EEG registration after cardiac arrest: On the way to plug and play?

The role of electroencephalography (EEG) for prognostication in patients admitted to the intensive care unit (ICU) after cardiac arrest has changed in the last decade. In the old days, it was recommended to start investigations for determination of the prognosis in patients who remained in coma after clearance of sedation administered during target temperature management. ^,  Somatosensory evoked potentials (SSEP) were advised as first diagnostic test and if the cortical N20 response was not bilaterally absent, an EEG was the next step. In daily practice, this meant the EEG was registered 48–96 h after cardiac arrest. Several recent publications have shown that the information obtained from an EEG is much more valuable when recorded in the first 12−24 h after cardiac arrest. ^{,  ,}  So, the general advice is shifting to a start of the EEG registration soon as possible after ICU admission, but this was not yet incorporated in the most recent official guidelines. This is easier said than done in most hospitals, especially as only 40% of all cardiac arrests occur during office hours. As the results of such an EEG registration may be used for decisions on life or death, e.g. continuation or withdrawal of supportive care, optimal recording and reliable assessment are crucial. This means that the department of (clinical) neurophysiology has to become a true 24/7 ‘partner in crime’. Even in centres where continuous EEG registration is well organised, it takes a mean of 11 h to start EEG recording. At 12 h after cardiac arrest, about 50% of the patients were hooked onto EEG. For start of EEG registration immediately after ICU admission, a dummy proof system that can be used by anyone at the bedside without extensive training, would be a very valuable solution. The study of Kortelainen et al. in this issue of Resuscitation is a fine example of the way to go. They investigated the prognostic value of frontal slow wave activity (SWA), recorded during propofol sedation in 100 patients after cardiac arrest. EEG recording was performed using the BrainStatus, a self-adhesive electrode which is placed on the forehead and behind the ears of the patient. This yields an EEG registration with a limited number of electrodes, encompassing frontal and temporal parts of the brain. Is this enough for a reliable registration? Studies on continuous EEG registration with a limited number of electrodes have shown promising results. Tanner et al. investigated a sub-hairline electrode set with 6 ECG electrodes for classification of EEG recordings in comatose ICU patients for seizure detection. They reported a high specificity but only moderate sensitivity suggesting that the type of brain injury, locally or more globally, can affect the results of limited electrode use. Brain injury after cardiac arrest is a global type of injury and it was shown that an EEG recording with less electrodes can suffice for reliable identification of the patient with severe brain injury and a poor outcome. The design of the BrainStatus electrode is so clear and straightforward that it can be applied by trained nurses enabling rapid EEG registration after ICU admission. Interpretation of the EEG recording at the bedside can be supported by using basic training programs for ICU team members. Another approach is “translation” of the EEG to an easy to understand figure. Ideally, in patients after cardiac arrest, the displayed figure gives a reliable prognosis of the outcome of the individual patient. Kortelainen et al. investigated the prognostic value of the frontal SWA displayed by the C-trend index. This index, developed by the same group, is based on medical software using several power and connectivity features in the EEG. The SWA is represented by a figure from 0 to 100. The results of the current study show that in the first 12 h the largest difference in frontal SWA is found between patients with a good or a poor outcome. This yields a specificity for prediction of a poor outcome of 94.7% (95% CI 83.4%–99.7%) using a cut-off C-trend index value of 20. Only 100 patients where included, so confirmation in a multicentre study in much more patients studies could lead to introduction of this technique on a really large scale. As the EEG pattern is known to recover over time, also in poor outcome patients, the prognostic accuracy was found to decrease in later parts of the EEG registration. Another index representing the EEG and translating it into a prognostic figure is the Cerebral Recovery Index (CRI) as developed by Tjepkema-Cloostermans et al. ^,  They also have reported very promising results for prognostic accuracy and outcome prediction. For optimal development of these indexes large datasets and computer techniques such as deep-learning and machine learning are necessary. With these tools, EEG monitoring on the ICU can be developed further. Not only in patients admitted to the ICU after cardiac arrest, but theoretically also in patients with other causes of acute brain injury or for early detection of brain injury during long-term sedation. Methods that enable plug and play EEG recording could improve early monitoring of the brain in all these ICU patients.