This trial provides evidence that administration of TXA to TBI patients within 3 hours of injury reduces head injury deaths, with no evidence of adverse effects or complications. There was a substantial reduction in head injury deaths with TXA in patients with mild or moderate head injuries, but no apparent reduction in those with severe head injuries. There was no increase in disability among survivors.
The effect of TXA on head injury death appears to depend on the time interval between injury and the initiation of the trial treatment, and on the severity of the TBI. Early treatment of patients with mild (GCS score of 13–15 and intracranial bleeding on baseline CT scan) and moderate head injury seems to confer the greatest mortality benefit. This is consistent with the hypothesis that TXA improves outcomes by reducing intracranial bleeding. As haemorrhage expansion occurs in the hours immediately after injury, treatment delay would reduce the potential for TXA to prevent intracranial bleeding.5,6 Patients with severe head injury may have less to gain from TXA treatment because they already have extensive intracranial haemorrhage prior to treatment, or other potentially life-threatening intracranial pathologies that are not affected by TXA. We anticipated in our statistical analysis plan that the effect of TXA would be greatest for head injury deaths occurring in the first few days after injury than for later head injury deaths, because early head injury deaths are more likely as a result of bleeding. Our data support this hypothesis, showing a substantial reduction in head injury deaths within 24 hours of injury (RR 0.72, 95% CI 0.56 to 0.92). Similar results were obtained in the CRASH-2 trial35 of TXA in traumatic extracranial bleeding, in which the effect of TXA on death from bleeding was greatest on the day of the injury (RR 0.72, 95% CI 0.60 to 0.86). However, thereafter, the benefit of TXA for head injury patients is slightly attenuated, probably as patients succumbed to non-bleeding-related pathophysiological mechanisms. This may explain why the effect of early TXA treatment on head injury death is slightly smaller than the effect of TXA on death due to bleeding seen in the CRASH-2 trial.35
Tranexamic acid did not appear to increase disability among survivors, and there was no evidence of any increased risk of adverse events. In particular, the risk of DVT, PE, stroke and MI was similar in the TXA and placebo groups. This is consistent with the results of the CRASH-2 trial35 in traumatic extracranial bleeding, which also recorded no increased risk of vascular occlusive events with TXA. Unlike the CRASH-2 trial,35 there was no evidence that administration beyond 3 hours of injury increased the risk of head injury death or any other adverse events. Indeed, given the absence of any adverse effects in this trial, the implications of wrongly concluding that TXA is ineffective are likely to be far more consequential than wrongly concluding that TXA is effective.
The CRASH-3 trial provides evidence that TXA is safe in TBI patients and that treatment within 3 hours of injury reduces head injury deaths.
Strengths and limitations
Our trial had several strengths but also some limitations. The method of randomisation ensured that participating clinicians had no foreknowledge of the treatment allocation, and the use of placebo control ensured that outcome assessment was blind to the intervention. Although the eligibility criteria required the recruiting doctor to be uncertain as to the appropriateness of TXA treatment, because TXA is not a recommended treatment for isolated TBI, almost all TBIs meeting the inclusion criteria were recruited. Baseline prognostic factors were well balanced and, because almost all randomly assigned patients were followed up, there is little potential for bias. The analysis was by intention to treat (176 patients did not receive any of the trial treatment). The primary outcome was head injury death as assessed by the responsible clinician. Although some misclassification of cause of death is inevitable, the assessment was blinded to the trial treatment. All-cause mortality combines causes of death that might be affected by TXA (e.g. head injury death due to intracranial bleeding) with causes that we do not expect to be affected by TXA (e.g. sepsis) and, therefore, would be biased towards the null. Although the CRASH-3 trial is one of the largest trials of TBI, the CIs were wide and compatible with a substantial reduction in head injury death and little or no benefit. On the other hand, when set in the context of all the available randomised trials of TXA in TBI, the possibility of no mortality benefit appears remote.21,22,58 When assessing outcome measures in clinical trials, provided that there are few false positives (high specificity), estimates of the RR are unbiased even when sensitivity is imperfect.59 For this reason a diagnosis of DVT or PE was recorded only if there was a positive result on imaging (e.g. ultrasound) or at post-mortem examination. As a result, although the trial may have underestimated the risk of DVT or PE, the RR estimates for this outcome should be unbiased.
