JC: Halt! It’s not time for TXA! Or is it? HALT-IT results at St Emlyn’s

Estimated reading time: 11 minutes

Cite this article as: Chris Gray, "JC: Halt! It’s not time for TXA! Or is it? HALT-IT results at St Emlyn’s," in St.Emlyn's, June 23, 2020, https://www.stemlynsblog.org/halt-it-st-emlyns/.

If you’re any kind of emergency care aficionado, you probably worship at the altar of tranexamic acid (TXA). With data showing more favourable outcomes from its use in traumatic bleeding​(1)​, post-partum haemorrhage​(2)​, maybe traumatic brain injury​(3)​, and potentially haemorrhagic stroke​(4)​, it’s easy to see why. It’s a cheap, safe drug, and quick to give.

So what about gastrointestinal haemorrhage then? Does it work there?

We’ve only been waiting nearly seven years but the results are finally here! The Haemorrhage Alleviation with Tranexamic Acid – Intestinal System​(5)​ (HALT-IT) trial started recruitment in July 2013 and initially aimed to recruit 8,000 patients by 2017. This was achieved but then extended to 12,000 patients and recruitment finally closed a year ago in June 2019.

Here at St Emlyn’s we’ve been excited to finally get our hands on the data and have a look for ourselves. Will this be yet another paper that gives us the go ahead to anoint more patients with TXA?

Spoilers: No.

Here’s the abstract but the paper is open-access in the Lancet​(5)​ so there is no excuse not to read and appraise it for yourself, and to then reflect on your own practice in light of this new evidence.

Talk trial types to me

If you’re playing critical appraisal bingo, you’re in for a treat, as this is an international, multi-centre, randomised, double-blind, placebo-controlled trial. All the boxes, right? However, there’s also another word in there that might not be on your card. Pragmatic.

Randomised controlled trials sit on a spectrum between explanatory trials, which aim to test the effectiveness of an intervention in ideal conditions, and pragmatic trials, which look at usual-care, or real-life conditions. HALT-IT describes itself as a pragmatic trial, which means that we should be able to generalise the results well to what we do day-to-day. But how do we know if it really is pragmatic?

Let’s look at their methods first.

Clinicians enrolled adult patients with a clinical diagnosis of significant upper or lower gastrointestinal bleeding into the study, with consent being taken prior to randomisation if appropriate, or afterwards if not. Block randomisation was used and blinded treatment packs contained either tranexamic acid, or placebo, with identical packaging and labelling other than the randomisation number. Patients received TXA or placebo “as soon as possible”, as a loading dose of 1g TXA or placebo added to 100ml bag of saline and infused over 10 minutes, followed by 3g TXA or placebo added to a litre bag of any isotonic solution infused over 24 hours.

Simple right, and seems pragmatic, but again, how can we tell?

The original PRECIS (PRagmatic Explanatory Continuum Indicator Summary) tool was developed in 2005-8 to help trialists ensure their design matches their intention for the study. This was developed further in 2015 into PRECIS-2​(6)​ and to cut a long story short there are nine domains in which you assess your trial design each on a range from explanatory to pragmatic to give you an overall picture. As far as I can tell, the HALT-IT trial was not assessed against the original PRECIS tool and PRECIS-2 came out after they had started so it hasn’t been put through these tools. You can read about the domains in greater detail in their paper, but in most they would seem to rate at the pragmatic end. For example, regarding the setting domain, this trial takes place in the usual clinical areas where you’d find such patients, and in the organisation domain, this trial doesn’t require any special equipment or drugs that wouldn’t ordinarily be there. From a flexibility point of view as well, if there were concerns that the drug/placebo could be making the patient worse, then they could be stopped, just like you would with any drug you give. There was no requirement to complete the dose if there was clinical concern.

However there are some parts that don’t make this trial completely pragmatic. The control arm received an infusion of placebo rather than usual care, but of course this was necessary from a blinding point of view. Also, the population recruited were adults, with significant gastrointestinal bleeding (at the clinician’s discretion), where the clinician was uncertain whether or not to use TXA. Seems pretty pragmatic, though that last criteria means we have left out patients for whom the clinician thought TXA would (or indeed would not) be beneficial.

What was the primary outcome?

