Central nervous system (CNS) involvement at diagnosis of pediatric acute myeloid leukemia (AML) is not considered as an independent prognostic factor. This study describes the prognostic value of pediatric AML with CNS involvement at diagnosis. Pediatric patients were treated for de novo AML in the French multicenter trial ELAM02. Lumbar puncture was carried out in the first week, and the treatment was adapted to the CNS status. No patient received CNS radiotherapy. The patients were classified into 2 groups: CNS+ and CNS-. Of the 438 patients, 16% (n=70) had CNS involvement at diagnosis, and 29% showed clinical signs. The patients with CNS disease were younger (40% were below 2 y old), had a higher white blood cell count (median of 45 vs. 13 G/L), and had M4 and M5 morphologies. The complete remission rate was similar at 92.8% for CNS+ and 88.5% for CNS-. There was no significant difference between the CNS+ and the CNS- group in overall survival (76% and 71%, respectively) and event-free survival (57% and 52%, respectively). Regarding the occurrence of first relapse, the CNS+ group had a higher combined relapse rate of 26.1% compared with 10% for the CNS- group. The results indicate that CNS involvement at diagnosis of pediatric AML is not an independent prognostic factor. Triple intrathecal chemotherapy combined with high-dose intravenous cytarabine should be the first-line treatment for CNS disease.