A prospective randomized and controlled clinical trial of prophylactic granulocyte transfusions was conducted in an attempt to reduce the mortality from infection in newly diagnosed adults with acute myelogenous leukemia. Granulocytes were harvested from normal donors by cytapheresis, and transfused patients received a median of 1.45 x 10(10) granulocytes (range 0.28--3.45 x 10(10)) on alternate days during marrow aplasia caused by initial induction chemotherapy. Transfusion were started if the absolute peripheral granulocyte count was less than 500 microliters. Thirteen patients received from one to 12 granulocyte transfusions, and 11 control patients received no granulocytes. No significant advantage was demonstrated in the transfused patients compared with controls with regard to the following: 1) deaths due to infection, 2) reduction in the frequency of febrile episodes, 3) delay in the onset of fever, 4) reduction in the length of febrile episodes, or 5) reduction in the frequency of proven infection.