Allogeneic bone marrow transplantation (BMT) is a treatment modality with the potential of curing otherwise lethal diseases. The predominant indications for BMT are haematological malignancies. In BMT alloreactivity plays a pivotal role for the outcome. Graft-versus-host disease (GvHD) and graft-versus-leukaemia (GvL) are correlated manifestations of alloreactivity. Severe GvHD is one of the main causes of morbidity and mortality post-BMT. In the absence of GvL the risk of relapse is high. The main effector cells are T lymphocytes. Donor leukocyte infusion (DLI) for treatment of leukaemic relapse after BMT can induce durable remissions. DLI causes GvHD in the majority of the responding patients. However, a GvL effect may be present without evidence of GvHD and vise versa. The importance of alloreactivity for the treatment outcome prompted the interest for a predictive test of alloreactivity. Interleukin 2 (IL-2) producing helper T lymphocyte precursor (HTLp) frequencies determined by limiting dilution analysis (LDA) in the graft-versus-host direction were explored. The HTLp assay was optimized and the sources of error minimized to ensure sensitive and reproducible results. The IL-2 dependent cell line, CTLL-2 was optimized to detect 0.6 pg IL-2. UV-B irradiation of the cells was demonstrated to effectively terminate proliferation of the responder cells and thus allow IL-2 to be detected in the whole culture volume. The design of the assay was explored by Monte Carlo simulations resulting in a design yielding frequencies with a coefficient of variation of 20% in the range of 1:20,000-1:1,000,000. The influence of autoreactivity of the donor and recipient cells was minimized as well as the risk of the stimulator cells producing IL-2. The HTLp frequencies correlated with the degree of human leukocyte antigen (HLA) disparity and the assays were able to detect minor histocompatibility antigen mismatches. The HTLp frequencies of 28 HLA-identical sibling BMT pairs and 20 HLA-matched unrelated and partially HLA-matched related BMT pairs were determined. HTLp frequencies from the HLA-identical sibling BMT pairs had a median of 1:557,362 (range 1:9.511 to < 1:2,500,000). The HTLp frequencies from the HLA-matched unrelated and partially HLA-matched related BMT pairs had a median of 1:88,110 (range 1:4.139-1:736,123). Analysis of the HLA-identical sibling BMT pairs in a high and a low HTLp frequency group above and below 1:500,000 showed a trend towards a higher risk of acute GvHD > or = grade II and a significantly higher risk of chronic GvHD in the high HTLp frequency group. This group had a significantly lower risk of relapse as well as a significantly better overall survival and leukaemia free survival. The HLA-matched unrelated and partially HLA-matched related BMT pairs were split evenly in a high and a low HTLp frequency group above and below 1:90,000. There was a significantly higher risk of acute GvHD > or = grade II and a trend towards a higher treatment related mortality (TRM) in the high HTLp frequency group. There were no differences in chronic GvHD, risk of relapse, overall survival and leukaemia free survival. Analyzing all 48 patients the risk of acute GvHD > or = grade II and TRM was significantly higher with HTLp frequencies > 1:100,000 and there was a trend towards a higher risk of relapse with low HTLp frequencies < 1:400,000. Patients in the intermediate HTLp frequency group 1:100,000-1:400,000 had a trend towards improved survival. The HTLp frequency seems to detect clinically significant differences in alloreactivity, that can be useful in donor selection, graft-engineering, T cell add-back and the pharmacological immunosuppression used after BMT.