Three hundred thirty newly diagnosed patients were studied to determine the frequency and type of chromosomal abnormalities in acute lymphoblastic leukemia (ALL) and their clinical significance. Analyses of banded chromosomes revealed clonal chromosomal abnormalities in 218 patients (66%), including all cases of B-ALL; and 70% of non-T, non-B ALL; but only 39% of T-ALL (p < 0.001). Patients were classified into 10 groups according to karyotype: no abnormalities (34%), one of the following recurring structural abnormalities [the Philadelphia chromosome (12%), t(4;11) (5%), t(8;14) (5%), 14q+ (4.5%), 6q— (4%)] or, in the remaining cases with abnormalities, the modal number [<46 (5%), 46 (12%), 47 to 50 (8%), >50 (9%)]. Response to treatment (achievement of complete remission and remission duration) and survival differed significantly among chromosome groups (p < 0.002). The best responses were seen in patients with a modal number >50; the poorest responses were found in patients with the t(4; 11) and t(8;14). Interestingly, survival for children and adults who had karyotypes with the same specific structural abnormalities [e.g., the Philadelphia chromosome or t(4; 11)] was identical. Multivariate analysis demonstrated that the karyotypic pattern was an independent prognostic factor even when age, initial leukocyte count, and French-American-British (FAB) type were considered. We conclude that banded chromosome studies should be performed in all patients with ALL at diagnosis to identify those patients who have a pattern associated with a poor prognosis who may require more aggressive therapeutic approaches such as marrow transplantation.