Purpose: We evaluated the correlation between quantitative background activities on electroencephalography (EEG) and serum neuron specific enolase (NSE) in patients with hypoxic-ischemic encephalopathy as well as a diagnostic value of prognostication.
Methods: This retrospective cohort study enrolled patients with return of spontaneous circulation after cardiac arrest from March 2010 to March 2020. The inclusion criteria were (1) older than the age of 16 years and (2) patients who had both EEG and NSE. The median time for EEG and NSE were 3 days (interquartile range 2-5 days) and 3 days (interquartile range 2-4 days), respectively. The quantification of background activity was conducted with the suppression ratio (SR). We used a machine learning (eXtreme Gradient Boosting algorithm) to evaluate whether the SR could improve the accuracy of prognostication.
Results: We enrolled 151 patients. The receiver operating characteristic analysis revealed a cut-off value of serum NSE and the SR for poor outcome, serum NSE (>31.9 μg/L, area under curve [AUC] = 0.88), and the SR (>21.5%, AUC = 0.75 in the right hemisphere, >34.4%, AUC = 0.76 in the left hemisphere). There was a significant positive correlation between the severity of SR and the level of NSE (ρ = 0.57, p < 0.0001 for the right hemisphere, ρ = 0.58, p < 0.0001 for the left hemisphere). The SR showed an excellent diagnostic value for predicting poor outcome (93% specificity, 60% sensitivity in the right hemisphere and 93% specificity, 58% sensitivity in the left hemisphere). With machine learning analysis, there was an increment in distinguishing the neurological outcome by adding SR on clinical factors.
Conclusions: The SR showed a positive correlation with the level of serum NSE. The diagnostic value of the SR for predicting poor outcome was excellent, suggesting that it can be a possible biomarker for neuroprognostication in patients with hypoxic-ischemic encephalopathy.
Copyright © 2023 by the American Clinical Neurophysiology Society.