The CRASH-2 trial: a randomised controlled trial and economic evaluation of the effects of tranexamic acid on death, vascular occlusive events and transfusion requirement in bleeding trauma patients

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 06/303/20. The contractual start date was in April 2007. The draft report began editorial review in January 2012 and was accepted for publication in August 2012. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors' report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

© Queen's Printer and Controller of HMSO 2013. This work was produced by Roberts et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Mechanisms

The haemostatic system helps to maintain the integrity of the circulatory system after severe vascular injury, whether traumatic or surgical in origin. 4 Major surgery and trauma trigger similar haemostatic responses and the consequent massive blood loss presents an extreme challenge to the coagulation system. Part of the response to surgery and trauma, in any patient, is stimulation of clot breakdown (fibrinolysis) which may become pathological (hyper-fibrinolysis) in some. 4 Antifibrinolytic agents have been shown to reduce blood loss in patients with both normal and exaggerated fibrinolytic responses to surgery, and do so without apparently increasing the risk of postoperative complications; most notably there is no increased risk of venous thromboembolism. 5

Existing knowledge

Tranexamic acid (TXA) is widely used in major surgery to prevent fibrinolysis and reduce surgical blood loss. A systematic review6 of randomised controlled trials of TXA in elective surgical patients identified 53 studies including 3836 participants. TXA reduced the need for blood transfusion by one-third [relative risk (RR) 0.61; 95% confidence interval (CI) 0.54 to 0.70]. TXA reduced the need for blood transfusion in cardiac surgery, orthopaedic surgery, liver surgery and vascular surgery. When the analysis was restricted to those trials that had good-quality allocation concealment, there was again a significant reduction in the need for blood transfusion (RR = 0.60; 95% CI 0.49 to 0.72). Although there was some statistical evidence of heterogeneity in the treatment effect, such that some trials showed a larger effect than others, the effect was consistent in as much as nearly all trials showed a reduction in the need for blood transfusion. There was also a non-significant reduction in the risk of death with TXA (RR = 0.60; 95% CI 0.32 to 1.12). The effect was large but imprecise, reflecting the fact that mortality is relatively rare in elective surgery. Importantly, there was no evidence of any increased risk of thromboembolic events with TXA. Specifically, there was no evidence of any increased risk of myocardial infarction, stroke, deep-vein thrombosis or pulmonary embolism. Taken together, the evidence that TXA reduces bleeding in surgical patients is reasonably strong. This evidence was an important motivation for the conduct of the Clinical Randomisation of an Antifibrinolytic in Significant Haemorrhage-2 (CRASH-2) clinical trial.

Tranexamic acid is a member of a class of drugs called antifibrinolytic agents. Other members of this class are aprotinin and aminocaproic acid. At the time that the CRASH-2 trial protocol was being developed there was also evidence that aprotinin reduced blood loss in surgery, again without compelling evidence of an increase in thromboembolic complications. Indeed, because there had been many more clinical trials of aprotinin in elective surgery than of TXA, the estimate of the effect of aprotinin on the need for blood transfusion was even more precise than for TXA (RR = 0.66; 95% CI 0.62 to 0.71). However, a major drawback of aprotinin is that it is considerably more expensive than TXA. Another disadvantage of aprotinin is that unlike TXA, which is a simple synthetic molecule, aprotinin is a bovine product with a consequent risk of allergic reaction and a hypothetical risk of disease transmission. Indeed, some sources recommended that before using aprotinin there should be a test dose to assess the presence or absence of allergic reactions.

At the time the CRASH-2 trial was being designed there was considerable interest in the potential of a drug called activated factor VII (factor VIIa) as a treatment for bleeding. Activated recombinant factor VII [rFVIIa or NovoSeven®(Novo Nordisk Inc., Princeton, NJ, USA)] is a pro-coagulant drug used to treat clotting disorders including haemophilia with inhibitors. It was suggested that it might also be useful in controlling non-haemophiliac bleeding. However, despite the considerable hype at the time, there was little reliable evidence that the benefits of factor VIIa outweighed the harms. One of the earliest reports of rFVIIa use in non-haemophiliac bleeding was published in The Lancet in 1999. 7 This report documented the use of rFVIIa in one Israeli soldier who suffered a thoracic gunshot wound. It reported that surgical attempts to stop bleeding had failed and the patient was near death but that then, ‘in a desperate attempt to control the bleeding’, he was given two doses of rFVIIa. Minutes later the bleeding stopped, allowing surgeons to repair the vessel. Although case reports such as this are a notoriously unreliable basis for treatment decisions, such narratives are remarkably influential. Following this early case report hundreds more were published, which may have influenced practitioners to take similar ‘desperate’ measures in comparable situations. However, the evidence base was weak. The scientific ‘prior’ for TXA as a treatment for bleeding trauma patients was much more compelling and, hence, TXA was selected for use in the CRASH-2 clinical trial.

Hypothesis

The extent to which a treatment effect observed in one clinical setting (elective surgery) might reasonably be generalised to another clinical setting (e.g. trauma) is a scientific judgement based largely on a consideration of whether or not the same mechanism of action is likely to apply in both cases. Of course, the most important criterion for any scientific generalisation is that the treatment effect to be generalised is valid and precise. The Cochrane systematic review6 of TXA in surgical patients had provided evidence of a treatment effect that was apparently valid – in as much as it was obtained from well-concealed randomised controlled trials and reasonably precise. TXA appeared to reduce bleeding in surgical patients and the effect was large. Is it possible, then, that it might also reduce bleeding in trauma patients? In both of these clinical situations patients experience tissue damage and bleed as a result. In both of these clinical situations the coagulation system is involved in achieving haemostasis. Antifibrinolytics such as TXA are believed to work by inhibiting clot breakdown. They work by inhibiting plasmin, an enzyme that causes clot breakdown. Because we judged that the mechanism of action of TXA on blood coagulation after injury was likely to be similar to the effect of TXA in surgery, we considered it possible that TXA might reduce blood loss, the need for transfusion and mortality following trauma. However, prior to the CRASH-2 trial there had been only one small randomised controlled trial of an antifibrinolytic agent in bleeding trauma patients (70 randomised patients: drug vs placebo: 0 vs 3 deaths). 8 As a result, there was insufficient evidence to either support or refute a clinically important treatment effect. Systemic antifibrinolytic agents have been used in the management of eye injuries, where there is some evidence that they reduce the rate of secondary haemorrhage. 9

Need for a trial

A simple and widely practicable treatment that reduces blood loss following trauma might prevent thousands of premature trauma deaths each year and could also reduce exposure to the risks of blood transfusion. Blood is a scarce and expensive resource and major concerns remain about the risk of transfusion-transmitted infection. Trauma is common in parts of the world where the safety of blood transfusion is not assured. A recent study in Uganda estimated that the population-attributable fraction of HIV acquisition as a result of blood transfusion to be around 2%, although some estimates are much higher. 10,11 Only 43% of the 191 World Health Organization (WHO) member states test blood for HIV and hepatitis C and B viruses. Every year unsafe transfusion and injection practices are estimated to account for 8–16 million hepatitis B infections, 2.3–4.7 million hepatitis C infections and 80,000–160,000 HIV infections. 12 A large randomised trial is therefore needed of the use of a simple, inexpensive, widely practicable antifibrinolytic treatment such as TXA, in a wide range of trauma patients who, when they reach hospital, are thought to be at risk of major haemorrhage that could significantly affect their chances of survival.

Dose selection

The systematic review of randomised controlled trials of antifibrinolytic agents in surgery showed that dose regimens of TXA vary widely. 6 Loading doses range from 2.5 mg/kg to 100 mg/kg and maintenance doses from 0.25 mg/kg/hour to 4 mg/kg/hour delivered over time periods of 1–12 hours. Studies examining the impact of different doses of TXA on bleeding and transfusion requirements showed no significant difference between a high dose and a low dose.

Studies in cardiac surgery have shown that a 10 mg/kg initial dose of TXA followed by an infusion of 1 mg/kg/hour produces plasma concentrations sufficient to inhibit fibrinolysis in vitro. 13 The dose–response relationship of TXA was examined by Horrow et al. ,14 who concluded that 10 mg/kg followed by 1 mg/kg/hour decreases bleeding after extracorporeal circulation and that larger doses did not provide any additional haemostatic benefit.

In this emergency situation, administration of a fixed dose would be more practicable, as determining the weight of a patient would be impossible. Therefore, a fixed dose within the dose range which has been shown to inhibit fibrinolysis and provide haemostatic benefit is being used for this trial. The fixed dose chosen would be efficacious for larger patients (> 100 kg) but also safe in smaller patients (< 50 kg), as the estimated dose/kg that the latter group would receive has been applied in other trials without adverse effects. The planned duration of administration allows for the full effect of TXA on the immediate risk of haemorrhage without extending too far into the acute phase response seen after surgery and trauma.

