Genotype and immunophenotype can be used to define biological species of acute lymphoblastic leukemia (ALL). The purpose of these two pilot studies, conducted between 1986 and 1994, was to explore the feasibility and acceptability of classifying ALL in this manner for selection of treatment rather than using conventional risk for failure factors such as age and initial white blood cell count. The possibility that conventional risk factors would be overcome and survival improved by this approach was also considered. Flow cytometry and chromosome analysis were used to classify the ALL of 150 children into one of five biologic categories as defined by cell surface antigens, DNA index and chromosome number and arrangement. Chemotherapy regimens depended on the assigned category. There was no provision for cranial irradiation and use of alkylating agents, anthracyclines and epipodophyllotoxins was restricted in order to reduce risk of late adverse sequelae. All patients are included in the analysis regardless of presenting condition or adherence to protocol. The majority of patients were Mexican-American or African-American. Eight-year event-free survival (EFS) is 60.7% (+/-4%) and 8-year overall survival (OAS) 72.6% (+/-3.7%). EFS and OAS varied significantly among the biologic categories despite differences in chemotherapy regimens. When the patients with B-precursor ALL were retrospectively classified by current Pediatric Oncology Group (POG) criteria, 8-year EFS was 82% (+/-7.3%) for the good risk group, 68.9% (+/-5.9%) for the standard risk and 48.8% (+/-7.6%) for the poor risk, all significant differences. However, when retrospectively classified according to the Rome/NCI prognostic criteria the 8-year EFS for standard risk patients was 69.1% (+/-5.1%) and for high risk 58.8% (+/-6.9%), not a statistically significant difference. Numbers of T cell and B cell patients are too few for comparison. Gender and ethnicity influenced survival as in treatment based on prognostic factors. Initial central nervous system (CNS) relapse occurred in five patients (3%) and combined CNS and hematological relapse in six (3%). Factors significantly associated with CNS and combined relapse were leukemic pleocytosis in the initial CSF sample, pro-B immunophenotype and DNA index <1.16, but not initial white blood cell count. Only three survivors appear to have serious late adverse sequelae, the only neurologic the result of asparaginase-induced cortical vein thrombosis. The results suggest that use of biologic species as defined by immunophenotype and genotype to select therapy of ALL is feasible and acceptable but under the conditions of these studies offered no apparent therapeutic advantage over conventional risk grouping. However, the introduction of molecular genotyping and novel gene targeted therapeutic agents justify further exploration of this approach.