INTRODUCTION: Experimental conditions are not mirrored by clinical practice. OBJECTIVE: To study the efficacy of vigabatrin in the usual conditions of everyday clinical practice. PATIENTS AND METHODS: Retrospective review of all epileptic patients treated with vigabatrin in a neuropediatric outpatient clinic. OUTCOME MEASURES: a) Persistent seizure frequency reduction > or = 50%, and total control of seizures. Potential predictors of response were studied by logistic regression. b) Duration of VGB therapy, studied by the Kaplan-Meier method and its associated log-rank test. RESULTS: 113 patients with: partial symptomatic epilepsy (38%), partial cryptogenic (25.6%), partial idiopathic (6%), West syndrome (14%), Lennox-Gastaut syndrome (6%), other syndromes (9.7%). Reduction of seizure frequency was attained by 60% of patients at 3 months, sustained during 12 months by 40%, and during 5 years by 14%. Total control of seizures was present in 33% of patients at 3 months, persisted 12 months in 18% and 5 years in 2%. Independent predictors of a poor outcome were generalized seizures (except infantile spasms) and cerebral palsy, among others. The probability of continuing vigabatrin (VGB) therapy was 78% at 6 months, 55% at 2 years and 32% at 5 years. Duration of therapy was modified by early therapeutic response and antecedent of status epilepticus, among others. Adverse events were recorded in 18.5%. Visual fields were not studied in these series. CONCLUSIONS: For the time being, VGB-treated patients belong to the difficult-to-treat group. Percentages of responders depend upon duration of follow-up.