Standardized EEG analysis to reduce the uncertainty of outcome prognostication after cardiac arrest

Abstract

Purpose Post-resuscitation guidelines recommend a multimodal algorithm for outcome prediction after cardiac arrest (CA). We aimed at evaluating the prevalence of indeterminate prognosis after application of this algorithm and providing a strategy for improving prognostication in this population. Methods We examined a prospective cohort of comatose CA patients (n = 485) in whom the ERC/ESICM algorithm was applied. In patients with an indeterminate outcome, prognostication was investigated using standardized EEG classification (benign, malignant, highly malignant) and serum neuron-specific enolase (NSE). Neurological recovery at 3 months was dichotomized as good (Cerebral Performance Categories [CPC] 1–2) vs. poor (CPC 3–5). Results Using the ERC/ESICM algorithm, 155 (32%) patients were prognosticated with poor outcome; all died at 3 months. Among the remaining 330 (68%) patients with an indeterminate outcome, the majority (212/330; 64%) showed good recovery. In this patient subgroup, absence of a highly malignant EEG by day 3 had 99.5 [97.4–99.9] % sensitivity for good recovery, which was superior to NSE < 33 μg/L (84.9 [79.3–89.4] % when used alone; 84.4 [78.8–89] % when combined with EEG, both p < 0.001). Highly malignant EEG had equal specificity (99.5 [97.4–99.9] %) but higher sensitivity than NSE for poor recovery. Further analysis of the discriminative power of outcome predictors revealed limited value of NSE over EEG. Conclusions In the majority of comatose CA patients, the outcome remains indeterminate after application of ERC/ESICM prognostication algorithm. Standardized EEG background analysis enables accurate prediction of both good and poor recovery, thereby greatly reducing uncertainty about coma prognostication in this patient population.