Added Prognostic Value of EEG Reactivity in Comatose Patients Following Cardiac Arrest
Sarah Caroyer1,2, Marie Obrecht1, Charlotte de Maeseneire1
1Department of Neurology, Université Libre De Bruxelles (ULB), Hôpital Universitaire de Bruxelles (HUB), CUB Hôpital Erasme, Bruxelles, Belgium.
Objectives:
To evaluate the added prognostic value of EEG reactivity for favorable outcome compared with background analysis during and after targeted temperature management (TTM).
Methods:
Prospective observational cohort study of comatose post-cardiac arrest patients admitted to a single academic center between 2017 and 2022, all undergoing continuous EEG monitoring. Reactivity was assessed using standardized definitions and stimulation protocols. EEG background was classified by continuity and amplitude according to consensus criteria. Favorable outcome was defined as CPC 1-2 at discharge or 3 months. Prognostic performance of background and reactivity, alone or combined in a multimodal process, was evaluated with sensitivity, specificity, likelihood ratios, accuracy, and balanced accuracy. A literature review was performed on the prognostic value of reactivity testing.
Results:
313 patients underwent reactivity testing during and 252 after TTM. Background assessment demonstrated very high sensitivity (98%-100%) but modest specificity (34%-58%). Reactivity alone had lower sensitivity (82%-93%) with slightly higher specificity. Combining background and reactivity increased specificity (53%-58% to 71%) and overall accuracy. In patients with non-highly malignant backgrounds, reactivity improved sensitivity (60% to 84%-90%) but reduced specificity (74% to 51%-55%). Patterns were consistent for survival and in the context of multimodal prognostication. Literature review showed similar findings despite heterogeneity in definitions and methods.
Conclusions:
EEG reactivity should not guide WLST decisions alone, but its presence reliably indicates a higher likelihood of a favorable outcome. Optimized interpretation modestly improves prognostic accuracy but does not substantially enhance multimodal prediction, highlighting the need for further investigation.
