Amplitude Integrated Electroencephalography: Simulated Assessment of Neonatal Seizure Detection in PICU Patients

Abstract

OBJECTIVES: Amplitude integrated electroencephalography (aEEG) is a mainstay of care in neonatal ICUs; however, knowledge gaps exist in relation to its accuracy for identifying seizures in older children. We aimed to review the diagnostic accuracy of existing neonatal seizure detection criteria for seizure detection in older children in hospital. DESIGN: Retrospective study. SETTING: PICU/Neurophysiology Department in Dublin. PATIENTS: One hundred twenty patients (2 mo to 16 yr old) were chosen from a database of formal 10–20 system, 21-lead electroencephalography recordings (2012–2020), comprising 30 studies with seizures, 90 without. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Electroencephalography studies containing electrographic seizures (ESzs) were annotated to describe number, duration, distribution, and spread. Two-channel aEEG (using leads C3–P3, C4–P4) recordings were generated and independently reviewed by a professional specialist in clinical neurophysiology blinded to outcome and without reference to the raw electroencephalography trace. Logistic regression was used to identify factors associated with correct seizure identification on aEEG. Median patient age was 6.1 years. Abnormal recordings featured 123 seizures. Status epilepticus (SE) was evident by electroencephalography in 10 cases. Using neonatal criteria, aEEG had a sensitivity of 70% and negative predictive value of 90% for identifying any ESz. Accurate detection of individual seizures was diminished when seizures were very short or occurred during waking. Sensitivity for individual seizures was 81% when seizures less than 1 minute were excluded. aEEG correctly identified SE in 70% of the 10 cases, although ESz were confirmed to be present in 80% of this subpopulation. CONCLUSIONS: aEEG criteria for neonatal seizure identification can be applied with caution to older children and should be supplemented by formal electroencephalography. Seizure identification is better for longer seizures and those arising from sleep. SE is not always recognized by aEEG among older children.