To compare single versus sequential high-dose chemotherapy (HDCT) as first or subsequent salvage treatment in patients with relapsed or refractory germ cell tumors (GCTs). Between November 1999 and November 2004, 230 patients were planned to be recruited in a prospective, randomized, multicenter trial comparing one cycle of cisplatin 100 mg/m2, etoposide 375 mg/m2, and ifosfamide 6 g/m2 (VIP) plus three cycles of high-dose carboplatin 1,500 mg/m2 and etoposide 1,500 mg/m2 (CE; arm A) versus three cycles of VIP plus one cycle of high-dose carboplatin 2,200 mg/m2, etoposide 1,800 mg/m2, and cyclophosphamide 6,400 mg/m2 (CEC; arm B). The study was stopped prematurely after recruitment of 216 patients as a result of excess treatment-related mortality in arm B. One hundred eleven (51%) of 216 patients were randomly assigned to sequential HDCT, and 105 (47%) of 216 patients were randomly assigned to single HDCT. Five (2%) of 216 patients had to be excluded because of non-GCT histologies at review. With a median follow-up time of 36 months, 109 (52%) of 211 patients were alive, and 91 (43%) of 211 patients were progression free. At 1 year, event-free, progression-free, and overall survival rates were 40%, 53%, and 80%, respectively, in arm A compared with 37%, 49%, and 61%, respectively, in arm B (P > .05 for all comparisons). Treatment-related deaths, mainly as a result of sepsis and cardiac toxicity, were less frequent in arm A (four of 108 patients, 4%) compared with arm B (16 of 103 patients, 16%; P < .01). We found no difference in survival probabilities between single HDCT using CE and sequential HDCT using CEC. Sequential HDCT was better tolerated and resulted in fewer treatment-related deaths.