Diagnosing refractory epilepsy would facilitate referral for specialist pharmacological review and early consideration of epilepsy surgery. An outcomes study was undertaken in an unselected cohort of newly diagnosed patients to determine the number of antiepileptic drug (AED) regimens needed to be failed before the epilepsy could be designated as pharmacoresistant. Between July 1982 and May 2001, 780 adolescents and adults prescribed their first AED at the Western Infirmary in Glasgow, Scotland provided longitudinal data suitable for analysis. Overall, 504 (64.6%) patients became seizure free for at least 12 months. Of these, 462 (59.2%) remained in remission, while 42 (5.4%) relapsed and subsequently developed refractory epilepsy. The relapse rate peaked at 10.4% after 8 years of follow-up. The other 276 (35.4%) patients were uncontrolled from the outset. Prognosis appeared better in seniors (85% remission, P < 0.001) and adolescents (65% remission, P < 0.01) than in the remainder of the population (55% remission). Overall response rates with the first, second and third treatment schedules were 50.4, 10.7 and 2.7%, respectively, with only 0.8% patients responding optimally to further drug trials. Patients not tolerating at least one AED schedule did better than those failing because of lack of efficacy. These data suggest that suitable patients failing two AED regimens should be referred for epilepsy surgery. Those who do not attain long-term seizure freedom with the first three treatment schedules are likely to have refractory epilepsy.