The prognosis for children with high-risk relapsed acute lymphoblastic leukemia (ALL) is poor. Here, we assessed the prognostic importance of response during induction and consolidation treatment prior to hematopoietic stem cell transplantation (HSCT) aiming to evaluate the best time to assess minimal residual disease (MRD) for intervention strategies and in future trials in high-risk ALL relapse patients. Included patients (n=125) were treated uniformly according to the ALL-REZ BFM (Berlin-Frankfurt-Münster) 2002 relapse trial (median follow-up time=4.8 years). Patients with MRD ⩾10(-3) after induction treatment (76/119, 64%) or immediately preceding HSCT (19/71, 27%) had a significantly worse probability of disease-free survival 10 years after relapse treatment begin, with 26% (±6%) or 23% (±7%), respectively, compared with 58% (±8%) or 48% (±7%) for patients with MRD <10(-3). Conventional intensive consolidation treatment reduced MRD to <10(-3) before HSCT in 63% of patients, whereas MRD remained high or increased in the rest of this patient group. Our data support that MRD after induction treatment can be used to quantify the activity of different induction treatment strategies in phase II trials. MRD persistence at ⩾10(-3) before HSCT reflects a disease highly resistant to conventional intensive chemotherapy and requiring prospective controlled investigation of new treatment strategies and drugs.