Best Evidence Topics
  • Send this BET as an Email
  • Make a Comment on this BET

The effect of administering tranexamic acid on mortality of adults with acute gastrointestinal bleeding

Three Part Question

In [adults with acute gastrointestinal bleed], does [administration of tranexamic acid] [improve mortality]?

Clinical Scenario

A 50 year old man is brought into the emergency department with signs of an acute gastointestinal bleed. An urgent endoscopy is arranged. As you resuscitate him and prepare for this you remember that tranexamic acid has proven useful in major haemorrhage and wonder if it will aid haemostasis in this man and increase his chance of survival.

Search Strategy

MEDLINE search. 1970-October 2020
EMBASE search. 1970- October 2020
Cochrane library search. 1970- October 2020
((tranexamic acid):ti,ab OR (TXA):ti,ab OR (antifibrinolytic):ti,ab) AND ((gastrointestinal bleed):ti,ab OR (gi bleed):ti,ab OR (gastrointestinal bleeding):ti,ab OR (gi bleeding):ti,ab OR (gastrointestinal haemorrhage):ti,ab OR (gi haemorrhage):ti,ab) AND ((acute):ti,ab OR (severe):ti,ab OR (emergency):ti,ab)

Search Outcome

93 papers were found when repeats were excluded. Of these, 8 were found to be relevant to the study question. A further 3 relevant papers were identified by searching the references of the 8 selected papers. This left 10 randomised controlled trials and one large retrospective cohort study.

Relevant Paper(s)

