Allogeneic stem cell transplantation is an effective treatment for high-risk myeloid malignancies, but relapse remains the major post-transplantation cause of treatment failure. Alloreactive natural killer (NK) cells mediate a potent antileukemic effect and may also enhance engraftment and reduce graft-versus-host disease (GVHD). Haploidentical transplantations provide a setting in which NK cell alloreactivity can be manipulated, but they are associated with high rates of GVHD. We performed a phase I study infusing escalating doses of NK cells from an HLA haploidentical-related donor-selected for alloreactivity when possible-as a component of the preparative regimen for allotransplantation from a separate HLA-identical donor. The goal of infusing third-party alloreactive NK cells was to augment the antileukemic effect of the transplantation without worsening GVHD and, thus, improve the overall outcome of hematopoietic transplantation. Twenty-one patients with high-risk acute myeloid leukemia (AML), myelodysplastic syndrome (MDS), or chronic myelogenous leukemia refractory or beyond first remission received a preparative regimen with busulfan and fludarabine followed by infusion of apheresis-derived, antibody-selected, and IL-2-activated NK cells. Doses were initially based on total nucleated cell (TNC) content and later based on CD56(+) cells to reduce variability. CD56(+) content ranged from .02 to 8.32 × 10(6)/kg. IL-2, .5 × 10(6) units/m(2) subcutaneously was administered daily for 5 days in the final cohort (n = 10). CD3(+) cells in the NK cell product were required to be < 10(5)/kg. Median relapse-free, overall, and GVHD-free/relapse-free survival for all patients enrolled was 102, 233, and 89 days, respectively. Five patients are alive, 5 patients died of transplantation-related causes, and 11 patients died of relapse. Despite the small sample size, survival was highly associated with CD56(+) cells delivered (P = .022) and development of ≥ grade 3 GVHD (P = .006). There were nonsignificant trends toward higher survival rates in those receiving NK cells from KIR ligand-mismatched donors and KIR-B haplotype donors. There was no association with disease type, remission at time of transplantation, or KIR content. GVHD was not associated with TNC, CD56(+), or CD3(+) cells infused in the NK cell product or the stem cell product. This trial demonstrates a lack of major toxicity attributable to third-party NK cell infusions delivered in combination with an HLA-compatible allogeneic transplantation. The infusion of haploidentical alloreactive NK cells was well tolerated and did not interfere with engraftment or increase the rate of GVHD after allogeneic hematopoietic transplantation. Durable complete remissions occurred in 5 patients at high risk for disease recurrence. This approach is being further developed in a phase I/II trial with ex vivo-expanded NK cells to increase the NK cell dose with the objective of reducing relapse and improving the outcome of allogeneic hematopoietic transplantation for AML/MDS.