Abstracts

RATIONALE: Neonatal EEG is not as powerful in predicting later occurrence of epilepsy as it is for neurologic outcome. EEG background abnormalities have been more useful outcome predictors than electrographic seizures. Burst suppression pattern, in particular, is associated with a poor outcome. However, this serious abnormality constitutes only a small proportion of discontinuous neonatal EEGs. The objective of our study was to determine whether any easily measurable parameters found in the larger population of discontinuous neonatal EEGs were predictive of the subsequent development of epilepsy. METHODS: We reviewed the EEGs and medical records of 43 term neonates with discontinuous EEGs treated at our institution from 1993 to 1998. We quantitated 10 clinical variables, and 10 EEG parameters relating to the bursts and the inter-burst intervals (IBI). These included burst amplitude and duration, maximum and mean IBI duration, IBI amplitude, presence of electrographic seizures, constant discontinuity, asynchrony, and sharp waves in the IBI. Univariate analyses were performed on all EEG and clinical variables against the presence or absence of subsequent epilepsy. The significant parameters were then examined in a multivariate logistic regression model. RESULTS: 86% of patients survived the neonatal period. Of those, 54 % developed epilepsy within a median follow up time of 2.5 years . In multivariate analysis, the presence of a mean IBI greater than 20 sec. was a highly significant predictor of epilepsy (p=0.003). The positive predictive value was 79 %, sensitivity was 65% and specificity was 85%. Clinical parameters such as abnormal neurologic exam and refractory seizures were unrelated to later epilepsy. CONCLUSIONS: A neonate with a discontinuous EEG containing a mean IBI longer than 20 sec. has a 79% probability of developing epilepsy. The presence of an easily measurable EEG predictor enhances the prognostic value of neonatal EEG.The fact that this interictal EEG parameter is the only predictor decreases the need for multiple recordings and facilitates prognosis prediction. Thus, EEG constitutes a better predictor of epilepsy than clinical data.