Effects of a high-dose 24-h infusion of tranexamic acid on death and thromboembolic events in patients with acute gastrointestinal bleeding (HALT-IT): an international randomised, double-blind, placebo-controlled trial

HALT-IT Trial Collaborators. Lancet 2020; 395; 1927-36. doi:10.1016/S0140-6736(20)30848-5

Clinical Question

  • In patients with gastrointestinal (GI) bleeding does high dose tranexamic acid (TXA) when compared to a placebo reduce mortality from GI bleeding?


  • The mortality rate in GI bleeding has been reported to be as high as 10%, and the risk of death is 4 times as high if re-bleeding occurs
  • The use of TXA (an anti-fibrinolytic) has been widely studied in multiple patient populations, with benefit shown if given early in traumatic and post-partum haemorrhage
  • A 2012 Cochrane Review for upper GI bleeding found a reduction in mortality for patients presenting with upper GI bleeding, however given the internal and external validity of the trials concluded that TXA use could not be recommended without additional trials


  • Randomised, placebo-controlled, double blind trial
  • Randomisation numbers generated by independent statistician and certified clinical trial service provider made uniquely numbered identical treatment packs
  • Randomisation was via blocks but not stratified
  • The lowest number treatment pack available was used once enrolled
  • Outcome data collected at first opportunity of death, discharge from randomising hospital or at 28 days
  • Sample size calculated initially based on all cause mortality as primary outcome, this was changed late in the trial (November 2018) to death due to bleeding within 5 days of randomisation
    • Based on amended primary outcome with an assumption that risk of death 4%, 12000 patients had 85% power (two sided α of 5%) to detect a 25% relative risk reduction in death due to bleeding
  • Subgroup analyses planned in 4 groups:
    • Time to treatment (<3 h or > 3 h); site of bleeding (upper vs lower); suspected variceal and co-morbid liver disease and clinical Rockall score
  • Cause of death ascertained by local investigators and reviewed by chief investigator
    • Diagnosis of thromboembolic events made on strict criteria consisting of clinical assessment, biomarkers and imaging
  • Funded by UK National Institute for Health Research Health Technology Assessment Programme and registered with ClinicalTrials.Gov and Current Controlled Trials


  • 164 hospitals in 15 countries
  • Data collected from July 2013 – June 2019


  • Inclusion:
    • Those enrolled needed to be an adult (age varied between 16 – 18 depending on country)
    • There must have been clinical equipoise as to whether TXA would be beneficial
    • Bleeding must be “significant” – this was determined clinically
      • Broad definition included patients  risk of bleeding to death, signs of shock, likely requirement for transfusion or urgent endoscopic / surgical intervention.
  • Exclusion
    • Nil criteria given
  • 12009 patients randomised
    • TXA: 5956
    • Placebo: 5981
  • Baseline characteristics similar
    • These included age, sex, time from onset to randomisation, suspected location of bleeding, suspected variceal bleeding, signs of haemodynamic compromise, Rockall score, co-morbidities and anticoagulant usage.


  • TXA
    • 1g in 100mls of 0.9% saline given over 1 hour
    • This was followed by 3g added to 1L of an isotonic solution and infused at 125mg/hour over 24 hours


  • Placebo
    • Placebo solution replaced TXA, otherwise infusion regime identical

Management common to both groups

  • No guidance on further management given
  • 223 given anti-fibrinolytic drugs outside protocol (105 in TXA group and 118 placebo)


  • Primary outcome:
    • Death due to bleeding within 5 days of randomisation – no significant difference
      • TXA 222 (3.7%) vs. Placebo 226 (3.8%)
      • Similar results when baseline covariates adjusted for and those who received anti-fibrinolytics outside trial protocol.
      •  No significant differences in the pre-specified subgroups
  • Secondary outcomes:
    • Multiple other outcomes listed
    • Only significant difference between TXA and placebo is rate of venous thromboembolic events (PE, DVT)
    • No significance between TXA and placebo groups in any of the following:
      • Death within 24h or 28 days
      • All cause mortality at 28 days
      • Re-bleeding rates at 24h, 5 and 28 days
      • Any thromboembolic event
      • Diagnostic and therapeutic endoscopic, surgical and radiological interventions
      • Transfusion of packed red blood cells, FFP or platelets

