Abstracts

Rationale: EEG features which distinguish non-convulsive status epilepticus (NCSE) and encephalopathy are controversial. Readers may use a number of EEG domains—discharge-complex morphology, frequency, rhythmicity/evolution and focality to conclude whether a particular record shows NCSE. The prioritization of these features, however, has not been formalized. We classified EEGs previously interpreted as showing NCSE into NCSE “probability grades”. We then determined the association between NCSE “probability grades”, EEG features, and patients’ clinical outcomes: history of seizures alone, acute cortical injury with a history of seizures, or acute injury only.Methods: We identified 49 EEG records with NCSE patterns on EEG summaries. Two electroencephalographers independently read and graded the EEGs, followed by a conference to reconcile any differences. Each record was assigned a NCSE probability grade on a scale of 1-4: 1 = borderline; 2 = possible; 3 = probable; and 4 = highly probable. The spectrum of features used to determine probability grades (low to high) were: waveform morphology (slow, broad sharp/triphasic, sharp/spike), rhythmicity (non-rhythmic, quasi-rhythmic, rhythmic), frequency (<2 cps, 2-2.5 cps, > 2.5 cps) and focaility (diffuse, asymmetric, focal). The readers independently determined final diagnosis for patients while blinded to EEG results. Based on AED response, clinical history and imaging, patients were classified as having: 1) seizures alone, 2) seizures and acute cerebral injuries and 3) acute cerebral injuries only. We then compared NCSE EEG “probability grades” and patients’ final clinical diagnosis. Results: High probability NCSE EEG grades were characterized by focal, frequent (>2.5cps), and very sharp or very rhythmic waves; diffuse patterns were >2.5cps, sharp and rhythmic. Lower probability grades were < 2.5 cps, slow wave morphology, and quasi-rhythmic or non-rhythmic. Nearly all (10 of 11) patients with “highly probable” NCSE on EEG had clinical NCSE (see table). Lower probability grades, however, had a variable association with seizures and acute injuries. An overall association between EEG probability grades and clinical NCSE and acute injuries was present only when patients with triphasic EEG patterns were excluded (Chi-square = 12.7, p=.0475). Most patients with triphasic waves (7/10) had acute injuries, e.g. ischemic injury. Conclusions: Patients with highly probable NCSE grades on EEG were strongly associated with clinical NCSE whereas patients with lower probability grades had a variable association with seizures and acute injuries. Triphasic waves were not a reliable marker of NCSE. Our study shows that a specific constellation of EEG features may distinguish patients with seizures from patients with encephalopathies only.