Prior data indicate similar outcomes after transplants from human leukocyte antigen (HLA)-haplotype-matched relatives, HLA-identical siblings and HLA-matched unrelated donors. We used our prospective data set to answer a clinically important question: who is the best donor for a person with acute leukaemia transplanted in first complete remission. Patients were randomly divided into training (n = 611) and validation (n = 588) sets. A total of 1199 consecutive subjects received a transplant from an HLA-haplotype-matched relative using granulocyte colony-stimulating factor and anti-thymocyte globulin (n = 685) or an HLA-identical sibling (n = 514); 3-year leukaemia-free survivals (LFSs) were 75 and 74% (P = 0.95), respectively. The multivariate model identified three major risk factors for transplant-related mortality (TRM): older donor/recipient age, female-to-male transplants and donor-recipient ABO major-mismatch transplants. A risk score was developed based on these three features. TRMs were 8%, 15% and 31% for subjects with scores of 0-1, 2 and 3, respectively, (P < 0.001). Three-year LFSs were 78%, 74% and 58%, respectively, (P = 0.003). The risk score was validated in an independent cohort. In conclusion, our data confirm donor source is not significantly correlated with transplant outcomes. Selection of the best donor needs to consider donor-recipient age, matching for gender and ABO incompatibility among persons with acute leukaemia receiving related transplants under our transplant modality.