Fentanyl, unlike morphine, is highly lipophilic and rapidly diffuses out of the epidural space. Respiratory depression is, therefore, unlikely when fentanyl is given epidurally. However, much of fentanyl's analgesic effect is mediated by systemic rather than spinal receptor binding. To test this hypothesis, we performed a prospective, double-blind, cross-over study comparing epidural and intravenous (IV) administration of fentanyl in 16 patients for the first 12 h after lower abdominal or lower extremity surgery. To allow direct comparison of these two routes of administration, patient-controlled analgesia was used so patients could self-titrate their analgesia. Patients were randomized to receive fentanyl initially by an epidural (group A, n = 8) or IV (group B, n = 8) catheter for 6 h, after which they were crossed-over to the alternate route by means of a hidden three-way stopcock. The degree of analgesia was subjectively evaluated by a visual analogue scale, and by an observer rating patient's comfort and sedation. Cumulative dosage of fentanyl was recorded, and plasma fentanyl concentrations were measured. The onset of analgesia and increase in plasma fentanyl concentrations were more rapid with intravenous fentanyl. However, after 60 min, analgesia (visual analogue scale 2–4 cm) or plasma fentanyl concentrations (0.3–0.7 ng/mL) did not differ between the two routes of administration. There were also no significant differences in the cumulative dosage of fentanyl within each group (epidural vs IV) or between the groups. Thus, the analgesic effects of epidural fentanyl appear largely mediated by systemic absorption. Intravenous fentanyl achieves a similar degree of analgesia and a more rapid onset of effect without the need for epidural catheterization.