For the treatment of patients with chronic refractory epilepsies, information about the long-term efficacy and safety profile of any new antiepileptic drug is crucial. Topiramate has been proven to be effective in patients with refractory chronic partial epilepsies in short-term controlled clinical trials, but the long-term retention, long-term efficacy, and long-term side-effect profile have not been sufficiently investigated. We analyzed all patients who had been treated with topiramate in the Epilepsy Centre Kempenhaeghe from the introduction of the drug in the spring of 1993 up to a final assessment point in mid-2002. In total, 470 patients were identified. The data show that the clinical dose achieved was about 200mg/day, reached after approximately 6 months of treatment. Further dose escalation in the survivors was slow, with a mean dose of about 300 mg/day after 24 months of treatment. Mean titration dose is 25mg/week, but titration strategy is mostly individual and responds to patient complaints. With respect to seizure frequency, 10-15% of the patients were seizure-free at the 6-month evaluation; 4 patients achieved a 2-year remission. Retention rate was 53% after 1 year, 45% after 2 years, 38% after 3 years, and 30% after 4 years. At 4 years, almost 70% of the patients had discontinued topiramate. The main reason was adverse events, which accounted for about 65% of the discontinuations. Behavioral side effects were dominant, with mental slowing (27.6%), dysphasia (16.0%), and mood problems (agitation: 11.9%) being the most frequently reported side effects. In about 10% of the patients side effects led to discontinuation despite the obvious favorable effects on seizure frequency. Comparisons between the patients who discontinued topiramate treatment and those who continued topiramate showed that discontinuation was associated with comedication (vigabatrin and lamotrigine). Our conclusion is that TPM is associated with a high incidence of side effects in clinical practice, affecting long-term retention. Meaningful prognostic factors that may help us in clinical decision making, i.e., to prevent the side effects or to help us identify those at risk, have not been found.