To determine the clinical significance of central nervous system (CNS) involvement at the time of diagnosis of pediatric acute myeloid leukemia (AML), we analyzed clinical features and outcomes of 290 patients treated consecutively on four institutional trials (AML80, AML83, AML87, and AML91). CNS status was classified as CNS1 (no blast cells in CSF; n ¼ 205), CNS2 (o5 WBC/ll CSF with blast cells; n ¼ 37), or CNS3 (X5 WBC/ll CSF with blast cells, or signs of CNS involvement; n ¼ 48). Patients with CNS3 status were significantly younger than others (P ¼ 0.016) and significantly more likely to have the favorable cytogenetic features t(9;11), t(8;21), or inv(16) (Po0.001). The CNS3 group had a significantly greater probability (7s.e.) of 5-year event-free survival (43.777.0%) than did the CNS1 (27.873.2%, P ¼ 0.015) and CNS2 (24.377.5%, P ¼ 0.032) groups. However, after adjustment for favorable genetic features, there was no significant difference in EFS between the CNS3 and the combined CNS1 þ CNS2 groups (P ¼ 0.075). In all, 10 of 151 patients treated on AML80 and AML83, but none of 139 treated on AML87 and AML91, had primary CNS relapse. CNS involvement had no adverse prognostic significance, and patients with CNS2 status had similar outcome to CNS1 patients in this large group of pediatric patients with AML, treated at a single institution.