OBJECTIVE: To report and analyse the pattern of end-of-life decision making for terminal Chinese cancer patients. DESIGN: Retrospective descriptive study. SETTING: A cancer clinical trials unit in a large teaching hospital. PATIENTS: From April 1992 to August 1997, 177 consecutive deaths of cancer clinical trial patients were studied. MAIN MEASUREMENT: Basic demographic data, patient status at the time of signing a DNR consent, or at the moment of returning home to die are documented, and circumstances surrounding these events evaluated. RESULTS: (...) DNR orders were written for 64.4% of patients. Patients in pain (odds ratio 0.45, 95% CI 0.22-0.89), especially if requiring opioid analgesia (odds ratio 0.40, 95% CI 0.21-0.77), were factors associated with a higher probability of such an order. Thirty-five patients were taken home to die, a more likely occurrence if the patient was over 75 years (odds ratio 0.12, 95% CI 0.04-0.34), had children (odds ratio 0.14, 95% CI 0.02-0.79), had Taiwanese as a first language (odds ratio 6.74, 95% CI 3.04-14.93), or was unable to intake orally (odds ratio 2.73, 95% CI 1.26-5.92). CPR was performed in 30 patients, none survived to discharge. CONCLUSIONS: DNR orders are instituted in a large proportion of dying Chinese cancer patients in a cancer centre, however, the order is seldom signed by the patient personally. This study also illustrates that as many as 20% of dying patients are taken home to die, in accordance with local custom. (shrink)