We anticipated that TBI patients with a GCS score of 3 and those with bilateral unreactive pupils prior to treatment would have little potential to benefit from TXA and that their inclusion in the analysis would bias the treatment effect towards the null. Most patients with bilateral unreactive pupils already have extensive intracranial haemorrhage and brain herniation and so it is unlikely that TXA could improve the outcome in these cases. We therefore prespecified a sensitivity analysis that excluded these patients. However, patients with unilateral unreactive pupils were not excluded, and because many of these patients have brain herniation their inclusion might also have diluted the treatment effect. Indeed, when patients with a GCS score of 3 and those with unilateral or bilateral unreactive pupils prior to treatment are excluded in a post hoc analysis, the treatment effect is noticeably larger (RR 0.85, 95% CI 0.74 to 0.96).
Cost-effectiveness
Although the cost of TXA treatment is low, providing treatment will still incur additional costs to the health service and, therefore, questions arise regarding whether or not this cost represents an efficient use of resources, based on the benefit associated with treatment. Our analysis shows that TXA is highly cost-effective in the UK for those with complicated mild and moderate TBI, and is also highly cost-effective for patients with both pupils reactive. These results were robust across sensitivity analyses, as probabilistic analyses showed that the intervention is 99% likely to be cost-effective in both model populations at the NICE willingness-to-pay threshold of £20,000 per QALY. Furthermore, all deterministic sensitivity analyses produced ICERs below the lower limit of the NICE cost-effectiveness threshold of £20,000.
The NICE guidelines included TXA for the pre-hospital care of patients with trauma, following the results of the CRASH-2 trial.60 Our analysis suggests that TXA should also be recommended for patients with complicated mild and moderate TBI and for patients with both pupils reactive, when treatment can be provided within 3 hours of injury, as treatment is highly cost-effective.
The cost-effectiveness analysis has some limitations. One limitation is that the trial followed patients for only 28 days post injury, leading to uncertainty about patient outcomes beyond this time. Furthermore, evidence on long-term outcomes post TBI in the literature is limited. To capture the long-term additional risk of death for these patients, we assumed that, after the trial period, the risk of death remained elevated compared with the risk of death in the general population (four times higher for the first year and two times higher thereafter). However, these estimates were derived from a case–control study performed in Scotland, and there is uncertainty as to whether or not the additional risk of death reported in this study is likely to be reflective of the patients in this trial. Sensitivity analyses were performed to consider the uncertainty in future outcomes, first, using higher discount rates (giving lower weighting to future events), and, second, performing a scenario excluding this additional mortality. Both had little impact on the estimated ICER, with TXA remaining cost-effective in both analyses, suggesting that this uncertainty is unlikely to affect the overall cost-effectiveness of the treatment.
In addition, the CRASH-3 trial did not collect direct utility estimates, meaning that they were estimated from the DRS outcomes at 28 days (or at time of discharge). Although just over half of all mild and moderate TBI patients had no disability at discharge, there was uncertainty in this estimation process of overall utility, as well as uncertainty regarding the long-term disability of patients compared with their status at discharge or 28 days, when some patients’ utility would be expected to improve over time. To address this, sensitivity analyses with lower utility values, and a lower utility value for those receiving TXA, were performed. Neither sensitivity analysis influenced the decision on cost-effectiveness.
Last, our analysis was performed from a health service perspective, and therefore did not capture the potential long-term costs that could be associated with caregiver burden or out-of-pocket medical payments that might be associated with those living with disability. However, it should be noted that the disability scores for survivors are similar between groups, and therefore any additional societal burden associated with TXA treatment would result from a higher proportion of patients surviving only, as the outcomes among survivors were comparable.
Despite the limitations stated above, we have used robust trial results and supporting evidence from the literature to show that TXA treatment is highly likely to be cost-effective for the treatment of patients with complicated mild and moderate TBI and for patients with both pupils reactive, when provided within 3 hours of injury.
Findings in context
Evidence before this study
Evidence from the CRASH-218 trial that administration of TXA within 3 hours of injury reduces death in patients with traumatic extracranial bleeding raised the possibility that it might reduce death from traumatic intracranial bleeding. Intracranial bleeding is common after TBI, and increases head injury death and disability. Prior to the CRASH-3 trial, we made a systematic search for all randomised trials of TXA in acute traumatic injury. We searched PubMed, Science Citation Index, National Research Register, Zetoc, System for Information on Grey Literature in Europe (SIGLE), Global Health, Latin American and Caribbean Health Sciences Literature (LILACS), Current Controlled Trials, the Cochrane Injuries Group Specialised Register, CENTRAL, MEDLINE and EMBASE to July 2010. Details of our search were published previously.61 We found two small randomised trials of TXA in TBI with a total of 510 patients. Meta-analysis of the two trials showed a statistically significant reduction (RR 0.63, 95% CI 0.40 to 0.99) in death with TXA. However, given the small size of the trials, we considered this evidence to be hypothesis generating requiring confirmation in larger randomised trials.
