Background Although rabbit anti-T-lymphocyte globulin (ATLG) is largely used for the prevention of immune-mediated complications in patients given allogeneic haemopoietic stem-cell transplantation (HSCT) from an unrelated donor, the optimum dose of this drug in children is still undefined. We aimed to test whether a higher dose of ATLG was superior to a lower dose for prevention of grade II–IV acute graft-versus-host disease (GVHD). Methods We conducted a multicentre, randomised, open-label, phase 3 trial in seven Italian centres comparing two different doses of ATLG (30 mg/kg vs 15 mg/kg, given intravenously over 3 days, from day −4 to −2) in children (aged 0–18 years) with haematological malignancies transplanted from an unrelated donor, selected using high-resolution typing for HLA-class I/II loci. All patients received a myeloablative regimen and cyclosporine-A plus short-term methotrexate as post-transplantation GVHD prophylaxis. Patients were randomly assigned (1:1) to either of the two groups and were stratified by the degree of HLA-compatibility with their donor, the source of haemopoietic stem cells used (bone marrow vs peripheral blood stem cells), and the disease risk category. The randomisation was open label; all investigators were aware of the treatment allocation. The primary endpoint of the study was 100-day cumulative incidence of grade II–IV acute GVHD. Statistical analyses were done according to the per-protocol principle. Other outcomes included cumulative incidence of chronic GVHD, non-relapse mortality, disease recurrence, and probability of overall survival and event-free survival. This study was registered with ClinicalTrials.gov, number NCT00934557. Findings Between Jan 15, 2008, and Sept 25, 2012, 89 patients were randomly assigned to the 30 mg/kg ATLG group and 91 to the 15 mg/kg ATLG group; 84 patients in the 30 mg/kg ATLG group and 88 in the 15 mg/kg ATLG group were included in the analysis. The median follow-up for the whole study population was 3·4 years (IQR 1·7–5·1). The 100-day cumulative incidence of grade II–IV acute GVHD was 36% (95% CI 28–48) in the 15 mg/kg ATLG group and 29% (20–40) in the 30 mg/kg ATLG group (hazard ratio [HR] 0·74, 95% CI 0·44–1·25; p=0·26). The cumulative incidence of non-relapse mortality was 9% (5–18) in the 15 mg/kg ATLG group and 19% (12–30) in the 30 mg/kg ATLG group (HR 2·08, 0·89–4·96; p=0·092). Cumulative incidence of disease recurrence was 15% (12–24): 14% (8–23) in the 15 mg/kg ATLG group versus 20% (13–31) in the 30 mg/kg ATLG group (HR 1·54, 0·74–3·21; p=0·25). The 5-year overall survival probability was 70% (62–77) for the whole study population: 78% (69–87) in the 15 mg/kg ATLG group versus 62% (50–73) in the 30 mg/kg ATLG group (HR 1·80, 1·01–3·20; p=0·045). The 5-year event-free survival was 77% for children in the 15 mg/kg ATLG group versus 61% in the 30 mg/kg ATLG group (HR 1·87, 1·07–3·28; p=0·028). Interpretation Children with haematological malignancies transplanted from unrelated donors selected through high-resolution HLA-typing benefit from the use of a 15 mg/kg ATLG dose in comparison with a 30 mg/kg ATLG dose. ATLG at 15 mg/kg should thus be regarded as the standard serotherapy regimen for unrelated donor allogeneic HSCT in this patient population. Future randomised studies will continue to aim to optimise patient outcome and strategies to prevent acute GVHD occurrence. Funding Fresenius/Neovii Biotech.