Abstracts

Rationale: Continuous EEG monitoring (cEEG) has become a standard practice in the neuroscience intensive care unit (ICU) for the management of patients (pts) with acute brain dysfunction. In non-neurological critically ill patients, acute brain dysfunction is a common but still poorly understood complication and the role of cEEG in this setting has not been elucidated. Methods: We retrospectively analyzed the data from pts in the medical ICU (MICU) at Columbia University Medical Center who underwent cEEG monitoring. Seizures (SZs) and periodic epileptiform discharges (PEDs, including both lateralized [PLEDS, BIPLEDs] and generalized [GPEDs]) on cEEG were quantified. We identified predictors of SZs and PEDs and analyzed their impact on outcome at hospital discharge. Univariate (using ANOVA) and multivariate (using a forward stepwise logistic regression model) analyses were performed.Results: 115 pts (median age, 63 [interquartile range, IQR: 53-73] years) underwent cEEG monitoring. Median time from MICU admission to cEEG was 1 day (IQR: 1-3 days) and median duration of cEEG was 3 days (IQR: 2-5 days). At the time of cEEG, 91% of pts were mechanically ventilated and 55% were comatose. The majority of pts (61%) had sepsis as their primary admission diagnosis. 28% of pts had clinical seizures prior to cEEG. Overall 27% of pts had either SZs or PEDs; 15% had SZs (11 non-convulsive, 6 convulsive) and 24% had PEDs (13 pts had both PEDs and SZs, 14 pts PEDs only and 4 pts SZs only). 65% (11/17) had SZs without clinical correlate. On multivariate analysis, coma, sepsis and clinical seizures were independent predictors of SZs or PEDs on cEEG. Of the 60 comatose pts, 38% had SZs or PEDs and of the 70 pts with sepsis 37% had these findings. At hospital discharge, 63/115 (55%) of pts had poor outcome (GOS 1-3= severely impaired, vegetative or dead). Among septic pts, 88% with SZs or PEDs had poor outcome, as compared to 54% without SZs or PEDs (p=0.005). Among the total cohort, 87% of pts with SZs or PEDs had poor outcome, as compared to 43% without (p<0.0001). On multivariate analysis, the presence of SZs or PEDs and the need of vasopressors were the only two independent predictors of outcome. When SZs and PEDs were analyzed separately, both were associated with worse prognosis on univariate analysis, but only PEDs independently correlated with poor outcome on multivariate analysis. Conclusions: Seizures and PEDs are common in MICU patients, seen in 27% of those undergoing cEEG at our center, particularly in patients with sepsis and coma, and are independent predictors of poor outcome. Fifteen percent of cEEG-monitored patients demonstrated electrographic seizure activity, and in two-thirds of cases there was no clinical correlate. Additional studies are needed to determine whether treatment or prevention of SZs and PEDs can improve outcome.