To determine the efficacy of pretreatment with amifostine in diminishing the hematologic and nonhematologic toxicities of cyclophosphamide and cisplatin in previously untreated patients with stage III/IV ovarian cancer, a multicenter randomized controlled trial of cyclophosphamide (1,000 mg/m(2)) and cisplatin (100 mg/m(2) with or without amifostine (910 mg/m(2)) was performed. Two hundred forty-two patients with stage III/IV epithelial ovarian cancer were enrolled. Following primary surgery, patients were stratified and randomized to either cyclophosphamide/cisplatin (CP; 120 patients) or amifostine plus CP (122 patients) every 3 weeks for six cycles. Patient characteristics were similar in both groups. Cytoprotective end points and tumor response were evaluated, including the need to delay or discontinue therapy because of toxicity and the incidence of febrile neutropenia with associated complications. Fourteen patients treated with CP discontinued protocol therapy because of hematologic or renal toxicity (eight hematologic and six renal). In contrast, only one patient treated with amifostine plus CP discontinued protocol therapy for hematologic or renal toxicity (P < .001). Forty-three percent of CP patients compared with 22% of amifostine plus CP patients had grade 4 neutropenia (P = .001); total days in hospital were reduced from 258 in the CP arm to 11 in the amifostine plus CP arm (P = .009, two-sided). Sixty-five percent of the CP patients and 41% of the amifostine plus CP patients (P = .004) had the next cycle of CP delayed because of an absolute neutrophil count below 1,500/microL at day 22. Platelet and red blood cell transfusion support were substantially reduced in the group that received amifostine. The serum creatinine failed to return to < or = to 1.5 mg/dL by day 22, requiring a delay in chemotherapy in 15% and 5%, respectively, of the CP and amifostine plus CP groups (P = .014). Over the six cycles, the incidence and severity of peripheral neuropathy were also significantly reduced in the amifostine-treated group (P = .029). Pathologic response rates and survival curves were equivalent. The significant reduction in the CP-induced acute and cumulative hematologic, renal, and neurologic toxicities by amifostine pretreatment with equivalent response and survival indicates selective cytoprotection. This selective effect has the potential to affect quality of life and medical economic considerations.