One hundred and seventy-one patients with advanced soft tissue sarcoma entered a randomized crossover phase II study comparing cyclophosphamide (CYCLO) with a new analogue, ifosfamide (IFOS), both administered as 24 h i.v. infusions every 3 weeks. The doses used were CYCLO 1.5 g/m2 and IFOS 5 g/m2, with provision for dose escalation. All patients received mesna 400 mg/m2 as an i.v. bolus 4 hourly X 9 doses, commencing at the start of the oxazophosphorine infusion. Patients who had received previous chemotherapy were eligible provided this did not include a classical alkylating agent. There were 22 patients who were ineligible, and response could not be evaluated in 12 additional patients. IFOS produced two complete and ten partial remissions, for an overall response rate of 18%. CYCLO was significantly (P = 0.04) less active, producing one complete and five partial remissions, an overall response rate of 9%. Stabilization of disease was similar in both arms (27% and 24% respectively), but fewer patients showed progression on IFOS. The response rate was higher (20% vs 5%) for patients who had not received previous chemotherapy, and also for female compared with male patients (21% vs 5%). When only patients who had not received previous chemotherapy were considered, the respective response rates for IFOS and CYCLO were 24% and 15%. There were no responses in previously treated patients receiving CYCLO. There were four partial responses in 33 patients crossing from CYCLO to IFOS, but no responses in 18 patients receiving CYCLO after IFOS. Leucopenia was significantly more pronounced (P = 0.0004) with CYCLO, both after the first course and throughout treatment, although the incidence of severe infections, 6%, was the same in both arms. Nausea and vomiting were more severe with IFOS (P = 0.022), but other toxicities were mild. Grade 1 or 2 bladder (haematuria) or renal (rise in serum creatinine) toxicity was slightly more frequent with IFOS (7 vs 3 patients) and was a reason for stopping treatment for one patient in each arm. Three episodes of mild to moderate drowsiness after IFOS were reported, but no severe encephalopathy. A higher response rate with less myelosuppression suggests that IFOS may have advantages over CYCLO in combination with such active agents as adriamycin.