Chemotherapy has become a routine part of the treatment of osteosarcoma. However, the precise role of preoperative chemotherapy remains in question. Between 1979 and 1982, a group of 40 patients were treated by multimodality therapy consisting of preoperative chemotherapy (intra-arterial cis-platinum and systemic adriamycin), surgery, and postoperative chemotherapy. Survival in this group is 64%, while continuous disease-free survival is 58%. Although age, sex, tumor size, site, and classification were found to be prognostic factors, histologic evidence of response to preoperative chemotherapy, measured as percent tumor necrosis, was found to be the most significant prognostic factor. When continuous disease-free survival was calculated as a function of tumor necrosis it was 91% in patients with greater than or equal to 90% tumor necrosis, while it was 14% in patients with less than 90% tumor necrosis. At initial presentation, 7% of patients were judged limb-salvage candidates. But due to the local effects of preoperative chemotherapy, 60% ultimately underwent limb-salvage surgery. Preoperative arteriograms were a reliable means of monitoring response to chemotherapy and served as an indicator of residual viable tumor. Using arteriogram directed planes of section, postchemotherapy, specimens were "mapped" and analyzed for chemotherapy effect. When present, residual viable tumor was preferentially found at the interface of tumor and normal anatomic structures; "sanctuary sites." It is necessary that standard methods for analyzing postchemotherapy specimens be developed; a technique is described.