The trial protocol,22 statistical analysis plan23 and results24 have been previously published and parts of these published articles are reproduced throughout this report. The protocol was published in Trials (reproduced from Dewan et al.22). This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The text below includes minor additions and formatting changes to the original text. The statistical analysis plan was published in Wellcome Open Research (© 2018 Roberts et al.23 This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text). The trial results were published in The Lancet [copyright © 2019 the CRASH-3 trial collaborators.24 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text].
The protocol and statistical analysis plan can be found at www.journalslibrary.nihr.ac.uk/programmes/hta/1419001/#/documentation (accessed November 2020).
Trial design
CRASH-3 is an international, multicentre, randomised, placebo-controlled trial of the effects of TXA on death and disability in patients with TBI. The trial protocol was peer reviewed and published in BioMed Central Trials journal as an open access article in 2012 (see the trial protocol).22
CRASH-3 is the third international, multicentre, randomised, placebo-controlled trial in trauma patients conducted by the London School of Hygiene & Tropical Medicine (LSHTM) trial co-ordinating centre. CRASH-1 investigated corticosteroid use in head injury and recruited 10,000 patients with TBI from across the world.7 CRASH-217 examined the effects of early administration of a short course of TXA in trauma patients. The trial recruited 20,211 patients from 274 hospitals in 40 countries.
Through these many years of collaboration, LSHTM has developed good working relationships with a large number of trauma doctors and an excellent global network of collaborating trauma hospitals.
CRASH-3 was undertaken in 175 hospitals in 29 countries. Suitable collaborating hospitals and investigators were assessed in terms of the trauma service that they provide and their ability to conduct the trial. Before the trial could begin at any site, the local principal investigator must have agreed to adhere to good clinical practice guidelines and all relevant national regulations. In addition, all relevant regulatory and ethics approvals were in place before the trial started at a site. See Appendix 1 for a list of the trial collaborators by country.
There is a wide spectrum of treatments for TBI. As the trial was conducted worldwide, each participating site was instructed to follow its own clinical guidelines for the treatment of TBI patients. No clinically indicated treatment was required to be withheld for the trial. TXA or placebo was provided as an additional treatment to the usual management of TBI.
Approvals
The Medical Research and Ethics Committee and Health Research Authority reviewed the protocol and supporting documents for the CRASH-3 trial and provided a favourable ethics opinion on 19 July 2012 (Research Ethics Committee reference 12/EE/0274). One substantial amendment to the protocol was submitted to limit the time window for eligibility from within 8 hours of injury to within 3 hours of injury. Favourable opinion was received on 6 September 2016. Two non-substantial amendments were submitted to extend the recruitment period, and were categorised on 1 August 2017 and 2 February 2018. The Medicines and Healthcare products Regulatory Agency authorised the CRASH-3 trial on 8 August 2012 (reference 17072/0007/001-0001). A favourable ethics opinion was received from the Observational/Interventions Research Ethics Committee at LSHTM on 17 November 2011 (reference 6060).
Participants (inclusion and exclusion)
Adults with TBI who were within 3 hours of injury and had a GCS score of ≤ 12 or any intracranial bleeding noted on their CT scan and no significant extracranial bleeding (i.e. not in need of immediate blood transfusion) were eligible. The time window for eligibility was originally within 8 hours of injury; however, in 2016, the protocol was changed to limit the time window for eligibility from within 8 hours to within 3 hours of injury. This change was made blind to the trial data, in response to external evidence suggesting that delayed treatment is unlikely to be effective.
The fundamental eligibility criterion was the responsible clinician’s ‘uncertainty’ about whether or not to use TXA in a particular patient with TBI. This is based on the uncertainty principle, which is a well-established approach for assessing trial eligibility.25 A patient can be enrolled if, and only if, the responsible clinician is substantially uncertain as to which of the trial treatments is most appropriate for that particular patient. A patient should not be enrolled if the responsible clinician or the patient (or his/her representative) is, for any medical or non-medical reasons, reasonably certain that one of the treatments that might be allocated is inappropriate for that particular individual (in comparison with either no treatment or some other treatment that could be offered to the patient in or outside the trial). Using the uncertainty principle allowed the process of this trial to be closer to what is appropriate in normal medical practice. The pragmatic design allowed us to find out how effective the treatment actually is in routine everyday practice.
