EEG and other early predictors of epilepsy remission: a community study

The course of seizures was reviewed in all 306 residents of Rochester, Minnesota who had epilepsy diagnosed between 1935 and 1978, lived in the region greater than or equal to 5 years after diagnosis, and had an electroencephalogram (EEG). The life-table probability of having achieved 5 years seizure-free (FYSF) by 20 years after diagnosis was 75%. In a multivariable proportional hazards model, these three variables predicted a significantly higher rate of achieving FYSF: no early-life brain damage (remission rate ratio = 2.27, p less than 0.01), no generalized epileptiform activity (rate ratio = 1.58, p less than 0.05), and never having had a generalized tonic-clonic seizure (rate ratio = 1.4, p less than 0.05). The same three variables, plus age at diagnosis, were descriptors of the rate of reaching FYSF off medication. Gender, family history, type of seizure, and EEG findings other than generalized epileptiform activity were not significantly related to either end point. The predictor set did 15% better than prior probabilities in forecasting FYSF. Although informative about group experience, these predictors are weak clinical discriminators, singly or in a set. FYSF occurred in high proportion of even those persons whose history, clinical examination, or EEG findings placed them in a less favorable stratum of one or more predictors.