Author, date and country Patient group Study type (level of evidence) Outcomes Key results Study Weaknesses
Roberts et al.
11952 adults admitted to hospital with clinically diagnosed significant (potentially life threatening) GI bleeding in cases where clinicians were unsure whether to use TXA. Randomised to receive either 1g IV TXA with 100mL 0.9% sodium chloride over 10 mins, followed by 3g IV TXA in 1L saline at 125mg/hour over 24 hours or a placebo with an identical routine.RCT (1b)Death due to bleeding within 28 daysTXA: 253 (4.2%) Placebo: 262 (4.4%) RR: 0.97 95% CI: 0.82-1.15 No significant difference No significant differences when measured at 24 hours or 5 days.The primary outcome changed 5 years into recruitment suggesting methodological flaw, but the change was justified, and the sample power calculation appropriately altered. Recording of death due to bleeding is susceptible to errors in clinical judgement. Strict VTE diagnostic criteria may have led to underreporting.
All-cause mortality within 28 daysTXA: 564 (9.5%) Placebo: 548 (9.2%) RR: 1.03 95% CI: 0.92-1.16 No significant difference
Rebleeding within 28 daysTXA: 410 (6.8%) Placebo: 448 (7.5%) RR: 0.92 95% CI: 0.81-1.05 No significant difference No significant differences when measured at 24 hours and 5 days.
Thrombo-embolic eventsTXA: 86 (1.4%) Placebo: 72 (1.2%) RR: 1.20 95% CI: 0.88-1.64 No significant difference. Risk of VTE was (0.8%) in TXA and (0.4%) in placebo (RR: 1.85, CI: 1.15- 2.98), statistically significant.
Death due to bleeding in 5 days in patients with upper and lower GI bleedsUpper- TXA: 212 (4.0%) Placebo: 220 (4.1%) RR: 0.97 CI: 0.81-1.17 Lower- TXA: 10 (1.5%) Placebo: 6 (0.9%) RR: 1.61 CI: 0.59-4.40 No significant differences
Miyamoto et al.
Database records of 78291 patients admitted to hospital in 2010-2018 with diverticular bleeds. Exclusion criteria: age <18 years, injuries due to trauma, did not have CT or colonoscopy on day of admission. Separated into two groups; those who received IV TXA on day of admission (mean dose of 961mg/day) as the treatment group and those who did not as the control. Propensity score matching was used to put patients into 30087 pairs. Cohort study (2b)In-hospital mortalityTXA: 0.7% Control: 0.7% OR: 1.07 95% CI: 0.88–1.30 p=0.52 No significant differenceNo sample size estimates, or power calculations but huge sample size so assume high power. Control group recruitment did not specify that patients had not received PO or NG TXA. Did not specify whether only acute bleeds were included. Data was from database shown to have 50-80% diagnostic sensitivity, but those who did not have diagnostic tests (CT/colonoscopy) were excluded to improve validity. No standard criteria for TXA dosage and timing. Average dose of 961mg is lower than the recommended treatment dose. Not an RCT but used propensity score matching to control other variables.
Severe bleeding TXA: 16.6% Control: 17.5% OR: 0.93 95% CI: 0.89–0.99 p=0.003 Significant difference
Blood transfusion within 7 daysTXA: 31.4% Control: 34.3% OR, 0.88; 95% CI, 0.84–0.92; P < 0.001 OR: 0.88 95% CI: 0.84–0.92 p<0.001 Significant difference
Mean costs of hospital stay (Japanese yen converted to US dollars)TXA: $4450 Control: $4683 Cost was $233 less in TXA group 95% CI: -$153 to -$314 p<0.001 Significant difference
Karadas et al.
157 adult patients presenting to ED with symptoms of upper GI bleeds were studied. Exclusion criteria: allergy to TXA, history of varices or thrombo-embolic disease, GI bleeding due to trauma, on renal replacement therapy or found to have variceal bleeding. Randomised to receive 1 dose of 2000mg, 5% TXA in 100mL saline or saline placebo, both via NG tube. RCT (1b)Mortality at 1 monthTXA: 8 (10.3%) Placebo: 10 (12.7%) p=0.637 No significant difference1-month data collected by phoning patients, not a reliable method. Sample size was powered to detect changes to rebleeding rate, not mortality. Inclusion criteria did not necessitate acute or severe GI bleeding. Exclusion of variceal bleeding after randomisation may have caused attrition bias.
Rebleeding in 1 monthTXA: 9 (11.5%) Placebo: 8 (10.1%) p=0.766 No significant difference
Need for endoscopic or surgical intervention within 1 monthEndoscopic- TXA: 11 (14.1%) Placebo: 7 (8.9%) p=0.303 Surgical- TXA: 3 (3.8%) Placebo: 3 (3.8%) p=1.000 No significant difference in either
Tavakoli et al.
410 adult patients admitted to hospital with significant acute upper or lower GI bleeding were studied. Exclusion criteria: age<18 years, pregnancy, lactation, OCP use, recent thrombo-embolic disease, malignancy, end-stage renal disease, anti-coagulation treatment, temperature >38oC, vision impairment, seizure, possibility of variceal bleeding, allergy to trial medications. Randomised to receive 1g IV TXA q6h or 1g stat NG TXA then systematic TXA or placebo through either route. RCT (1b)Mortality (recorded at 72 hours and 1 month)72 hours- IV TXA: 1 (0.7%) NG TXA: 2 (1.5%) Placebo: 6 (4.3%) p=0.886 1 month- IV TXA: 9 (7.8%) NG TXA: 11 (9.6%) Placebo: 18 (15.1%) p=0.172 No significant differenceSample size calculation based on power to detect endoscopic findings, not powered for study outcomes. IV TXA group had 133 patients and sample size estimates necessitated 135 per group. Not blinded to difference between IV and NG treatment groups. Double-dummy technique not used to compensate. Did not specify dosage of systematic TXA in NG group, or duration of treatment in either. 61 patients lost to 1-month follow up.
Rebleeding within 72 hoursIV TXA: 9 (6.5%) NG TXA: 11 (8.3%) Placebo: 13 (9.4%) p= 0.682 No significant difference
Surgery within 72 hoursIV TXA: 4 (2.9%) NG TXA: 3 (2.3%) Placebo: 2 (1.4%) p= 0.708 No significant difference
Urgent endoscopy rateIV TXA: 20 (14.49%) NG TXA: 14 (10.52%) Placebo: 42 (30.21%) p<0.001 Difference significant
Saidi et al.
131 patients admitted with haematemesis, melaena or both were studied. Exclusion criteria: no gastric or duodenal lesion on endoscopy, pregnant or lactating, GI malignancy, history of thrombo-embolism, end stage renal disease, TXA allergy, on anticoagulants, coagulopathy. Randomised to receive 1g NG TXA in 250ml saline within 30 mins of admission or saline placebo by the same route. RCT (1b)Mortality within 30 daysTXA: 4 (5.97%) Placebo: 9 (14.06%) p=0.150 No significant difference111 patients were randomised and treated but excluded after endoscopy with no follow up or intention to treat analysis so likely attrition bias. Sample power calculation did not specify the size of change it was powered to detect. Not powered to record mortality.
