A 2012 Cochrane review analyzed 19 randomized controlled trials (RCTs), enrolling 6,264 patients overall, that compared restrictive versus liberal transfusion strategies in different clinical settings (surgery, acute blood losses, and/or trauma and critical care units) [3]. Results showed a reduction of the transfusion rates in the restrictive group compared to liberal group (risk ratio (RR) 0.61, 95 % confidence interval (CI) 0.52–0.72; p < 0.00001; I ² = 93 %). Such results were not confirmed in the vascular surgery sub-analysis (RR 0.91, 95 % CI 0.77–1.08; p < 0.3). Heterogeneity among trials for this outcome was statistically significant (Chi2 = 238.95, df = 16, p < 0.00001, I ² = 93 %). Furthermore, hospital mortality was 23 % lower in the restrictive group (RR 0.77, 95 % CI 0.62–0.95; p < 0.018; I ² = 0 %). Thirty-day mortality, hospital length of stay, and complications (cardiac events, myocardial infarction, pulmonary edema, cerebrovascular accidents/stroke, pneumonia, and infection) were not different in the two groups.

In 2015, an update of the Cochrane review and meta-analysis [4] of both single-center and multicenter RCTs (overall 9,813 patients of both surgical and medical settings) confirmed the reduction of the transfusion need with the restrictive strategy versus the liberal one (RR 0.54, 95 % CI 0.47–0.63; p < 0.001; I ² = 95 %). Even in this case, no difference in terms of mortality, myocardial infarction, overall morbidity, and adverse events (cardiac complications, renal failure, thromboembolic stroke, transient ischemic attack, or hemorrhage) was observed between groups. Only a possible association between the restrictive strategy and a reduced rate of infections was noted (RR 0.73, 95 % CI 0.55–0.98; p = 0.03; I ² = 53 %).

Recently, three RCTs performed in perioperative period demonstrated benefits from a liberal transfusion strategy [5–7]. Two of them were not included in the 2015 update of the Cochrane review.

De Almeida and colleagues [5] studied 198 patients undergoing major abdominal surgery for cancer and who required postoperative ICU stay for at least 24 h. Patients were randomly assigned to either the restrictive or the liberal transfusion strategy group. Patients received erythrocyte units each time the hemoglobin level decreased below 7 g/dL (restrictive group) or 9 g/dL (liberal group) during their ICU stay. The authors observed, in the liberal strategy group, a lower 30-day (8[8.2 %] vs. 23[22.8 %]; p = 0.005) and 60-day (11[11.3 %] vs. 24[23.8 %]; p = 0.022) mortality rate compared to the restrictive strategy group. Furthermore, a lower incidence of overall major cardiovascular events in the liberal group (5[5.2 %] vs. 14[13.9 %]; p = 0.038) and a higher incidence of intra-abdominal infections in the restrictive group (15[14.9 %] vs. 5[5.2 %]; p = 0.024) were observed.

The Transfusion Indication Threshold Reduction (TITRe2) trial (17 cardiac surgery centers in the United Kingdom) randomized 2003 patients undergoing non-emergency cardiac surgery to the restrictive-threshold group (hemoglobin level of 7.5 g/dL) or to the liberal-threshold group (hemoglobin level of 9 mg/dL). Results of this trial showed a higher mortality rate within 3 months in the restrictive group than in the liberal one (4.2 % vs. 2.6 %; hazard ratio 1.64; 95 % CI 1.00–2.67; p = 0.045). No differences were found between the two groups with regard to the other outcomes (infections, ischemic events, ICU, high-dependency unit, and hospital length of stay) [6].

The Transfusion Requirements in Frail Elderly (TRIFE) trial, conducted by Gregersen and colleagues, is a single-center trial that enrolled 284 patients, aged ≥65 years, undergoing surgery for unilateral hip fracture, coming from nursing homes or from sheltered housing facilities. Patients in the restrictive strategy group received transfusions if their postoperative hemoglobin levels were lower than 9.7 g/dL, while patients in the liberal strategy group received transfusions if postoperative hemoglobin levels were lower than 11.3 g/dL. The authors found no difference for the primary outcome (recovery from physical disabilities at 10, 30, and 90 days after surgery) between the two groups. Concerning the secondary outcome (30-day and 90-day mortality), no difference was found between groups analyzing the data by the intention-to-treat, while a higher 30-day mortality was observed in the restrictive group using the per-protocol analysis (HR 2.4; 95 % CI 1.1–5.2; p = 0.03). Subgroup analysis showed a higher 90-day mortality in the nursing home patients of the restrictive group with both the intention-to-treat (HR 2.0; 95 % CI 1.1–3.6; p = 0.01) and the per-protocol (HR 1.9; 95 % CI 1.0–3.4; p = 0.04) methods [7].

These three RCTs raised the possibility that a more restrictive transfusion strategy may be associated with a higher mortality. For this reason and since the above-mentioned Cochrane reviews considered RCTs conducted in different contexts (e.g., surgery, ICUs, etc.) without distinguishing among them and considered both adults and children, Fominskiy et al. performed a new meta-analysis of RCTs [8]. The authors considered RCTs that enrolled only adults (age ≥18 years). Furthermore, they analyzed separately studies performed in perioperative settings (17 studies of which 9 in orthopedic surgery, 5 in cardiac surgery, 1 in vascular surgery, 1 in cancer surgery, and 1 in obstetrics) and studies performed in critically ill contexts (10 studies) for a total of 11,021 patients. Fourteen trials were multicenter; 18 trials included more than 100 patients and 2 studies more than 1,000 patients. Results of this meta-analysis showed that, in perioperative setting, mortality for all causes is reduced with the liberal transfusion strategy groups compared with the restrictive transfusion strategy groups (Odds Ratio 0.81, 95 % CI 0.66–1.00; p = 0.05; I ² = 25 %). In the critically ill setting, there was no difference in all-cause mortality between liberal and restrictive groups (RR 1.10, 95 % CI 0.99–1.23; p = 0.07; I ² = 34 %). No differences were also found in all-cause mortality between liberal and restrictive strategies when considering together the perioperative and the critically ill settings (Odds Ratio 0.96, 95 % CI 0.78–1.18; p = 0.68). This is a further step forward in understanding the importance of tailoring the best transfusion strategy on each clinical setting.