• 
• 

Three Part Question

In an [adult patient presenting to the emergency department with an upper gastrointestinal bleed], does using [tranexamic acid] improve outcomes [in terms of mortality, transfusion requirements, re-bleeding rates and need for surgery]?

Clinical Scenario

A 40-year-old man who is having a haematemesis is brought into the emergency department. He has not attended the department before but admits to chronic alcohol dependency. It is not known if he has liver disease or varices. You resuscitate him and arrange an endoscopy. Being aware of the use of tranexamic acid in trauma patients you wonder if using it would reduce this man's bleeding and improve his chance of survival?

Search Strategy

MEDLINE search (1950-Dec, 2011)
EMBASE search (1980-Dec, 2011)
Cochrane Database of Systematic Reviews

MEDLINE: [exp TRANEXAMIC ACID or tranexamic AND acid.ti,ab or exp ANTIFIBRINOLYTIC AGENTS or antifibrinolytic.ti,ab] and [exp GASTROINTESTINAL HEMORRHAGE or gastrointestinal AND hemorrhag*.ti,ab or gastrointestinal AND haemorrhag*.ti,ab or gastrointestinal AND bleed*.ti,ab]

EMBASE: [exp TRANEXAMIC ACID or tranexamic AND acid.ti,ab or exp ANTIFIBRINOLYTIC AGENT or antifibrinolytic.ti,ab] and [exp GASTROINTESTINAL HEMORRHAGE or gastrointestinal AND hemorrhag*.ti,ab or gastrointestinal AND haemorrhag*.ti,ab or gastrointestinal AND bleed*.ti,ab]

Cochrane Database of Systematic Reviews [Tranexamic acid]

Search Outcome

MEDLINE search: 250 papers
EMBASE search: 603 papers
Cochrane Database of Systematic Reviews: 22 reviews

Total: 753 papers identified once duplicates removed. Of these seven hundred and twenty were excluded as irrelevant or on methodological grounds (not randomised controlled trials) and two were excluded because they were duplicate reports on trials published elsewhere. This left seven randomised controlled trials and two systematic reviews. One Cochrane Systematic review was identified. This had already appeared as a systematic review in 2008 and identified no new trials.

Relevant Paper(s)

