Improvements in therapy for childhood acute lymphoblastic leukemia (ALL) have led us to reevaluate the prognostic significance of lymphoblast characteristics at diagnosis. From application of univariate and multivariate statistical methods, we determined the relationship of five blast cell features to treatment outcome in 250 patients who were enrolled in two clinical trials at this center from May 1979 through April 1982. Karotype ploidy, lymphoblast morphology, and immunophenotype were each significantly related to prognosis as measured by time to failure, while periodic acid-Schiff reactivity and glucocorticoid receptor number lacked prognostic implication for this patient population. In addition, clinical features of initial WBC count, age, and race were also significant independent variables in predicting treatment response. By multivariate analysis, both ploidy and morphology contributed prognostic information to a clinical model based on WBC count, age, and race. If the model was adjusted for impact of ploidy, however, French-American-British morphology no longer contributed additional prognostic information. Our findings suggest that many traditional biological features used to estimate prognosis in ALL can be discarded in favor of clinical features (leukocyte count, age, and race) and cytogenetics (ploidy) for planning of future clinical trials.