EEG response during sedation interruption complements behavioral assessment following severe brain injury

Abstract

Accurate assessment of level of consciousness and potential to recover in severe brain injury patients underpins crucial decisions in the intensive care unit but remains a major challenge for the clinical team. The neurological wake-up test (NWT) is a widely used assessment tool, but many patients' behavioral response during a short interruption of sedation is ambiguous or absent, with little prognostic value. This study assesses the brain's electroencephalogram response during an interruption of propofol sedation to complement behavioral assessment during the NWT to predict survival, recovery of consciousness, and long-term functional outcome in acute severe brain injury patients. We recorded 128-channel EEG of 41 severely brain-injured patients during a clinically indicated NWT. The Glasgow Coma Scale (GCS) was used to assess behavioral responsiveness before and after interruption of sedation (GCS_observed). During the NWT, nine patients regained responsiveness, 13 patients showed ambiguous responsiveness and 19 patients were not responsive. Brain response to sedation interruption was quantified using EEG power, spatial ratios and the spectral exponent. We trained a linear regression model to identify brain patterns related to regaining behavioral responsiveness. We then applied this model to patients whose behavioral responses were ambiguous or absent, using their NWT brain responses to predict a change in behavioral response (Delta_GCS_predicted). Prognostic value of the Delta_GCS_predicted was assessed using the Mann-Whitney-U test and group-separability. The patients' survival, recovery of responsiveness, and functional outcomes were assessed up to 12 months post-recording. EEG patterns during interruption of sedation reliably predicted the GCS_observed in patients who regained responsiveness during the NWT. Electrophysiological patterns of waking-up were observed in some patients whose behavioral response was ambiguous or absent. Compared to the GCS_observed, the Delta_GCS_predicted improved separability of prognostic groups and significantly distinguished patients according to survival (U = 87, p<0.05). The EEG-trained model outperformed outcome predictions of the patients' attending physician and predictions based on the patients' APACHE score. EEG can complement behavioral assessment during the NWT to improve prognostication, inform clinicians, family members and caregivers, and to set realistic goals for treatment and therapy.