PURPOSE: The Pediatric Oncology Group (POG) conducted a randomized phase II study to evaluate the activity of carboplatin and iproplatin in children with progressive or recurrent brain tumors. PATIENTS AND METHODS: The study was designed to evaluate the activity of these agents and to compare the toxicities associated with their use. Treatment consisted of carboplatin 560 mg/m2 at 4-week intervals or iproplatin 270 mg/m2 at 3-week intervals. RESULTS: The major toxicity observed was myelosuppression, particularly thrombocytopenia, for both agents. Ototoxicity (grade 1 or 2) was seen in 2.5% of patients treated with carboplatin and 1.3% of patients treated with iproplatin. The majority of patients with low-grade astrocytic neoplasms treated with carboplatin (nine of 12 patients) or iproplatin (eight of 12 patients) demonstrated tumor response or prolonged stable disease that persisted off-therapy. The duration of stable disease produced by carboplatin was particularly striking, ranging from 2 months to 68 + months (median, 40 + months). Neither drug demonstrated appreciable activity in the treatment of medulloblastoma (two of 26 responses to carboplatin, one of 14 responses to iproplatin), ependymoma (two of 17 responses to carboplatin, none of seven responses to iproplatin), high-grade glioma (two of 19 responses to carboplatin, one of 14 responses to iproplatin), or brain-stem tumors (one of 23 responses to carboplatin, none of 14 responses to iproplatin). CONCLUSION: Carboplatin is active against low-grade gliomas. Further evaluation of the role of carboplatin in the preirradiation treatment of children with low-grade gliomas of the optic pathway is currently underway in a clinical trial.