Consent
Owing to the nature of their injury, most TBI patients are unable to provide prior informed consent to participate in a clinical trial. As acknowledged in the Declaration of Helsinki,26 patients who are incapable of giving consent are an exception to the general rule of informed consent in clinical trials. In the CRASH-3 trial, consent was usually sought from the patient’s relative or a legal representative. If no such representative was available, the study proceeded with the agreement of two clinicians (one independent of the trial). If and when the patient regained capacity, they were told about the trial and written consent was sought to continue participation. If the patient or their representative declined consent, participation stopped. If patients were included in the trial but did not regain capacity, consent was sought from a relative or legal representative. We adhered to the requirements of the local and national ethics committees. See Appendix 2 for an overview of the consent procedure.
Randomisation and blinding
Sites were advised to randomise patients who were eligible for inclusion as soon as possible. The entry form was used to assess eligibility and collect baseline information. Following confirmation of eligibility, patients were randomly allocated to receive TXA or matching placebo (0.9% sodium chloride) by intravenous (i.v.) infusion. An independent statistician from Sealed Envelope Ltd (London, UK) prepared the randomisation codes and gave them to the drug packers so that treatment packs could be prepared. After baseline information was collected on the entry form, the lowest numbered treatment pack remaining was taken from a box of eight treatment packs. If the treatment ampoules were intact, the patient was considered randomised. Entry form data were entered into a secure online database by the trial investigators. Both participants and study staff (site investigators and trial co-ordinating centre staff) were masked to allocation. An emergency unblinding service was available for use in those rare situations when the clinician believed that clinical management depended on knowledge of whether the patient received TXA or placebo.
The TXA (Cyklokapron® injection) was manufactured by Pfizer Ltd (Sandwich, UK). The Torbay and South Devon NHS Foundation Trust prepared the 0.9% sodium chloride placebo. Ampoules and packaging were identical in appearance. The blinding was done by Bilcare GCS (Europe) Ltd (Crickhowell, UK). This entailed removal of the manufacturer’s label and replacement with the trial label and treatment pack number. Pack label texts were identical for TXA and placebo. We checked the coding of the blinded ampoules by randomly testing each batch of treatments and doing high-performance liquid chromatography to determine the contents.
Trial intervention
Patients were randomly allocated to receive a loading dose of 1 g of TXA infused over 10 minutes, started immediately after randomisation, followed by an i.v. infusion of 1 g over 8 hours, or matching placebo. Every patient was assigned a treatment pack with a unique number, which contained four ampoules of either 500 mg of TXA or placebo, one 100-ml bag of 0.9% sodium chloride (to use with the loading dose), a syringe and needle, stickers with the trial details and randomisation number (for attaching to the infusion bags, forms and medical records), and instructions. We separately provided information for patients and representatives, consent forms and data collection forms. The stickers, instructions, leaflets and forms were in local languages.
Dose selection
Tranexamic acid has been used to reduce bleeding in elective surgery for many years. A systematic review of randomised trials of TXA in surgery shows that dose regimens of TXA vary widely.16 Loading doses range from 2.5 mg/kg to 100 mg/kg and maintenance doses range from 0.25 mg/kg/hour to 4 mg/kg/hour delivered over periods of 1 to 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.16,27 In emergency situations, the administration of a fixed dose is more practicable because weighing patients in such situations is difficult. In the CRASH-3 trial, a fixed dose of 1-g loading dose of TXA, followed by a 1-g maintenance dose over 8 hours was selected. This fixed dose is within the dose range that has been shown to inhibit fibrinolysis and provide haemostatic benefit. It should be efficacious for heavier patients (> 100 kg) but also safe for lighter patients (< 50 kg), as the estimated dose/kg that the latter group would receive has been used in other trials without adverse effects. Furthermore, this fixed dose was used for 20,211 patients enrolled in the CRASH-2 trial and was found to be both effective and safe.17 The same fixed dose was also used in two studies of TXA in TBI patients, again with no evidence of adverse effects.20,21
Sites
We recruited patients with TBI from 175 hospitals in 29 countries. We enrolled the first patient on 20 July 2012 and the last patient on 31 January 2019. We stopped recruiting when the trial treatment expired. See Appendix 3, Table 17, for the total number of randomisations by geographical region.
