Upper GI bleed evidence

Welcome back to the tasty morsels of critical care podcast.

Today we’re going to look at the red stuff – blood, and when to give it. This will cover some of Oh’s Manual chapter 97 covering blood transfusion. But we’ll have a focus on transfusion targets. There’s a nice narrative of evidence here over the past 20 years that has given us a relatively robust evidence base for practice in this area, something quite novel in critical care.

Blood is expensive and unlike fossil fuels currently remains a renewable resource in the healthy population, it is obviously quite limited and nations frequently experience shortage of various blood groups and products that can have significant impacts on health care delivery. The red cells we give undergo a number of changes in the donation process with “storage lesions” becoming more prevalent over the duration of storage. A list of potential problems with stored red cells might run as follows:

• red cells change in shape biconvave to spherocytes (echinocytes) losing flexibility

• change in red cell membrane leading to sticking to the endothelium (esp in activated states like sepsis)

• 2,3 DPG depletion (which means Hb holds onto Oxy)

• reduced NO

• progressive increase in K+

• acidosis

The ABO reactions of transfusion should be dealt with by good governance of your transfusion service but fevers and other reactions are still an issue. The wonderfully named TRALI and TACO are also well described and space precludes a detailed discussion of these in this post.

Now that we know giving red cells is not an entirely benign intervention we are left with the question that all competitive limbo dancers are faced with on a daily basis – how low can you go. What would be an appropriate Hb target for a critically ill patient.

So let me tell you a little story… back in the late 90s when i was binging on OK Computer some Canadians led by Paul Hebert produced a large observational cohort of ICU patients called the TRICC trial suggesting that those with lower Hb did poorly and those who got more transfusions did better. But they were good empiricists and acknowledged that this could all be confounded by unmeasured factors. The only way to deal with that is randomisation and so 2 years later, Paul Hebert was at it again producing the TRICC 2 trial. This time an 800 pt multicentre randomised trial looking at Hb of 7 v 10. The headline result here was that the restrictive group did at least as well and probably better than the liberal transfusion group. This was a major trial and I’m pretty sure triggered a major change in practice. The caveats to this were as expected – those with ischaemic heart disease should probably have a higher target.

Things went quiet for a few years but in 2010 we saw the TRACS trial from Brazil looking at one of the sacred cows of transfusion targets – cardiac surgery. Can we lower the Hb target in those with dodgy coronaries? They looked at Hb 9 vs 10.5 and found no difference.

Villaneueva in 2013 took on upper GI bleeds. They smartly excluded the unstable active bleeders but in 500 patients randomised to 7 v 9, the lower target won out.

The trials started to come thick and fast now with TRISS trial in 2014 taking on sepsis. The problem in sepsis is oxygen delivery so surely more Hb is good. But yet again, in 1000 pts with sepsis there was no benefit in targeting 9 vs 7

2015 brought the TRIGGER trial (hopefully you’re starting to see the unofficial naming convention here…) looking again at UGIB and again finding no benefit to the higher target

2017 brought the TRICS 3 trial, looking at 5000 patients undergoing cardiac surgery. Again randomised, this time 7.5 v 9.5, again no advantage to the higher target

in 2021 they took on ACS patients in the REALITY trial, the most obviously ischaemic group and randomised 8 v 11 and no benefit to the higher target

Most recently in 2025 the TOP RCT looked at vasculopaths having vascular surgery and in 3000 pts there was no benefit to the higher target.

Phew… that’s a lot of trials but I think you’re starting to get the point that in general the answer to the question “what is your Hb target” is going to be 7-8

There are of course caveats to throw in at this stage.

Firstly, it’s important to note that none of these trials looked at the exsanguinating patient where you should be targeting physiology like HR and BP and perfusion rather than Hb. Restrictive Hb targets are in general questions for the daily ward round rather than the massive transfusion protocol.

Finally, in the past couple of years we’ve seen 2 RCTs looking at critically ill patients with sick brains. One looking at TBI and the other looking at SAH. Both suggest that if you have a sick brain you probably should be targeting a higher Hb of 9 or so. When you look at their outcomes the differences do not reach statistical difference in either trial but the trends are clearly to my eye towards more blood leading to better outcomes.

Reading:

LITFL has a lovely written summary of all the major trials

I have included the two neuro trials here as they’re not noted in the LITFL summary

• Turgeon, A. F. et al. Liberal or Restrictive Transfusion Strategy in Patients with Traumatic Brain Injury. N. Engl. J. Med. (2024) doi:10.1056/nejmoa2404360.
• English, S. W. et al. Liberal or Restrictive Transfusion Strategy in Aneurysmal Subarachnoid Hemorrhage. N. Engl. J. Med. 392, 1079–1088 (2025).