We studied recurrence risks and predictive factors of relapse after antiepileptic drug (AED) discontinuation in a prospective analysis of 425 children with epilepsy who had not had a seizure for at least 2 years (follow-up after withdrawal 1.6-12 years, mean 8 years). Factors closely related by multivariate analysis to relapse were neurologic abnormalities, mental retardation, seizure type (infantile spasms, absence seizures), and appearance or persistence of EEG abnormalities during the course of the illness and before discontinuation. When multivariate analysis was performed to evaluate outcome of patients with a first relapse (isolated vs. multiple relapses), the variables closely related to a poor prognosis were etiologic factors, first relapse characterized by more than one seizure in a 24-h period, seizure-free period less than 4 years, unchanged seizure type at first relapse, more than one AED for seizure control, and abnormal EEG before the first relapse. In itself, resumption of therapy did not influence outcome. At the study cutoff point, 88% of patients with relapse were again seizure-free. We conclude that AEDs can safely be discontinued if predictive factors are considered to individualize the risk of relapse for each patient.