Study design

Clinical Randomisation of an Antifibrinolytic in Significant Haemorrhage-2 is a large placebo-controlled trial of the effects of early administration of a short course of TXA on death, vascular occlusive events and the receipt of blood transfusion. The trial protocol was peer-reviewed and published on The Lancet website in 2005. 15

Although the effect of TXA on blood transfusion in surgical patients was reasonably large, we felt that it would be unreasonable to expect a large reduction in mortality with TXA in bleeding trauma patients. On the other hand, a modest reduction in all-cause mortality was scientifically plausible. However, to detect such a modest treatment effect the trial would need to be large and would have to include many thousands of randomised trauma patients. It would clearly have to be an international clinical trial, as it would take many years to enrol such a large number of trauma patients in the UK alone. The trial co-ordinating centre at the London School of Hygiene and Tropical Medicine (LSHTM) had experience in conducting large international clinical trials. Indeed, the CRASH-2 trial followed on closely from the Corticosteroid Randomisation After Significant Head Injury (CRASH-1) trial of corticosteroids in head injury. The Medical Research Council (MRC)-funded CRASH-1 trial had enrolled 10,000 patients with significant traumatic brain injury from many countries worldwide. As a result of the CRASH-1 clinical trial, the LSHTM trial co-ordinating centre had established an excellent global network of collaborating trauma hospitals. Over many years of collaboration, LSHTM had developed good working relationships with a large number of trauma doctors around the world. The trial network thus had extensive experience in conducting large-scale clinical trials and it was a critically important human resource when it came to the conduct of the CRASH-2 trial. Planning for the CRASH-2 trial was already under way before the end of the CRASH-1 trial. Indeed, the draft protocol of the CRASH-2 trial was presented for discussion at the closure meeting of the CRASH-1 clinical trial. There was considerable enthusiasm within the trial network for the conduct of the CRASH-2 clinical trial. This was a major asset in ensuring rapid recruitment in the CRASH-2 trial.

The trial was undertaken in 274 hospitals in 40 countries. The main criteria for selecting participating hospitals were that the hospitals provide definitive trauma care for a sufficiently large number of trauma patients; the hospital doctors are substantially uncertain as regards the effect of TXA in the management of bleeding trauma patients; and the hospitals have the necessary research infrastructure to conduct the trial. As an example of the hospitals having the necessary research infrastructure to conduct the trial, it is essential that the trial co-ordinating centre has a reliable means of communication with all the participating hospitals and reliable e-mail and telephone communication with the principal investigator at the participating hospital.

Although some clinicians believe that standardisation of clinical care is a prerequisite for the conduct of multicentre randomised trials, the CRASH-2 investigators did not stipulate how trauma patients at participating hospitals should be managed. Providing that a trial is large enough, randomisation will ensure that the intervention and control groups are identical with regard to both known and unknown confounders. It is of course conceivable that the size of the intervention effect may vary a little depending on the other aspects of care given, but not the direction of the effect. Patients in the future will almost certainly receive different forms of care from those given today and treatments shown to be effective today may be more or less effective in the future, but the direction of the effect is likely to be the same. Rather than standardise care, the CRASH-2 trial investigators believed that it was much more important to make sure that the CRASH-2 trial was large enough to detect reliably moderate but clinically important treatment effects. Indeed, if the trial was large enough, it might have sufficient statistical power to assess the overall treatment effect and examine how the effect varies according to other factors. One of the major obstacles to conducting large clinical trials is that triallists attempt to collect too much information about other interventions. The more information participating doctors are required to collect, the more burdensome the trial becomes and so they tend to recruit fewer patients. In other words, the best way to ensure that patients in the treatment and comparison group have a similar prognosis, apart from the treatment, is not to collect more information on potential confounders but to collect much less and instead randomise a greater number of patients.

Study patients

The first patient was enrolled in May 2005. The last patient was enrolled in January 2010. Adult trauma patients with significant haemorrhage [systolic blood pressure (BP) < 90 mmHg, heart rate > 110 beats per minute or both] or were considered to be at risk of significant haemorrhage and who were within 8 hours of injury, were eligible for the trial. Patients were included if the responsible doctor was substantially uncertain about whether or not to treat with TXA (i.e. entry was governed by the uncertainty principle). 16 Patients in whom the responsible doctor considered that there was a clear indication for TXA were not randomly assigned. Similarly, patients in whom there was considered to be a clear contraindication to TXA treatment were not randomly assigned. However, when the responsible doctor was substantially uncertain whether or not to treat with this agent, these patients were eligible for randomisation. The use of simple entry criteria and the uncertainty principle, as recommended in the context of large trials, allows participating doctors to use clinical judgement when deciding whether or not to enrol patients into the trial, just as in normal medical practice. This is particularly appropriate in the context of traumatic haemorrhage, where it is necessary to evaluate a range of clinical signs (also taking into account remedial measures such as fluid resuscitation) when establishing the presence or absence of major haemorrhage. For example, patients with haemorrhagic hypovolaemia can maintain a reasonable BP by vasoconstriction and about one-third of patients with traumatic haemorrhage will present with bradycardia. Although the use of clinical judgement is very likely to result in variation in the types of patients entering the trial, this heterogeneity is a scientific strength, not a weakness. If a wide range of patients are randomised (the relevant clinical characteristics will have been carefully recorded), then it may be possible for a large trial such as this one to help determine which (if any) particular types of patient are most likely to benefit from treatment. In clinical trials in trauma care it is particularly important that the inclusion criteria are simple and straightforward. Acute severe trauma is a medical and surgical emergency and the responsible doctor's primary responsibility is to provide urgent clinical care for the bleeding trauma patient. It would be inappropriate to ask doctors in this situation to consider a long list of inclusion and exclusion criteria. Fortunately, there were few absolute contraindications to the administration of TXA and so it was possible to allow a wide range of patients to be enrolled without a long list of exclusions.

It was expected from the outset that a proportion of the patients with significant haemorrhage enrolled in CRASH-2 would also have head injuries. Indeed, because haemorrhagic hypotension is an important risk factor for poor outcome in head injury, if TXA reduces haemorrhage it could improve outcome in these patients. Early TXA administration may also prevent, or limit the extent of, delayed intracranial bleeding, an important treatable cause of secondary damage after head injury. On the other hand, up to one-third of patients with head injury have computerised tomography (CT) evidence of traumatic subarachnoid haemorrhage and randomised controlled trials of TXA in patients with aneurysmal subarachnoid haemorrhage had shown that, although a 6-week course of TXA reduced the rate of rebleeding by approximately 40%, there was no overall clinical benefit because of an increase in cerebral ischaemia. Of course, the duration of TXA treatment in the CRASH-2 trial (8 hours of treatment in the acute phase while the patient is bleeding) was much shorter than the 6 weeks used in aneurysmal subarachnoid haemorrhage, in which case this should be less of a concern. Nevertheless, we elected to collect data on the occurrence of head injury in order to examine the effect of TXA administration in patients with and without head injury. We also conducted a CT scan substudy to examine the effect of TXA administration on neuroradiological and clinical outcomes in trauma patients with significant haemorrhage and traumatic brain injury. These data are reported in a separate Health Technology Assessment journal publication. 17 Data on the effects of TXA in patients with and without head injury were also shown to the Data Monitoring Committee so that the overall effect and the effect within the head injury subgroup could be monitored.

Consent procedures at participating hospitals were established by local regulation and the appropriate ethics committees. Informed consent was obtained from patients if physical and mental capacity allowed. If patients could not give consent, proxy consent was obtained from a relative or representative. If a proxy was unavailable, then, if permitted by local regulation, consent was deferred or waived. When consent was deferred or given by a proxy, the patient was informed about the trial as soon as possible and consent obtained for use of the data collected if needed.

Randomisation and masking

After eligibility had been confirmed and the locally approved consent procedures had been completed, patients were randomly assigned. Randomisation was balanced by centre, with an allocation sequence based on a block size of eight, generated with a computer random number generator. In hospitals in which telephone randomisation was not practicable, we used a local pack system that selected the lowest-numbered treatment pack from a box containing eight numbered packs. Apart from the pack number, the treatment packs were identical. The pack number was recorded on the entry form which was sent to the international trial co-ordinating centre in London, UK. Hospitals with reliable telephone access used the University of Oxford's Clinical Trial Service Unit (CTSU) telephone randomisation service. The randomisation service used a minimisation algorithm balancing for sex, age, time since injury, type of injury (blunt or penetrating), Glasgow Coma Scale (GCS) score, systolic BP, respiratory rate, central capillary refill time and country, taking into account what packs were available at that hospital. Once the treatment pack number was recorded, the patient was included in the trial whether the treatment pack was opened or the allocated treatment started. Both participants and study staff (site investigators and trial co-ordinating centre staff) were masked to treatment allocation.

Tranexamic acid and placebo ampoules were indistinguishable. TXA was manufactured by Pharmacia (Pfizer, Sandwich, UK) and the placebo by St Mary's Pharmaceutical Unit, Cardiff, UK. The treatment packs were prepared by an independent clinical trial supply company (Bilcare, Crickhowell, UK). Correct blinding and coding of ampoules was assured by independent random testing of each batch by high-performance liquid chromatography to confirm the contents. Emergency unblinding was available by telephoning CTSU.

Procedures

Patients were randomly allocated to receive a loading dose of 1 g of TXA infused over 10 minutes, followed by an intravenous (i.v.) infusion of 1 g over 8 hours or matching placebo (0.9% saline). Every patient was assigned a uniquely numbered treatment pack that contained four ampoules of either 500 mg TXA or placebo, one 100-ml bag of 0.9% saline (for use with the loading dose), a syringe and needle, stickers with the trial details and randomisation number (for attaching to infusion bags, data forms and patient medical records), and instructions. Each box contained information leaflets for patients and their representatives, consent forms and data collection forms. The stickers, instructions, leaflets and forms were in local languages.

