Appropriateness and safety of blood transfusion

For a decade, huge sums have been spent on reducing infectious risks of blood transfusion in the United Kingdom. Since it seems paradoxical to finesse the safety of a medication when it is often used unnecessarily, we need to invest in understanding when transfusion will be effective.¹ Transfusion can avert death from haemorrhage, prevent bleeding when the marrow is suppressed, or raise the haemoglobin count rapidly when it is dangerously low. However, well designed clinical trials seem to have disproved some widely accepted indications for transfusion.

About 2.5 million units are supplied annually in the United Kingdom, of which 40-50% are used in elective and emergency surgery² and 5% in intensive care. A single large randomised trial showed that critically ill patients did as well or better with less transfusion.³ A recent small trial found that two units of red cells had no effect on oxygenation in patients in intensive care units whose haemoglobin concentration was below 80 g/l.⁴ In neonates, a liberal transfusion regime was no more effective than a conservative one.⁵

About 300 000 units of fresh frozen plasma are supplied annually. A systematic review assessed 57 trials: only three were of adequate design and power.⁶ One showed that fresh frozen plasma failed to prevent neonatal intracranial haemorrhage. The second showed no benefit in the treatment of acute pancreatitis. The third showed that plasma exchange with fresh frozen plasma was more effective than its infusion in treating thrombotic thrombocytopenic purpura.

A high quality trial of human albumin showed that it was equivalent to saline in resuscitation,⁷ resolving a debate that has exercised BMJ correspondents since 1998.

The United Kingdom haemovigilance study, SHOT (serious hazards of transfusion), collated reports of serious transfusion hazards for 1996-2003, during which time 23 million units of blood components were supplied. The incidence of serious adverse reactions (per 100 000 units of blood supplied) was death 0.2 and major morbidity 1.1 (of which 0.6 was transfusion related acute lung injury and 0.2 was infection—mostly bacterial).⁸ The chance, in the United Kingdom, that a unit of blood might transmit one of the viruses for which blood is tested is lower and estimated at HIV 0.014, hepatitis C virus 0.024, and hepatitis B virus 0.176 per 100 000 units (Eglin R, Davison K, personal communication, 2004). Considerably more patients may also be put at risk by receiving an unintended blood component (six per 100 000) including ABO incompatible blood (one per 100 000).

In 1995 it cost £250m (at 2004 sterling values) to produce blood components and associated services. A similar quantity now costs the NHS £500m. Put another way, in 1998 a unit of red cells cost £47, a platelet dose £90, and a unit of fresh frozen plasma £13. These have risen to £120, £198, and £31, mainly owing to newer antimicrobial tests and processes, some of which increase safety only marginally. An example is the hepatitis C virus RNA test, introduced in 1999 to detect infection missed by the routine hepatitis C virus antibody screening of all blood donations. This test has detected one additional positive sample per 1.4 million donations tested, at a cost of £8.3m per detection. A suggested cost of £7.9m per quality of life year gained is therefore almost certainly an underestimate.⁹

In the United Kingdom, 17 blood donors are known to have later developed variant Creutzfeldt-Jakob disease. Fifty recipients of blood from these donors have been identified, of whom 18 are still alive. Two of the 50 recipients are known to have been infected. One has since died with clinical vCJD; the other had histological signs of infection but died of unrelated causes.¹⁰ Precautionary measures to mitigate the risk of transmission of vCJD include removing leucocytes from all blood components at an annual cost of £70m, enough to fund the blood supplies of several African countries. Leucodepletion may remove about half of the infectious agent,¹¹ but on current estimates of infectivity levels in blood, leucocyte removal alone is unlikely to remove the risk of transmission. New processes soon to be available may reduce vCJD infectivity by perhaps 10 000 fold, but could add £100m per year to the blood services bill. The cost of avoiding an infection cannot be estimated but is certain to be enormous.

Other anti-infection technologies that are in the pipeline for platelets and red cells would add yet more costs and could also introduce new and unpredicted risks.¹² For example, a process that inactivates viruses in plasma also reduces the concentrations of natural anticoagulants (proteins S and C), increasing the risk of thrombosis in some patients. Some processes use DNA cross linking chemicals, and it may be difficult to prove that even tiny residual quantities could not be immunogenic or mutagenic. Furthermore, these processes may reduce the efficacy of blood. For example, losses owing to tests and safety measures now reduce the red cell content of a blood pack by 10%, and some patients will therefore require more units, adding to the risk.

We think that these large and recurring expenditures on blood safety should be balanced against the costs of the clinical trials still needed to provide an adequate evidence base for the use of transfusion, alternatives, and avoidance strategies. The decisions should involve a well informed public and be understood, and accepted, by them.