Despite reported activity in many other solid tumors, high-dose ifosfamide produces few objective responses in recurrent pediatric brain tumors. Alkylating agents such as cyclophosphamide (CYCLO) possess good activity in many of solid tumors, including brain tumors. Although Ifosfamide (IFOS), a structural congener of CYCLO, has been suggested to have greater activity in several tumors, its activity in brain tumors is uncertain. We conducted a phase II trial of every-other day IFOS (3 gm/M2/qod x 3) in 87 recurrent pediatric brain tumors. Responses were evaluable in 71 patients. Partial responses occurred in 1/6 patients with low grade astrocytoma, 1/16 with malignant glioma, 1/16 with medulloblastoma, 1/3 with pineoblastoma and 1/12 patients with ependymoma. No responses occurred among 10 patients with brain stem gliomas or 8 patients with other brain tumors. Despite the poor objective response rate, 23/71 patients were clinically and imaging stable for periods of 8-62 weeks. There was no relationship between prior CYCLO treatment and subsequent response or failure with IFOS. The predominant toxicity was myelosuppression. Although generally reversible, prolonged suppression and sepsis were responsible for the deaths of 3 heavily pretreated patients. Renal toxicity was uncommon; 2 patients had grade III, and one grade IV renal tubular dysfunction. One patient had grade IV hematuria. Neurotoxicity was less common than in studies of daily ifosfamide; only 1 patient had grade IV neurotoxicity. Three patients had grade III or IV IFOS related hyponatremia. Despite the good stable disease rate, the poor rate of objective response suggests that IFOS monotherapy possesses little clinically meaningful activity in brain tumors.