Outcome measures and prespecified subgroup analyses

The primary outcome was death in hospital within 4 weeks of injury. Cause of death was described by the following categories: bleeding, vascular occlusion (myocardial infarction, stroke and pulmonary embolism), multiorgan failure, head injury and other. Secondary outcomes were vascular occlusive events (myocardial infarction, stroke, pulmonary embolism and deep-vein thrombosis), surgical intervention (neurosurgery, thoracic, abdominal and pelvic surgery), receipt of blood transfusion and units of blood products transfused. Dependency was measured at hospital discharge, or on day 28 if still in hospital, with the five-point Modified Oxford Handicap Scale. The scale was dichotomised into dead, dependent (fully dependent requiring attention day and night or dependent but not needing constant attention) or independent (some restriction in lifestyle but independent, minor symptoms or no symptoms). 18 Data for the use of rFVIIa and for gastrointestinal bleeding as a complication were also collected. Because the expected complications of the trial treatment were collected on the outcome form, only adverse events that were serious, unexpected and suspected to be related to the study treatment were reported separately. Outcomes were recorded if they occurred while the patient was still in hospital for up to 28 days after randomisation. Data were sent to the co-ordinating centre either electronically (by encrypted electronic data forms that could be sent by e-mail or uploaded to a secure server) or by fax, and were entered onto a central database at the trial co-ordinating centre in London, UK. We monitored the quality of the trial data using a combination of centralised statistical data checking and site visits at which patient outcome forms were compared with clinical case notes. 19

We planned to report the effects of treatment on the primary outcome subdivided by four baseline characteristics: (1) estimated hours since injury (< 1, 1–3, 3–8 hours); (2) systolic BP (≤ 75, 76–89, > 89 mmHg); (3) GCS score (severe 3–8, moderate 9–12, mild 13–15); and (4) type of injury (penetrating only or blunt, which included blunt and penetrating).

Statistical analyses

The statistical analysis plan was sent to all ethics committees and regulatory agencies before unblinding. As the risk of death might be around 20%, and even a 2% survival difference (corresponding to an RR of death with TXA of 0.9) would be important, a trial of 20,000 patients was planned, which would then have an 85% chance of achieving a two-sided p-value of < 0.01 and a 95% chance of a two-sided p-value of < 0.05. All analyses were undertaken on an intention-to-treat basis. For each binary outcome we calculated RRs and 95% CIs, and two-sided p-values for statistical significance. The RR gives the number of times more likely (RR > 1) or less likely (RR < 1) an event is to happen in the TXA group than in the placebo group. For analysis of the prespecified subgroups (primary outcome only) we calculated RRs with 99% CIs with two-sided p-values. Heterogeneity in treatment effects across subgroups was assessed with chi-squared tests. We prespecified that, unless there was strong evidence (p < 0.001) against homogeneity of effects, the overall RR would be considered the most reliable guide to the approximate RRs in all subgroups. Means and standard deviations (SDs) were estimated for count outcomes, and we calculated two-sided p-values of the difference in means of logarithms. A complete case analysis, including only cases for which the relevant outcome data were available, was undertaken. There was no imputation for missing data. During the study, unblinded interim analyses were supplied by an independent statistician to the Data Monitoring Committee.

Role of the funding source

The trial was funded by the UK National Institute for Health Research Health Technology Assessment programme, Pfizer, the Bupa Foundation and the J P Moulton Charitable Foundation. The funders of the study had no role in study design, data collection, data analysis, data interpretation or writing of the report. The Writing Committee had full access to all data in the study and had final responsibility for the decision to submit for publication.

Introduction

The CRASH-2 trial was motivated by the evidence that TXA reduces bleeding in patients undergoing elective surgery, and the hypothesised mechanism was inhibition of fibrinolysis leading to improved clinical effectiveness of haemostasis. 20 However, as seen in the previous chapter, no significant difference was recorded in transfusion requirements between the TXA and placebo groups, and the CRASH-2 trial did not measure the effect of this drug on fibrinolytic assays. Thus, an alternative hypothesis is that TXA might act by reducing the pro-inflammatory effects of plasmin rather than by improving haemostasis. 21 In order to examine this issue, we conducted the same prespecified subgroup analyses as we did for all cause mortality, but for the outcome that we hypothesised would be most affected by TXA, specifically mortality due to bleeding.

Methods

We examined the effect of the trial treatment on death due to bleeding subdivided by four baseline characteristics: (1) time from injury to treatment (≤ 1, > 1 to ≤ 3 and > 3 hours); (2) severity of haemorrhage as assessed by systolic BP (≤ 75, 76–89 and > 89 mmHg); (3) GCS score (severe 3–8, moderate 9–12 and mild 13–15); and (4) type of injury (penetrating only, blunt plus blunt and penetrating). These were the same subgroup analyses that were reported in the previous chapter, but for the outcome of death due to bleeding rather than for all-cause mortality.

Heterogeneity in treatment effects across subgroups was assessed by a chi-squared test. We had prespecified that, unless there was strong evidence against the null hypothesis of homogeneity of effects (i.e. p < 0.001), the overall RR would be considered the most reliable guide to the approximate RRs in all subgroups. To test the independence of any observed treatment interactions we ran a logistic model including all possible interactions in the four prespecified baseline characteristics and treatment subgroups.

A logistic regression was estimated with death due to bleeding as the dependent variable, with treatment group and time to treatment as explanatory factors. We included an interaction parameter to allow for a proportional change in the odds ratio (OR) as time to treatment increases. ORs and 95% CIs were estimated for different times to treatment. CIs were calculated with a logistic model with time as a continuous term and an interaction term between time and TXA. We also ran a model with an interaction term for time to treatment squared to allow for a non-constant proportional change in the OR.

Results

Of the 3076 deaths from all causes, death due to bleeding accounted for 1063 (35%). The risk of death due to bleeding was significantly reduced with TXA. A total of 489 of 10,060 (4.9%) patients died because of bleeding in the TXA group compared with 574 of 10,067 (5.7%) patients in the placebo group (RR 0.85; 95% CI 0.76 to 0.96; p = 0.0077). We noted no significant effect on the risk of death for all other (non-bleeding) causes combined (Table 4).

Table 5 shows the baseline characteristics of patients according to time to treatment.

Figure 3 shows the results of the subgroup analyses for death due to bleeding. Time to treatment was unknown in nine participants. Treatment given ≤ 1hour from injury significantly reduced the risk of death due to bleeding [198 out of 3747 patients (5.3%) died in the TXA group vs 286 out of 3704 patients (7.7%) in the placebo group: RR 0.68; 95% CI 0.57 to 0.82; p < 0.0001]. Treatment given between 1 and 3 hours also reduced the risk of death due to bleeding [147 out of 3037 patients (4.8%) died in the TXA group vs 184 out of 2996 patients (6.1%) in the placebo group: RR 0.79; 95% CI 0.64 to 0.97; p = 0.03]. Treatment given > 3 hours after injury significantly increased the risk of death due to bleeding [144 out of 3272 patients (4.4%) died in the TXA group vs 103 out of 3362 patients (3.1%) in the placebo group: RR 1.44; 95% CI 1.12 to 1.84; p = 0.004].

FIGURE 3.

Mortality due to bleeding by subgroups.

We recorded strong evidence that the effect of TXA on death due to bleeding varied according to time from injury to treatment (p < 0.0001). The evidence for interaction remained strong even after adjustment for interactions between the other prespecified baseline characteristics and treatment (p < 0.0001; data not shown).

The estimated OR of TXA on death due to bleeding when given immediately after injury was 0.61 (95% CI 0.50 to 0.74). We estimated that this OR is multiplied by 1.15 (95% CI 1.08 to 1.23) for every hour that passes since the injury. Figure 4 shows how the OR and 95% CIs vary with time to treatment. The interaction term for time to treatment squared was not significant (OR = 0.99; p = 0.38).

FIGURE 4.

Effect of TXA on death due to bleeding by time to treatment. Shaded area represents 95% CI.

We recorded no evidence of heterogeneity for the subgroup analyses according to systolic BP, GCS score at randomisation or type of injury (see Figure 3). We detected no evidence of heterogeneity in the effect of TXA on the risk of non-bleeding deaths (see Table 4).

Introduction

Although the primary objective of the trial was to assess the clinical effectiveness of TXA, there is a widespread recognition that, should TXA prove to be a clinically effective intervention, information is also required regarding its cost-effectiveness. Given that TXA appears to be clinically effective, its cost-effectiveness might appear obvious given that it is a relatively inexpensive intervention. However, in countries with limited health-care budgets, it is very important to establish the ratio of incremental cost to incremental effect. In the absence of any evidence of heterogeneity in the treatment effect between countries, both the cost of TXA and the effectiveness of TXA [in terms of life-years (LYs) gained] will still depend on the setting of analysis. In low-income countries (LICs) and in middle-income countries (MICs) life expectancies are shorter than in high-income countries (HICs). Consequently, any gain in LYs associated with TXA will tend to be lower than in HICs. The incremental cost of a policy of giving TXA will also tend to be higher in HICs than in both MICs and LICs (see Appendix 6).

The objective of the economic evaluation is to assess the cost-effectiveness of TXA for the treatment of significant haemorrhage following trauma in three different health-care settings: Tanzania [LIC with gross domestic product (GDP) per capita $509], India (MIC with GDP per capita $1134) and the UK (HIC with GDP per capita $35,165). 22

Methods

The cost-effectiveness of TXA was measured in terms of the additional costs per LY gained. A lifetime Markov model was developed in order to estimate the LYs gained as a result of the intervention. In each cycle, assumed to last 1 year, patients are either alive or dead. LYs gained were discounted at an annual rate of 3.5% in accordance with the National Institute for Health and Clinical Excellence recommendation [and broadly consistent with the 3% rate used for disability-adjusted life-years (DALYs)]. The number of LYs saved for each death averted by TXA depends on the age and sex of the patients. In the CRASH-2 trial, consistent with the statistics reported by Norton et al. ,23 the majority of the trauma victims were aged between 16 and 35 years, and only a small proportion were female. In order to estimate the incremental LYs gained by TXA, five age groups were identified, each containing about one-fifth of the trial participants: 16–20 years, 21–25 years, 26–34 years, 35–50 years and > 50 years. The mean age of trial participants in each age group was calculated and the model was then run for cohorts with these starting ages: 18 years, 22 years, 30 years, 42 years and 75 years old. The overall number of LYs saved in each country was then estimated as a weighted average of the results for the five age groups using the age distribution for trial participants in countries of the relevant income group (low, middle or high). All analyses were carried out using Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) and Stata 11 (StataCorp LP, College Station, TX, USA).

