Performance of the ERC/ESICM-recommendations for neuroprognostication after cardiac arrest: Insights from a prospective multicenter cohort

Abstract

Aim: To investigate the performance of the 2021 ERC/ESICM-recommended algorithm for predicting poor outcome after cardiac arrest (CA) and potential tools for predicting neurological recovery in patients with indeterminate outcome.

Methods: Prospective, multicenter study on out-of-hospital CA survivors from 28 ICUs of the AfterROSC network. In patients comatose with a Glasgow Coma Scale motor score ≤3 at ≥72 h after resuscitation, we measured: (1) the accuracy of neurological examination, biomarkers (neuron-specific enolase, NSE), electrophysiology (EEG and SSEP) and neuroimaging (brain CT and MRI) for predicting poor outcome (modified Rankin scale score ≥4 at 90 days), and (2) the ability of low or decreasing NSE levels and benign EEG to predict good outcome in patients whose prognosis remained indeterminate.

Results: Among 337 included patients, the ERC-ESICM algorithm predicted poor neurological outcome in 175 patients, and the positive predictive value for an unfavourable outcome was 100% [98-100]%. The specificity of individual predictors ranged from 90% for EEG to 100% for clinical examination and SSEP. Among the remaining 162 patients with indeterminate outcome, a combination of 2 favourable signs predicted good outcome with 99[96-100]% specificity and 23[11-38]% sensitivity.

Conclusion: All comatose resuscitated patients who fulfilled the ERC-ESICM criteria for poor outcome after CA had poor outcome at three months, even if a self-fulfilling prophecy cannot be completely excluded. In patients with indeterminate outcome (half of the population), favourable signs predicted neurological recovery, reducing prognostic uncertainty.

Keywords: Cardiac arrest; Coma; Electroencephalogram (EEG); Neuron-Specific Enolase (NSE); Prognosis; Short-latency Somatosensory Evoked Potentials (SSEPs).