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study Weaknesses
Cormack et al. 1973 United Kingdom	150 patients presenting with frank haematemesis or melaena that required hospital admission. Randomised to receive either tranexamic acid 1.5g po 8 hourly for 7 days or placebo tablet.	Randomised controlled trial (RCT) (2b). Jadad score 2.	Death, need for surgery, continuing or recurrent bleeding. If patients met any of these end points they were classified as ‘treatment failures’.	15/76 in the tranexamic group and 20/74 in the control group were classified as treatment failures. Difference did not meet statistical significance. Subgroup analysis excluding patients with hiatus hernia or varices showed 7/62 treatment failures in tranexamic acid group and 17/63 in the treatment group. Statistically significant, p <0.05.	No explanation of randomisation method. Inadequate explanation of blinding. No power calculation. Ad hoc subgroup analysis performed. More patients considered to be in ‘poor condition’ on arrival were found in the control group than in the treatment group (10 vs 6).
Biggs et al. 1976 Australia	200 patients presenting to the Emergency Department with witnessed haematemesis or melaena who required hospital admission. Exclusions: renal failure, pregnancy, a thromboembolic event or vascular surgery in the preceding twelve months. Randomised to receive either tranexamic acid 1g iv and 1g orally 8 hourly for 48 hrs followed by 1g orally eight hourly for additional 72 hrs or placebo.	Randomised controlled trial (2b). Jadad score 3.	Transfusion requirements.	Full information not given. No significant difference between groups.	No power calculation. No explanation of how randomisation was performed. Impression that no dropouts or withdrawls occurred but not specifically stated.
			Need for surgical intervention.	21/97 in control group vs 7/103 in treatment group. Statistically significant with p < 0.005.
			All case mortality.	2/103 in tranexamic group vs 4/97 in control group. No significant difference.
			Side effects (poorly defined).	None identified.
Engqvist et al. 1979 Sweden	149 patients with witnessed haematemesis or melaena and circulatory embarrassment (pulse >100, systolic BP < 100, blood loss over 1800ml, Hb loss > 3g in first two days) admitted to an Intensive Care Unit. Patients with history of thromboembolic disease or renal failure were excluded. Randomised to either receive tranexamic acid 1g iv six times a day for up to 3 days followed by 1.5g orally 4 times a day for up to four days or placebo.	Randomised controlled trial (2b). Jadad score 2	Transfusion requirements.	Mean of 4.26 units in tranexamic acid group vs 5.22 units in placebo group in first four days. Not statistically significant.	No explanation of how randomisation was performed. No power calculation. Ad hoc subgroup analysis performed. A large number of patients originally entered into the study (55/ 204) were later excluded (13 as judged to not have had severe bleeding and 37 as judged to have had inadequate therapy). Their outcomes are not described. Not analysed on an intention to treat basis.
			Need for surgery.	10/76 patients in treatment group vs 18/73 in treatment group, p = 0.10.
			Mortality.	11/76 deaths in treatment group vs 12/73 in placebo group. Not statistically significant.
			Recurrent or ongoing bleeding.	23/76 in treatment group vs 29/73. Not statistically significant.
Bergqvist et al. 1980 Sweden	50 patients presenting with massive upper gastrointestinal haemorrage; defined as haematemesis and/or melaena and circulatory involvement. Randomised to either receive tranexamic acid 2g orally via NG tube 4 hourly for 2 days or to receive placebo.	RCT (2b). Jadad score 2.	Transfusion requirements (mean units).	6 units in control vs 8.1 units in treatment group (no statistical analysis provided).	Method of randomisation not provided, inadequate explanation of blinding. Small study with no power calculation. No statistical analysis attempted to judge the significance of outcomes. Not analysed on intention to treat basis. Incomplete recording of results (7 patients excluded and no results provided).
			Need for surgery.	7 patients in each group.
			Death.	5 in the placebo group vs 3 in the treatment group. No statistical analysis undertaken.
Barer et al. 1983 UK	775 patients presenting with haematemesis or melaena. Excluded if bleeding so severe as to require immediate surgery; if other serious medical conditions primarily affected management and made “blind” treatment inappropriate; or if bleeding was not considered serious. Randomised to 3 groups: (1) Tranexamic acid 1g iv 6 hourly for 48 hrs followed by 1g orally 6 hourly (256 patients). (2) Cimetidine 400 mg iv 6 hourly for 48 hours followed by 400 mg oral 6 hourly (259 patients). (3) Similar schedule with placebo injections and tablets (260 patients). All treatment was continued for one week or until discharge or death.	RCT (1b). Jadad score 4.	Death.	16 (6.3%; 95% CI: 3.7-10.1%) in tranexamic acid group (TA). 35 (13.5 %; 95% CI: 9.7-18.4%) in placebo group (PG). Difference between tranexamic acid and placebo group significant; p = 0.0092.	Large study with power calculation (strength). A double dummy methodology was not used, resulting in incomplete blinding. While mortality was decreased in the tranexamic acid group rates of re-bleeding, transfusion rates and need for surgery were no different. This raises questions about whether it is plausible to ascribe the mortality differences to the use of tranexamic acid or whether an unknown difference between the two groups (or simple chance) has caused the results.
			Transfusion rates (mean units).	5.7 tranexamic acid vs 5.7 in placebo.
			Re-bleeding rates.	58 in tranexamic acid group vs 51 in placebo group. No significant difference between groups.
			Need for surgery.	47 in tranexamic acid group vs 40 in placebo group. No significant difference between groups.
