By Kyle Hobbs, MD and Sara Stern-Nezer, MD
Coma secondary to acute brain injury is extremely common in the neurocritical care unit, but prognosis after coma is challenging. Electroencephalography provides a window into the comatose brain, revealing brain activity even in the absence of clinical responses. The reviewed studies below investigate the role of EEG in comatose patients.
Electroencephalographic reactivity as predictor of neurologic outcome in postanoxic coma: A multicenter prospective cohort study.
Ann Neurol 2019; 86:17-27
Read the article.
Prognostication in post-anoxic coma is of the utmost importance after cardiac arrest. EEG reactivity (EEG-R) for prognosis has been studied in brain injury, but with variability in methods and analysis. This prospective, multicenter cohort study aimed to determine the prognostic value of EEG-R using a strict replicable protocol, assessing both poor and good neurologic outcomes. A total of 160 comatose adult patients after cardiac arrest (CA) were enrolled if continuous EEG (cEEG) was started within 24 hours. Patients were treated with TTM for 24 hours, and cEEG was continued for 3 days unless patient regained consciousness. EEG-R was assessed using a standardized protocol of various stimuli applied for 5 seconds consecutively three times. EEG-R was assessed by three blinded readers. EEG-R was present if any of the stimuli induced a change in EEG amplitude or frequency at least twice. In the case of uncertainty, EEG-R was considered present. The primary outcome was the best neurologic outcome within 6 months (Cerebral Performance Category score). Some 345 EEG-R assessments were available for analysis. Seventy patients (47%) had poor outcome (CPC 3-5), which was associated with longer time to ROSC, nonshockable initial rhythm, absent 72-hour brainstem reflexes and unfavorable EEG patterns at 12 and 24 hours. Specificity of EEG-R was 95% for poor outcome and 80% for good outcome. Absence of EEG-R predicted poor outcome with specificity of 82% (95% CI=74-91) and sensitivity 73% (95% CI=62-83). Presence of EEG-R predicted good outcome with 73% specificity (95% CI=62-83) and 82% sensitivity (95% CI=74-91). Addition of absent EEG-R to other predictors of poor outcome (EEG background at 24 hours, brainstem reflexes, or SSEPs) resulted in increased specificity (99%) and decreased sensitivity (54%). Addition of present EEG-R to EEG background at 12 hours post CA or brainstem reflexes increased specificity for good outcome to 89% and decreased sensitivity to 66%.
EEG-R alone was not reliable in predicting poor or good neurologic outcome after cardiac arrest. Absence of EEG-R added to other predictors of poor outcome did not result in a significant increase in specificity (98-99%), but the presence of EEG-R added to other predictors of good outcome increased specificity for good outcome. This study suggests that EEG-R has a limited role for prediction of neurologic outcome after CA, even when integrated with other known prognosticating. Strengths of this study include a protocolized assessment of EEG-R, as well as the multicenter nature of the study. Future studies of EEG-R may benefit from quantitative analysis of EEG-R, as well as inclusion of spontaneous variability of the EEG.
Detection of brain activation in unresponsive patients with acute brain injury.
New Engl J Med 2019; 380:2497-505
Read the article.
Clinically unresponsive patients have been shown to have brain activation in response to spoken commands (cognitive-motor dissociation), but the prognostic significance of this is unclear. This single-center prospective cohort study used a machine-learning technique to analyze EEG evidence of brain activation in unresponsive patients. All adult patients with acute brain injury undergoing EEG monitoring who were in a coma, vegetative state or minimally conscious state-minus (unresponsive with preserved visual fixation or pursuit, or localization to noxious stimuli) were enrolled. Functional outcomes were assessed with dichotomized GOS-E levels (>4 = good outcome). Ten healthy volunteers underwent EEG monitoring using the same EEG protocol. Patients underwent spoken command instructions during EEG recording, consisting of alternations between “keep opening and closing your right/left hand” and “stop opening and closing your right/left hand.” Six blocks (3 right, 3 left), each with eight consecutive trials of either command were recorded. For each EEG recording, a machine-learning algorithm was trained using calculated power in predefined frequency ranges to distinguish between EEG responses following these commands. EEG recordings with ROC AUC of > 0.5 were considered to show brain activation if temporally concordant with spoken commands. A total of 104 patients were enrolled, from which 240 EEG recordings were obtained at a median of 6 days post-injury. All healthy volunteers had evidence of brain activation to motor commands. Eight of 16 (50%) patients with cognitive-motor dissociation on at least 1 recording improved to being able to follow spoken commands by discharge, and 2 additional patients were able to follow commands after hospital discharge. Twenty-three of 104 patients (26%) patients without cognitive-motor dissociation were able to follow commands by discharge. Good outcome at 12-months occurred in 44% of patients with cognitive-motor dissociation and 14% without. Cognitive-motor dissociation remained predictive of GOS-E >4 even when patients with withdrawal of life-sustaining therapy were removed from the analysis.
Brain activation as measured on EEG occurred in 15% of clinically unresponsive acute brain injury patients, and in this small, single-center study, was associated with increased chance of good functional recovery. The prognostic implications of cognitive-motor dissociation are noteworthy, and merit further exploration. This study was limited by the small sample size; future studies should investigate this phenomenon in a larger population. In addition, this study analyzed patients with heterogeneous brain injury; further study in specific disease states may be revealing.