mapping at 72 hours and beyond after ROC or CT scan showing loss of gray-white matter differentiation within 24 hours after ROC may be considered as part of multimodal testing to predict poor neurological outcome in children after cardiac arrest (good practice statement).

 

Insights and Implications

The new Red Cross guidelines are informed by the ILCOR treatment recommendations and systematic review. The CoSTR (Scholefield et al.et al. 2024b; Scholefield et al.2025, S116) authors note that for the MRI studies of regional brain abnormalities, the false positive rate was low, but there were few cases and wide confidence intervals. In addition, similar to other studies of tests for prediction of poor neurological outcome, the studies had a high risk of bias, were limited in number and enrolled patients, lacked blinding, had variation in test assessment and performance, and had variability in outcome measurement. Definitions of findings or cutoff values on MRI and CT were inconsistent between studies. Further research utilizing multimodal assessments—including clinical evaluations, biomarkers, electrophysiological studies and neuroimaging—is essential to identify the most accurate strategies for predicting poor neurological outcomes, thereby improving clinical decision-making in critical care.