Ifosfamide (IFOS) 5 g/m2 and its parent analog Cyclophosphamide (CYCLO) 1.5 g/m2 were studied in a randomized phase II study, accruing 171 patients with advanced soft tissue sarcoma. Both drugs were administered as 24 hr infusions, every 3 weeks, with comcomitant Mesna 400 mg/m2 i.v. bolus 4 hourly X 9 doses. Twenty-four patients were ineligible and 12 were not evaluable. The groups were well matched for age, previous chemotherapy (42% of the total) or radiotherapy, the presence of distant metastases and performance status, but there were more females (59% vs. 45%) in the IFOS arm. Among the 68 evaluable patients receiving IFOS, there were 2 CR, 10 PR (overall response 18%), 27 SD and 29 PD. For CYCLO, the corresponding results in 67 patients were 1 CR, 4 PR (overall response 8%), 23 SD and 39 PD. Using the chi-square test the P values for response rate and linear trend were 0.13 and 0.04 respectively. Response rates were higher for females (20% vs. 5%, P = 0.01) and patients who had not received previous chemotherapy (19% vs. 4%, P = 0.01). Fourteen of the 17 responses came from a group of 43 females, who had not received previous chemotherapy, for whom the overall response rate was 37.5%. Remissions were noted in only 4 histological subtypes (centrally reviewed material), i.e., 5 of 17 synovial sarcomas, 7 of 13 mixed mesodermal sarcomas and 2 of 7 fibrosarcomas. One of the 31 leiomyosarcomas responded to Cyclophosphamide. Durations of response did not differ significantly between the 2 arms--median 26, range 10-81+ weeks. Leucopenia was significantly more severe on CYCLO, particularly in patients who had received previous chemotherapy (P = 0.007). Serious infections occurred in approx. 7% of patients with no difference between the two drugs, although there was one toxic death on CYCLO. Nausea and vomiting were significantly worse on IFOS and alopecia, related in extent to dose, was seen in both arms. Other side-effects, such as hematuria or rises in serum creatinine and encephalopathy, were infrequent and mild. A higher response rate with less myelosuppression suggests that IFOS may have advantages over CYCLO in combination therapy.