Epidural administrations of midazolam have been empirically used to enhance epidural analgesia with little understanding of its pharmacokinetics. The present study compared the plasma concentration of midazolam after an epidural administration and that after an intravenous administration, and observed the analgesic effect of epidural midazolam on the postoperative pain. 45 ASA physical status I and II patients, aged 31 to 58 years, undergoing propofol intravenous anesthesia. Patients were randomly assigned into three groups: received a pre-operative injection of 10 ml midazolam (0.05mg/kg) either into (1) epidural space (Mepi group), (2) intravenously (Miv group), (3) or an epidural injection of 10 ml normal saline as control (Ctr group) for the assessment of analgesia. Plasma concentrations of midazolam in the epidural group and in the intravenous group were measured by high performance liquid chromatography, and the elimination half-life of midazolam was calculated. Analgesic effects were evaluated by Visual Analog Scale pain scores at rest, time to first request for analgesic, and morphine consumption during the initial post-operative time of 48 hours. Midazolam plasma concentration peaked at 20 minutes after injection in Mepi group, but it peaked immediately in Miv group. The Mepi group had lower plasma concentrations of midazolam at peak time and a longer elimination half-life of midazolam than the Miv group. Visual Analog Scale pain scores or time to first request for analgesia are similar among three groups, but patients in the epidural group consumed less morphine than patients in the control group. The results suggest epidural injections of midazolam produce different changes in the plasma concentration of midazolam from those of intravenous injections. In addition, the pre-operative epidural administration of midazolam reduces the need for analgesics after surgery.