Abstract WMP110: Effectiveness of EEG Ischemia Monitoring in Clinical Practice for the Detection of Delayed Cerebral Ischemia after Subarachnoid Hemorrhage

Abstract

Background: Quantitative EEG trend monitoring has shown high sensitivity in pilot studies for early detection of delayed cerebral ischemia (DCI) after aneurysmal SAH. Little is understood, however, about its performance in the prospective clinical environment. We hypothesized that clinical confounders and mimics of ischemic EEG patterns may result in high false alarm rates. Methods: Patients undergoing EEG ischemia monitoring within 4 days of SAH were identified from a prospectively documented clinical database and evaluated for any EEG ischemic signs (EEG-IS) denoted by the clinical neurophysiologist (NPhys-EEG-IS), including deterioration by one grade in percent alpha variability (PAV) over 8 hours, >50% decrement in alpha-delta ratio (ADR) exceeding 1 hour, >10% decrement in ADR exceeding 5 hours, or new focal EEG slowing on the raw EEG. We separately performed a wholly computational retrospective assessment for EEG-IS (Comp-EEG-IS). DCI was retrospectively ascertained as an unexplained focal clinical or radiologic decline. Results: 19 patients underwent EEG monitoring (median 10 days, IQR 9-11). Late epileptiform or rhythmic patterns emerged in 14 patients (74%). There was a strong trend towards NPhys-EEG-IS being associated with DCI (univariate OR 8.4, p=0.15), which persisted (multivariate OR 7.5, p=0.08) after adjusting for the association of Hunt-Hess grade >4 and DCI (multivariate OR 3.4, p=0.27). Of 11 patients (58%) with NPhys-EEG-IS alarms, 6 developed DCI (Se 86%, Sp 58%, PPV 55%). DCI was not associated with longer NPhys-EEG-IS alarm duration (5.0 vs. 4.8 alarm-days; NS). The Comp-EEG-IS method was problematic, resulting in multiple false alarms per patient-day; we identified several clinical variables contributing to false alarms. Conclusions: When applied purely quantitatively, published EEG trend analysis methods result in unacceptably high false alarm rates. Neurophysiologist integration of quantitative and raw EEG data modestly improves the predictive value of EEG ischemic signs and affords relatively high sensitivity. Better understanding of specific EEG signatures of confounding variables, and real-time integration of parallel clinical data are required for more accurate EEG DCI detection in clinical practice.