The importance of donor-recipient human leukocyte antigen (HLA)-DPB1 matching for the clinical outcome of unrelated hematopoietic stem cell transplantation (HSCT) is controversial. We have previously described an algorithm for nonpermissive HLA-DPB1 disparities involving HLA-DPB1*0901,*1001,*1701,*0301,*1401,*4501, based on T-cell alloreactivity patterns. By revisiting the immunogenicity of HLA-DPB1*02, a modified algorithm was developed and retrospectively tested in 621 unrelated HSCTs facilitated through the Italian Registry for oncohematologic adult patients. The modified algorithm proved to be markedly more predictive of outcome than the original one, with significantly higher Kaplan-Meier probabilities of 2-year survival in permissive compared with nonpermissive transplantations (55% vs 39%, P = .005). This was the result of increased adjusted hazards of nonrelapse mortality (hazard ratio [HR] = 1.74; confidence interval [CI], 1.19-2.53; P = .004) but not of relapse (HR = 1.02; CI, 0.73-1.42; P = .92). The increase in the hazards of overall mortality by nonpermissive HLA-DPB1 disparity was similar in 10 of 10 (HR = 2.12; CI, 1.23-3.64; P = .006) and 9 of 10 allele-matched transplantations (HR = 2.21; CI, 1.28-3.80; P = .004), both in early-stage and in advanced-stage disease. These data call for revisiting current HLA matching strategies for unrelated HSCT, suggesting that searches should be directed up-front toward identification of HLA-DPB1 permissive, 10 of 10 or 9 of 10 matched donors.