Restrictive or Liberal Red-Cell Transfusion for Cardiac Surgery

Mazer. 2017 DOI 10.1056/NEJMoa1711818

Clinical Question

  • In patients undergoing cardiac surgery, who have a moderate-to-high risk of death, does a restrictive transfusion strategy compared with a liberal strategy impact on a composite outcome of death, myocardial infarction, stroke and acute kidney injury requiring dialysis


  • Transfusion thresholds in critically ill patients is a long debated topic
  • There have been several high quality randomised controlled trials investigating transfusion
  • The TRISS trial investigated liberal versus restrictive transfusion in patients with septic shock and demonstrated no significant difference in 90 day mortality
  • Patients with acute upper gastrointestinal bleeding had lower mortality with a restrictive strategy (Villanueva, NEJM 2013;368(1)11-21)
  • The TRICC trial assessed a liberal versus restrictive strategy in 838 ICU patients expected to stay >24 hours and showed no mortality difference. However, patients who were admitted after cardiac surgery were excluded from this study
  • The evidence in post-cardiac surgery patients is limited by small, poor quality trials. The concern has been that a restrictive strategy may increase the risk of anaemia-induced tissue hypoxia


  • Randomised controlled trial
  • 1:1 ratio of restrictive vs liberal strategy
  • Allocation concealment (central web-based allocation)
  • Permuted blocks of varying size, stratified to centre
  • Unblinded, although the outcomes were evaluated by those unaware of the trial-group assignment
  • The primary analysis was a per-protocol analysis; excluding patients who were withdrawn from the trial by treating physician, patients who withdrew consent and those that had a protocol adherence of less than 90%.  Non-adherence was measured by counting the number of non-adherent trigger events and non-adherent red blood cell (RBC) transfusions, as a proportion of the total number of hemoglobin measurements recorded in the electronic case report form (eCRF). Adherence is 1- non adherence
  • A modified intention-to-treat analysis was also conducted
  • Power calculation: The event rate of the primary composite outcome was assumed to be 10%. Partway through the trial, the power of the trial was increased from 85% to 90%, which required a sample size of 5000 patients. The aim was to detect non-inferiority at a one-sided alpha of 0.025, with a non-inferiority margin of 3%


  • 19 countries, 73 sites
  • Randomisation from January 20th 2014 to March 20th 2017


  • Inclusion: patients 18 years or older undergoing cardiac surgery with a pre-operative EuroSCORE I of 6 or higher (the EuroSCORE I is a validated scoring system that is predictive of in-hospital mortality)
  • Exclusion
    • unable to receive or declined blood products
    • pre-operative autologous donation program
    • heart transplant surgery
    • surgery for ventricular assist device
    • pregnant or lactating
  • 5243 were randomized; 54 did not have surgery, 63 did not have cardiopulmonary bypass, 34 withdrew before surgery to leave 5092 in the modified intention-to-treat analysis
  • 4860 patients were included in the per protocol analysis (excluding 8 in each group where intervention was discontinued or consent withdrawn AND 117 in the restrictive group and 99 in the liberal group where adherence to the protocol was <90%)
  • Baseline demographics similar in both groups


  • Patients in the restrictive group received a red-cell transfusion if the haemoglobin concentration was less than 7.5g/dL intra-operatively or post-operatively


  • Patients in the liberal group received a red-cell transfusion if the haemoglobin concentration was less than 9.5g/dL intra-operatively or post-operatively or less than 8.5g/dL in the non-ICU ward

Management common to both groups

  • Haemoglobin level was measured pre-operatively, intra-operatively and post-operatively at specified intervals. If the haemoglobin fell below the threshold, 1 unit of red cells was administered followed by reassessment of the level
  • All other aspects of care were directed by the treating medical teams


  • Primary outcome: A restrictive transfusion threshold was non-inferior to the liberal transfusion group for a composite outcome of death, myocardial infarct, new onset dialysis requiring renal failure, during the hospitalisation (from the start of surgery until either hospital discharge or 28 days after surgery (whichever came first)
    • Per-protocol analysis:11.4% in the restrictive group compared to 12.5% in liberal group (absolute risk difference = -1.11%; 95% CI, -2.93 to 0.72, OR 0.90; 95% CI -2.93 to 0.72)
    • Modified intention to treat analysis (which included patients even if there was <90 % non-adherence to the protocol) also showed non-inferiority in the restrictive group (12.3% vs 12.9%)
  • Secondary outcome: No statistical differences in:
    • Each of the components of the composite outcomes
    • LOS in ICU and hospital
    • Duration of mechanical ventilation
    • Prolonged cardiac output state
    • Infection
    • Bowel infarction
    • Acute Kidney Injury
    • Seizure
    • Delirium
    • Encephalopathy
  • In the restrictive group, 52.3% were transfused and those transfused averaged 2 units. In the liberal group, 72.6% were transfused and those transfused averaged 3 units of red-cells.
  • Subgroup analysis based on age (>75 years) showed the restrictive-threshold group was associated with a lower risk of the composite primary outcome than the liberal-threshold group (OR 0.70; 95% CI 0.54-0.89 p=0.004)

Authors’ Conclusions

  • The restrictive transfusion strategy is non-inferior to the liberal transfusion strategy in cardiac surgery patients with a moderate-to-high risk of death with regards to a composite outcome of death and major disability


  • Multi-centre
  • Allocation concealment
  • Complete follow-up
  • Per-protocol and modified intention-to-treat analysis
  • Blinding of investigators assessing primary outcome, patients were not actively told about their treatment assignment
  • Patients were randomised before surgery meaning that the effect of two transfusion strategies was examined for the whole of their journey as a cardiac surgery patient
  • The sample population were a sick group of cardiac surgery patients, so treatment effect, if any would likely be more obvious than low-risk cardiac surgery patients


  • Non-blinding of clinical staff administering the blood
  • The per-protocol analysis excluded patients where adherence to the protocol was <90%. This could potentially pull potentially sicker patients out of the analysis of the primary outcome, impacting the power of the study. The authors did a modified intention to treat analysis which did include these patients though and still demonstrated non-inferiority in the restrictive strategy
  • This trial is in patients with moderate-to-high risk of death during cardiac surgery and is not necessarily generalisable to those with low-risk cardiac surgery
  • It is difficult to know the significance of the subgroup analyses from this trial but it is certainly hypothesis generating, particularly that older patients may fair worse than younger patients with liberal transfusion
  • The trial does not provide information as to whether a transfusion threshold even lower than 7.5g/dL would be safe

The Bottom Line

  • This trial adds to the growing body of evidence that restrictive transfusion protocols are not harmful in critically ill patients
  • This study demonstrates that post-cardiac surgery patients can safely be managed with haemoglobins as low as 7.5g/dL

External Links


Summary author: Celia Bradford – @celiabradford
Summary date: November 22nd 2017
Peer-review editor: Steve Mathieu

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.