Minimal residual disease (MRD) during early chemotherapy is a powerful predictor of relapse in acute lymphoblastic leukaemia (ALL) and is used in children to determine eligibility for allogeneic haematopoietic stem cell transplantation (HSCT) in first (CR1) or later complete remission (CR2/CR3). Variables affecting HSCT outcome were analysed in 81 children from the ANZCHOG ALL8 trial. The major cause of treatment failure was relapse, with a cumulative incidence of relapse at 5 years (CIR) of 32% and treatment-related mortality of 8%. Leukaemia-free survival (LFS) and overall survival (OS) were similar for HSCT in CR1 (LFS 62%, OS 83%, n = 41) or CR2/CR3 (LFS 60%, OS 72%, n = 40). Patients achieving bone marrow MRD negativity pre-HSCT had better outcomes (LFS 83%, OS 92%) than those with persistent MRD pre-HSCT (LFS 41%, OS 64%, P < 0·0001) or post-HSCT (LFS 35%, OS 55%, P < 0·0001). Patients with B-other ALL had more relapses (CIR 50%, LFS 41%) than T-ALL and the main precursor-B subtypes including BCR-ABL1, KMT2A (MLL), ETV6-RUNX1 (TEL-AML1) and hyperdiploidy >50. A Cox multivariate regression model for LFS retained both B-other ALL subtype (hazard ratio 4·1, P = 0·0062) and MRD persistence post-HSCT (hazard ratio 3·9, P = 0·0070) as independent adverse prognostic variables. Persistent MRD could be used to direct post-HSCT therapy.