Source: CRASH-3 trial collaborators. Effects of tranexamic acid on death, disability, vascular occlusive events and other morbidities in patients with acute traumatic brain injury (CRASH3): a randomised, placebo controlled trial. Lancet. 2019;394(10210):1713-1723.
Yokobori S, Yatabe T, Kondo Y, Kinoshita K; Japan Resuscitation Council (JRC) Neuroresuscitation Task Force and the Guidelines Editorial Committee. Efficacy and safety of tranexamic acid administration in traumatic brain injury patients: a systematic review and meta-analysis. J Intensive Care. 2020;8:46.
Study Population: 7 trials comprising 10,044 patients with traumatic brain injury
Efficacy Endpoints: Mortality, poor neurologic outcome, hemorrhagic complications
Harm Endpoints: Ischemic or thromboembolic complications
Narrative: Traumatic brain injury (TBI) is a significant cause of morbidity and mortality. In 2014, there were approximately 2.5 million emergency department visits for TBI in the United States, with 56,800 deaths attributed directly to TBI.1 Among survivors it is estimated millions more live with long-term disability.2 While one large study demonstrates a mortality benefit (NNT=67) for tranexamic acid (TXA) in multi-system trauma,3 no primary outcome benefit was demonstrated for TXA in the WOMAN trial4 of post-partum hemorrhage or in the HALT-IT trial5 of gastrointestinal bleeding. Here we summarize results from the CRASH-3 trial,6 as well as from a systematic review combining this trial with data from 6 smaller trials of TXA for TBI.7
CRASH-3 was a multinational randomized placebo-controlled trial enrolling 12737 subjects with TBI, with a mean age of 42 years.6 Included patients suffered TBI <8 hours prior to potential administration of study drug, and had either a Glasgow Coma Scale (GCS) score ≤12 or intracranial hemorrhage on brain imaging, with no major extracranial bleeding. Patients were randomly assigned to either TXA 1g intravenously over 10 minutes followed by 1g over 8 hours, or placebo infusion. The author-reported primary outcome was head injury-related in-hospital mortality within 28 days of injury among the subgroup receiving treatment within 3 hours. Secondary outcomes included, among others, all-cause mortality, disability, vascular occlusive events, seizures, neurosurgical interventions, and other complications.
For the outcome of head injury-related mortality in patients treated within 3 hours, there was no significant difference, with a mortality rate of 18.5% in the tranexamic acid group and 19.8% in the placebo group (Relative risk [RR]: 0.94 95% CI 0.86 to 1.02). There was also no difference in the key secondary outcomes of neurologic or functional disability and all-cause mortality. For subjects receiving TXA within 3 hours the authors do, however, report a lower risk of head injury-related death among those with mild or moderate head injury. This is based on post-hoc subgroup analysis (5.8% versus 7.5%; RR: 0.8; 95% confidence interval [CI]: 0.6 to 0.95, Absolute risk difference [ARD]: 1.7%; Number needed to treat [NNT]: 59). Risk of vascular occlusion was not different between groups.
A systematic review incorporating these data and six additional randomized trials (total n=10,044) further compares TXA with placebo or no intervention for patients with TBI.7 This systematic review included 9202 subjects from CRASH-3 who received TXA within 3 hours of injury and 842 subjects from additional trials administering TXA within 8 hours. The primary outcome was mortality, while secondary outcomes included poor neurologic outcome, ischemic or thromboembolic complications, and hemorrhagic complications.
The mean age of participants in the included studies ranged from 32-42 years, and 9-25% were women. There were no differences between TXA and control groups in mortality, neurologic outcome, ischemic or thromboembolic complications, or hemorrhagic complications.
Caveats: The CRASH-3 trial is the largest to date evaluating the utility of TXA in patients with TBI.6 CRASH-3 and a separate meta-analysis of seven trials both find no mortality or disability benefit from administering TXA in TBI patients.6, 7 The authors of CRASH-3, however, assert a mortality benefit in subgroups based on TBI severity and recommend that patients receive TXA within 3 hours of injury to reduce “head injury-related death,” but there are several significant issues with this interpretation.6 First, the authors executed more than half of CRASH-3 with a primary outcome of all-cause mortality for those treated within 8 hours, but in 2016 they changed their protocol. This change resulted in head injury-related mortality replacing all-cause mortality as the primary endpoint, and the authors excluded those treated after 3 hours from their new primary outcome. Therefore, more than 3500 enrolled subjects were removed from their primary analysis.
This transition has other considerations as well. All-cause mortality is a more patient-centered outcome. Additionally, head injury-related death is not specifically defined, and deaths in the period immediately following an isolated TBI could all be considered head injury-related. Randomization and blinding should neutralize any biasing effect of ostensibly non-head injury deaths, though an exception would be if drug-related deaths occur. Thus, all-cause mortality is critical, because using head injury-related deaths may remove deaths potentially due to TXA, such as vaso-occlusive fatalities. By our calculations from the data provided in the supplementary online appendix of CRASH-3,6 we noted a statistically significant increase in combined fatal stroke, myocardial infarction, and pulmonary embolism (the adverse event measure reported in CRASH-2),3 in the TXA group (RR: 2.0, 95% CI: 1.04 to 3.7). These deaths, however, were removed from the CRASH-3 primary outcome analysis by the change to ‘head injury-related death’.
Third, CRASH 3 posit a subgroup benefit with TXA in patients with GCS 9-15 (i.e., mild-tomoderate head injury). Based on their 2018 statistical analysis plan paper,8 they planned to evaluate three discrete severity subgroups: mild, moderate, and severe. However, in their 2019 paper, they combine mild and moderate into one subgroup.6, 8 Moreover, in their statistical analysis paper the investigators state they expected outcomes with TXA not to vary based on severity, and thus select a heterogeneity calculation threshold of p <0.001 for presenting their subgroup analysis results.8 The 2019 paper then reports a p-value of 0.03 for this calculation, suggesting that according to their protocol they should not have presented these subgroup analyses.6 Finally, subgroup analyses reduce external validity in general, create challenges in application, and should be considered hypothesis generating at best. In this case, it is particularly unreliable as the primary analysis found no benefit.
The systematic review and meta-analysis that incorporated CRASH-3 also has several important limitations.7 First, CRASH-3 represents 92% of the total number of patients included in the analysis, but uses a different enrollment window (0-3 hours) than the other 6 trials. Additionally, GCS score varied between studies, and there was limited description of severity or mechanism of injury. Moreover, many patients had polytrauma, and it is unclear the degree to which this contributed to death. Finally, the authors did not assess important complications such as seizures or cognitive deficits.
A more recent RCT released in September 2020 included out-of-hospital patients with TBI with GCS of 12 or less and a systolic blood pressure of greater than 90 mm Hg.9 Of the 966 patients included, TXA administration within 2 hours of injury compared with placebo did not improve 28-day mortality or 6-month neurologic outcome.9
Based on the above data, TXA does not show any significant benefit in patients with TBI. Administration of TXA after TBI also does not seem to offer any benefit in disability or neurologic functional outcomes. In addition, there may be a small increase in the rate of thromboembolic fatalities associated with this treatment. Therefore, we have assigned a color recommendation of Red (No benefit) to this intervention.
The original manuscript was published in Academic Emergency Medicine as part of the partnership between TheNNT.com and AEM.
Author: Brit Long, MD; Michael Gottlieb, MD
Supervising Editors: Shahriar Zehtabchi, MD
Published/Updated: December 1, 2020
The title bar is color-coded with our overall recommendation.