Temozolomide (TMZ) has demonstrated activity and acceptable toxicity for the treatment of recurrent high-grade gliomas in prospective phase II studies. Limited information is available on TMZ when prescribed outside a clinical trial. We conducted a retrospective study to evaluate the activity and safety of TMZ that was prescribed for the treatment of recurrent glioma in the context of a compassionate use program in Belgium. Data were obtained on 117 adult patients (from five hospitals) who received TMZ as first or second line chemotherapy. The recommended starting dose of TMZ was 200 mg/m2 x5d q28d for chemonaive patients and 150 mg/m2 x5d q28d for pre-treated patients. Toxicity was generally mild. Thrombocytopenia was the most frequent treatment related adverse event (grade 3/4 in 17% of patients). Its occurrence was correlated with a starting dose of 200 mg/m2/d and stresses the need to monitor toxicity. The overall objective response rate (complete and partial response) was 29 and 34% of patients achieved an objective disease stabilization. The median progression-free survival was 104 days (95% CI: 85-123) and the median overall survival was 215 days (95% CI: 161 269). In multivariate analysis a 'deep localization' of the glioma (as opposed to a cortico-subcortical localization) and 'the preceding history of a low-grade glioma' were respectively identified as a negative and positive independent prognostic variable for survival. No significant difference in terms of response or median survival was observed between patients with anaplastic astrocytoma or oligo-astrocytoma and chemonaive glioblastoma multiforme. This retrospective study indicates that the reported activity and toxicity profile of TMZ for the treatment of patients with recurrent glioma is reproducible outside the setting of a prospective clinical trial.