Since the CRASH-2 trial recorded data up to 28 days or death, a parametric survival function was fitted to extrapolate mortality experience over the 12 months following injury. Four different parametric survival functions (Weibull, Gompertz, log-logistic and log-normal) were fitted to data from the placebo arm and the best-fitting model was selected using the Akaike information criterion (AIC). Cox–Snell residuals were plotted as a confirmatory test. 24 Both the AIC test and the Cox–Snell residuals suggest that the Gompertz parametric model fits the data best. The AIC scores were 10,088 (Weibull), 10,065 (log-logistic), 9,954 (log-normal) and 9,425 (Gompertz). The four functions and the Cox–Snell plots are reported in Appendix 6.

The Gompertz survival function,24 which has been extensively used in medical research to model mortality data, has the following cumulative hazard rate:

where γ determines if the hazard function increases with time (if γ is positive) or decreases (if γ is negative). If γ is equal to zero, the hazard function follows an exponential model. t is the time frame over which the cumulative probability of dying is estimated.

In the analyses, three covariates were explored: age, sex and GDP. When covariates are considered in the analysis, λ is given by the following equation:24

Using CRASH-2 data the cumulative hazard rate is:

As expected, γ (–0.20; 95% CI –0.21 to –0.18) is negative, implying that after trauma the hazard rate decreases over time. The probability of dying increases with age (age coefficient = 0.020; 95% CI 0.016 to 0.024), whereas sex was not found to be influential for the hazard rate (–0.06; 95% CI –0.22 to 0.11). A GDP per capita was assigned to each country in the trial according to the latest World Bank estimates and two binary variables, x₂ and x₃, were constructed to estimate whether or not the likelihood of death changes according to the GDP. 22x₂ took the value 1 for LICs and 0 otherwise. Similarly, x₃ took the value of 1 for MICs and 0 otherwise. GDP coefficients (β₂ = −0.31; β₃ = −0.61) were found to be highly significant. Thus, the function for λ is:

Figure 5 shows the hazard rate function during the first year after trauma observed in the placebo group. The hazard rate decreases almost to zero in the first 40 days after hospital admission and remains constant for the rest of the year. This finding is consistent with previous studies which suggest that the majority of trauma-related deaths occur within a few weeks after injury. The figure also validates the model assumption that, after 1 year, the baseline risk of death is the same observed in the general population. 25

FIGURE 5.

Gompertz regression. Hazard rate function during the first year after trauma observed in the placebo group.

Risk of death during the first year following trauma in the TXA group was estimated by multiplying the cumulative hazard for the placebo group by the RR reduction in all-cause mortality estimated by the CRASH-2 trial (RR = 0.87; 95% CI 0.81 to 0.95). Beyond 12 months, the risk of death is assumed to be equal whether or not the patient received TXA, and is set equal to the risk of death for the relevant age–sex group in the general population.

The evaluation was conducted from a health services' perspective. 26 Resource use in the TXA and placebo arms was compared. As a consequence, two cost items were considered in the present study: (1) the cost of administering TXA to bleeding trauma patients; and (2) the incremental cost of non-intensive care hospital stay. There was no evidence from the trial of any other differences in resource use, such as time spent in intensive care, the frequency of transfusion or the volumes transfused. Costs have been converted from national currencies into international dollars ($) using purchasing power parities (Organisation for Economic Co-operation and Development and Penn World Table). 27 Where necessary, the US Consumer Price Index was used to inflate prices. 28 All costs are expressed in 2008–9 values. Costs were not discounted, as the costs associated with giving TXA occur within the year following trauma.

One-way sensitivity analysis was undertaken in order to assess the impact of uncertainty regarding the input parameters [RR of death with TXA, cost of TXA, cost of additional non-intensive care unit (ICU) stay and cost per non-ICU day] and the choice of cumulative hazard function. The incremental cost per LY saved for TXA compared with no TXA was estimated using the 95% confidence limits for the RR of death with TXA observed in the trial (0.81 and 0.95). 29 In order to evaluate the effect of variation in the price of TXA on the incremental cost-effectiveness ratio (ICER), the price was assumed to vary between $2.57/g (from Casati et al. 30) and $45.67/g (from Eaton31). One-way sensitivity analysis was also conducted on the increase in non-ICU hospital stay following TXA administration, again using the confidence limits from the trial (0.007 and 0.08). For the cost of a non-ICU day, the lowest estimates, for both Tanzania and India, were the cost per day in a primary health centre ($9.84 and $18.75, respectively) whereas the highest estimates ($18.69 and $35.41) were obtained from tertiary hospitals. 32,33 For the UK, the lower and the upper estimates of cost per non-ICU day ($90–784) were taken from UK reference costs. 34

A further deterministic sensitivity analysis was performed to investigate how the results change if different parametric distributions are adopted (Weibull, log-normal and log-logistic) when estimating survival beyond 28 days.

In the probabilistic sensitivity analysis, uncertainty in the data was captured by fitting probability distributions to each parameter (see Appendix 6 for details). Beta and log-normal distributions were used for binomial data and RR parameters, respectively. For costs a ‘right-tailed’ γ distribution was adopted. Finally, to account for correlation among the parameters of the Gompertz survival function a Cholesky decomposition of the covariance matrix was performed. For each country, 1000 second-order Monte Carlo simulations of the expected incremental cost, LYs saved and net benefits of TXA were produced using the following formula:

where willingness to pay (WTP) is the WTP per LY saved, ΔE is the incremental number of LYs saved and ΔC is the incremental cost of the TXA. Cost-effectiveness acceptability curves (CEACs) were drawn for each of the three countries to show the probability that TXA is cost-effective for WTP values from $0 to $250 per LY saved.

Economic evaluation results

Probabilistic sensitivity analysis

Figure 6 presents the CEACs for Tanzania, India and the UK. The CEACs are similar for India and the UK, whereas Tanzania is more likely to be cost-effective even for lower WTP values because of the lower incremental cost of TXA. For example, when WTP is $100 per LY saved, about 80% of simulations produce a positive net benefit in India and in the UK, whereas in Tanzania the corresponding probability is close to 100%.

FIGURE 6.

Willingness to pay per LY saved.

Can the result of the CRASH-2 trial be generalised?

There is widespread misunderstanding about the scientific basis for generalisation of the result of medical research. Some clinicians believe that the results of medical research can be generalised only to populations that are similar to those that were studied. This leads to a misplaced emphasis on the extent to which the study population is representative of the target population, to which the results are to be generalised. So, for example, when considering whether or not the results of the CRASH-2 trial, in which a large proportion of the randomised patients were recruited from MICs, can be applied to patients in the NHS, they consider whether or not the clinical care patients received in the trial was similar to that of patients in the NHS. As it is reasonably easy to imagine at least some differences, it is tempting to conclude that the results cannot be generalised. If this argument were valid, it would have dramatic implications for medical care. Since medical care is constantly changing as new forms of care are introduced, it would mean that all clinical trials would need to be repeated at regular intervals to update our medical knowledge to take account of the changing clinical situation. Fortunately, at least in most cases, this is not the case.

The first prerequisite for scientific generalisation is that the results to be generalised are valid and appropriately precise. Results from large-scale clinical trials, such as the CRASH-2 trial, that provide precise estimates of modest treatment effects meet this first criterion. Next, one has to make a scientific judgement about how the treatment worked and whether or not the mechanism of action is likely to be modified by other factors. In the case of the CRASH-2 trial, TXA is a drug that reduces fibrinolysis by inhibition of the blood enzyme plasmin. At least at present, there is no scientific reason to expect that TXA would work differently in different populations of the same species. The basic pathophysiology of fibrinolysis is going to be more or less the same in all members of the human species and it is likely that TXA would inhibit plasmin in a similar way. Of course, we cannot be completely sure that TXA worked through the inhibition of fibrinolysis, but then again we have to bear in mind that any proposed mechanism of action is a causal theory. At least at the moment, inhibition of fibrinolysis seems to be the most reasonable causal theory. In summary, we observe that TXA reduces bleeding in surgical patients most of whom are from HICs, we conduct a large trial and find that TXA reduces mortality in a heterogeneous group of bleeding trauma patients, we suppose that the mechanism of action of TXA is something to do with the inhibition of the enzyme plasmin, that all humans share, so we can reasonably conclude that TXA is likely to work in trauma patients in, for example, the USA, even though there were no patients from the USA included in the CRASH-2 trial.

Of course, at any given level of injury severity, the risk of death following severe trauma in the UK or USA might be different to the risk of death following trauma in, say, Tanzania. But this is not the point. The point is whether or not TXA can reasonably be expected to reduce the risk of death in all these settings. Experience from other large-scale international trials has shown that RRs appear to be remarkably homogeneous even when baseline risk varies. In other words, a treatment that reduces the risk of death by about one-third would reduce the risk of death from 30% to about 20% and from 3% to about 2%. Following the publication of the results of the CRASH-2 trial, several clinicians asked us to examine the extent to which the effect of TXA on the risk of death due to bleeding varied by geographical region. In response, the hospitals participating in the CRASH-2 trial were grouped into four geographical regions: (1) Africa, (2) Asia, (3) Europe, North America and Australia, and (4) South America. We examined the extent to which the effect of TXA on death due to bleeding varied across continents. We found no evidence for heterogeneity in the effect of TXA by region (χ² = 1.445; p = 0.70).

