Purpose To determine the prognostic significance of blasts, and of white and red blood cells, in CSF samples at diagnosis of acute lymphoblastic leukemia (ALL), a uniform CSF and risk group classification schema was incorporated into Children's Oncology Group B-cell ALL (B-ALL) clinical trials. Methods CSF status was designated as follows: CNS1, no blasts; CNS2a to 2c, < 5 WBCs/μL and blasts with/without ≥ 10 RBCs/μL or ≥ 5 WBCs/μL plus blasts, with WBCs ≥ 5 times the number of RBCs; CNS3a to 3c, ≥ 5 WBCs/μL plus blasts with/without ≥ 10 RBCs/μL or clinical signs of CNS disease. CNS2 status did not affect therapy; patients with CNS3 status received two extra intrathecal treatments during induction and augmented postinduction therapy with 18 Gy of cranial radiation. Results Among 8,379 evaluable patients enrolled from 2004 to 2010, 7,395 (88.3%) had CNS1 status; 857 (10.2%), CNS2; and 127 (1.5%), CNS3. The 5-year event-free and overall survival rates were, respectively, 85% and 92.7% for CNS1, 76% and 86.8% for CNS2, and 76% and 82.1% for CNS3 ( P < .001). In multivariable analysis that included age, race/ethnicity, initial WBC, and day-29 minimal residual disease < 0.1%, CSF blast, regardless of cell count, was an independent adverse predictor of outcome for patients with standard- or high-risk disease according to National Cancer Institute criteria. The EFS difference reflected a significant difference in the incidence of CNS, not marrow, relapse in patients with CNS1 versus CNS2 and/or CNS3 status. Conclusion Low levels of CNS leukemia, regardless of RBCs, predict inferior outcome and higher rates of CNS relapse. These data suggest that additional augmentation of CNS-directed therapy is warranted for CNS2 disease.