As I mentioned earlier, the trial was extended by two years, and before now I hadn’t been able to find any details on why this was. Initially the sample size calculation was based on all-cause mortality within 28 days, however whilst underway, they noticed that over half of all deaths were from non-bleeding causes. Because TXA doesn’t seem to have an effect on reducing deaths not relating to bleeding, they changed the primary outcome to death due to bleeding within five days. Plugging this into their power calculations meant recruiting 4000 more patients for them, and two more years of waiting for the rest of us!

Now, I care more about the overall dying bit than the dying of bleeding bit, so I think the change in primary outcome is interesting. Whether they died of bleeding or something else is also very subjective (especially compared to the binary option of dead or not dead). However, it did increase the numbers rather than decrease, so you could argue it was still powered to detect a difference in all-cause mortality, so not the end of the world.

They also had a lot of secondary outcomes which you can read in the paper, but these included 28 day all-cause mortality having been bumped from the primary outcome spot (so they did still look into it), together with thromboembolic events.

Who did they study?

As already stated, patients recruited had to meet three criteria:
– Adults (either over 16 or over 18 depending on the country)
– Significant gastrointestinal bleeding (clinically diagnosed)
– The clinician was substantially uncertain whether to use tranexamic acid

The top two here are exceptionally pragmatic, though I just want to touch on this last point again. If we don’t know whether or not tranexamic acid works in GI bleeding, surely we’d want to study everyone. In fact the only exclusion criteria listed in the study protocol are specifically related to this:
– Patients for whom the clinician feels there is a clear indication for TXA
– Patients for whom the clinician feels there is a clear contraindication for TXA

This muddies the trial results a little, as now how do we know whether our gestalt is correct? Did TXA actually benefit those patients who the clinician thought it would benefit? We’ll never know.


Ok, you’ve come this far, so let’s go through the results.

12,009 patients were enrolled. 64% were male and the mean age was 58 years old. 72% had co-morbidities and 9% were on anticoagulants. 89% were suspected to have upper GI bleeding. 43% had signs of shock.

For the primary outcome of death due to bleeding within 5 days, there was no significant difference between the groups – TXA 222 (3.7%) vs placebo 226 (3.8%), RR 0.99 (95% CI 0.82-1.18)

The original primary outcome of all-cause mortality at 28 days was also not significantly different with TXA 9.5% vs placebo 9.2%, RR 1.03 (0.92-1.16).

In fact, there were no statistically significant differences between the groups relating to mortality despite the group looking at the data from an intention-to-treat perspective, as well as per-protocol and after adjusting for baseline covariates. They tried, but it’s still a no.

There were a couple of secondary outcomes where tranexamic acid did come out worse, these were:
– Venous thromboembolic events as a whole – TXA 0.8% vs placebo 0.4%, RR 1.85 (1.15-2.98) – though looking individually at PE or DVT these were not statistically significant and in a smaller subgroup analysis it appeared higher in patients with suspected variceal bleeding or liver disease
– Seizures – TXA 0.6% vs placebo 0.4%, RR 1.73 (1.03-2.93)

One of the other results to talk about is the time from onset to randomisation, for which the mean was 21.4 hours in the TXA group and 22.5 hours in the placebo group. Only 16% were randomised within 3 hours, and 58% were randomised more than 8 hours after onset of symptoms.


Overall this is a really well conducted study, with a good question to be answered about a drug that we seem to be using more and more, and as such gives us a much needed answer. With only 14 protocol violations and three consent violations and excellent follow up, the numbers are good and the researchers should be congratulated for this.

They conclude that tranexamic acid does not reduce death from gastrointestinal bleeding, and advise against its use in these patients outside of the context of further randomised trials.

So sadly it looks like that’s the end of TXA in GI bleeding. Or is it?

Let’s not forget that group that were excluded from enrolment where the clinician thought TXA was clearly indicated. This is a likely explanation for the low number of patients who were clinically shocked at the time of randomisation, and suggests that there were some sicker patients out there who may have been given TXA anyway, and were not considered for the trial.

The time to randomisation was also very high, and with some evidence that earlier tranexamic acid is better, maybe we didn’t see a difference here as the patients just presented too late. Don’t forget though, that this is all part of the pragmatic trial design – there’s no point seeing if it works for GI bleeds if given within 3 hours, if our patients don’t make it to us by then.

The bottom line

Tranexamic acid should not be used routinely in the acute management of patients with gastrointestinal bleeding. However, there may still be a role for it as part of your major haemorrhage protocol if it is activated for such patients.

We’d be very interested to hear your thoughts, were you using TXA in GI bleeds before? What will you do now?



