Journal Club: HALT-IT Trial
Updated: Jun 22
Tranexamic Acid (TXA) has been touted as a wonder-drug that cures many bleeding pathologies. Multiple studies have purported the benefits of TXA for significant bleeding from trauma, epistaxis, post-tonsillectomy bleeds, hemoptysis, postpartum hemorrhage, and even mild to moderate traumatic brain injury. Many small trials have been conducted to assess the effect of TXA on gastrointestinal bleeds and a Cochrane Database systematic review and meta-analysis combined multiple trials and determined that TXA appears to have a beneficial effect on mortality (Bennett 2014). A systematic review with meta-analysis published just this year concluded that moderate-quality evidence shows that TXA is superior to placebo for the reduction in mortality in patients with upper GI bleeding (Twum-Barimah 2020). The much-anticipated HALT-IT trial was finally published in the Lancet this week and the results just might surprise you…
Also, check out our Journal Club review of the Cal-PAT Trial, which studied prehospital TXA use for trauma.
Article: Roberts I et al. 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. Lancet 2020.
What did they do?
International, randomized, double blind, placebo-controlled trial in 164 hospitals in 15 countries (UK, Pakistan, Nigeria, Egypt, Malaysia, Georgia, Romania, Nepal, Sudan, Saudi Arabia, Spain, Ireland, Albania, Papua New Guinea, and Australia)
Above the minimum age considered an adult in whichever country the patient resided
Responsible clinician was substantially uncertain whether to use TXA
Patient had significant GI bleeding as determined by the treating clinician (significant bleeding was defined by having a risk of bleeding to death and included patients with hypotension, tachycardia, or signs of shock, or those likely to need transfusion or urgent endoscopy or surgery).
There were no official exclusion criteria mentioned
Intervention and Comparison
Eligible patients were randomly assigned to receive TXA or placebo as soon as possible and treatment was started immediately.
Loading dose: 1g TXA or placebo (0.9% sodium chloride) was added to a 100mL infusion bag of 0.9% sodium chloride and infused over 10 minutes
Maintenance dose: 3g TXA or placebo was added to 1L of any isotonic IV solution and infused at 125mg/hr for 24 hours (ie. 1g over 8 hours for 3 doses)
Death due to bleeding within 5 days of randomization.
Death due to bleeding within 24h and within 28 days of randomization
Rebleeding within 24 hours, 5 days, and 28 days
Surgery or radiological intervention
Blood product transfusion
Thromboembolic events (DVT, PE, stroke, and myocardial infarction)
Other complications (other significant cardiac event, sepsis, pneumonia, respiratory failure, renal failure, liver failure)
Days in ICU
Functional status at hospital discharge or at 28 days (by Katz Index of Independence in Activities of Daily Living)
Main analyses were performed on a modified intention-to-treat basis, after excluding patients who received neither dose of the allocated treatment and those for whom outcome data on death were unavailable.
The investigators performed subgroup analyses stratified by the following baseline characteristics:
time to treatment (≤3hrs, >3hrs)
site of bleeding (upper vs lower GI)
suspected variceal bleeding and comorbid liver disease compared with other or unknown causes of bleeding and by clinical Rockall Score
They also performed a post-hoc subgroup analyses to examine the effects of TXA stratified by World Bank country income level, anticoagulant use, and systolic blood pressure.
Patients were enrolled between July 4, 2013 and June 21, 2019. The mean time from onset of bleeding to randomization was about 22 hours.
12,009 patients were enrolled and randomly assigned to TXA or placebo groups. After exclusions, the number of patients that received allocated treatment and were included in the primary analysis were:
Both groups had similar baseline characteristics.
Primary outcome: Death due to bleeding within 5 days (RR 0.99, 95% CI 0.82-1.18)
TXA: 222 (3.7% of 5956)
Placebo: 226 (3.8% of 5981)
Similar results after adjusting for baseline covariates, performing per-protocol analysis, and removing patients who received open-label antifibrinolytics.
Prespecified subgroup analyses also showed no evidence of interaction.
