We estimated the prevalence and incidence of epilepsy in The Netherlands using drug-dispensing information from the PHARMO database, containing medication histories of nearly 300,000 individuals. An algorithm based on antiepileptic drug prescription records was used to identify patients with epilepsy requiring medication for seizure control. The algorithm was validated by comparing positive algorithm identifications to medical diagnoses from general practitioners and hospital records. In 1990-1991, the algorithm revealed 1158 patients with "certain" epilepsy, and 451 patients with "probable" epilepsy. Epilepsy was present in 93% of patients on polytherapy, and 58% on monotherapy. Clonazepam monotherapy was non-specific for epilepsy. The use of carbamazepine monotherapy for epilepsy was age-dependent. After correcting the algorithm for these drugs, and standardizing to the Dutch population, the point prevalence of epilepsy was 4.8/1000 (95% CI: 4.5-5.0). The incidence rate was 0.72/1000 person-years (95% CI: 0.65-0.79). Using drug-dispensing data for epilepsy medication, it is possible to make valid estimations of the number of epilepsy patients requiring drug therapy.