The rapidity of response to induction therapy is emerging as an important prognostic factor in children and adolescents with acute lymphoblastic leukemia (ALL). We studied the relationship between rapidity of reduction in peripheral blood blast count and treatment outcome in children with T cell ALL (T-ALL). Initial systemic chemotherapy included prednisone, vincristine, doxorubicin and cyclophosphamide. A Cox analysis evaluated the correlation between the length of time that the peripheral blood absolute blast count (ABC) remained above 1000/mm3 following the start of treatment and event-free survival (EFS). Data were available for 281 patients. Patients for whom the ABC remained >1000/mm3 for 3 or more days following administration of intensive therapy had an estimated 5-year EFS of 34.2% (s.e. = 7.2) vs 58.3% (3.5) for those whose ABC was <1000/mm3 within 0-2 days, with a hazard ratio (HR) of failure of 2.03 (95% CI = 1.35-3.06, P < 0.001) for the slower responding patients. Pre-treatment of some type (usually with prednisone) occurred in 128 patients (average duration 1.7 days). When this was accounted for, patients with an ABC >1000/mm3 for 5 or more days following the start of treatment of any kind had a HR for failure of 2.27 (95% CI = 1.38-3.72, P < 0.001) compared to those responding within 0-4 days. Inclusion of other clinical and biological factors in a multivariate analysis did not alter the prognostic importance of slower blast clearance. Pediatric patients with T-ALL who have a circulating blast count >1000/mm3 at diagnosis and a relatively slower response to initial treatment are at increased risk of treatment failure. Rapidity of response may therefore be a clinically useful prognostic factor for patients with T-ALL.