The medical records of 102 autopsies performed in Royal Prince Alfred Hospital on patients treated by a medical oncology unit between January 1977 and May 1979 were analysed. Organ failure was the recorded cause of death in 42%; infection in 23%; carcinomatosis in 18%; haemorrhage in 15%; and metabolic derangement in 2%. 27% of patients had autopsy evidence of severe coexistent non-neoplastic disease, predominantly widespread atheroma and coronary artery disease. The immediate cause of death was unrelated to cancer in 19% and treatment-related in 19%. In 13%, death accompanied a high intake of narcotic analgesia during the pre-terminal 24 hours. A terminal care policy was adopted for 56 (55%) patients during their last month and three-quarters of narcotic-related deaths occurred in this group. 55% of the patients received some form of aggressive anti-tumour therapy in their terminal month and 30% of these died of treatment-related causes. 25% of the patients underwent an invasive investigation during their terminal month. Correlation of death certificate, clinical and autopsy causes of death showed the death certificate to be wrong in 41% of cases, with 29% of these errors being of potentially epidemiological importance. Pre-mortem assessment of the cause of death was in error in 26% of patients.