Most clinical practice guidelines recommend restrictive red cell transfusion practices, with the goal of minimising exposure to allogeneic blood (from an unrelated donor). The purpose of this review is to compare clinical outcomes in patients randomised to restrictive versus liberal transfusion thresholds (triggers). To examine the evidence for the effect of transfusion thresholds on the use of allogeneic and/or autologous blood, and the evidence for any effect on clinical outcomes. Trials were identified by: computer searches of the Cochrane Central Register of Controlled Trials (the Cochrane Library Issue 3, 2009), OVID MEDLINE (1966 to August 2009), Current Contents (1993 to November 2004), and the Web of Science (2004 to August 2009). References in identified trials and review articles were checked and experts contacted to identify any additional trials. Controlled trials in which patients were randomised to an intervention group or to a control group. Trials were included where intervention groups were assigned on the basis of a clear transfusion 'trigger', described as a haemoglobin (Hb) or haematocrit (Hct) level below which an RBC transfusion was to be administered. Relative risks of requiring allogeneic blood transfusion, transfused blood volumes and other clinical outcomes were pooled across trials, using a random effects model. The risk of bias was assessed. Seventeen trials involving a total of 3746 patients were identified. Restrictive transfusion strategies reduced the risk of receiving a red blood cell (RBC) transfusion by a relative 37% (RR=0.63; 95% CI 0.54 to 0.74). This equates to an average absolute risk reduction (ARR) of 33% (95% CI 21% to 45%). The volume of RBCs transfused was reduced on average by 0.75 units (95% CI 0.20 to 1.30 units). However, heterogeneity between trials was statistically significant (P<0.001; I2≥74%) for these outcomes. Restrictive transfusion strategies did not appear to impact on the rate of adverse events compared to liberal transfusion strategies (i.e. mortality, cardiac events, myocardial infarction, stroke, pneumonia and thromboembolism). Restrictive transfusion strategies were associated with a statistically significant reduction in the rates of infection (RR=0.76; 95% CI 0.60 to 0.97). The use of restrictive transfusion strategies did not reduce hospital or intensive care length of stay. The existing evidence supports the use of restrictive transfusion triggers in patients who are free of serious cardiac disease. The effects of conservative transfusion triggers on functional status, morbidity and mortality, particularly in patients with cardiac disease, need to be tested in further large clinical trials. In countries with inadequate screening of donor blood, the data may constitute a stronger basis for avoiding transfusion with allogeneic red cells.