Recent enthusiasm for intraoperative autotransfusion has overshadowed critical assessment of its potential risks. In this study, adult mongrel dogs underwent controlled intraperitoneal hemorrhage of twice their estimated blood volume over a 4-hour period. The blood was replaced by an equal volume of banked blood (Group I,n= 5), or collected and reinfused via the Sorenson® System (Group II,n= 6), or the Haemonetics Cell Washing Device® (Group III,n= 6). Acid citrate dextrose was the local anticoagulant for Groups I and II, and heparin for Group III. Pulmonary capillary wedge pressure and cardiac output were maintained at baseline values with crystalloid infusion. Core temperature, pO2, and systemic pH remained normal throughput the 4 hours of evaluation.Red blood cell recovery was efficient in all animals, and the 2,3 DPG levels remained normal in the autotransfused dogs. Thrombocytopenia, however, developed uniformly and was more pronounced after autotransfusion. Platelet numbers decreased nearly 45% in the Sorenson as well as Haemonetics animals. Additionally, platelet dysfunction occurred after one blood volume exchange as evidenced by prolonged bleeding times and loss of the secondary wave on Sonoclot profiles. Coagulation studies revealed progressive consumptive coagulopathy and fibrinolysis in autotransfused dogs. The P.T., P.T.T., and T.T. lengthened, and levels of factors II, V, VIII, and fibrinogen fell.Autotransfusion clearly eliminates the infectious and incompatibility problems of banked homologous blood. Despite advances in technique, however, consumptive coagulopathy, fibrinolysis, and platelet dysfunction occur. The risk of these complications may outweigh the benefits of autotransfusion in the critically injured patient with multiple potential bleeding sites.