It has remained unclear whether epidural opioid analgesia permits better recovery of postthoracotomy pulmonary function than an optimal method of systemic opioid administration. Lumbar epidural meperidine infusions were compared with intravenous patient-controlled analgesic (PCA) meperidine infusions in a prospective randomized unblinded study for 72 hours postthoracotomy. Before induction of general anesthesia, patients received a bolus of meperidine, 1 mg/kg, and an infusion of meperidine, 0.33 mg/kg/hr, was started via either a lumbar epidural or intravenous catheter. Postoperatively, the meperidine infusion rates were titrated as needed for analgesia. In addition, the intravenous group received meperidine, 10 mg per dose, as required, from a patient-controlled analgesia pump. No other opioid was administered during the study period. Patients were studied for recovery of spirometric tests of pulmonary function, visual analog pain scores, sedation, arterial blood gases, meperidine dose requirements, radiographic pulmonary complications, and neurologic signs and symptoms. A subgroup of 10 patients (5 from each group) had venous blood samples drawn every 24 hours for 96 hours and assayed for serum meperidine and normeperidine concentrations. Epidural meperidine analgesia was associated with improved postthoracotomy pulmonary function, better analgesia scores, and lower meperidine dose requirements than intravenous PCA meperidine. There were no differences between the epidural versus intravenous PCA subgroups with respect to serum meperidine or normeperidine levels. Normeperidine levels greater than 300 ng/mL were associated with an increased incidence of shakiness and/or tremors. Meperidine provides satisfactory postthoracotomy analgesia via a lumbar epidural infusion. This analgesia is associated with improved recovery of postoperative pulmonary function when compared with an intravenous PCA meperidine infusion.