The HLAMatchmaker program is based on the donor/recipient comparison of the polymorphic triplet amino-acid sequences of the antibody-accessible regions on the human leucocyte antigen (HLA) molecule. The previous reports on its predictive value for renal allograft outcomes are conflicting. We conducted a retrospective study in a predominantly African-American (AA) cohort (N = 101, 94% AA). HLA typing was performed by molecular methods and triplet matching using HLAMatchmaker. Study end points included graft survival and incidence of acute rejection. The relationship between the number of triplet mismatches (TMM) and the degree of HLA antigen MM was evaluated using Pearson's correlation coefficient. Logistic regression models were used to examine the association between triplet matching and the study end points. Kaplan-Meier and Cox proportional hazard models were used for graft survival analysis. The strongest relationship between the number of TMM and HLA antigen MM was observed for HLA-DQ (r = 0.88). The association between triplet matching at HLA-A, -B, -DR and -DRw HLA loci and the study end points was not statistically significant. However, after grouping, the unadjusted estimates of graft survival for those with more than 10 Class I TMM were significantly worse than the others (p = 0.03). Adjusting for the effect of donor source, recipient characteristics and the immunosuppressive regimen did not change this association (hazard ratio = 0.2, confidence interval = 0.04-1.1). We conclude that triplet matching using HLAMatchmaker can provide useful prognostic information in kidney transplantation and that more than 10 donor/recipient Class I HLA TMM is predictive of worse graft outcome.