Authors’ Conclusions

  • TXA does not reduce death from GI bleeding and should not be used as part of a uniform approach to treat GI bleeding


  • Pragmatic design to allow access to trial drug as soon as feasibly possible
  • Good Internal and External Validity
    • Very large study across multiple international sites
    • Well balanced baseline characteristics
    • Randomisation and protocols for blinding maintained
    • Minimal loss to follow up
    • Small number of protocol violations
  • Sensible pre-specified subgroup analyses – in particular looking at the early treatment (< 3 hours) given previous work showing importance of early TXA use
    • Although 3 hours may seem like a long time for TXA use in haemorrhage, this is from symptom onset so shorter time frames may be hard to achieve given logistical constraints of getting into hospital and time taken to make decision to seek medical attention
    • Modelling time since bleeding onset as a continuous variable is clinically sensible (no interaction was shown) given that time is not a dichotomous variable


  • Amendment to primary outcome seems like a methodological weakness
    • This was done 6 months prior to the end of recruitment, although the rationale seems sensible and the power calculations for the new primary outcome showed an appropriate number of patients recruited
  • How heterogenous is the trial population?
    • This trial included all presentations of upper and lower GI bleeding. Would TXA have the same effect if it is given to a lower GI bleed from a diverticulum compared an upper GI bleed from severe oesophagitis?
    • Data surrounding pathological cause of bleed following intervention would be useful
  • The requirement that the treating clinician be uncertain as to the potential benefit of TXA could mean that some patients with large GI bleeds may have been given TXA and not randomised on the basis those involved may have thought it would help (especially given previous trials on TXA use albeit not in this patient population and previous low risk of harm shown)
    • This may be echoed by the fact that only 69% required transfusion of any sort and  87% had a SBP > 90 and only 30% had a HR > 107.
    • The converse argument is that trials should be used to answer questions for which clinical equipoise exist
  • Only small numbers of VTE were found in a very large cohort of patients
    • No mention is made as to whether only those that were symptomatic were screened (this would still be important as a patient-focused outcome) but may under-estimate occurrence
    • Study not powered to detect this
    • ATACAS TXA was a large trial (~5000 patients) which used higher doses of TXA (50-100mg/kg) and there was no increase in PE (15 in each arm)
    • No data provided surrounding time-frames of VTEs.
      • Can you solely put a DVT on day 27 down to TXA received on day 1? It would be plausible that a late VTE following a prolonged hospital stay (potentially without anticoagulation for a period of time) would have multifactorial causes

The Bottom Line

  • Despite the weaknesses above this is a very well designed, large trial. A “negative” trial of this magnitude is clinically important and following this I won’t be using TXA in my management of GI bleeding.
  • Future trials (PATCH) will provide further information on TXA and it’s side effect profile

External Links


Summary author: George Walker
Summary date: 20th June 2020
Peer-review editor: Adrian Wong

One comment

  • John Bowery, D.O.

    As an ER physician at a small critical access hospital in a rural area with limited resources (lack of emergent endoscopy, limited blood bank, no inpatient hemodialysis, lack of interventional radiology, and significant transport time to a higher level of care, etc.) I am always looking for ways of at least temporarily stabilizing my patients with significant GI hemorrhage while I arranging for transport for definitive care.

    I can’t see that the results of the HALT-IT trial rule out potential benefit in patients for whom other treatments will be unavailable or significantly delayed.

    The Covid-19 pandemic which has caused even greater delays in transport to a higher level of care, has greatly exacerbated this issue.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.