Stael Von Holstein et al. 1987 Sweden	154 patients admitted to hospital with confirmed upper gastrointestinal bleeding from benign gastric or duodenal lesions (confirmed by endoscopy). Exclusions: thromboembolic disease, renal failure, pregnancy, coagulopathy or taking an anticoagulant. Randomised to receive either tranexamic acid 1g iv four hourly for three days followed by oral tranexamic acid 1.5g 6 hourly for further 3 days and antacid (novaluzid 10ml 2 hourly) or placebo and antacid.	Randomised controlled trial (1b). Jadad score 3.	Need for surgery.	3 in treatment group vs 15 in control, p = 0.01	328 patients with suspected upper GI bleed initially entered the trial but 174 were excluded from final evaluation (majority because a gastric or duodenal bleeding source was not found). Outcomes in patients excluded were not described. Primary analysis not performed on an intention to treat basis. Significant withdrawals (48 patients) due to “misunderstandings and technical problems”. Death not included as outcome. Inadequate explanation of randomisation.
			Re-bleeding rates.	10 in tranexamic acid group vs 19 in placebo group. No significant difference found.
			Blood transfusion requirements.	Mean requirement of 2.2 units in treatment group vs 3.2 in control group, p = 0.018
			Death (not included as outcome in study).	2 in treatment group vs 4 in control.
Hawkey et al. 2003 UK	414 patients presenting with an upper gastrointestinal bleed. Patients were excluded if the bleeding was considered so severe as to require immediate surgical intervention, they were on a palliative pathway, were pregnant or lactating or had active thromboembolism or coagulopathy or had renal impairment, were on phenytoin or had known adverse reaction to trial drugs. Patients were randomised to receive either: (1) Lansoprazole 60mg po stat and 30mg po qds (2) Tranexamic acid 2 g po stat followed by 1g po qds (3) Both drugs (4) Placebo. All treatments were given for 4 days or until a clinical end-point was reached or a patient was discharged.	RCT (1b). Jadad score 4.	Blood in stomach at time of endoscopy.	Placebo group = 53 %, lansoprazole group (PPI) = 25.9 % (OR 0.22, 95% CI: 0.07-0.63, p = 0.005). Tranexamic acid (TA) group = 33.3% (OR 0.27, 95% CI: 0.09 -0.81, p=0.019). Both drugs = 25.9% (OR 0.26, 95% CI: 0.09-0.80, p = 0.013).	Main outcome measure was a surrogate endpoint of blood in stomach. No measure of inter-operator agreement in endoscopy findings. Not of sufficient power to detect any differences in clinical endpoints between different treatment groups. 61 patients judged not to have had a bleed after endoscopy and not included in analysis of endoscopy findings (no intention to treat). Inadequate explanation of randomisation.
			Need for blood transfusion.	Placebo = 60/103, PPI = 67/102, TA = 58/103, both =64/106. No significant difference.
			Re-bleeding rates.	Placebo = 10/103, PPI = 10/102, TA = 9/103, both =10/106. No significant difference.
			Surgery.	Placebo = 6/103, PPI = 3/102, TA = 5/103, both = 2/106. No significant difference.
			Death.	Placebo = 5/103, PPI = 2/102, TA = 3/103, both = 3/106. No significant difference.
Henry and O’Connell 1989	Patients from six randomised controlled trials. 1267 patients presenting with an upper GI bleed in Sweden, UK and Australia between 1973 and 1987.	Meta-analysis (1a).	Frequency of recurrent haemorrhage.	Pooled odds ratio 0.72; 95% CI: 0.46-1.10, p = 0.14. P value for heterogeneity = 0.09 (no statistically significant heterogeneity).	Included additional data provided by authors for two trials (Bergqvist et al and Stael Von Holstein et al). Results skewed by one large trial (Barer et al), which showed significant reduction in mortality but lacked internal consistency. 2 studies (Bigg et al and Bergqvist et al) had incomplete data on the endpoint of “re-bleeding’, this was replaced with data from patients who had re-bleeding severe enough to require surgery. Heterogeneity in patients, treatment and outcomes between trials.
			Need for surgery.	Pooled odds ratio 0.58; 95% CI: 0.29-1.80, p = 0.18. P value for heterogeneity = 0.008 (significant heterogeneity).
			All case mortality.	Pooled odds ratio 0.60; 95% CI: 0.39-0.90, p = 0.02. P value for heterogeneity 0.78 (no significant heterogeneity).
Gluud et al. 2008	Seven RCTs. 1754 patients in the UK, Sweden and Australia presenting with upper GI bleeds between 1973 and 2003.	Meta-analysis (1a).	All case mortality.	Tranexamic acid 5% vs placebo 8%, RR: 0.61; 95% CI: 0.42-0.89. Chi-square for inter-trial heterogeneity (ITH): p = 0.87 (no significant heterogeneity).	High degree of heterogeneity for outcomes of recurrent haemorrhage and need for surgery. Adverse events inconsistently recorded across trials. 21% of patients randomised were excluded after randomisation, data for these patients was missing (did not acquire the additional information obtained by Henry et al). When worst-case scenario analysis was performed no significant difference in mortality between tranexamic acid and placebo. Results for endpoint of needing a transfusion only available for 62% of patients.
			Bleeding (re-bleeding or continued bleeding).	Tranexamic acid 14% vs placebo 16%, RR: 0.88; 95% CI: 0.66-1.18. Chi-square for ITH: p = 0.05 (significant heterogeneity).
			Bleeding related mortality.	Tranexamic acid 3% vs placebo 5%, RR 0.66; 95% CI: 0.40-1.10. Chi-square for ITH: p = 0.54 (no significant heterogeneity).
			Surgery.	Tranexamic acid 10% vs placebo 14%. RR: 0.62; 95% CI: 0.35-1.09. Chi-square for ITH: p = 0.008 (significant heterogeneity).
			Need for transfusion.	Tranexamic acid 56% vs placebo 57%, RR: 1.0; 95% CI: 0.93-1.11. Chi-square for ITH: p = 0.59 (no significant heterogeneity).
			Serious adverse thromboembolic events (PE, MI, CVA).	Tranexamic acid = 5/ 522, placebo = 4/526, RR: 1.4; 95% CI: 0.36-5.28. Chi-square for ITH: p=0.36 (no significant heterogeneity).

