Abstracts

Rationale: Extracorporeal membrane oxygenation (ECMO) is a life-saving procedure for children requiring cardiorespiratory support. The clinical scenario that requires ECMO (e.g. cardiac arrest) and process of cannulation often leads to cerebral injury. Quantitative EEG (qEEG) enables detection of subtle EEG background changes over extended time periods. Despite its utility, the use of qEEG has been mostly limited to seizure detection in pediatrics, while it is often used to detect delayed cerebral ischemia in adults with subarachnoid hemorrhage. Accordingly, we aimed to measure the qEEG capacity to detect cerebral injury in children on ECMO, while establishing normative values in pediatric patients.  Methods: Data were obtained retrospectively from 24 patients (71% male), age 44 weeks to 18 years (median 2.5 years) who were on ECMO from 2017 to 2018. All patients were monitored with continuous EEG (cEEG) for at least 24 hours. Clinical details extracted from the medical record included ECMO type, reason for ECMO, and time from ECMO cannulation to cEEG. We excluded patients with seizures. Consecutive EEGs were reviewed using PERSYST trending software with right and left hemisphere alpha delta ratios (ADR) calculated. The mean ADR between the right and left hemisphere was calculated per hour and examined over the monitoring session. Control recordings were obtained from patients without known parenchymal cerebral injury prior to ECMO cannulation who had negative imaging during ECMO or post decannulation. Patients with acquired injury during ECMO were defined as those with no prior history of cerebral injury, a first normal image after cannulation, but unambiguous brain injury on subsequent imaging. Patients with established injury were those with known injury prior to ECMO cannulation.  Results: Cardiac  arrest (58% (14/24) was the most common indication for ECMO. cEEG was started 11.25 (IQR 5.17 – 24.72) hours after cannulation to ECMO. cEEG was classified as slow disorganized in 83% (20/24), with asymmetry in 38% (9/24). Thirteen patients were controls. Established injury prior to cannulation was seen in 24% (8/24) and 13%, (3/24) had acquired injury. Focal injury was seen in all patients with acquired injury (3/3) and 63% (5/8) patients with established injury. The median ADR between the left and right hemisphere over the monitoring session for the controls was 0.01 (IQR 0.01 – 0.02), for established was 0.04 (IQR 0.03 – 0.05), and for acquired was 0.16 (IQR 0.12 – 0.17). There was a significant difference when comparing the median ADR of control patients to both those with acquired and established injury prior (0.01 vs 0.16 p= 0.01) (0.01 vs 0.04, p=0.005).   Conclusions: We show that qEEG signals may be useful in pediatric ECMO, and our data indicate normative values that mark a threshold above which injury is likely to have occurred. With further study this threshold may act as a target for interventions to prevent injury and decrease morbidity due to cerebral injury in patients on ECMO.  Funding: No funding