The CD34 antigen was detected on > or = 10% of the blast cells in 235 (70%) of 335 cases of newly diagnosed childhood acute lymphoblastic leukemia (ALL) treated in two consecutive chemotherapy trials. By immunophenotype, the distribution of positive cases favored early pre-B ALL (83%; n = 180) followed by pre-B ALL (61%; n = 89) and then T-cell ALL (46%; n = 61) (P < .001). Among the B-lineage cases, CD34 expression was significantly associated with favorable presenting features: age 1 to 10 years, white race, absence of central nervous system (CNS) leukemia, low serum lactate dehydrogenase level, CD10 expression, and leukemic cell hyperdiploidy (> 50 chromosomes or DNA index > or = 1.16). Event-free survival was clearly superior for patients with CD34+ leukemia (P = .01), with an estimated 83% +/- 6% (SE) of the cohort remaining free of adverse events at 5 years post diagnosis, as compared to 63% +/- 10% of the group without this feature. Multivariate analysis showed that the prognostic influence of the antigen was independent of age, leukocyte count, and other well- recognized factors, suggesting that it would add discriminatory power to current systems of risk assignment. Findings in T-cell ALL were the reverse: CD34 expression showed positive correlations with initial CNS leukemia and CD10 negativity but not with any good-risk presenting characteristics. Log-rank analysis indicated no adverse effect on treatment outcome by CD34 antigen expression, although additional patients with need to be studied to obtain a definitive answer. The opposed clinical associations of CD34 expression in B- and T-lineage ALL may reflect fundamental biologic differences between these leukemia species.