From 1982 to 1987, a randomized phase III trial was performed in order to determine the long-term effect of induction chemotherapy before standard pelvic irradiation in stage IIb-N1, III squamous cell carcinomas of the cervix. Patients were randomized to either chemotherapy and radiotherapy (C + R group) vs radiotherapy alone (R group). Radiotherapy for all patients consisted of 50 Gy in the pelvis with a boost by external irradiation or by brachytherapy (cumulative dose of 68 Gy). The chemotherapy regimen was an association of methotrexate (10 mg/m2, D2-4), chlorambucil (4 mg/m2, D1-5), vincristine (0,7 mg/m2, D1), cisplatin (80 mg/m2, D5), given every 3 wks; at least 2 courses were to be given before assessing efficacy and 2 more courses were given to patients who responded. One hundred and fifty-one patients were fully evaluable, after a mean follow-up of 38 mths (range 2-7 years), 76 in the R arm and 75 in the C + R arm. The response rate (greater than 50%) to chemotherapy was 42.5%. After completion of treatment, the complete response rate was 86.8% in the R arm and 86.3% in the C + R arm. The 3 year disease-free survival was 58.7% in the C + R group and 54.5% in the R group, and the median survival was 39.5% and 47 months respectively (NS). The survival of patients with a complete response at the end of radiotherapy was significantly better in the C + R group (when chemotherapy had been active) than in the R group (p = 0.04). Although radiotherapy was not modified whether patients had initial chemotherapy or not, tolerance was not significantly different between the 2 groups. The data collected in this study indicate that: 1) efficacy of induction chemotherapy is the only available predictive test for long-term results, 2) tolerance to treatment is crucial for optimal chemotherapy delivery, 3) higher dose intensity of chemotherapy in cervical carcinoma is associated with a better tumor reduction, and probably a better survival.