In a prospective study, patients undergoing abdominal cancer surgery were randomly allocated to receive either general anesthesia with fentanyl intravenously and postoperative analgesia with parenteral morphine (GA group) or general anesthesia combined with epidural bupivacaine and epidural morphine for postoperative pain relief (EP group). Analgesia was tested on a visual pain scale. Pulmonary complications were evaluated by clinical complications, blood gas analysis, x-ray film changes, and pulmonary volumes (vital capacity, forced expiratory volume in 1 second). Measurements were performed on the day before the operation and on the first 5 postoperative days. In the EP group the pain relief was significantly better on the first day (p less than 0.03). Whatever the criteria used, the rates of pulmonary complications were similar in the two groups: clinical complications 21% versus 26%, radiologic complications 50% versus 64% for GA and EP groups, respectively. Postoperative PaO2 and spirometric values were similar in the two groups. Postoperative epidural analgesia may improve the patient's comfort but does not decrease the incidence of pulmonary complications.