Comment(s)

In combination the findings were of significant improvements in death rates (number needed to treat of 33) but no significant improvement in re-bleeding rates, need for transfusion, or surgery when tranexamic acid was used for upper gastrointestinal haemorrhage. In addition, no significant difference in adverse thromboembolic events was found when tranexamic acid was used. Issues around internal and external validity of the studies discussed mean that these findings should be interpreted with caution. The majority of studies, including the largest (Barer et al) are over 24 years old. Changes in the aetiology, prognosis and management of patients with upper gastrointestinal bleeds have occurred in this time, which limits the external validity of these early studies. Of particular relevance is the observation that all but one of the studies (Hawkey et al) pre-date the use of proton pump inhibitors (PPIs) and early endoscopic treatment of upper gastrointestinal bleeds. The one study that incorporated PPIs suggested that both tranexamic acid and PPIs reduced the amount of blood found in the stomach at endoscopy but that using both agents together did not provide any additional benefit over using a single agent. This has important implications given that PPIs are commonly used in non-variceal upper gastrointestinal bleeds.

Editor Comment

There is insufficient evidence to recommend using tranexamic acid as part of the routine management of patients presenting with a gastrointestinal bleed. It may be appropriate to consider using it in cases in which patients are having a major haemorrhage on the basis of a possible benefit combined with lack of evidence of harm. We suggest that an adequately powered randomised controlled trial in which tranexamic acid is combined with PPIs and early endoscopic treatment is still required before more firm conclusions can be made.

CVA, cerebrovascular accident; GI, gastrointestinal; ITH, inter-trial heterogeneity; MI, myocardial infarction; PE, pulmonary embolism; PPI, proton pump inhibitor; RCT, randomised controlled trial; SBP, systolic blood pressure; TA, tranexamic acid.

Clinical Bottom Line

The authors have presented the seven clinical trials and the resulting two meta-analyses. The first summarises the six trials published up to 1987 and the second updates this by including the study by Hawkey et al. This approach allows the reader to assess the data that have gone in to the meta-analyses and to see whether the publication of an additional trial, with 414 patients, changes the overall conclusions. The review by Gluud et al forms the basis of the Cochrane review ‘Tranexamic acid for upper gastrointestinal bleeding’ published in 2012 (doi: 10.1002/14651858.CD006640.pub2).

References

1. Cormack F, Chakrabarti RR, Jouhar AJ, et al. Tranexamic acid in upper gastrointestinal haemorrhage. Lancet 1973 Jun; 1: 1207-8
2. Biggs JC, Hugh TB, Dodds AJ. Tranexamic acid and upper gastrointestinal haemorrhage – a double blind trial. Gut 1976; 17: 729-34
3. Engqvist A, Brostrom O, Feilitzem F, et al. Tranexamic acid in massive haemorrhage from the upper gastrointestinal tract: a double-blind study. Scand J Gastroenterol 1979; 14: 839-44
4. Bergqvist D, Dahlgren S, Hessman Y Local inhibition of the fibrinolytic system in patients with massive upper gastrointestinal hemorrhage. Upsala J Med Sci 1980; 85: 173-8
5. Barer D, Ogilvie A, Henry D, et al. Cimetidine and tranexamic acid in the treatment of acute upper-gastrointestinal-tract bleeding. N Engl J Med 1983 Jun; 308: 1571-5
6. Stael Von Holstein CC, Eriksson SB, Kallen R. Tranexamic acid as an aid to reducing blood transfusion requirements in gastric and duodenal bleeding. BMJ 1987; 294: 7-10
7. Hawkey GM, Cole AT, McIntyre AS, et al. Drug treatment in upper gastrointestinal bleeding: value of endoscopic findings as surrogate end points. Gut 2001; 49: 372 – 379
8. Henry DA, O’Connell DL. Effects of fibrinolytic inhibitors on mortality from upper gastrointestinal haemorrhage. BMJ 1989 April; 298: 1142-6
9. Gluud LL, Klingenberg SL, Langholz SE. Systematic review: tranexamic acid for upper gastrointestinal bleeding. Alimentary Pharmacology & Therapeutics 2008 May; 27(9): 752-8