The TRICC Trial

Hébert PC, Wells G, Blajchman MA, Marshall J, Martin C, Pagliarello G, Tweeddale M, Schweitzer I, Yetisir E. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. Transfusion Requirements in Critical Care Investigators, Canadian Critical Care Trials Group. N Engl J Med. 1999 Feb 11;340(6):409-17. Erratum in: N Engl J Med 1999 Apr
1;340(13):1056. PubMed PMID: 9971864. [Free Full Text]

• In patients with a critical illness does a restrictive transfusion strategy (to a target of 70 – 90 g/L, transfusion trigger <70 g/L) produce equivalent mortality outcomes to a liberal transfusion strategy (target of 100 – 120 g/L, transfusion trigger <100 g/L)?

TYPE OF STUDY

• Multi centre randomised controlled trial (22 tertiary level and 3 ‘community’ ICUs in Canada)
• Computer randomised using permuted blocks of 4 or 6, and stratified by center APACHE score of >15 and 15 or less
• Non-blinded
• See CONSORT diagram

POPULATION

• Inclusion criteria:
  • Expected to stay in ICU > 24 hours
  • Hb concentration of 90 g/L or less within 72 hours of admission to ICU
  • Considered to be euvolaemic following initial treatment by ICU team
• Exclusion criteria:
  • Age < 16 years
  • Inability to receive blood products
  • Active bleeding at time of enrolment (evidence of ongoing blood loss and a decrease in Hb concentration of 30 g/L in the preceding 12 hours OR a requirement for at least 3 units of packed RBCs in the same period)
  • Chronic anaemia
  • Pregnancy
  • Brain death
  • Imminent death (<24 hours)
  • Consideration of withdrawal of treatment/DNR order
  • Admission for a routine cardiac surgical procedure
  • Previous transfusion
  • Enrolment in other study

INTERVENTIONS/ COMPARISONS

• Restrictive transfusion strategy (to a target of Hb 70 – 90 g/L, transfusion trigger Hb <70 g/L); n = 418
• Liberal transfusion strategy (to target of Hb 100 – 120 g/L, transfusion trigger Hb <100 g/L); n = 420

OUTCOMES

• Average daily Hb concentrations differed significantly
  • Restrictive 85 +/- 7 g/L, Liberal 107 +/- 7 g/L
  • Number of units of RBCs transfused was significantly reduced (RRR 0.54) in the restrictive group (2.6 +/- 4.1 units) compared with the liberal group (5.6 +/- 5.3 units) (p < 0.01)
• Primary outcome
  • rate of death from all causes at 30 days
    • No significant difference
    • Restrictive 18.7%, Liberal 23.3% (95%CI -0.84 – 10.2%, p = 0.11)
• Secondary outcomes
  • Mortality rates during hospitalisation were lower in the restrictive group, but there was no difference in ICU mortality or 60 day mortality
    • inpatient mortality: 22.2% vs. 28.1% (ARR 5.8%; P=0.05)
    • ICU mortality: 13.9% vs. 16.2% (ARR 2.3%; P=0.29)
    • 60-day mortality: 22.7% vs. 26.5% (ARR 3.7%; P=0.23)
  • There were slightly lower adjusted multiple-organ dsyfunction scores in the restrictive group (10.7 +/- 7.5 vs 11.8 +/- 7.7; p=0.03)
• Subgroup analyses
  • When analysed by sub-group baseline characteristics remained similar between treatment arms
  • No differences in survival when adjusted for:
    • trauma
    • cardiac disease
    • severe infections or septic shock
  • There was lower mortality in the restrictive group for these a priori determined subgroups:
    • APACHE II ≤20: 8.7% vs 16.1% (95% CI 1.0% – 13.6%, p = 0.03)
    • Age <55 years: Restrictive 5.7%, Liberal 13.0% (95% CI 1.1% – 13.5%, p = 0.02)

Adjusted multiple-organ dysfunction scores were statistically significantly lower in the restrictive group. compared with the liberal group, in the APACHE <20 and age < 55 subgroups

• Adverse cardiac events (e.g. cardiac ischaemia, pulmonary oedema, cardiac arrest) were more common in the liberal group: 13.2% vs. 21% (ARR 7.8%; P<0.01)
• All of the liberal group received blood transfusion, whereas 33% of the restrictive group did not receive transfusion

Commentary

• This is a landmark practice-changing classic of ICU research, but it is not without its flaws.
• The study raised concerns about the harms of blood transfusion and questioned the presumed benefit of increasing oxygen delivery – at least by transfusion – in the critically ill.
• Practice guidelines today generally support a transfusion trigger of Hb 70 g/L in critically ill patients as a result, except for selected patient subgroups (e.g. cardiac patients)

• Subsequent studies have supported similar targets in patients with upper GI haemorrhage and septic shock
• There is considerable controversy about the optimal Hb target in individual ICU patients and there are concerns over the external validity and power of the TRICC trial

Strengths

• Randomised, multicentre controlled trial
• Appropriately chosen primary/secondary outcomes
• Power calculation performed, but…
  • calculated to detect a 5% difference in 30d mortality
  • target sample size of 1620 patients was determined by an interim analysis suggesting a higher than expected mortality rate
  • see criticisms below
• used intention to treat analysis
• Relevant population group (i.e. APACHEII scores consistent with critical illness)
• Comparable healthcare setting to the Australian setting
• Highlighted how a simple, inexpensive to institute intervention could significantly reduce costs and lead to apparently superior patient outcomes

Criticisms

• The total number of patients enrolled (n=838) resulted in the study being underpowered and thus prone to type II error
• Only ~12.9% of total assessed patients were enrolled in the trial, with a large number of them being excluded due to physician refusal (selection bias)
• APACHE II score subgroup stratification was altered after the fact (<15 changed to <20) – would a significant difference favouring restrictive blood transfusion in the less severely ill group have been found otherwise?
• This study may have been subject to practice misalignment
  • a survey by Hebert et al (1998) showed that standard practice was not to transfuse patients to Hb >100 g/L and most clinicians would use a target higher than 70 g/L in patients with ischaemic heart disease. Allocation of these patients to the corresponding treatment arms might have been expected to cause harm.
• External validity is limited by the following:
  • Exclusion of patients following cardiac surgery removed a large population of critically ill patients that may require significant volumes of blood and blood products – this is markedly different to the Alfred ICU population
  • The study not include long-stay ICU patients who required transfusion later (i.e. onset of anaemia had to occur within 72 hours)
  • There was a significant difference in the percentage of patients with cardiovascular disease between those excluded from the trial and those enrolled (20% vs 26% respectively) – excluded pateints also tended to be older – decreasing ability to apply results across a broader population
  • 297 patients were excluded for having had a transfusion prior to ICU that increased their Hb to >90 g/L – may reduce applicability to certain patient groups who may receive frequent or be at high likelihood of transfusion
  • Study conducted prior to routine use of leucodepleted PRBCs in Canada – would repeating the trial with leucodepleted PRBCs diminish the significant differences?

• The TRICC study is a landmark trial that supports the use of a restrictive transfusion strategy targeting Hb> 70 g/L in ICU patients. However, the optimal Hb target in different patient subgroups (e.g. cardiac disease) remains uncertain.
• Concerns remain over the external validity of the study and the fact that it was underpowered.