Assessing both early and late EEG patterns improves prediction of outcome after cardiac arrest

Previously proposed "synchronous EEG patterns" predict poor outcome within 24h after cardiac arrest (CA). We investigate the prognostic performance of these early EEG predictors in addition to the late EEG predictors (>24h) recommended in the European post-resuscitation guidelines. Observational substudy of the TTM2-trial including consecutive comatose resuscitated patients. Continuous EEG-monitoring (cEEG) was blindly assessed using the American Clinical Neurophysiology Societýs standardised EEG terminology and categorised into early EEG predictors (burst-suppression with identical or highly epileptiform bursts, or suppression with generalised periodic discharges) and late EEG predictors (heterogenous burst-suppression or suppression). Poor outcome was defined as modified Rankin Scale 4-6 at six months. Of 191 included patients, 53% had poor outcome. Early EEG predictors had 100%[CI 96-100] specificity at all time-points and maximal sensitivity 30%[CI 21-40] before 24h. Late EEG predictors had 100%[CI 96-100] specificity beyond 24h with maximal sensitivity 32%[CI 21-43]. Using both early and late EEG predictors, and gradually adding cEEG-information from consecutive time-epochs, sensitivity increased to 49%[CI 39-59] up to 36h after CA (p=0.001). A continuous background within 12h predicted good outcome (sensitivity 61%[CI 50-71]; specificity 87%[CI 79-93]). Searching for both early EEG predictors (e.g. identical burst-suppression) and late EEG predictors (e.g. heterogenous burst-suppression >24h) significantly improved sensitivity of poor outcome prediction without false positive survivors in this cohort. A self-fulfilling prophecy may have affected our results. cEEG during the first two days after CA identified half of the patients with a long-term poor outcome and half of the patients with a good outcome.