The Seattle group has demonstrated that bone marrow transplantation (BMT) using partially HLA-mismatched related donors is feasible in principal. However, it was unclear whether these results can also be achieved at smaller-sized BMT units. Therefore a matched pair analysis enrolling 52 BMTs from partially HLA-mismatched relatives and 52 control BMTs from HLA-identical siblings was performed at the University Hospital of Essen. Overall survival (OS) and incidence of graft-versus-host disease (GVHD) did not differ significantly after study and control BMTs (OS: 52% v 63% 1 year, 46% v 48% 5 years post transplant; acute GVHD: 37% v 35%, chronic GVHD: 67% v 55%). After study BMTs, however, therapy-related mortality (P=0.018) and incidence of graft failure (P=0.002) were increased, whereas relapse rate was reduced (11% v 27%). Two or three mismatches in HVG direction implied the same risk of graft failure as one mismatch. Therefore, (i) OS after BMT from one HLA antigen mismatched relative and from HLA-identical siblings does not differ significantly even when performed at a 'smaller' centre; (ii) two or even three HLA mismatches in host-versus graft (HVG) direction might be acceptable in family BMT for leukaemia.