High frequency oscillations may improve somatosensory evoked potential detection of good outcomes in cardiac arrest survivors

Abstract-Text (inkl. Referenzen und Bildunterschriften)

Background:

Somatosensory evoked potentials (SEPs) are widely considered the most specific predictor of poor prognosis in coma when cortical responses are absent, but are nonspecific for a good prognosis when present. High-frequency oscillations (HFOs) within the SEP waveform appear to predict neurologic recovery in animal models, but are difficult to measure in an electrically complex clinical environment. In a retrospective study of patients with neurologic injury resulting in coma, we sought to develop a measure of HFOs from clinical diagnostic SEPs and determine if the addition of this measure improved the predictive value of a normal SEP result.

Materials and Methods:

We identified all adult patients (N=58) who underwent diagnostic SEPs from 2020-2022 at the Johns Hopkins Hospital for prognostication in coma, including all etiologies of coma. To capture short-term neurologic changes, the primary outcome measure was the final in-hospital Glasgow Coma Scale (GCS). Using SEP waveform data from only the patients with bilaterally present N20s, we developed a measure of HFO power using the normal time-frequency space (NTFS) and subsequent evoked-to-spontaneous ratios (ESRs) to account for background noise, with the median value for binary classification. We additionally trained a support vector machine (SVM) model on the HFO-ESRs as a secondary predictive model.

Results:

The mean age was 55 (+/-17) years, 31% female. Primary injuries resulting in coma were post-cardiac arrest hypoxic-ischemic (56%), primary respiratory hypoxia (18%), intracranial hemorrhage (12%), hypotension (6%), traumatic brain injury (3%), and other etiologies (6%). 34 patients had bilaterally present N20s; of these, 14 had final GCS>=9, and 20 had final GCS<=8, with 17 patients above and below median HFO-ESR. In those patients with N20s, the addition of averaged HFO-ESRs greater than the median value predicted final GCS>=9 with 55% specificity. Use of the SVM model further improved specificity to 65%.

Conclusions:

In this novel investigation of SEP-HFOs in multiple etiologies of coma, we developed a new measure of HFOs detectable in the electrically complex ICU environment. Incorporation of this measure in patients with positive N20s may improve the specificity of the SEP to predict in-hospital GCS outcomes, although this would require further exploration and validation in larger prospective studies.