The prognosis of acute leukemia refractory to induction chemotherapy or immunotherapy is dismal. Salvage allogeneic hematopoietic stem cell transplantation (HSCT) is widely used option for these patients, but only 10% to 15% of patients are cured by the procedure. Preclinical studies indicate that substitution of post-transplantation cyclophosphamide with bendamustine (PTB) in a prophylaxis regimen may be associated with an augmented graft-versus-leukemia (GVL) reaction. The aim of this study was to establish the optimal dose of PTB and evaluate the antileukemic effect of HSCT with this type of graft-versus-host disease (GVHD) prophylaxis. In the prospective trial (NCT02799147), PTB was administered in doses of 140, 100, and 70 mg/m2 on days +3 and +4. Myeloablative conditioning with fludarabine and oral busulfan was provided to all patients. The first 12 patients received single-agent PTB, and subsequent patients received combination therapy with tacrolimus and mycophenolate mofetil (MMF). Inclusion criteria were acute myelogenous leukemia (AML) or acute lymphoblastic leukemia (ALL) refractory to at least one induction course of chemotherapy or target therapy and ≥5% clonal blasts in the bone marrow. The study cohort comprised 22 patients with AML and 5 with ALL. Seven patients were enrolled in the 140 mg/m2 group (due to a stopping rule), and 10 each were enrolled in the 100 mg/m2 and 70 mg/m2 groups. Primary refractory disease was documented in 41% of the patients, and secondary refractory was documented in 59%. The median blast count in the bone marrow at the start of the conditioning was 18% (range, 6% to 97%). Transplantation was performed with a matched sibling donor in 5 patients, a matched or mismatched unrelated donor in 15, and a haploidentical donor in 7. Engraftment was documented in 93% of the patients, including 89% with complete remission and 63% without measurable residual disease. After PTB prophylaxis, we observed an unusual complication, a cytokine release syndrome (CRS), in 70% of the patients, including grade 3 to 5 CRS in 44%. The most frequent clinical symptoms included high fever in 67% of patients, abnormal liver function tests in 67%, pancreatitis in 63%, skin vasculitis in 56%, enterocolitis in 48%, inflammation of oral mucosa in 37%, disseminated intravascular coagulation in 37%, and central nervous system toxicity in 26%. The development of CRS was associated with use of an HLA-mismatched donor (75% versus 20%; P=.0043). Classic acute GVHD was documented in 44% of the patients. Grade II-IV acute GVHD was associated with grade 3 to 5 CRS (67% versus 25%; P=.031). Moderate and severe chronic GVHD in the 100-day survivors were more often observed after single-agent PTB than after the combination immunosuppression (100% versus 18%; P=.002). A relatively low relapse rate was observed for this patient population. Three-year overall survival was 28% (95% confidence interval [CI], 13% to 46%), and event-free survival was 29% (95% CI, 13% to 46%). Nonrelapse mortality was 46% (95% CI, 25% to 64%), and the cumulative incidence of relapse was 26% (95% CI, 11% to 44%). No relapses were documented after day +100. There were no statistically significant differences among the dose groups (P=.3481); however, survival was higher in the 100 mg/kg group. Survival was higher in patients with AML compared with those with ALL (35% versus 0%; P=.0157). PTB represents a promising option to augment the GVL effect in refractory AML; however, the high CRS-associated mortality necessitates additional studies to reduce the risk of this complication. Thus, routine clinical application of PTB cannot be currently recommended. Combination immunosuppression with tacrolimus and MMF partially ameliorates these complications, at least in the setting of HLA-matched allografts. Biological mechanisms of CRS and GVL after PTB require further elucidation.