TRICC Investigators: A Multicenter, Randomized, Controlled Clinical Trial of Transfusion Requirements in Critical Care

Herbert et al for the TRICC Investigators. N Engl J Med 1999; 340:409-417.

Clinical Question

• In patients admitted to ICU does a restrictive vs. liberal blood transfusion strategy alter mortality?

Design

• RCT, computer generated randomisation, opaque envelopes
• Stratified by centre and APACHE score
• Non-blinded

Setting

• 22 tertiary and 3 community ICUs in Canada
• November 1994–7

Population

• 6541 adult patients were assessed of whom 838 randomised
• Euvolaemic patients as judged by attending physician
• Expected stay in ICU >24hrs
• Acute anaemia with Hb ≤9 g/dl within 72hrs of admission
• Excluded patients with active blood loss; chronic anaemia (Hb<9 g/dl more than one month prior to admission); admission after routine cardiac surgery

Intervention

• Restrictive strategy: Transfused when Hb<7 g/dl (target 7-9 g/dl)

Control

• Liberal strategy: Transfused when Hb<10 g/dl (target 10-12 g/dl)

Outcome

• Primary outcome: Overall 30 day mortality was similar in both groups. 18.7% in restrictive vs. 23.3% in liberal, (95% C.I. -0.84 to 10.2, P=0.11)
• Secondary outcomes:
  • Hospital mortality higher in liberal transfusion group: 22.2% in restrictive vs. 28.1% in liberal, (95% C.I. -0.3 to 11.7, P=0.05)
  • More than 3 organ failure no difference
• Subgroup analyses:
  • Mortality rate was significantly lower in the restrictive group in the following circumstances:
    • In the less acutely ill patients (APACHE II score ≤ 20), 30-day mortality was 8.7% vs. 16.1%, 95% C.I. for absolute difference 1-13.6%, P=0.03
    • In the patients who were < 55 years of age (5.7% vs. 13%, P=0.02)
  • Mortality was similar in restrictive compared with the liberal groups in patients with known cardiac disease  (20.5% vs. 22.9%, p=0.69)
  • Cardiac events (MI, pulmonary oedema, angina, cardiac arrest) more common in liberal (21%) vs. restrictive group (13.2%), P<0.01

Authors’ Conclusions

Strengths

• Multi-centre
• Randomisation appropriate design and method
• Audited opaque envelopes to check none missing
• Appropriate primary outcome
• Appropriate power calculation
• Intention to treat analysis
• Use of consort style diagram
• The average APACHE II score was 21 indicating a reasonably sick population
• 33% of the restrictive group did not receive any blood, but all of the liberal group received blood, so the transfusion rate was reasonably high, making the findings more impressive. A lower hospital mortality with an average transfusion requirement difference of 2 units between the groups is a powerful challenge to the previously held belief that high haemoglobin levels are always necessary in the critically ill

Weaknesses

• Selection bias with large number of patients not included in the trial, including 598 following physician refusal. Only 13% patients were included in the trial
• Changed subgroups post-hoc from APACHE score <15 to <20 (did they do some data dredging to get their ‘significant’ result?)
• Only included patients with anaemia within 3 days of admission
• Inadequate sample size. Study powered to rule out an absolute difference in 30-day mortality of 5.5% between treatment groups with 1620 patients. However study only randomised 838 patients therefore potential for Type 2 error
• No subgroup analysis for patients with traumatic brain injury and it has been argued that these patients should have a higher Hb level
• Not necessarily a weakness, but worth raising, is the issue surrounding the transfusion trigger for patients with IHD. The authors recommended caution in those patients with significant cardiac disease, in particular those with acute myocardial infarction and unstable angina. This was based primarily on 2 cohort studies (Carson 1996; Hebert, TRICC & Canadian Critical Care Trial Investigators 1997) which demonstrated that increased level of anaemia in the presence of ischaemic heart disease resulted in increased mortality. Interestingly, this study does not support this finding and hypothesised whether ‘the apparent discrepancy between our results and those of previous studies may be the result of confounding or an inability to document the negative effects of transfusion in the observational studies’