OBJECTIVE To evaluate the influence of the route of administration [epidural, intravenous (IV), or transdermal] on onset and quality of analgesia and to evaluate the pharmacokinetics of continuous administration of fentanyl. METHODS Randomized, open, single-dose, prospective study. METHODS Postanesthesia care unit of a university hospital. METHODS 54 ASA physical status I-III patients scheduled for lower major abdominal, gynecologic, or urologic surgery. METHODS Patients were assigned to 1 of 3 comparable groups, to receive, for 72 hours postoperatively, an IV (Group IV) or epidural (Group EP) constant-rate infusion of fentanyl (loading dose 1.5 micrograms/kg, infusion rate 1 micrograms/kg/hr), or a transdermal patch (Group TTS), applied preoperatively, delivering fentanyl 75 micrograms/hr. RESULTS Pain intensity, vital signs, blood gas status, and plasma fentanyl concentration (Cp) were measured for up to 96 hours postoperatively. Onset of analgesia was delayed in Group TTS. Analgesic efficacy was similar for all 3 routes of administration except during the first 4 hours in Group TTS. Initially, the patients in Groups TTS and IV needed significantly more rescue morphine than those in Group EP. The time course for fentanyl Cp in Groups TTS and IV evolved to similar plateau levels. However, fentanyl Cp continued to rise throughout the study period in Group EP, reaching concentrations that elicited hypoxemia after 48 hours. A significant negative relationship existed between fentanyl Cp and oxygen saturation in this group. All 3 routes of administration showed similar frequencies of side effects (i.e., nausea, vomiting, pruritus, and ileus). CONCLUSIONS The epidural, transdermal, and IV administration of identical doses of fentanyl given at a constant rate provided almost equivalent degrees of analgesia. But continuing epidural administration produced a steady rise in systemic fentanyl concentrations into the ventilatory-depressant range, affecting the hypoxemic regulation of breathing.