The safety and efficacy of patient-controlled analgesia for the long-term control of cancer pain was tested prospectively. Respiratory rates, mental status, and pain relief were recorded at baseline and compared with those during the study period. Patients had a lower analgesic demand (i.e., self-administered less morphine during the nighttime); specifically, dosing declined 48% from the daytime level. Respiratory rates did not change appreciably during the study and no cases of significant respiratory depression were encountered. Patients self-administered sufficient morphine to produce adequate but not complete pain relief in almost all trials. Pain relief was safely achieved by both intravenous and subcutaneous routes of administration in both the inpatient and outpatient settings. Mean 24-h morphine use stayed relatively constant even for patients receiving more than 2 weeks of treatment. In conclusion, patient-controlled analgesia is effective and safe therapy for the long-term control of severe cancer pain.