One hundred patients of median age 34 years (range, 14-53) received bone marrow transplants from unrelated donors serologically matched for human leukocyte antigen HLA-A, HLA-B, and HLA-DR using tacrolimus and minimethotrexate for prevention of acute graft-versus-host disease (GVHD). Sixty-eight patient-donor pairs had allelic matches at HLA-DRB1 and HLA-DQB1, 20 pairs had a single mismatch at HLA-DRB1 or HLA-DQB1, and 12 were mismatched at both HLA-DRB1 and HLA-DQB1. Minimum follow-up time was 6 months. Grades 2 to 4 GVHD occurred in 43% of patients with matched donors, 69% with single allele-mismatched donors, and 71% with double allele-mismatched donors; grades 3 to 4 GVHD occurred in 22%, 43%, and 64%, respectively. On multivariate analysis, the relative risk of grades 2 to 4 GVHD was 2.2 (95% CI, 1.1-4.5; P = .03) with a single allele mismatch and 2.7 (95% CI, 1.2-6.0; P = .02) with a double allele mismatch. The relative risks of grades 3 to 4 GVHD were 3.0 (95% CI, 1.2-7.6; P = .02) and 5.0 (95% CI, 1.9-12.6; P = .001), respectively. Day 100 treatment-related mortality was also adversely affected by allelic mismatching, occurring in 21% of those with matched donors, 50% with single allele-mismatched donors, and 42% with double allele-mismatched donors (P = .02), but overall survival at day 180 did not differ significantly among the 3 groups. Tacrolimus does not abrogate the adverse impact of allele mismatching at HLA-DRB1 and HLA-DQB1 on the risk of moderate-to-severe acute GVHD.