We have examined the toxicity and overall outcome of the Medical Research Council UKALL R1 protocol for 256 patients with relapsed childhood acute lymphoblastic leukaemia (ALL). Second remission was achieved in over 95% of patients. Two patients died during induction and seven patients died of resistant disease. The overall actuarial event-free survival (EFS) at 5 years for all patients experiencing a first relapse was 46% (95% CI 40–52). Duration of first remission, site of relapse, age at diagnosis and sex emerged as factors of prognostic significance. Five-year EFS was only 7% for children relapsing in the bone marrow within 2 years of diagnosis, but was 77% for those relapsing without bone marrow involvement > 2.5 years from diagnosis. All analyses in this report are by treatment received. For those receiving chemotherapy alone, the 5-year EFS was 48%; for autologous bone marrow transplantation (BMT), the 5-year EFS was 47%; for unrelated donor BMT, it was 52%; and for related donor BMT, the 5-year EFS was 45%. The groups, however, were not comparable with respect to risk factor profile, and therefore direct comparison of EFS is misleading. Adjustment for time to transplant and prognostic factors was used to reduce the effects of biases between treatment groups, but did not suggest benefit for any particular treatment. There was failure of our planned randomization scheme in this trial with only 9% of those eligible being randomized, which highlights the difficulties in running randomized trials especially in patients who have relapsed from a previous trial. The optimal treatment for relapsed ALL therefore remains uncertain. Alternative approaches are clearly needed for those with early bone marrow relapse if outcome is to improve.