Cost-effectiveness

Early administration (within 3 hours) of TXA would cost $48, $66 and $64 per LY saved in Tanzania, India and the UK, respectively. According to the WHO Commission on Macroeconomics, health-care interventions costing less than GDP per capita per DALY averted should be considered ‘very cost effective’. 51 DALYs are a measure of the years of life lost from disease and years lived with a disability. According to the World Bank classification, GDP per capita in LICs ranges from $380 to $975, in lower MICs between $976 and $3855, in upper MICs between $3856 and $11,905, and in HICs > $11,906. 52 Thus, if the LYs saved by TXA are spent in perfect health (one LY saved is equal to one DALY averted) then TXA is a highly cost-effective intervention in all the countries considered. The sensitivity analyses presented reinforce this conclusion.

However, a number of limitations should be considered when interpreting these results. It was necessary to model survival over 12 months using data for the first 28 days following the trauma, and different statistical models will produce different estimates of benefit. However, different models were explored and the one selected, as well as providing the best fit to the data, also produced more realistic estimates of the impact on long-term survival. In this evaluation, those predicted to survive the first year following the trauma are assumed then to have the same life expectancy as members of the general population of similar age and sex.

A related limitation is that the measure of incremental cost-effectiveness is cost per LY gained rather than cost per quality-adjusted life-year (QALY; a year in perfect health is considered equal to one QALY) gained or DALY averted. Depending on the extent to which these patients do not enjoy perfect health, the cost per QALY gained or the cost per DALY averted will be higher than the cost per LY saved.

A further potential limitation is that the analysis does not allow for future health service savings. The CRASH-2 trial showed that after 28 days the proportion of patients reporting no symptoms at discharge was significantly higher in the TXA group than in the placebo group. 29 If TXA patients are more likely to survive without disability, then this study undervalues the potential cost saving arising from the administration of TXA as healthier people will have lower future utilisation of health-care services.

Identifying cost-effective interventions to decrease the number of injury-related deaths is a major public health challenge. According to the Disease Control Priority Project, in 2001 unintentional injuries alone accounted for 6% of all deaths worldwide. 23 The majority of unintentional injury-related deaths, > 90%, occur every year in LICs and MICs, posing a further financial burden on these countries' economies and underfinanced health-care systems. 23

This analysis suggests that TXA is a highly cost-effective intervention in three very different settings. Given that TXA is clinically effective and it is of relatively low cost, it is not surprising that it is cost-effective in high-income settings. However, it was important to demonstrate that it was likely to be cost-effective in countries with much more limited health-care budgets, such as in Tanzania and India, where, owing to the high numbers of trauma victims, this simple intervention can avert thousands of deaths every year. Further research is needed to evaluate the effects of TXA on the quality of life of trauma patients.

Future clinical research

The knowledge that TXA is a highly cost-effective treatment to reduce the risk of death from traumatic bleeding raises the possibility that it might also be effective in other situations in which bleeding can be life-threatening or disabling. Traumatic brain injury is commonly accompanied by intracranial bleeding, which can develop or worsen after hospital admission. Traumatic intracranial haemorrhage is associated with an increased risk of death and disability, and, irrespective of location, haemorrhage size is strongly correlated with outcome. If TXA reduced intracranial bleeding after isolated traumatic brain injury, then patient outcomes might be improved. The Clinical Randomisation of an Antifibrinolytic in Significant Head injury-3 (CRASH-3) trial will assess the effect of TXA on patient outcomes after traumatic brain injury (http://crash3.lshtm.ac.uk/).

Tranexamic acid might also have a role in bleeding conditions apart from traumatic injury. Postpartum haemorrhage is a leading cause of maternal mortality, accounting for about 100,000 maternal deaths every year. 53 Although evidence suggests that this drug reduces postpartum bleeding, the quality of the existing trials is poor and none have been large enough to assess the effect of TXA on end points that are important to women. 54 A large trial is currently under way to assess the effect of TXA on the risk of death and hysterectomy in women with postpartum haemorrhage. 53

Contribution of authors

Ian Roberts (Professor of Epidemiology) was involved in the design, conduct, analysis and reporting phases.

Haleema Shakur (Senior Lecturer in Clinical Trials) was involved in the design, conduct, analysis and reporting phases.

Tim Coats (Professor of Emergency Medicine) was involved in the design, conduct, analysis and reporting phases.

Beverley Hunt (Professor of Haematology) was involved in the design, conduct, analysis and reporting phases.

Eni Balogun (Assistant Trial Manager) was involved in the acquisition of data, analysis and reporting phases.

Lin Barnetson (Senior Data Manager) was involved in the acquisition of data, analysis and reporting phases.

Lisa Cook (Assistant Trial Manager) was involved in the acquisition of data, analysis and reporting phases.

Taemi Kawahara (Trial Manager) was involved in the acquisition of data, analysis and reporting phases.

Pablo Perel (Clinical Lecturer) was involved in the conduct, analysis and reporting phases.

David Prieto-Merino (Lecturer) was involved in the analysis and reporting phases.

Maria Ramos (Senior Trials Administrator) was involved in the design, conduct, analysis and reporting phases.

John Cairns (Professor of Health Economics) conducted the economic analysis.

Carla Guerriero (Research Fellow) conducted the economic analysis.

Department of Health disclaimer: the views and opinions expressed herein are those of the authors and do not necessarily reflect those of the Department of Health.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.

Publications

CRASH-2 trial collaborators. Effects of tranexamic acid on death, vascular occlusive events, and blood transfusion in trauma patients with significant haemorrhage (CRASH-2): a randomised, placebo-controlled trial. Lancet 2010;376:23–32.

CRASH-2 trial collaborators. The importance of early treatment with tranexamic acid in bleeding trauma patients: an exploratory analysis of the CRASH-2 randomised controlled trial. Lancet 2011;377:1096–101.

Steering Committee

Professor Ian Franklin (chairperson), University of Glasgow and Scottish Blood Transfusion Service.

Ms Brigitte Chaudhry, RoadPeace.

Professor Tim Coats, University of Leicester.

Dr Charles Deakin, Southampton General Hospital.

Dr Steve Goodacre, University of Sheffield.

Dr Beverley J Hunt, Guy's and St Thomas' Hospital NHS Trust.

Dr David Meddings, World Health Organization.

Professor Sir Richard Peto, University of Oxford.

Professor Ian Roberts, LSHTM.

Professor Peter Sandercock, University of Edinburgh.

Data Monitoring Committee

Professor Rory Collins (chairperson), Professor Adrian Grant and Professor John A Myburgh.

Trial Coordinating Centre Team

Ian Roberts (clinical co-ordinator, chief investigator), Haleema Shakur (trial manager), Pablo Perel (regional co-ordinator), Lin Barnetson (data manager), Maria Ramos (trial administrator), Lisa Cook (assistant trial manager, regional co-ordinator from 2007), Taemi Kawahara (assistant trial manager, regional co-ordinator from 2007), Eni Balogun (regional co-ordinator from 2006), Matthew Berle (trial assistant from 2007), Collette Barrow (assistant administrator from 2008), Tony Brady (programmer to 2006), Chris Rubery (data assistant from 2009), Jackie Wayte (UK nurse co-ordinator from 2008) and Cynthia To (data assistant from 2007 to 2009).

Steering Committee – Ian Franklin (chairperson), Brigitte Chaudhry, Tim Coats, Charles Deakin, Steve Goodacre, Beverley Hunt, David Meddings, Richard Peto, Ian Roberts and Peter Sandercock.

CRASH-2 trial collaborators by country

Albania National Trauma Centre Hospital: Fatos Olldashi, Mihal Kerçi, Tefik Zhurda and Klotilda Ruçi; Spitali Civil Durres: Arben Banushi.

Argentina Hospital Ángel Cruz Padilla: Mario Sardón Traverso and Juan Jiménez; Hospital Regional Rio Grande: Jorge Balbi; Hospital ‘4 de Junio’ Dr Ramon Carrillo: Christian Dellera; Hospital Castro Rendón: Silvana Svampa; Hospital San Martín de La Plata: Gustavo Quintana; Hospital Municipal de Agudos ‘Dr Leonídas Lucero’: Gustavo Piñero; Hospital Interzonal General de Agudos ‘Dr Oscar Alende’: Jorge Teves.

Australia Nepean Hospital: Ian Seppelt; Sir Charles Gairdner Hospital: David Mountain; John Hunter Hospital: Zsolt Balogh.

Bangladesh United Hospital Limited: Maniruz Zaman.

Belgium Sint-Vincentius Hospital: Patrick Druwé and Robert Rutsaert; Centre Hospitalier Regional de Namur: Guy Mazairac.

Cameroon Tombel District Hospital: Fogang Pascal, Zognou Yvette and Djeuchon Chancellin; St Theresa's Catholic Medical Centre: Patrick Okwen; Bamenda Provincial Hospital: Jules Djokam-Liapoe; Bali District Hospital: Ernest Jangwa; Bafut District Hospital: Lawrence Mbuagbaw; Fundong District Hospital: Ninying Fointama; St John of God Medical Centre: Nguemo Pascal.

