Our purpose was to compare the initial efficacy and retention at 5 years of lamotrigine (LTG) and vigabatrin (VGB) in patients with severe childhood-onset epilepsies, especially with regard to loss of efficacy. Add-on therapy in 95 young patients with severe epilepsies in a neuropaediatric department was prospectively followed in open label studies for up to 5 years. VGB group: 56 patients (mean age: 11.1 years). LTG group: 39 patients (mean age: 13.6 years). Definition of initial responders: more than 50% reduction in seizure-frequency after 6 weeks of VGB treatment or after 4 months of LTG treatment. Definition of retention at 5 years: percentage of patients still taking LTG or VGB after 5 years. Definition of loss of efficacy: return to the baseline seizure frequency. The results were: VGB group: after 6 weeks 18 of 56 patients (32%) were initial responders. The retention at 5 years was five of 56 patients (8.9%). One patient (1.8%) was still seizure free. A loss of efficacy occurred in 10 of the 18 initial responders, usually within the first 9 months after the initial response. In the LTG group, after 4 months, 11 of 39 patients (28%) were initial responders. The retention at 5 years was 10 of 39 patients (25.6%). Five patients (12.8%) were still seizure free. The rate of adverse events was equal in both groups (41%). All but one occurred within the first 6 months of treatment. Our study in patients with difficult to treat childhood epilepsies suggests that in clinical practice patients were less likely to be discontinued from LTG than from VGB within 5 years, after similiar initial efficacy. This was mainly due to a loss of efficacy in VGB treatment early after initial response.