This study assessed sedation in terminal cancer patients in terms of three characteristics: frequency; relationship to intractable symptoms; and the extent to which medical staff, family, and patients found sedation to be ethically acceptable and efficacious. Two hundred seventy-six consecutive patients, who were admitted to the palliative care unit of National Taiwan University Hospital in Taiwan between August 1998 and the end of May 1999, were enrolled. A recording form was completed every day. This included demographic data, pain and common symptom scores, and the use of sedation in the terminal phase. Seventy (27.9%) of 251 patients who died received sedation. Sedation was administered to relieve agitated delirium in 40 (57.1%), dyspnea in 16 (22.8%), severe pain in 7 (10%) and insomnia in 5 (7.2%). The drugs used for sedation were haloperidol in 35 (50%), midazolam in 17 (24.3%), and rapidly increasing dosage of morphine in 9 (12.9%). In fewer than half (42.9%) of the patients, sedation was with the consent of both patient and family, and half (50%) had the consent of family alone. The overwhelming majority of medical staff and family felt the decision to use terminal sedation was ethically acceptable. There was no significant difference in survival time between sedated and non-sedated patients (28.49 vs. 24.71 days, t = -0.791, P = 0.430). Positive ethical acceptability and higher satisfaction with symptom control with terminal sedation were found in both medical staff and family in this study. Further work is needed to find the most appropriate time of intervention and to improve management of refractory symptoms in dying patients.