Despite the use of partially T cell-depleted grafts, 20% of the recipients of an HLA-identical sibling marrow graft develop aGVHD > or = II. This indicates that the current method for selecting a sibling donor, ie serological typing for HLA-A, B and DR, and a mixed lymphocyte culture (MLC) or molecular typing for HLA-DRB/DQB, is not predictive for aGVHD. In order to optimise our selection procedure, we retrospectively analysed patients who developed aGVHD > or = II by means of sequencing based typing for HLA-DPB and frequency analysis of alloreactive helper and cytotoxic T lymphocyte precursors (HTLp-f and CTLp-f). Patients who did not develop aGVHD or developed aGVHD grade I served as controls. Retrospective typing for HLA-DPB revealed only a single disparity in the group with aGVHD > or = II, indicating that mismatches for antigens other than HLA are the major cause of aGVHD in these patients. Furthermore, in our patient group, neither HTLp-f nor CTLp-f were predictive for development of aGVHD indicating that these assays in their current set-up are insufficiently sensitive to predict aGVHD in BMT with a partially T cell-depleted graft. We conclude, that HLA-identical siblings can be identified by means of serological typing for HLA-A and B and intermediate resolution molecular typing for DRB and DQB, but that for the prediction of aGVHD cellular tests with higher sensitivity and specificity as compared to the currently used HTLp-f and CTLp-f assays need to be developed.