The outcome of pediatric ependymomas is difficult to predict based on clinical and histological parameters. To address this issue, we have performed a comparative genomic hybridization screen of 42 primary and 11 recurrent pediatric ependymomas and correlated the genetic findings with clinical outcome. Three distinct genetic patterns were identified in the primary tumors and confirmed by hierarchical cluster analysis. The first group of structural tumors, showed few, mainly partial imbalances (n = 19). A second numerical group showed 13 or more chromosome imbalances with a nonrandom pattern of whole chromosome gains and losses (n = 5). The remaining tumors (n = 18) showed a balanced genetic profile that was significantly associated with a younger age at diagnosis (P < 0.0001), suggesting that ependymomas arising in infants are biologically distinct from those occurring in older children. Multivariate analysis showed that the structural group had a significantly worse outcome compared to tumors with a numerical (P = 0.05) or balanced profile (P = 0.02). Moreover genetic group and extent of surgical resection contributed significantly to outcome whereas histopathology, age, and other clinical parameters did not. We conclude that patterns of genetic imbalances in pediatric intracranial ependymomas may help to predict clinical outcome.