Monday, 20 October 2014

Transfusion restriction in Septic Shock - the TRISS trial

Paper Link

Holst L et al for the TRISS trials group and the Scandinavian Critical Care Trials Group
New England Journal of Medicine (2014); 371:1381-91

At a glance

  • Intervention - restrictive transfusion threshold (<7g/dL) in septic shock patients
  • Randomised - Yes
  • Controlled - Yes.  Control group transfused to 9g/dL
  • Placebo used - no
  • Blinded - Partly.  Clinicians were not blinded, though this would be very difficult.  Data collectors were blinded to study arm.
  • Population / Sample size calculation - Powered to detect 9% absolute risk reduction, requiring 1000 patients
  • Groups same at randomisation - Yes
  • Groups treated the same after randomisation - Yes
  • Intention to treat analysis - Yes
  • Follow-up - 90 day.  Achieved 99.3% completion of follow up
  • Significant primary endpoint - 90 all cause mortality was no different between the 2 groups
  • Precision of results - The confidence interval for point estimate was quite wide, suggesting the paper was underpowered.
  • Harms analysed - All relevant harms were considered.


The concept of restrictive transfusion practice is not new.  The first major trial in critically ill patients was the much discussed TRICC trial.  Since the release of this somewhat seminal paper, similar results have been produced in other patient groups including cardiac surgery, paediatric ICU, and orthogeriatric surgery (such as the FOCUS trial), though conflicting papers do exist.

The mechanisms by which blood transfusion may worsen outcomes in critically ill patients are becoming more defined.  The storage of blood for a period of days to weeks can induce the so-called "Storage Lesion", in which degradation of red cells can lead to pro-inflammatory byproducts.  When infused, these mediators can result in immune system and organ dysfunction, particularly in patients with critical illness.  

Furthermore, in some studies, the transfusions used contained non-leuko-reduced red cell concentrates.  The transfusion of donor white cells had been linked to viral transmission and to additional pro-inflammatory effects.  Most modern blood services now issue leukodepleted red cells as a standard, and this is thought to potentially impact on the outcomes of those studies.  However, it should be noted that the impact of leukodepleted blood on a broad basis has yet to be defined.

Red blood cells of course become deactivated during storage due to depletion of 2,3-DPG.  Consequently, the transfusion of red blood cells may not improve oxygen delivery as intended, and simply expose the patient to further risks.

The available evidence led the Surviving Sepsis Campaign to recommend transfusion in the resuscitation of septic patients to "maintain a haematocrit of 30% or greater in the first 6 hours", and thereafter maintain a Hb of >70, and less than 100 in patients with specific risk factors.  However, these recommendations appear to be based largely on the results of TRICC, which contained patients with sepsis, but was not limited to this diagnosis.  The precise role of restrictive transfusion in septic shock is therefore only inferred.

This concept is important because over 50% of patients in ICUs receive blood transfusions, and 90% are for non-bleeding reasons.  The potential to reduce costs, conserve a precious resource and reduce harm is therefore very high.

The Study

Anaemic adult patients with septic shock admitted to intensive care in Scandinavia were recruited to the study.  They were randomised to either a liberal transfusion trigger (i.e., they were transfused if their haemoglobin fell below 90g/dL) versus a restrictive strategy (transfused if <7g/dL).  Leukodepleted red cell units were transfused, one unit at a time, with a haemoglobin repeated at 3 hours to assess if the trigger had been met.  Patients with acute haemorrhage or an acute coronary syndrome were excluded.

Enrolment of patients was stratified by the study site and by the presence or absence of acute haematological malignancy.

998 patients were recruited from 32 intensive care units in Scandinavia.  Patients were a combination of surgical (elective and emergency) and medical, and both large and small hospitals were well represented.  40% had abdominal sepsis, 50% had respiratory sepsis, though only one third of patients had a confirmed microbiological diagnosis.  70% were mechanically ventilated and 10-15% required renal replacement therapy.

Clinicians were permitted to suspend the allocated strategy if the patient required ECMO, suffered a significant bleeding event or developed an ischaemic syndrome.  This resulted in transfusion outside the triggers in 6% of the restrictive arm patients, versus 2.2% of the liberal arm.

The study achieved good separation in treatment between the groups with respect to their Hb management.  The total number of transfusions used in the restrictive arm was less than half of that used in the liberal group.  The median number of transfusions was 1 unit in the restrictive group, while that of the liberal group was 4.  The average daily Hb of the patients was 1.5g/dL less in the restrictive group compared with the control arm.  36% of the restrictive arm did not receive any transfusion, compared with 1% of the liberal arm.

The primary endpoint of 90 day all cause mortality was not different between the groups - 43% in the intervention arm versus 45% in the control arm.  The wide confidence limits of this point estimate suggest however that the paper was underpowered to detect such a large absolute risk reduction.

Similar results were seen in the secondary endpoints, including significant reactions, ischaemic events or use of life support at 5, 14 and 28 days.

Additionally, there were no changes in the result when the analysis was made on a per-protocol basis, or in subgroups based on chronic cardiac disease, age >70 or severity of illness.

The results would appear quite generalisable for a number of reasons :
  • The population studied reflects a broad landscape of critical care patients - surgical, medical, large hospital, small hospital, predominantly respiratory and abdominal sepsis etc.
  • The vast majority of patients screened were ultimately included in the study
  • The design is relatively pragmatic in nature
For just on 500 patients in the intervention arm, nearly 1500 units of blood were saved, a rough saving of nearly $USD 150,000, or around $3000 per patient.  Applying this strategy more widely would have an obvious impact on health care expenditure.


The adoption of a restrictive strategy for transfusion in patients with septic shock does not appear to be associated with harm, and reduces exposure of patients to potential harms of transfusion, even with the use of leukodepleted blood.  Furthermore, this strategy potentially conserves a precious resource and can result in significant cost savings.

The study does not specifically address the role of blood as part of sepsis resuscitation in the first 6 hours.

Reviews, Resources and Editorials

  • Lars Holst, lead investigator of the TRISS trial

No comments:

Post a Comment