We report the prognostic significance of the pre-B-cell immunophenotype and other presenting features, including blast cell karyotype, in a randomized clinical trial conducted from 1981 to 1986 for children with early pre-B (n = 685) or pre-B (n = 222) acute lymphoblastic leukemia (ALL). Patients greater than or equal to 1 year and less than or equal to 21 years of age who attained complete remission were stratified by conventional risk criteria and immunophenotype and then randomized to receive continuation therapy with either of two regimens of intensive chemotherapy, designated S (standard) and SAM (standard plus intermediate-dose methotrexate, 1 g/m2 every 8 weeks). The proportions of subjects achieving complete remission in the two phenotypically defined subgroups were identical, 96%. At a median follow-up time of 42 months, the overall probability of 4-year event-free survival (+/- SE) was 63% +/- 2% (pre-B = 51% +/- 5% and early pre-B = 66% +/- 3%). Children with pre-B ALL had significantly shorter durations of continuous complete remission (P = .0004); this association included both bone marrow and CNS remissions (P = .0004 and P = .02, respectively). In a univariate Cox regression analysis of potentially important prognostic factors, the pre-B immunophenotype was significantly related to a poorer outcome, as were other recognized biologic and clinical features (eg, pseudodiploidy, older age, male sex, black race, and a higher WBC). It retained its prognostic strength in a multivariate model based on age, WBC, ploidy, and sex. The risk of failure at any point in the clinical course of a child with the pre-B immunophenotype was 1.8 times as great as that in a patient lacking this feature but otherwise having an equivalent risk status. It should be stressed that the predictive value of any of the significant characteristics identified in this study could diminish in the context of another, more effective treatment program. Nevertheless, our major conclusion, that children with pre-B ALL fare worse than those with early pre-B disease in a contemporary clinical trial has implications for stratified randomization of patients and the design of risk-specific treatment protocols.