Secondary outcomes: Most had no statistically significant difference, except for:
1) Venous events (DVT, PE) (RR 1.85, 95% CI 1.15-2.98)
TXA: 48 (0.8%)
Placebo: 26 (0.4%)
Fragility Index of 6
2) Seizures (RR 1.73, 95% CI 1.03-2.93)
TXA: 38 (0.6%)
Placebo: 22 (0.4%)
Fragility Index of 1
Randomization and blinding were performed by a Good Manufacturing Practice certified clinical trial service provider. The trial had a formal inspection process and the interventions came in a nicely packaged treatment pack with everything the clinician would need. The investigators performed appropriate education and retraining when needed.
Outcome data was collected even if the treatment was not given.
The authors had a clear primary outcome (though it was perhaps not a patient-oriented outcome; see limitations section below).
Funders had no role in study design, data collection, data analysis, data interpretation, or writing of the report.
The study had a very large sample size and only a small proportion were not included in the analysis. Out of the 12009 patients who were randomized into the trial, 11937 (99.4%) received the allocated treatment and were included in the primary analysis. This size is large enough to sufficiently detect adverse events, a feat that many other TXA trials did not achieve.
The baseline characteristics were remarkably similar in both groups.
The primary outcome changed part way through the trial; the sample size calculation was initially based on all-cause mortality as the primary outcome. As the trial proceeded, the investigators observed that over half of all deaths were due to non-bleeding causes. The sample size was increased from 8,000 to 12,000 to retain sufficient power for all-cause mortality.
The authors used a disease specific mortality instead of all-cause mortality. This can be problematic because the cause of death is whatever the clinician writes on the death certificate and is not always diagnosed by autopsy. Disease specific mortality is an arbitrary outcome that is not a patient-oriented outcome, and therefore offers dubious clinical benefit. ***In Justin Morgenstern’s review of the CRASH-3 Trial, he put it simply when he wrote: “More importantly, if overall mortality is unchanged, disease specific mortality is nothing more than a futile exercise in accounting. Patients care whether or not they die. They don’t care what we write on their death certificates.” Read point #2, then read the rest of the very well-written article about the WOMAN trial.
The intervention was different compared to the usual TXA protocols (1g over 10 minutes, then 1g over 8 hours), making it difficult to compare to other trials. See the discussion below about this.
The study analyzed data on a modified intention-to-treat basis, excluding patient s who received neither dose of the allocated treatment and those for whom outcome data on death were unavailable.
The study population does not appear to be a very sick cohort of patients. Only about 45% of patients showed signs of shock. 62% had Rockall scores of 4 or less. The average number of days in the ICU were 0.4 in both groups. Only about 9% of patients were taking anticoagulants. It is difficult to say whether the results would be different if the trial focused on a sicker population of patients.
Overall, this trial was very well-conducted and must have taken a tremendous effort by numerous collaborators. With 11,937 patients included in the final analysis, it is the largest trial to date and even exceeds the Cochrane Review that combined multiple smaller studies (Bennett 2014). It had a solid randomization process, with appropriate allocation concealment and blinding.
However, it is challenging to determine the exact onset of bleeding with GI bleeding. Unlike trauma or epistaxis, it is problematic to know exactly when the patient started bleeding. Moreover, not all patients come to the hospital after their first episode of melanic stool; this is demonstrated in the trial by 57% of patients being randomized in the trial greater than 8 hours after onset of bleeding. Only 16% of patients presented within the first 3 hours of onset of bleeding. Prior studies have supported early administration of TXA, which would clearly be difficult to accomplish in GI bleeds. However, the study population does seem to represent the typical presentation of GI bleeding for all comers, as opposed to CVA studies, which have very narrow selection criteria and excludes large percentages of presenting patients. As time of onset is challenging to determine, attempting to generate a narrow criteria of who needs TXA based on time of onset is likely a fool’s errand, and therefore the study is clinically applicable the way it is designed.
