Matching for HLA has been the gold standard in bone marrow donor selection. But, with the ever increasing number of identified HLA alleles, it is becoming more difficult to find a fully HLA-identical donor other than a sibling. Retrospective analysis revealed that HLA mismatches do not necessarily give rise to acute graft-versus-host-disease (GVHD). However, we have no means of defining these 'permissible' mismatches before bone marrow transplantation (BMT). Thus, we set out to establish whether functional matching by means of helper and cytotoxic T-lymphocyte precursor frequency analysis (HTLp-f and CTLp-f respectively) can be applied to this end. Fifty-five recipient-donor pairs other than HLA-identical siblings, the recipient of which received a T-cell-depleted graft, were analysed by high-resolution HLA typing and/or HTLp-f/CTLp-f analysis. The predictive value of the CTLp-f assay for development of acute GVHD was confirmed. More importantly, our data indicate that the CTLp-f assay was able to discriminate permissible from non-permissible HLA-A, -B or -Cw mismatches, but not for DRB/DQB mismatches. The absolute number of alloreactive CTLs present in the graft correlated with the risk of acute GVHD. Although HTLp-f and CTLp-f together had a high negative predictive value, HTLp-f outcome by itself was not correlated with acute GVHD. As we have no evidence yet that HTLp-f or CTLp-f can identify permissible DRB/DQB mismatches, high-resolution matching for these antigens remains the best option. The combination of high-resolution DRB/DQB typing and the CTLp-f assay would enable the accurate prediction of the risk of acute GVHD while extending the pool of potential donors. Furthermore, it would enable adjustment of the number of T- cells in the graft accordingly to improve clinical outcome.