Reducing the need for blood transfusions is desirable for several reasons. Since 2000 in the United Kingdom it has been mandatory to remove all white cells from donated blood to reduce the small but theoretical risk of prion disease (variant Creutzfeldt-Jakob disease). This has trebled the cost of providing donated blood. Transmission of hepatitis B, hepatitis C, and HIV by transfusion occurs in 1 in 300 000 cases, despite screening programmes.1 However, non-fatal but serious transfusion errors occur in 1 in 16 000 transfusions.1

Critically ill patients are now known to do just as well with a lower haemoglobin concentration than previously thought, thus reducing the need for top-up transfusions.2 There is also some evidence that homologous blood transfusions increase the rates of recurrence of some cancers (tumours of the bowel and oesophagus, in particular) and can increase the incidence of wound infections.3 It is unclear why these phenomena occur.

A number of mechanical methods have been developed to help reduce the need for postoperative blood transfusions. In the United States erythropoetin injections or autologous blood donations (or both), given preoperatively, are commonly used. Both require the exact date of surgery to be known—but neither process is free from human error in labelling, storing, and administration.

Perioperative dilution and intraoperative blood salvage techniques (such as those described in this paper) are gaining credence, particularly for patients undergoing cardiac and orthopaedic surgery. But neither of these processes is suitable for patients with infection or malignant disease.

After surgery, devices are available to collect blood from wound drains, which can then be retransfused back into the patient. Such techniques reduce the formation of haematomas, but few studies of their efficacy are available, and the techniques are not in general use.4