PURPOSE Patients undergoing hematopoietic stem-cell transplantation (HSCT) or induction chemotherapy for acute leukemia (AL) generally receive a low-bacterial neutropenic diet (ND) to minimize infection from ingested microbes. Recent studies suggest that a more liberalized diet (LD) might be safe. However, those studies lacked robust designs to demonstrate noninferiority and did not address whether an LD improves caloric intake or nutritional status. This trial aimed to determine whether the rate of major infections in LD patients was noninferior to that in ND patients and whether an LD improved caloric intake and nutritional status. METHODS In this phase III noninferiority trial, patients were randomly assigned to either an ND or LD, containing fresh fruits and vegetables. The primary end point was the development of any major infection during neutropenia. The noninferiority margin was set at a difference of ≤10% in the major infection rate. RESULTS The trial was halted at the second interim analysis after enrolling 214 evaluable patients, because of the LD arm's major infection rate surpassing the predefined stopping boundary. In the LD arm, major infections occurred in 31.4% of patients compared with 20.2% of patients in the ND arm, a difference of 11.2% ([95% CI, –0.4 to 22.9]; P = .58). Furthermore, the caloric intake in the LD arm was not improved and nearly two thirds of patients on both diets reported critical nutritional problems. There was no advantage in symptoms, quality of life, or survival for the LD arm. CONCLUSION These findings suggest that LD is not a safe alternative to ND in HSCT patients and patients with AL because of increased infection risk without nutritional or other benefit. Further dietary strategies are needed to improve nutrition without compromising safety.