Canada Hamilton General Hospital: Frank Baillie.

China Renji Hospital: Ji-yao Jiang, Guo-yi Gao and Yin-hui Bao.

Colombia Hospital Universitário San Vicente de Paúl, Universidad de Antioquia: Carlos Morales, Juan Sierra, Santiago Naranjo, Camilo Correa and Carolina Gómez; Hospital Universitário San Jose de Popayan: Jorge Herrera, Liliana Caicedo, Alexei Rojas, Henry Pastas and Hugo Miranda; Hospital Pablo Tobon Uribe: Alfredo Constaín, Mayla Perdomo, Diego Muñoz, Álvaro Duarte and Edwin Vásquez; Hospital San Andrés de Tumaco: Camilo Ortiz, Bernardo Ayala, Hernán Delgado, Gloria Benavides and Lorena Rosero; Fundación Clínica Valle del Lili: Jorge Mejía-Mantilla, Ana Varela, Maríaisabel Calle, José Castillo and Alberto García; Clínica las Americas: Juan Ciro, Clara Villa and Roberto Panesso; Hospital General de Medellin: Luz Flórez and Argemiro Gallego; Hospital San Felix ESE: Fabián Puentes-Manosalva, Leonor Medina and Kelly Márquez; Hospital Universitário del Caribe: Adalgiza Reyes Romero, Ricardo Hernández and Julio Martínez; Hospital Universitário San Jorge: Wilson Gualteros; Hospital San Rafael Tunja: Zulma Urbina and Julio Velandia; Clínica La Estancia SA: Federico Benítez and Adolfo Trochez; Fundación Hospital San José de Buga: Andrés Villarreal and Pamela Pabón; Hospital Civil de Ipiales: Hernán Delgado; Hospital Universitário Departamental Nariño: Héctor López; Hospital Universitário del Valle: Laureano Quintero; Hospital Universitário de Neiva: Andrés Rubiano; Hospital Manuel Uribe Ángel: Juan Tamayo.

Cuba Hospital Clínico-Quirúrgico Docente ‘Saturnino Lora’: Marjoris Piñera, Zadis Navarro, Deborah Rondón and Bárbara Bujan; Hospital General Universitário ‘Carlos Manuel de Céspedes’: Leonel Palacios, Daymis Martínez, Yalisa Hernández and Yaimara Fernández; Hospital Provincial Docente ‘Manuel Ascunce Domenech’: Eugenio Casola; Hospital Universitário ‘Arnaldo Milián Castro’: Rodolfo Delgado, Carlos Herrera, Migdacelys Arbolaéz and Mario Domínguez; Hospital Universitário ‘Dr Gustavo Aldereguía Lima’: Marcos Iraola, Omar Rojas and Alba Enseñat; Hospital Abel Santamaría Cuadrado: Irene Pastrana, Daniel Rodríguez and Sergio Álvarez de la Campa; Hospital Miguel Enríquez: Thorvald Fortún; Hospital General Calixto García: Martha Larrea; Hospital Antonio Luaces Iraola: Lensky Aragón; Hospital Provincial Docente VI Lenin: Aida Madrazo.

Czech Republic Research Institute for Special Surgery and Trauma: Petr Svoboda.

Ecuador Hospital Luis Vernaza: Mario Izurieta, Alberto Daccach, Mónica Altamirano, Antonio Ortega, Bolívar Cárdenas and Luis González; Hospital José Carrasco Arteaga: Marcelo Ochoa, Fernando Ortega, Fausto Quichimbo and Jenny Guiñanzaca; Hospital de Niños Dr Roberto Gilbert Elizalde: Ines Zavala and Sayra Segura; Hospital Naval Guayaquil: Johnny Jerez; Hospital Alcivar: Daniel Acosta; Hospital ‘Dr Rafael Rodríguez Zambrano’: Fabián Yánez; Clínica De Especialidades Medicas ‘San Gregorio’: Rubén Camacho.

Egypt Mataria Teaching Hospital: Hussein Khamis, Hossam Shafei, Ali Kheidr, Hani Nasr, Moetaz Mosaad and Safwat Rizk; Suez Canal University: Hesham El-Sayed, Taha Moati and Emad Hokkam; Aswan Teaching Hospital: Mamdouh Amin, Hany Lowis, Medhat Fawzy, Nabil Bedir and Mohamed Aldars.

El Salvador Hospital Nacional Rosales: Virginia Rodríguez, Juan Tobar and Jorge Alvarenga.

Georgia Tbilisi State University Clinical Hospital ‘I Javakhishvili’: Budu Shalamberidze, Elza Demuria, Nikoloz Rtveliashvili, Gocha Chutkerashvili and David Dotiashvili; Tbilisi First Hospital, University Clinic, Neurosurgery Center: Tamar Gogichaishvili, George Ingorokva, David Kazaishvili, Besik Melikidze and Natia Iashvili; Tbilisi City Hospital #1: Gia Tomadze, Manana Chkhikvadze, Leri Khurtsidze, Zviad Lomidze and Diana Dzagania; Tbilisi State Medical University ER Department: Nikoloz Kvachadze, Giorgi Gotsadze and Vakhtang Kaloiani; Institute of Critical Care Medicine: Nino Kajaia.

Ghana Korle Bu Teaching Hospital: Jonathan Dakubo, Simon Naaeder and Priscilla Sowah; Nyinahin Government Hospital: Adamu Yusuf and Alhaji Ishak; Sogakope District Hospital: Paul Selasi-Sefenu; Methodist Hospital Wenchi: Ballu Sibiri; Effia Nkwanta Regional Hospital: Sampson Sarpong-Peprah; Saint Theresa's Hospital: Theodore Boro.

India Medical Trust Hospital Kochi: Kanjithanda Bopaiah, Kishore Shetty, Raja Subbiah, Lukman Mulla and Anand Doshi; Christian Medical College Ludhiana: Yashbir Dewan, Sarvpreet Grewal, Pradipta Tripathy, Jacob Mathew and Bharat Gupta; Aditya Neuroscience Centre: Anil Lal and Majulie Choudhury; Sri Sai Hospital: Sanjay Gupta, Smita Gupta and Arun Chug; Care Hospital: Venkataramana Pamidimukkala, Palaniappan Jagannath, Mohan Maharaj, Ramaraju Vommi and Naresh Gudipati; North Bengal Neuro Research Centre: W H Chhang; Sheth VS General Hospital and NHL Municipal College: Pankaj Patel, Nilay Suthar, Deepa Banker and Jyotish Patel; LTM Medical College and General Hospital: Satish Dharap, Ranjeet Kamble, Shraddha Patkar and Sushil Lohiya; Government Medical College and Associated Hospitals Jammu: Rakesh Saraf, Dinesh Kumar, Satish Parihar and Rahul Gupta; MKCG Medical College: Rasananda Mangual, Alagumuthu, Don Kooper and Chinmaya Mohapatra; Christian Medical College Hospital Vellore: Suresh David and Wesley Rajaleelan; KLE Hospital and Medical Research Centre: Ashok Pangi, Vivek Saraf and Santhosh Chikareddy; NKP Salve Institute of Medical Sciences and Lata Mangeshkar Hospital: Sushil Mankar, Anil Golhar, Rahul Sakhare and Nilesh Wagh; Sanjivani Diagnostics and Hospital: Anil Lal, Dhiman Hazarika; Parkar Hospital: Pratyush Chaudhuri; Jeevan Jyoti Hospital and Research Centre: Prakash Khetan; Mansarovar Hospital: Govindbhai Purohit, Yogesh Purohit and Mandakini Pandya; Postgraduate Institute of Medical Science Rohtak: Rakesh Gupta, Shashi Kiran and Saurab Walia; Goyal Hospital Jalna: Sonam Goyal, Sidhant Goyal and Satish Goyal; Government Medical College Chandigarh: Sanjay Gupta, Ashok Attri and Rajeev Sharma; Oberai Hospital: Ashok Oberai, Mahesh Oberai and Supriya Oberoi; Rajeev Gandhi Memorial Hospital and Research Centre: Gajendra Kant Tripathi; Calicut Medical College Hospital: Vijayan Peettakkandy, Premkumar Karuthillath and Pavithran Vadakammuriyil; Krishnamai Medical and Research Foundation's NIKOP Hospital: Jalindar Pol, Sunita Pol and Manisha Saste; St Stephen's Hospital: Subrat Raul, Shashi Tiwari and Neileino Nelly; Government Rajaji Hospital: M Chidambaram; Medical College Trivandrum: Viswanathan Kollengode and Sam Thampan; Sanjeevani Hospital: Sunder Rajan and Sushrut Rajan; Kamineni Hospital: Subodh Raju and Renuka Sharma; Sri Sakthi Hospital: Subbiah Venkatesh Babu and Chellappa Sumathi; Bhattacharya Orthopaedic and Related Research Centre: Protyush Chatterjee and Alok Agarwal; Sushrut Hospital: Hemant Magar and Meera Magar; All India Institute of Medical Sciences: Manmohan Singh and Deepak Gupta; GM Hospital (P), Ltd: Anil Lal and Kamal Haloi; Government Medical College and Superspeciality Hospital Nagpur: Varsha Sagdeo and Pramod Giri; Government Medical College New Civil Hospital: Nimesh Verma, Ravi Jariwala and Ashish Goti; Chikitsa Hospital: Aman Prabhu-Gaonkar and Sagar Utagi; Apollo Health City: Mahesh Joshi and Ruchit Agrawal; Apex Neurotrauma and Superspeciality Hospital: Gopal Sharma and Gurvinder Saini; Neuro Centre Gola Ghat: Vinod Tewari; NSCB Medical College: Yad Yadav and Vijay Parihar; BGS Global Hospital: Neelam Venkataramana and Shailesh Rao; Chettinad Hospital and Research Institute: Narayana Reddy and SG Chander; Sir Sayajirao General Hospital and Medical College Baroda: Virsing Hathila; Goyal Hospital and Research Centre Jodhpur: Vithal Das; Krishna Surgical Hospital and Trauma Care Centre: Kantibhai Agaja; Nizam's Institute of Medical Sciences: Aniruddh Purohit; Niramay Hospital: Akilesh Lahari; Apex Hospital Bhopal: Rajesh Bhagchandani; Dr Jeyasekharan Medical Trust: Bala Vidyasagar; Himalayan Institute of Medical Sciences: P K Sachan; Apollo Gleneagles Hospitals: Tanmoy Das; Civil Hospital Gandhinagar: Sharad Vyas; Sukhdev Raj Soin Hospital: Sujoy Bhattacharjee; Sancheti Institute for Orthopaedics and Rehabilitation: Parag Sancheti; St James Hospital: T Manoj; Al Shifa Hospital: Mubarak Moideen; Anant Institute of Medical Sciences: Kailash Pansey; Vinayaka Mission Hospital: V P Chandrasekaran; Gauhati Medical College and Hospital: Kabul Saikia; Krishna Hospital and Medical Research Centre: Hoshedar Tata; Ruby Hall Clinic: Sanjay Vhora; Shreejee Hospital: Aniket Shah; Nazareth Hospital: Gordon Rangad; Ganga Hospital: S Rajasekaran; Makurdi: Andrea Jogo and Terna Bitto; Nnamdi Azikiwe University Teaching Hospital: Stanley Anyanwu and Okechukwu Mbonu; Lagos State University Teaching Hospital: Mobolaji Oludara and Michael Somoye; Usmanu Danfodiyo University Teaching Hospital: Bello Shehu and Nasir Ismail; National Orthopaedic Hospital Enugu: Amechi Katchy; University of Calabar Teaching Hospital: Rowland Ndoma-Egba and Ngim Grace-Inah; University of Abuja Teaching Hospital: Zumnan Songden and Abdulrahman Abdulraheem; University of Uyo Teaching Hospital: Akpan Out and Timothy Nottidge; Federal Medical Centre, Yenagoa: Domingo Inyang and David Idiapho; Seventh Day Adventist Hospital: Herb Giebel; Federal Medical Centre Birnin-Kebbi: Ramatu Hassan; Abia State University Teaching Hospital: Adeyinka Adisa; Wesley Guild Hospital: Akinbolaji Akinkuolie; Federal Medical Centre, Umuahia: Kalu Okam; University of Maiduguri Teaching Hospital: Abubakar Musa; National Orthopaedic Hospital, Igbobi: Ignatius Falope; University of Nigeria Teaching Hospital Enugu: John Eze.