Rebleeding within 4 weeksTXA: 4 (6%) Placebo: 12 (18.8%) OR: 3.635 95% CI: 1.106-11.943 p=0.033 Difference significant
Emergency endoscopy within 4 weeksTXA: 6 (9.0%) Placebo: 14 (21.9%) OR: 2.847 95% CI: 1.019-7.949 p=0.040 Difference significant
Mean transfusion requirementsTXA: 1.77 (±1.08) Placebo: 2.90 (±1.61) p<0.001 Difference significant
Hawkey at al.
414 patients admitted to hospital with suspected upper GI bleed. Exclusion criteria: need urgent surgery, on palliative care, pregnant, lactating, active thromboembolism or coagulopathy, creatinine level above 250µmol/l, using phenytoin, known allergies to TXA or PPI. Randomised to receive either 2g stat PO TXA then 1g q4h, 60mg stat PO lansoprazole then 30mg q4h, both drugs with the above doses, or a placebo. All drugs were taken for up to 4 days or until discharge.RCT (1b)Death within 30 daysPlacebo: 5/103 Lansoprazole: 2/102 TXA 4/103 Both: 5/106High exclusion rate, but clinical outcomes were measured on intention to treat basis. Small trial powered to accurately detect 50% reduction of blood in stomach due to interventions. Not powered to detect difference in mortality. Sparse information given on randomisation process. No statistical analysis of mortality, rebleeding or transfusion rates.
Rebleeding ratePlacebo: 10/103 Lansoprazole: 10/102 TXA 9/103 Both: 10/106
Transfusion ratePlacebo: 60/103 Lansoprazole: 67/102 TXA 58/103 Both: 64/106
Blood in stomach at endoscopyPlacebo: 53.7% Lansoprazole: 25.9% OR: 0.22, CI: 0.07-0.63, p=0.005 TXA: 33.3% OR: 0.27, CI: 0.09-0.81, p=0.019 Both: 25.9% OR: 0.26, CI: 0.09-0.80, p=0.013 Differences are significant
Von Holstein et al.
154 patients admitted to hospital with suspected upper GI bleed. Exclusion criteria: no gastric or duodenal bleeding on endoscopy, thromboembolic disease, bleeding or coagulation defects, on anticoagulants or pregnant. Randomised to receive 1g IV TXA q4h for 3 days then 1.5g PO q6h for 3 days. Or placebo.RCT (1b)MortalityTXA: 2/82 Placebo: 4/72174 of the patients entered were excluded, mostly due to endoscopic findings, with no follow up or intention to treat analysis which could cause attrition bias. Insufficient explanation of randomisation process. No statistical analysis of mortality rate. Not powered to detect differences in mortality, surgery or rebleeding.
Rebleeding rateTXA: 10/82 95% CI: 7-25 Placebo: 19/72 95% CI: 17-38 p=0.097 No significant difference
Emergency surgery rateTXA: 3/82 95% CI: 1-12 Placebo: 15/72 95% CI: 11-28) p=0.010 Difference significant
Mean transfusion requirementsTXA group: 2.2 units Placebo: 3.2 units p=0.018 Difference significant
Barer et al.
775 patients urgently admitted with haematemesis, melaena or both. Exclusion criteria: immediate need for operation, medical conditions seriously affected their management or bleeding not serious. Randomised to receive either 400mg of IV cimetidine q6h for 48 hours then 400mg PO q6h for 5 days, or 1g of IV TXA q6h for 48 hours then 1g PO q6h for 5 days, or a placebo following same regime.RCT (1b)MortalityPlacebo: 13.5% 95% CI: 9.7-18.4 TXA: 6.3% 95% CI 3.7-10.1 p=0.0092 Cimetidine: 7.7% 95% CI 4.9-11.9 p=0.045 Differences significantLarge trial but sample size only had 50% power to detect halving of mortality caused by interventions. No minimum age of inclusion so may include paediatric patients. Cimetidine and TXA tablets were not identical and the double dummy method was not used to compensate, so blinding was not adequate. No statistical analysis for outcomes other than mortality.
RebleedingPlacebo: 50/219 TXA: 55/225 Cimetidine: 49/232
Mean transfusion requirementsPlacebo: 5.7 units TXA: 5.7 units Cimetidine 5.3 units
Rate of surgeryPlacebo: 35/219 TXA: 44/225 Cimetidine: 32/232
Bergqvist et al.
43 patients with massive upper GI bleed (haematemesis or melaena and circulatory involvement) were randomised to receive 2g NG TXA q4h for 2 days or a placebo.RCT (2b)Percentage mortalityTXA: 12.3% Placebo: 22.7%No set exclusion criteria beforehand. Small sample size and no power calculation. No explanation of randomisation or blinding process. No reporting of side effects. Unclear how long patient were monitored for. No attempts at statistical analysis. Exact patient data not given.
Mean number of transfusionsTXA: 6.0 Placebo: 8.1
Number of surgical interventionsTXA: 7/21 Placebo: 7/22
Engqvist et al.
149 patients admitted to ICU with massive upper GI haemorrhage (haematemesis or melaena with circulatory issues). Received 1g IV TXA q4h for 3 days then 1.5g PO q6h for 4 days (discontinued if they went into surgery). Or a placebo in the same regime.RCT (2b)MortalityTXA: 11/76 Placebo: 12/73Small trial with no sample size estimates or power calculation. Patients were given treatment drugs for varying amounts of time. 37 patients were excluded as “treatment not given according to protocol” which suggests poor study implementation. Outcomes for excluded patients were not recorded and no intention to treat analysis. Inadequate information on randomisation and blinding methods.
Rate of surgical interventionTXA: 10/76 Placebo: 18/73 0.10>p>0.05 No significant difference
Mean transfusion requirements across the first 4 daysOnly significant difference on day 2 after treatment when TXA: 0.81 units and placebo: 1.26 units p<0.05
Biggs et al.
200 adult patients admitted to hospital for GI bleeding (confirmed by gastric aspiration and examination of stool for melaena). Exclusion criteria: pregnant, chronic renal impairment, previous vascular surgery, or history of a thromboembolic episode. Randomised to receive either 1g IV TXA and 1g PO q8h for 24 hours, then 1g PO q8h for a further 72 or a placebo with the same regime.RCT (2b)MortalityTXA: 3% Placebo 3%No explanation as to how blinding and randomisation took place. No power calculation or sample size estimates. No statistical analysis of mortality or overall transfusion requirement. Results for mean daily transfusion rates not reported in full. No information on whether any patients were excluded.
Rate of surgeryTXA: 7/103 Placebo: 21/97 p<0.005 Difference significant
Mean transfusion rateFull details not provided but authors stated no overall significant difference.


