Discussed in Oxford 24th October 2014 – report by Akshay Shah
Holst LB, et al. Lower versus Higher Hemoglobin Threshold for Transfusion in Septic Shock. New England Journal of Medicine 2014; 371 (15): 1381-91.
The Surviving Sepsis Campaign recommends red blood cell (RBC) transfusion should occur if the haemoglobin (Hb) concentration decreases to <7 g.dL-1 and to target an Hb of 7-9 g.dL-1 once tissue hypoperfusion has resolved. During this period of hypoperfusion, practice has been guided by the results of the trial by Rivers et al where the transfusion threshold was a haematocrit of 30% (which approximately equates to an Hb of 9-10 g.dL-1. The landmark TRICC trial concluded that a restrictive transfusion strategy was at least as effective as and possible superior to a liberal transfusion strategy expecting in patients with unstable angina and acute myocardial infarction – two areas in which the evidence base is still unclear to this day. Since then, studies in different settings (cardiac surgery, orthopaedic surgery, acute GI haemorrhage) and a Cochrane systematic review in 2012 consisting of 6294 patients have all shown that a restrictive transfusion strategy is associated with less RBC transfusions without any apparent harm.
These investigators set out to investigate the effects of a lower versus higher transfusion threshold on mortality in patients with septic shock in critical care. They conducted a multi-center randomized, stratified, parallel controlled trial involving 32 Scandinavian general ICUs over a two-year period. Randomization was performed using a centralized computer-generated assignment sequence in a 1:1 ratio with permuted blocks of varying sizes. The investigators assessing mortality, data and safety monitoring committee and trial statistician were all blinded. Due to the nature of the intervention, it was not possible to ‘blind’ the patient or attending clinician. They performed an intention-to-treat analysis. Their power calculation was based on the ‘severe infection’ subgroup of the TRICC trial (relative risk reduction of 20% in restrictive group) and they calculated that 1000 patients would be needed to show a 9% reduction in the absolute risk assuming a mortality of 45% in the liberal group (based on regional cohort data).
The PICO analysis is as follows:
Inclusion criteria were adult patients (>18 years old) with septic shock and Hb <9 g.dL-1. Exclusion criteria were reasonable and excluded patient groups in whom the transfusion evidence-base is not quite clear/limited – life threatening bleeding and acute coronary syndromes. A total of 1,224 patients were assessed for eligibility and eventually 503 patients were randomized to the ‘lower threshold’ group and 497 to the ‘higher threshold’ group. Baseline characteristics were similar in both groups. Subgroup analyses were carried out in the following groups – presence/absence of chronic cardiovascular disease, age less than/greater than 70, and a SAPS II score of above 53 versus below 53.
Single unit of cross-matched prestorage leukoreduced red cell transfusion if Hb concentration <7 g.dL-1.
RBC transfusion if Hb concentration <9 g.dL-1.
Primary outcome – 90-day mortality – data was available for 998/1005 patients and showed no significant difference between both groups (per-protocol populations). 216 patients (43%) in the lower threshold group and 223 patients (45%) in the higher threshold group had died (RR 0.94, 95% CI 0.78 – 1.09, p = 0.44). There was also no significant difference in secondary outcomes – use of life support (vasopressor or inotropic therapy, mechanical ventilation, or RRT) at days 5, 14 and 28 after randomization, incidence of adverse and ischaemic events, percentage of days alive without life support and percentage of days alive and out of the hospital. There was a significantly lower proportion of RBC transfusions in the restrictive group compared to the liberal group (1545 units vs. 3088 units, p<0.001). In addition, a total of 176 patients (36%) did not receive RBC transfusion at any point. There were no significant differences in the subgroup analyses either.
There are many strengths of this trial. The randomization process appears robust. The blinding is acceptable given the nature of the intervention. There is a high recruitment rate – 1005 out of 1224-screened patients ended up being randomized. The statistics and power calculation are reasonable – using the data from the TRICC trial, albeit old, is probably justified given the lack of evidence in this area. Primary outcome data was available for 99.2% of patients and the investigators’ went to good lengths to try obtaining primary outcome data even when patients were discontinued from the study for whatever reason. Another potential strength is the short time between trial completion (December 2013) and publication (2014). The authors also comment that the trial was pragmatic as it did not affect any other aspects of care.
This trial does not specifically answer the question of what the transfusion threshold should be during the EGDT window i.e. during the first 6 hours whilst in A&E and little information is available about what happened during that time period. The ARISE and PROCESS trials hopefully shed some light on this. There was a significant delay from time of randomization to first day of treatment – median times were 15 and 14 hours in the lower threshold and higher threshold groups respectively. Protocol violations occurred in both groups – 10% of patients in the ‘lower threshold’ group received transfusion despite being above the transfusion threshold. However, there was still a significant difference in the daily Hbs in both groups and the authors carried out a per protocol analysis which showed no overall difference between both groups. It is worth questioning why their mortality rate of patients with septic shock is 40 – 45% – this is in contrast to mortality data from the ARISE and PROCESS trials which quote a mortality of approximately 20%.
Will this change my practice?
Probably not. A transfusion threshold of 7 g.dL-1 when compared with 9 g.dL-1 results in no difference in mortality and morbidity in critical care patients suffering from septic shock. These findings are in keeping with other trials in other medical/surgical settings. This work also supports the clinical practice of majority of the clinicians who attended the journal club meeting. The results of trials in high-risk groups e.g. ACS are awaited.