Primary mediastinal seminoma is an uncommon neoplasm, the optimal management of which is still debated. Radiotherapy produces a 65% disease-free survival rate. We assess whether these results have been improved with the advent of cisplatin-based chemotherapy. Data from 14 patients treated at the Institut Gustave-Roussy were reviewed. 9 had received cisplatin-based chemotherapy (Group 1): their outcome was compared with that of 5 patients treated with radiotherapy without chemotherapy (Group 2). We also reviewed data from the English literature using strict criteria, and report results concerning patients who received cisplatin-based chemotherapy and those who received radiotherapy. 8 of the 9 patients (89%) in Group 1 are long-term disease-free survivors and only 3 of 5 patients in Group 2. The patient who died in Group 1 was the only one who refused surgical resection of residual masses after chemotherapy. The review of the literature revealed that 59 of 68 (87%) patients initially managed with cisplatin- or carboplatin-based chemotherapy and for whom sufficient data are available, are long-term survivors and free of disease. Some of these patients had also received radiotherapy. Only 64 of 103 (62%) treated with thoracic radiotherapy without chemotherapy were long-term disease-free survivors. The disease-free survival rate of 51 patients who received cisplatin-based chemotherapy (excluding those who received carboplatin) was 86%. The difference in survival between patients administered cisplatin-based chemotherapy and those who underwent radiotherapy is apparently not due to unbalanced prognostic factors, the effect of time or non-specific medical management. We conclude that cisplatin-based chemotherapy allows long-term disease-free survival in approximately 85% of patients. These results seem to be higher than those obtained without cisplatin-based chemotherapy. However, a randomised study is required for definitive conclusions, but it is very unlikely that such a study will be performed due to the rarity of this neoplasm. Another alternative would be a meta-analysis based on individual data.