Neurophysiologic Features Predicting Brain Injury During Pediatric ECMO Support

Abstract

Abstract Background/Objective: Extracorporeal membrane oxygenation (ECMO) provides life-saving support to critically ill patients who experience refractory cardiopulmonary failure but carries high risk of acute brain injury. We aimed to identify characteristics predicting acute brain injury in children requiring ECMO support.Methods This is a prospective observational study from 2019–2022 of pediatric ECMO patients undergoing neuromonitoring including continuous electroencephalography, cerebral oximetry, and transcranial Doppler ultrasound (TCD). Primary outcome was acute brain injury. Clinical and neuromonitoring characteristics were collected. Multivariate logistic regression was implemented to model odds ratios (OR) and identify the combined characteristics that best discriminate risk of acute brain injury using the area under the receiver operating characteristic curve (AUROC).Results Seventy-five pediatric patients requiring ECMO support were enrolled in this study. Of the seventy-five patients enrolled, nineteen experienced acute brain injury (25.3%), including seven (36.8%) with arterial ischemic stroke, four (21.1%) with hemorrhagic stroke, seven with hypoxic ischemic brain injury (36.8%) and one (5.3%) with both arterial ischemic stroke and hypoxic ischemic brain injury. Univariate analysis demonstrated acute brain injury to be associated with maximum hourly seizure burden (p = 0.026), epileptiform discharges (p = 0.020), electroencephalographic suppression ratio (p = 0.014), increased interhemispheric differences in electroencephalographic total power (p = 0.023) and amplitude (p = 0.011), and increased differences in TCD Thrombolysis in Brain Ischemia (TIBI) scores between bilateral middle cerebral arteries (p = 0.049). Best subset model selection identified increased seizure burden (OR = 3.86, partial R-squared 0.40, p = 0.013), increased quantitative electroencephalographic interhemispheric amplitude differences (OR = 2.69, partial R-squared 0.18, p = 0.007), and increased interhemispheric TCD TIBI score differences (OR = 4.97, partial R-squared 0.22, p = 0.005) to be independently predictive toward acute brain injury (AUROC = 0.92).Conclusions Increased seizure burden, increased interhemispheric differences in quantitative electroencephalographic amplitude and increased TCD TIBI scores each independently predict acute brain injury in children undergoing ECMO support.