Added value of this study
The CRASH-3 trial included 9202 TBI patients who were within 3 hours of injury with either a GCS score of ≤ 12 or any intracranial bleeding on CT scan and no major extracranial bleeding. The risk of head injury death was lower with TXA, particularly when patients who had a GCS score of 3 and those with bilateral unreactive pupils at baseline were excluded as prespecified in the statistical analysis plan (RR 0.89, 95% CI 0.80 to 1.00). There was no evidence of any increase in disability among survivors. The risk of vascular occlusive events was similar in both groups.
Implications of all the available evidence
An updated search for randomised trials of the early administration of TXA in patients with TBI identified one randomised trial in addition to the CRASH-3 trial. This was a randomised trial of pre-hospital TXA in 967 patients with TBI, which was funded by the US National Institutes of Health and sponsored by the University of Washington. The dose of TXA was the same as in the CRASH-3 trial and it also excluded patients with a GCS score of 3 and those with unreactive pupils at baseline. When the two trials are pooled (Figure 11), there is a reduction in head injury death with TXA (RR 0.89, 95% CI 0.80 to 0.99) and no evidence of an increased risk in vascular occlusive events (RR 0.89, 95% CI 0.71 to 1.13) or seizures (RR 1.08, 95% CI 0.89 to 1.31). When the results of all available randomised trials are combined there is a reduction in head injury death with TXA (RR 0.88, 95% CI 0.79 to 0.97). Early administration of TXA should be considered in patients with TBI.
FIGURE 11 Summary of (a) previous evidence and (b) current evidence on the effect of TXA on head injury death. (Figure not reproduced here; see the original article.)
Implications for practice in the NHS
Based on the CRASH-218 trial results, TXA was included in guidelines for the pre-hospital care of trauma patients. Box 1 shows the TXA trauma guideline from the Joint Royal Colleges Ambulance Liaison Committee (JRCALC).63
Treatment of known or suspected severe traumatic internal or external haemorrhage as soon as clinically possible on arrival at the scene and within 3 hours of bleeding starting in adults and children who are considered to be at risk of significant haemorrhage. This may be demonstrated by one or more of:
- SBP of < 90 mmHg or absent radial pulse or heart rate of > 110 b.p.m. believed to be due to bleeding in adults. In children this may be demonstrated by changes in the normal physiological parameters for age (see Joint Royal Colleges Ambulance Liaison Committee page for age).
- Any patient where haemostatic gauze, arterial tourniquet(s), chest dressing(s) or pressure dressing(s) have been applied.
- Patient who has suffered a traumatic cardiac arrest.
Contraindications
- Known previous anaphylactic reaction to TXA.
- Bleeding started > 3 hours ago.
- Obvious resolution of haemorrhage.
- Isolated head injury.
- Critical interventions required [must be given only after critical interventions have been performed (i.e. airway managed, control or splinting of major haemorrhage, etc.), and if administration does not delay transfer, noting that it may be administered en route].
As can be seen, patients with isolated TBI are specifically excluded. The CRASH-3 trial provides evidence that TXA is safe in TBI patients and that treatment within 3 hours of injury reduces head injury deaths.24 In the light of this evidence, the exclusion of patients with isolated TBI from TXA treatment guidelines seems unnecessary.
The effect of TXA on head injury-related death appears to depend on the time interval between injury and the initiation of the trial treatment and on the severity of the TBI. Early treatment of patients with mild (GCS score of 13–15 and intracranial bleeding on baseline CT scan) and moderate head injury seemed to confer the greatest mortality benefit. This finding is consistent with the hypothesis that TXA improves outcomes by reducing intracranial bleeding. Haemorrhage expansion occurs in the hours immediately after injury and, therefore, treatment delay would reduce the potential for TXA to prevent intracranial bleeding. Patients with mild or moderate head injury have the most to gain from TXA treatment because, if intracranial haemorrhage can be prevented, these patients are less likely to die from other life-threatening intracranial pathologies such as generalised brain swelling, which may not be affected by TXA.
