The effect on survival of including HLA-DPB1 in a 12-allele matching strategy was retrospectively evaluated in 130 patients with acute leukaemia and myelodysplasia undergoing T-cell-depleted PBSC transplantation using unrelated donors. Patients received alemtuzumab in vivo T-cell depletion as part of a myeloablative (MA; n ¼ 61) or reduced-intensity conditioning regimen (n ¼ 69). No difference in OS was seen with single-locus mismatching (mm) when 10 conventional alleles (HLA-A, B, C, DRB1 and DQB1) were considered. However, the addition of HLA-DPB1 matching data proved highly discriminatory. Mismatches were identified in 87% of patients previously considered fully matched (1DPmm ¼ 49pts: 2DPmm ¼ 28pts), and in the 9/10 group 22 patients were reclassified as double and 16 as triple mismatches. In 10/10 transplants, there was a distinct trend to poorer OS with double DPB1 mm. If all 12 loci were considered, 98% of single mm were at HLA-DPB1. Furthermore, cumulative mm at two or more loci was associated with significantly poorer 3-year OS (34% vs 48%, P ¼ 0.013: hazard ratio 1.8 (95% confidence interval 1.14-3.06; P ¼ 0.017), although his detrimental effect was only apparent using MA conditioning, in which reduced OS was associated with increased chronic GVHD (61% vs 16%, P ¼ 0.018) and nonrelapse mortality (30% vs 9%, P ¼ 0.039).