Peru Hospital Regional Docente de Trujillo: José Caballero, Wenceslao Azabache and Oscar Salirrosas; Hospital Nacional Hipolito Unanue: Alonso Soto, Elfi Torres, Gloria Ramírez and Mónica Pérez; Clinica Santa Ana: Cesar Malca; Hospital La Caleta: Juan Velez; Hospital Nacional Sergio E Bernales: Raul Yepez; Hospital de Apoyo de Sullana: Hernan Yupanqui; Hospital IV Essalud Huancayo: Pedro Lagos; Hospital Nacional Arzobispo Loayza: Diana Rodriguez; Hospital Municipal Los Olivos: Jorge Flores; Hospital Jose Cayetano Heredia: Anselmo Moya; Hospital Nacional Carlos Alberto Seguin Escobedo: Alejandro Barrionuevo; Hospital Nacional Dos De Mayo: Marco Gonzales-Portillo; Hospital Nacional Cayetano Heredia: Edgar Nunez.

Saudi Arabia King Khalid University Hospital: Abdelazeem Eldawlatly, Mohammed Al Naami and Bilal Delvi; King Khalid National Guard Hospital: Walid Alyafi.

Serbia Klinicki Centar Srbije: Branko Djurovic.

Singapore National Neuroscience Institute: Ivan Ng.

Slovakia FNsP Ružinov: Aktham Yaghi; NsP Poprad: Anton Laincz; NsP JA Reiman Hospital: Stefan Trenkler; Faculty Hospital FD Roosevelta: Jozef Valky.

South Africa Dr George Mukhari Hospital: Mphako Modiba, Peter Legodi and Thomas Rangaka; George Provincial Hospital: Lee Wallis.

Spain Hospital Universitário Virgen del Roció: Ángeles Muñoz; Hospital Ramón y Cajal de Madrid: Ana Serrano; Hospital Universitário Germans Trias i Pujol: Maite Misis; Hospital Torrecardenas: Martin Rubi; Hospital Universitário Virgen de la Victoria: Victoria de la Torre.

Sri Lanka National Hospital of Sri Lanka: Ranjith Ellawala, Samitha Wijeratna, Lukshrini Gunaratna and Crishantha Wijayanayaka.

Tanzania Muhimbili Orthopaedic Institute: Kitugi Nungu, Billy Haonga and Grenda Mtapa.

Thailand Khon Kaen Regional Hospital: Surakrant Yutthakasemsunt, Warawut Kittiwattanagul, Parnumas Piyavechvirat, Tawatcahi Impool and Santipong Thummaraj; Pattani Hospital: Rusta Salaeh; Suratthani Hospital: Sakchai Tangchitvittaya; Bhumibol Adulyadej Hospital: Kamol Wattanakrai, Chatchai Soonthornthum and Teerasak Jiravongbunrod; Lampang Hospital: Surasak Meephant; Rayong Hospital: Pusit Subsompon; Roi-Et Hospital: Phaiboon Pensuwan; Phrae Hospital: Wicheanrat Chamnongwit.

Tunisia Hospital Habib Thameur: Zouheir Jerbi and Abderraouef Cherif.

UK University Hospital of North Staffordshire: Mark Nash; Royal London Hospital: Tim Harris; Leicester Royal Infirmary: Jay Banerjee; Nottingham University Hospitals NHS Trust: Ramzi Freij; Frenchay Hospital: Jason Kendall; Countess of Chester Hospital: Stephen Moore; Hull Royal Infirmary: William Townend; Royal Sussex County Hospital: Rowland Cottingham; Derby Hospitals NHS Trust: Dan Becker; Bedford Hospital NHS Trust: Stuart Lloyd; Royal Liverpool University Hospital: Peter Burdett-Smith; Colchester General Hospital: Kazim Mirza; Royal Lancaster Infirmary: Andrew Webster; Worthing Hospital: Suzanne Brady and Amanda Grocutt; Darent Valley Hospital: John Thurston; Hope Hospital: Fiona Lecky; Northern General Hospital: Steve Goodacre.

Zambia University Teaching Hospital, Lusaka: Yakub Mulla and Dennis Sakala; Nchanga North General Hospital: Charles Chengo.

Writing committee main analyses Haleema Shakur (chairperson), Ian Roberts (chief investigator), Raúl Bautista (Mexico), José Caballero (Peru), Tim Coats (UK), Yashbir Dewan (India), Hesham El-Sayed (Egypt), Tamar Gogichaishvili (Georgia), Sanjay Gupta (India), Jorge Herrera (Colombia), Beverley Hunt (UK), Pius Iribhogbe (Nigeria), Mario Izurieta (Ecuador), Hussein Khamis (Egypt), Edward Komolafe (Nigeria), María-Acelia Marrero (Cuba), Jorge Mejía-Mantilla (Colombia), Jaime Miranda (Peru), Carlos Morales (Colombia), Oluwole Olaomi (Nigeria), Fatos Olldashi (Albania), Pablo Perel (UK), Richard Peto (UK), PV Ramana (India), RR Ravi (India) and Surakrant Yutthakasemsunt (Thailand). National coordinators – Jonathan Dakubo (Ghana), Tamar Gogichaishvili (Georgia), Nyoman Golden (Indonesia), Mario Izurieta (Ecuador), Hussein Khamis (Egypt), Edward Komolafe (Nigeria), Jorge Loría-Castellanos (Mexico), Jorge Mejía-Mantilla (Colombia), Jaime Miranda (Peru), Ángeles Muñoz (Spain), Vincent Mutiso (Kenya), Patrick Okwen (Cameroon), Zulma Ortiz (Argentina), María Pascual, CENCEC (Cuba), RR Ravi (India), April Roslani (Malaysia), Stefan Trenkler (Slovakia), Annalisa Volpi (Italy) and Surakrant Yutthakasemsunt (Thailand).

Writing committee exploratory analyses Ian Roberts (UK) (chairperson), Haleema Shakur (UK), Adefemi Afolabi (Nigeria), Karim Brohi (UK), Tim Coats (UK), Yashbir Dewan (India), Satoshi Gando (Japan), Gordon Guyatt (Canada), B J Hunt (UK), Carlos Morales (Colombia), Pablo Perel (UK), David Prieto-Merino (UK) and Tom Woolley (UK).

