In a prospective study, antiepileptic drugs were discontinued in 264 children with epilepsy after a mean seizure-free interval of 2.9 years. They were then followed for a mean of 58 months to ascertain whether seizures recurred. Seizures recurred in 95 (36%) of the children. Etiology was a significant predictor of outcome (relative risk [RR] = 1.81). On multivariable analysis, significant factors in the idiopathic group included age at onset above 12 years (RR = 5.4), a family history of seizures (RR = 3.1), the presence of slowing on the electroencephalogram prior to medication withdrawal (RR = 2.4), and a history of atypical febrile seizures (RR = 2.8). Specific epileptic syndromes such as juvenile myoclonic epilepsy and benign rolandic epilepsy were also significant predictors of outcome. In the remote symptomatic group, significant predictors of outcome included age at onset older than 12 years (RR = 3.6), moderate to severe mental retardation (IQ < 50) (RR = 2.8), a history of atypical febrile seizures (RR = 2.0), and a history of absence seizures (RR = 0.4). The majority of children with epilepsy in remission while on antiepileptic drug therapy will remain seizure free when medications are withdrawn. A few readily available parameters distinguish those with a good prognosis from those in whom seizures are likely to recur. These data provide the framework for the clinical decision making for withdrawal of medications in these children.