Purpose To determine the response rate (RR) of neuroblastoma (NB) in children to temozolomide (TMZ), and evaluate the duration of response and tolerance of the drug in this patient population. Patients and Methods A multicenter, phase II evaluation of an oral, daily schedule of TMZ (200 mg/m²/d × 5 days every 28 days) was undertaken in children with refractory or relapsed high-risk NB (metastatic or localized with Myc-N amplification). Response assessment was based on imaging with two-dimentional measurement of disease and meta-iodobenzylguanidine (MIBG) score. Activity was defined by a reduction in lesion size or isotope uptake at anytime. Methodology included a two-step design using Fleming’s method with a first step of 15 patients and a second of 10 additional patients if two to four responses had been observed in the first cohort. All data was centrally reviewed by a panel. Results Twenty-five assessable patients were recruited over a 14-month period in 14 centers and received 94 cycles of chemotherapy. Twenty-three patients had metastatic NB either refractory (n = 9) or in relapse (n = 14). Grade 3 or 4 thrombocytopenia was the most frequent toxicity (16% of cycles). Myelosuppression resulted in treatment delays and dose reductions (24% and 21% of cycles, respectively). Response (complete response, very good partial response, or partial response) was observed in five patients (RR = 20% ± 8%) with a median duration of 6 months and an objective or mixed response in five additional patients. Conclusion Temozolomide shows activity in heavily pretreated patients with NB, and deserves further evaluation in combination with another drug.