1. Background

Introduction: For people at ages 5 to 45 years, trauma is second only to HIV/AIDS as a cause of death. Each year, worldwide, about three million people die as a result of trauma, many after reaching hospital. 1 Among trauma patients who do survive to reach hospital, exsanguination is a common cause of death, accounting for nearly half of in-hospital trauma deaths. 2 Central nervous system injury and multi-organ failure account for most of the remainder, both of which can be exacerbated by severe bleeding. 3

Mechanisms: The haemostatic system helps to maintain the integrity of the circulatory system after severe vascular injury, whether traumatic or surgical in origin. 4 Major surgery and trauma trigger similar haemostatic responses and the consequent massive blood loss presents an extreme challenge to the coagulation system. Part of the response to surgery and trauma, in any patient, is stimulation of clot breakdown (fibrinolysis) which may become pathological (hyper-fibrinolysis) in some. 4 Antifibrinolytic agents have been shown to reduce blood loss in patients with both normal and exaggerated fibrinolytic responses to surgery, and do so without apparently increasing the risk of post-operative complications, most notably there is no increased risk of venous thromboembolism. 5

Existing knowledge: Systemic antifibrinolytic agents are widely used in major surgery to prevent fibrinolysis and thus reduce surgical blood loss. A recent systematic review6 of randomised controlled trials of antifibrinolytic agents (mainly aprotinin or tranexamic acid) in elective surgical patients identified 89 trials including 8,580 randomised patients (74 trials in cardiac, eight in orthopaedic, four in liver, and three in vascular surgery). The results showed that these treatments reduced the numbers needing transfusion by one-third, reduced the volume needed per transfusion by one unit, and halved the need for further surgery to control bleeding. These differences were all highly statistically significant. There was also a statistically non-significant reduction in the risk of death (RR = 0.85: 95% CI 0.63 to 1.14) in the antifibrinolytic treated group.

Hypothesis: Because the coagulation abnormalities that occur after injury are similar to those after surgery, it is possible that antifibrinolytic agents might also reduce blood loss, the need for transfusion and mortality following trauma. However, to date there has been only one small randomised controlled trial (70 randomised patients, drug versus placebo: 0 versus 3 deaths) of the effect of antifibrinolytic agents in major trauma. 8 As a result, there is insufficient evidence to either support or refute a clinically important treatment effect. Systemic antifibrinolytic agents have been used in the management of eye injuries where there is some evidence that they reduce the rate of secondary haemorrhage. 9

Need for a trial: A simple and widely practicable treatment that reduces blood loss following trauma might prevent thousands of premature trauma deaths each year and secondly could reduce exposure to the risks of blood transfusion. Blood is a scarce and expensive resource and major concerns remain about the risk of transfusion-transmitted infection. Trauma is common in parts of the world where the safety of blood transfusion is not assured. A recent study in Uganda estimated the population-attributable fraction of HIV acquisition as a result of blood transfusion to be around 2%, although some estimates are much higher. 10,11 Only 43% of the 191 WHO member states test blood for HIV and hepatitis C and B viruses. Every year unsafe transfusion and injection practices are estimated to account for 8–16 million Hepatitis B infections, 2.3–4.7 million Hepatitis C infections and 80,000–160,000 HIV infections. 12 A large randomised trial is therefore needed of the use of a simple, inexpensive, widely practicable antifibrinolytic treatment such as tranexamic acid (aprotinin is considerably more expensive and is a bovine product with consequent risk of allergic reaction and hypothetically transmission of disease), in a wide range of trauma patients who, when they reach hospital are thought to be at risk of major haemorrhage that could significantly affect their chances of survival.

2. Study design

3. Organisation

Protocol Appendix 1a – Patient information sheet and consent form

Protocol Appendix 1b – Representative information sheet and consent form

Protocol Appendix 2 – Entry form

Protocol Appendix 3 – Outcome Form

Protocol Appendix 4 – Protocol Summary

Authors

Ian Roberts, David Prieto-Merino and Haleema Shakur, Clinical Trials Unit, LSHTM; Iain Chalmers, James Lind Initiative; Jon Nicholl, School of Health and Related Research, University of Sheffield.

Clinical trials are important in improving the safety and effectiveness of emergency care. Many such trials seek to assess the effects of time-critical treatments for life-threatening disorders such as traumatic brain injury, severe haemorrhage or respiratory distress. In general, before patients can be enrolled in such trials, current regulations require that they or their legal representatives provide written informed consent. 63,64 Although the requirement for written informed consent can sometimes be waived65 (e.g. if the patient is unconscious, treatment is urgent, and no relative is available), written consent is usually required in emergency-care research, despite the delays to treatment that this will usually entail.

Analysis of data from the MRC CRASH-1 trial,62 a multicentre randomised controlled trial of corticosteroid administration in acute severe head injury, provides an estimate of the delay associated with the requirement for written consent. On average, compared with hospitals that waived the need for consent, initiation of treatment was delayed by 1.2 (95% CI 0.7 to 1.8) hours in hospitals where written consent from relatives was required. 66 This delay would not occur if the treatment was given (or withheld) outside the context of a clinical trial, in normal clinical practice.

The delay can be life-threatening. The CRASH-2 trial29 showed that giving TXA to trauma patients with bleeds results in a significant and clinically important reduction in overall mortality (RR 0.91; 95% CI 0.85 to 0.97), as well as in mortality specifically ascribed to bleeding (RR 0.85; 95% CI 0.76 to 0.96). Further analyses have shown that these beneficial effects depend importantly on the promptness with which treatment with TXA is started. Taking account of the average delay (1.2 hours) associated with the need to obtain written consent in the MRC CRASH-1 trial, we used CRASH-2 data to provide an estimate of the consequences for survival of a (more conservative) 1-hour delay resulting from the requirement to obtain written consent.

We used a logistic regression model, with mortality due to bleeding as the outcome variable and treatment group, time to treatment, and an interaction term, as explanatory variables. The interaction term from this model estimates how the effect of treatment (OR) changes with time to treatment. We then calculated the risk of death for patients by treatment group, according to time to treatment.

The results are shown in Figure 7. The lines in part (b) of the figure give the risk of death in treated and untreated patients, respectively. The effect of a treatment delay in treated patients was estimated by applying the OR corresponding to a 1-hour treatment delay to the risk of death in the untreated group. The dashed line gives the estimated risk of death in patients in whom treatment is delayed. The point at which the ‘no delay’ and ‘if delay’ lines intersect the ‘risk of death if no intervention’ line gives the time when the trial treatment no longer provides patient benefit.

Using these data and data from the CRASH-2 trial on the proportion of patients who arrive at hospital within a given time since injury, we estimate that a 1-hour treatment delay reduces the proportion of patients who benefit from the trial treatment from 63% to 49%. Whereas the RR of death from bleeding with TXA was estimated in the CRASH-2 trial as 0.85 (95% CI 0.76 to 0.96), the corresponding RR in the presence of a 1-hour delay is 0.96 (95% CI 0.86 to 1.08).

FIGURE 7.

Proportion of patients benefitting from treatment.

The delay from consent rituals in emergency situations has important consequences. First, it results in avoidable mortality and probably morbidity in participants in the trial. Indeed, far from protecting the interests of patients participating in research, requirements for written informed consent and the resultant delay in starting treatment could be lethal. Second, the delay in starting treatment can obscure a real treatment benefit from the administration of a time-critical treatment. In the CRASH-2 trial, the requirement for written informed consent probably means that the trial has underestimated the beneficial effect of TXA in trauma patients with bleeds, which would be given without delay in normal clinical practice.

In the context of research involving people who are incapable of giving informed consent, the Declaration of Helsinki67 states that, if no patient representative is available and the research cannot be delayed, the study may proceed without informed consent provided that the specific reasons for involving patients with a disorder that renders them unable to give informed consent have been stated in the research protocol and that the study has been approved by a research ethics committee.

We argue that the need for an urgent trial treatment, even in patients who are conscious and whose relatives are available, by itself excludes the possibility of fully informed consent. If consent rituals delay the start of a trial treatment such that the treatment effect could be reduced or obscured, we maintain that seeking consent is actually unethical. There might be other treatments whose benefits have been missed or underestimated as a result of insistence on the rituals of informed consent, against the precepts of the Declaration of Helsinki, with resultant avoidable harm to patients. There is little evidence that widely promoted forms of research regulation do more good than harm. 68 Informed consent procedures, like other well-intentioned public health interventions, should be assessed rigorously. The lethal effects we have shown might have been found decades ago had the research ethics community accepted a responsibility to provide robust evidence that its prescriptions are likely to do more good than harm.

Why share data?

Data generated through participation of patients and the public should be put to maximum use by the research community and, whenever possible, translated to deliver patient benefit. Data sharing benefits numerous research-related activities: reproducing analyses; testing secondary hypotheses; developing and evaluating novel statistical methods; teaching; aiding design of future trials; meta-analyses; and hopefully helping to prevent error, fraud and selective reporting.

Data sharing achieves many important goals for the scientific community, such as:

• reinforcing open scientific inquiry

• encouraging diversity of analysis and opinion

• promoting new research, testing of new or alternative hypotheses and methods of analysis

• supporting studies on data collection methods and measurement

• facilitating education of new researchers.

Free Bank of Injury and emergency Research Data

At the end of the CRASH-2 trial, additional funding was obtained from the NIHR HTA programme for the purpose of launching a data sharing facility for injury and emergency-related research data.

The Free Bank of Injury and emergency Research Data (freeBIRD) website (http://freebird.Lshtm.ac.uk) allows investigators to upload and share data from such trials. Making clinical trial data sets available to investigators beyond the original research team can improve patient care, advance medical knowledge and provide better value for money from health research.

The freeBIRD website aims to facilitate data sharing in the area of injury and emergency research in a timely and responsible manner. It has been launched by providing open access to anonymised data on > 30,000 injured patients (the CRASH-1 and CRASH-2 trials).

We hope that other trial investigators will also upload their datasets and that freeBIRD becomes a valued resource for all those committed to improving injury and emergency care.