Over 80% of stem cell transplantations (SCT) from unrelated donors (UD) are performed across allelic HLA-DPB1 disparities which have been associated with acute graft-versus-host disease (aGvHD) and, in 10/10 matched pairs, protection from disease relapse, while no significant correlation with overall survival (OS) could be revealed so far. We have previously developed an algorithm for non-permissive HLA-DP disparities involving an immunogenic T cell epitope (TCE) encoded by DPB1*0901, *1001 and *1701 (group 1), and, in a weaker form, by DPB1*0301, *1401 and *4501 (group 2) (Zino et al, Blood 2004). Here we report on the analysis of this algorithm in 627 UD SCT facilitated through the Italian Bone Marrow Donor Registry (IBMDR) and the Gruppo Italiano Trapianto di Midollo Osseo (GITMO) between 1999 and 2006, performed for malignant disorders including acute myeloid or lymphoid leukaemia (n=320; 51%) and Hodgkin's or non-Hodgkin's lymphoma (n=86; 13.7%). 242 pairs (38.5%) were matched at the allelic level for HLA-A, B, C, DRB1 and DQB1 (10/10 allele-matched pairs), while the remaining 385 pairs (61.5%) presented at least one mismatch at either of these loci (<10/10 matched pairs). Non-permissive DPB1 TCE disparities were determined by either considering both group 1 and group 2 DPB1 alleles (“2-group model”), or group 1 alleles only (“1-group model”). In line with DPB1 allelic frequencies in the Italian population, non-permissive DPB1 TCE disparities were present in 40% of pairs according to the 2-group model, and in 16% of pairs according to the 1-group model. In the 10/10 matched pairs, the Cox regression probability of engraftment was significantly reduced by the presence of non-permissive TCE disparities in the 2-group model (HR=0.23; C.I. 0.07–0.70, p=0.01) and in the 1-group model (HR=0.25; C.I. 0.08–0.75, p=0.01). Importantly, in the 10/10 matched pairs, the adjusted Cox regression hazards of transplant related mortality (TRM) were significantly increased by the presence of non-permissive TCE disparities in the 2-group model (HR=1.72; C.I. 1.09–2.70, p=0.02) and in the 1-group model (HR=1.71; C.I. 1.01–2.92, p=0.05), resulting in a significant increase in the hazards of overall mortality in the 1-group model (HR=1.66; C.I. 1.07–2.58, p=0.02) which was marked but not statistically significant in the 2-group model (HR=1.38; C.I. 0.96–1.98, p=0.08). These effects of non-permissive TCE DPB1 disparities were completely abrogated by the additional presence of one or more mismatches at either of the other five HLA loci in the <10/10 matched group. Interestingly, the estimated cumulative incidence probability of TRM and OS was similar in the 10/10 matched pairs with non-permissive DPB1 TCE disparities according to the 2-group model and in the <10/10 matched pairs regardless of DPB1 matching status, and worse in the presence of DPB1 TCE disparities according to the 1-group model. No impact of non-permissive DPB1 disparities on the hazards of aGvHD grade 2–4 or disease relapse were observed in neither of the two models. Taken together, our data show, for the first time, a significant association between HLA-DP matching status and survival in UD SCT, suggesting that the avoidance of non-permissive DPB1 TCE disparities according to the 2-group model and, even more so, according to the 1-group model, could significantly improve the clinical success of this treatment by reducing transplant mortality.