To measure the incidence of sudden unexplained death in treated persons with epilepsy (SUDEP) and to identify risk factors for SUDEP, a cohort of 6,044 persons aged 15-49 years with more than four prescriptions for antiepileptic drugs (AEDs) was identified from the Saskatchewan Health prscription drug file. To exclude subjects whose sudden deaths (SUDs) might be misattributed to another chronic underlying disease, subjects with hospitalizations for cancer or heart problems were excluded. To exclude subjects without epilepsy, subjects with > 2-year AED treatment followed by AED-free time and subjects receiving < 1 U/day were excluded. The final cohort consisted of 3,688 subjects. Follow-up was started at the first AED prescription listed in the prescription drug file and ended at the earliest of the following: age 50 years, death, or last registration in the Saskatchewan Health. For 153 of 163 deaths occurring in the cohort, copies of anonymized death certificates were obtained and copies of anonymized autopsy reports of potential SUDEP cases were examined. There were 18 definite/probable SUDs and 21 possible SUDEPs, yielding a minimum incidence of 0.54 SUDEP per 1,000 person-years and a maximum of 1.35 SUDEP per 1,000 person-years. SUDEP incidence increased with male sex, number of AEDs ever prescribed, and prescription of psychotropic drugs and was highest in males with a history of treatment with three or more AEDs and four or more psychotropic drug prescriptions. Poisson regression showed a 1.7-fold increase in risk of SUDEP for each increment in maximum number of AEDs administered, a likely surrogate for severity and persistence of seizures.