Review of the papers found only one with evidence of mortality reduction from tranexamic acid (Barer et al.) which was outweighed by the null results of larger and more recent studies. Therefore, there is currently no evidence to suggest that that tranexamic acid reduces mortality when used generally in the treatment of adults with acute gastrointestinal bleeding. Route of tranexamic administration does not seem to effect its impact on mortality. When evaluating papers based on the site of bleeding, this review found that the evidence for tranexamic acid in lower gastrointestinal haemorrhage is sparse, likely due to its relatively low prevalence compared to bleeds of the upper tract. There may be a need for large double blind randomised controlled trials assessing the effects of the drug on mortality, transfusion rate and other outcomes in patients with lower gastrointestinal bleeding to fill this gap in the evidence base before conclusions on its beneficence can be confidently made. There was no apparent benefit when tranexamic acid was used generally in upper GI tract bleeding. However, there may be a need for randomised controlled trials to assess its use in individual pathologies such as peptic ulcers and varices if a solution can be found with regards to how to balance the relatively slow diagnostic workup to endoscopy with tranexamic acid’s short window of effectiveness in order to both avoid attrition bias and allow effective translation into clinical practice.

Clinical Bottom Line

Tranexamic acid does not reduce mortality in adults with gastrointestinal bleeding. Further studies assessing TXA in individual GI bleeding pathologies are recommended.


  1. Roberts 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) Lancet 2020 Pages 1927-1936
  2. Miyamoto et al. Effect of tranexamic acid in patients with colonic diverticular bleeding: A nationwide inpatient database study Journal of Gastroenterology and Hepatology 2020
  3. Karadas et al. A randomized controlled trial of the effects of local tranexamic acid on mortality, rebleeding, and recurrent endoscopy need in patients with upper gastrointestinal hemorrhage European Journal of Gastroenterology & Hepatology 2020 Pages 26-31
  4. Tavakoli et al. Comparison of the efficacy of intravenous tranexamic acid with and without topical administration versus placebo in urgent endoscopy rate for acute gastrointestinal bleeding United European Gastroenterology Journal 2018 Pages 46-54
  6. Hawkey at al. Drug treatments in upper gastrointestinal bleeding: value of endoscopic findings as surrogate end points Gut 2001 Pages 372-379
  7. Von Holstein et al. Tranexamic acid as an aid to reducing blood transfusion requirements in gastric and duodenal bleeding British Medical Journal 1987 Pages 7-10
  8. Barer et al. Cimetidine and Tranexamic Acid in the Treatment of Acute Upper-Gastrointestinal-Tract Bleeding New England Journal of Medicine 1983 Pages 1571-1575
  9. Bergqvist et al. Local Inhibition of the Fibrinolytic System in Patients with Massive upper Gastrointestinal Hemorrhage Upsala Journal of Medical Sciences 1980 Pages 173-178
  10. Engqvist et al. Tranexamic acid in massive haemorrhage from the upper gastrointestinal tract: A double-blind study Scandinavian Journal of Gastroenterology 1970 Pages 839-844
  11. Biggs et al. Tranexamic acid and upper gastrointestinal haemorrhage-a double-blind trial Gut 1976 Pages 729-734