However, the need to rapidly treat the large number of patients who attend emergency departments with mild or moderate TBI presents challenges for implementation in the NHS. Each year, about 1.4 million people attend emergency departments in England and Wales with a recent head injury.64 Around 95% of these patients present with a normal or minimally impaired consciousness level (GCS score of < 12) and are classified as having mild TBI.64 It is unlikely that all patients attending hospital with mild TBI would be treated with TXA because the inclusion criteria of the CRASH-3 trial included only those patients with mild TBI with evidence of intracranial bleeding on their CT scan. Although patients with intracranial bleeding on their CT scan represent only about 5–10% of patients with mild TBI, this is still a large number of patients. The indications for TXA treatment in mild TBI are clearly a matter for discussion between clinicians and policy-makers and will need to take into account considerations of practicality and cost-effectiveness.
Implications for research in the NHS
The CRASH-217 and CRASH-324 clinical trials have shown that i.v. administration of TXA significantly reduces mortality in trauma patients; however, patients must be treated urgently. Many deaths occur on the day of the injury and treatment delay reduces the survival benefit from TXA. Immediate TXA treatment improves survival but the treatment benefit decreases by about 10% for every 15 minutes of treatment delay until 3 hours, after which there is no benefit.19 To reduce delay, TXA is increasingly given by paramedics at the scene of injury. Trauma audit data for England and Wales (2016) show that when TXA is given by paramedics the median time to treatment is 50 minutes, compared with 110 minutes when TXA is given in hospital.65
One of the main obstacles to further reducing treatment delay is the need for an i.v. injection. Securing i.v. access at the injury scene can be difficult, particularly for trapped patients. Moreover, on-scene i.v. cannulation increases pre-hospital times, thus delaying definitive surgical control of bleeding. If TXA could be given by intramuscular (i.m.) injection, this might reduce the time to TXA treatment and pre-hospital times. It would also facilitate the more rapid treatment of the large number of patients with mild TBI seen in UK emergency departments. If, for example, mild TBI patients could be rapidly triaged to identify those who would benefit from TXA treatment, nursing staff could administer an i.m. injection of TXA while the patient was waiting to see an emergency physician. Although absorption of TXA from muscle tissue would involve some delay, the available pharmacokinetic data suggest that an immediate i.m. injection might achieve therapeutic TXA levels faster than a delayed i.v. injection. As TXA has a wide therapeutic index, i.m. TXA injection can be followed by an i.v. injection.
The British military also has a strong interest in i.m. TXA use in trauma and is in the early stages of developing an i.m. TXA auto-injector for combat use.66 An easy-to-use TXA auto-injector would allow soldiers to administer i.m. TXA to themselves or their colleagues as soon as possible after wounding to maximise survival. Such a device could also have implications for civilian trauma, particularly mass casualty events (the UK incidence is three or four events per year67), as it would facilitate rapid treatment of a large number of trauma patients. An easy-to-use auto-injector would also raise the possibility of use by non-medical first aiders.
Studies of i.m. TXA in healthy volunteers show that therapeutic levels (plasma TXA > 10 mg/l) are reached within 30 minutes of i.m. injection of 500 mg of TXA.68 Administration of 1000 mg (the dose used in trauma) would achieve therapeutic levels even sooner.69 If absorption was as rapid in trauma patients, this would strongly suggest the i.m. route as an alternative to i.v. use. The main uncertainty is the impact of bleeding on muscle absorption of TXA. Acute blood loss leads to compensatory cardiovascular responses that maintain blood flow to the vital organs at the expense of the peripheral tissues.70 Skin and skeletal muscle are major targets for these responses, with significant reductions in muscle blood flow. This could reduce the rate of absorption of TXA from muscle. Studies of i.m. atropine in animal shock models show that hypovolaemia significantly reduces absorption, although the reductions are modest (10 minutes).71 In most cases, on-scene i.m. injection would be given before shock onset, as only the most severely bleeding patients have early shock and shock is rare in patients with isolated TBI. Furthermore, because low-risk patients greatly outnumber high-risk patients, early treatment of low-risk patients prevents more deaths. To resolve this uncertainty, studies of the pharmacokinetics of i.m. TXA in a spectrum of trauma patients to assess the time taken to reach therapeutic levels would be a research priority.
To determine whether i.m. TXA has the potential to improve the care of trauma patients, we need to understand the pharmacokinetics of TXA following i.m. use. If we find that i.m. TXA is well absorbed, with therapeutic TXA levels achieved in a timely manner, i.m. TXA would provide a rapid alternative to i.v. use when immediate i.v. administration is not possible. This would expand the treatment options available to UK paramedics at the scene of a crash and facilitate the development of a TXA auto-injector for use on the battlefield and during mass casualty events. It would also facilitate early treatment of the large numbers of patients with mild or moderate TBI seen in UK emergency departments. Because TXA safely reduces mortality after trauma, this research would provide immediate benefits to patients.