In this prospective, randomized, double-blind study, we have investigated the effect of an antifibrinolytic agent, tranexamic acid (Cyklokapron), on blood loss and transfusion requirements associated with total knee arthroplasty. Twenty-nine patients were allocated randomly to receive either tranexamic acid 15 mg kg-1 or an equal volume of placebo a few minutes before a tourniquet was deflated. Blood loss during surgery, in the recovery room and on the surgical ward was recorded, together with the number of units of blood transfused in hospital. Mean blood loss during surgery was 428 (SD 254) ml in the tranexamic acid group (n = 15) compared with 415 (244) ml in the placebo group (n = 13). In the recovery room the tranexamic acid group lost 127 (95) ml and the placebo group 576 (245) ml (P < 0.001). On the ward the respective volumes were 293 (200) ml and 558 (293) ml (P < 0.01). Total blood loss was 847 (356) ml in the tranexamic acid group and 1549 (574) ml in the placebo group (P < 0.001). During the hospital stay the treatment group received 1.5 (1.3) units of blood compared with 3.3 (1.8) in the control group (P < 0.005). Two patients in the placebo group experienced a thrombotic complication compared with none in the treatment group. We conclude that tranexamic acid reduced perioperative blood loss and transfusion requirements associated with total knee arthroplasty.