Further learning

For more on the HALT-IT trial, go over to first10EM and have a look at Justin Morgenstern‘s review, or if you prefer learning by listening, try The Resus Room podcast, where they interview lead author Ian Roberts!

We’re sure there’s more coming and we’ll update this as more FOAMed appears!

For more resources around gastrointestinal haemorrhage, check out our blog post and podcast. Or, to run your own teaching session, Iain has put together a lesson plan for trainees, as well as an induction lesson plan which is tailored more to those new to working in the ED. And don’t forget to reflect on your learning and link it in to your portfolio!


  1. 1.
    Carley S. TXA for everyone [Internet]. St Emlyn’s. 2012 [cited 2020]. Available from: https://www.stemlynsblog.org/tranexamic-acid-for-everyone-st-emlyns/
  2. 2.
    Shakur H, Roberts I, Fawole B, Chaudhri R, El-Sheikh M, Akintan A, et al. Effect of early tranexamic acid administration on mortality, hysterectomy, and other morbidities in women with post-partum haemorrhage (WOMAN): an international, randomised, double-blind, placebo-controlled trial. The Lancet [Internet]. 2017 May;2105–16. Available from: http://dx.doi.org/10.1016/S0140-6736(17)30638-4
  3. 3.
    Effects of tranexamic acid on death, disability, vascular occlusive events and other morbidities in patients with acute traumatic brain injury (CRASH-3): a randomised, placebo-controlled trial. The Lancet [Internet]. 2019 Nov;1713–23. Available from: http://dx.doi.org/10.1016/S0140-6736(19)32233-0
  4. 4.
    Sprigg N, Flaherty K, Appleton JP, Salman RA-S, Bereczki D, Beridze M, et al. Tranexamic acid for hyperacute primary IntraCerebral Haemorrhage (TICH-2): an international randomised, placebo-controlled, phase 3 superiority trial. The Lancet [Internet]. 2018 May;2107–15. Available from: http://dx.doi.org/10.1016/S0140-6736(18)31033-X
  5. 5.
    Roberts I, Perel P, Prieto-Merino D, Shakur H, Coats T, Hunt BJ, et al. Effect of tranexamic acid on mortality in patients with traumatic bleeding: prespecified analysis of data from randomised controlled trial. BMJ [Internet]. 2012 Sep 11;e5839–e5839. Available from: http://dx.doi.org/10.1136/bmj.e5839
  6. 6.
    Loudon K, Treweek S, Sullivan F, Donnan P, Thorpe KE, Zwarenstein M. The PRECIS-2 tool: designing trials that are fit for purpose. BMJ [Internet]. 2015 May 8;h2147–h2147. Available from: http://dx.doi.org/10.1136/bmj.h2147

Cite this article as: Chris Gray, "JC: Halt! It’s not time for TXA! Or is it? HALT-IT results at St Emlyn’s," in St.Emlyn's, June 23, 2020, https://www.stemlynsblog.org/halt-it-st-emlyns/.

3 thoughts on “JC: Halt! It’s not time for TXA! Or is it? HALT-IT results at St Emlyn’s”

  1. Any indication why they decided to include only patients where the clinician was “substantially uncertain” whether to use TXA? It just seems like a counter-productive sample limitation that only serves to unnecessarily cast doubt on their conclusions.

    I’ll give it a closer read later when I have time, but I didn’t see anything on quickly skimming of the paper.

    1. That’s a characteristic of a clinical trial and also a pragmatic approach that would reflect future practice. If for one reason you knew that TXA was not to be used e.g. allergy then you would not put them into the trial. If you knew that it had to be given (known clotting disorder treated by TXA) then you would give it.

      It’s also a requirement from ethics when some people have strong a priori beliefs and feel that their opinion should override the science. They are wrong, but sometimes that caveat has to stay in there.


      1. It’s a standard of all trials that there is equipoise as to whether or not the intervention does or does not have an effect.

        We assume that if you have equipoise and participate as a trial site, you include all patients who meet the criteria. We further assume that HALT-IT is therefore contaminated by the fact that the clinician could choose to include them or not based on their individual equipoise, somehow making this trial less robust

        However, there is the potential for this in all trials. Regardless of the criteria used, some patients would have been deemed “too sick”, excluded from the trial and given TXA anyway. We’ll never know.

Thanks so much for following. Viva la #FOAMed

Scroll to Top