In most studies of IV TXA, an initial 1g is given over 10 to 15 minutes and an additional 1g is given over 8 hours (TICH-2, CRASH-2, CRASH-3, and others). However, in this study, the authors gave an initial 1g bolus over 10 minutes followed by 3g over 24 hours. In an interview of the lead author, Ian Roberts, on The Resus Room podcast, he discussed that the risk of rebleeding is highest in the first 24 hours. Since TXA has such a short half-life (approximately 2 hours), the investigators aimed to cover patients through this high-risk period. However, seizures have been previously linked to TXA administration in a dose-dependent manner (Lecker 2016, Zhang 2016). It is possible that the incidence of seizures in this trial was due to the higher dose of TXA given over the first 24 hours. Using a lower dose, while less likely to induce seizure, still remains futile and should probably not be given anyways. However, with a Fragility Index of 1 and a NNH of 370 in this study, seizure may not be a meaningful concern.
“In summary, we found no evidence that tranexamic acid decreases the risk of death in patients with gastrointestinal bleeding…Because gastrointestinal bleeding is a licensed indication for tranexamic acid, our results could have regulatory implications.”
Based on this large and well-conducted trial, TXA (even at higher doses and a longer duration than prior studies) does not seem to decrease the risk of death or rebleeding in GI bleed, and it seems to be associated with increased rates of DVT, PE, and seizures. I will not be giving TXA to my next GI bleed patient in the ED.
This study is remarkable and relatively unprecedented in size and results. The limitations of this study (well-discussed in detail by Dr. Murali above) mostly increase the risk of exaggerating benefit of the therapy. Despite this, it is a resoundingly negative trial. I applaud the authors for presenting it as a purely negative trial rather than attempting to statistically play with the subgroups to eke out a positive result. TXA to me has usually served as a “sure why not?” medication that I will toss in after I have established appropriate resuscitative care and known beneficial treatment options. This study does give me some pause to that practice given the, albeit slight, increased risk of harm demonstrated.
In relation to GI bleeding; more and more evidence is piling up that many of our common therapeutic options just aren’t all that helpful in improving outcomes. PPIs have been of dubious value for a long time in acute UGIB presentations; overall mortality and other patient centered outcomes are not impacted with its use but PPIs do seem to reduce the need for emergent endoscopic intervention. Some European medical societies don’t even start PPI until after endoscopy. Octreotide may have benefit but only in very specific presentations that are difficult to be certain of prior to endoscopy. Antibiotics (specifically ceftriaxone) has demonstrated clear benefit but only in the variceal bleeding subset due to prevention of translocation of gut bacteria to the portal venous system. Essentially, aggressive resuscitative efforts including early use of blood products for shock presentations and emergent to urgent endoscopic intervention remains our best treatments in a very challenging presenting complaint. It’s not terribly surprising that TXA does not confer significant benefit in this population as many of them, specifically patients with known liver disease, have baseline coagulation pathway derangement leading to both increased uncontrolled bleeding events refractory to coagulation therapies as well as thromboembolic events.
Other FOAMed Resources to Check Out:
2. TXA in GI Bleeds, HALT-IT; Roadside to Resus via The Resus Room (includes an interview with the primary author)
1. Bennett C et al. Tranexamic Acid for Upper Gastrointestinal Bleeding. Cochrane Database Systematic Reviews 2014. PMID: 25414987
2. Twum-Barimah E et al. Systematic Review With Meta-Analysis: The Efficacy of Tranexamic Acid in Upper Gastrointestinal Bleeding. Aliment Pharmacol Ther 2020. PMID: 32363690.
3. Roberts I et al. 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. Lancet 2020.
4. Lecker I et al. Tranexamic acid–associated seizures: Causes and treatment. Annals of Neurology 2016. PMID: 26580862.
5. Zhang L et al. Tranexamic Acid-Associated Seizures: A Meta-Analysis. Seizure 2016. PMID: 26967164.
Post peer-reviewed by:
Alex Dzurik, MD, FAAEM; Assistant Program Director, Mercy Health St. Vincent Medical Center Emergency Medicine Residency
Michael Plewa, MD, FACEP; Research Director, Mercy Health St. Vincent Medical Center Emergency Medicine Residency