Between June 1973 and November 1980, 1,171 patients with metastatic breast cancer were treated with various doxorubicin-containing regimens at our institution (M.D. Anderson Hospital and Tumor Institute, Houston). Retrospective analysis of all 233 cases (20%) with liver metastases was done to correlate various clinical and biochemical characteristics with response to treatment, survival, and causes of death. A similar analysis was performed for 58 consecutive patients with liver metastases treated at this hospital between December 1955 and December 1957 with hormone therapy or single-agent chemotherapy. Objective responses were observed in 132 of 233 patients (57%) treated with combination chemotherapy. The median survival was 14 months in the 1970s and 5 months in the 1950s. Among patients who had liver metastases at the time of initial diagnosis of breast cancer, survival was longer for the group treated with combination chemotherapy. All cases were classified according to the number of organ sites involved by metastases. Patients with only liver metastases, or liver plus bone lesions had the longest survival. Other clinical and biochemical factors that correlated significantly with longer survival were: no prior chemotherapy, performance status of 1 to 2, absence of ascites, normal bilirubin and lactic dehydrogenase (LDH), SGOT less than or equal to 2 times normal and albumin greater than 4.5 g/dL. The main cause of death was multiorgan failure, with only 20% of patients dying of liver failure. The present study shows that the presence of liver metastases in breast cancer is not by itself an ominous factor. Most patients respond to therapy, and significant palliation with extended survival is possible for several prognostic subgroups. Further improvement in length and quality of survival is expected with earlier diagnosis.