Data collection
Baseline data
The trial entry form was used to collect baseline information including age, sex, time since injury, systolic blood pressure (SBP), GCS score, pupil reaction and, if relevant, the location of intracranial haemorrhage.
Outcome data
An outcome form was required to be completed 28 days after randomisation, or at death or hospital discharge if either event had already occurred. Once randomised, outcome data were collected even if the trial treatment was interrupted or not actually given. Short-term disability was assessed on the outcome form using the Disability Rating Scale (DRS). This scale measures the level of disability in six diagnostic categories of (1) eye opening, (2) best verbal response, (3) best motor response, (4) self-care ability for feeding, grooming and toileting, (5) level of cognitive functioning and (6) employability, and it can be used across the span of recovery. The maximum score a patient can obtain is 29, which represents an extreme vegetative state. A person without disability would score zero.28 Specific patient-orientated outcomes were also assessed. These measures were identified from the literature and then considered and agreed by patient representatives from RoadPeace (London, UK), the UK national charity for those killed or injured in road crashes.
Monitoring
As the trial was assessed as low risk (TXA is widely used and the trial was considered to have a low risk of bias), central trial monitoring and central statistical monitoring were used in conjunction with investigator training, meetings and written guidance. Trial investigators and their institutions provided direct access to the source data for trial-related monitoring, audits and regulatory inspections. We planned to monitor about 10% of patient records on site; however, after changing the primary outcome, we expanded our monitoring plan to include patients enrolled within 3 hours of injury who subsequently died. We monitored 2436 (19%) patient records on site or remotely (using video call or telephone). This included 1161 (67%) of the patients who died from head injury (the primary outcome). The team of monitors worked alongside local trial teams to verify data from the source data, including pre-hospital ambulance cards, admission registers, emergency department notes, CT scans, surgery notes, blood transfusion registers, death registers and death certificates.
Outcome measures
Primary outcome
The primary outcome was head injury death in hospital within 28 days of injury in patients randomised within 3 hours of injury. The primary end point was originally 8 hours but, in 2016, the protocol was changed to patients treated within 3 hours of injury. Cause of death was assessed by the responsible clinician.
Secondary outcome
Secondary outcomes were early head injury death (within 24 and 48 hours after injury), all-cause and cause-specific mortality, disability, vascular occlusive events [myocardial infarction (MI), stroke, deep-vein thrombosis (DVT), pulmonary embolism (PE)], seizures, complications, neurosurgery, days in intensive care unit and adverse events within 28 days of randomisation. 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.
Adverse events
Tranexamic acid has a well-documented safety profile. Although the summary of product characteristics29 suggests that rare cases of thromboembolic events might be associated with TXA administration, there is no evidence that the TXA treatment regimen used in this trial is associated with an increased risk of vascular occlusive events. Nevertheless, data on vascular occlusive events and seizures were collected as secondary outcomes and presented to the independent Data Monitoring Committee for unblinded review.
Change to the protocol
In September 2016, in response to evidence external to the trial indicating that TXA is unlikely to be effective when initiated beyond 3 hours of injury, the Trial Steering Committee (TSC) amended the protocol to limit recruitment to within 3 hours of injury.18,30,31 Consequently, the primary end point was changed to ‘head injury death in hospital within 28 days of injury for patients treated within 3 hours of injury’.
To ensure that the trial would be large enough to reliably confirm or refute an early (< 3 hours) treatment benefit, the sample size was increased from 10,000 to 13,000 patients with the aim of enrolling 10,000 patients within 3 hours of injury.
The changes were made without reference to the unblinded trial data. The Data Monitoring Committee was not consulted about the change. The change was therefore not driven by the unblinded trial data seen by the Data Monitoring Committee, but instead driven by accumulating evidence external to the trial. The trial was conducted in accordance with International Conference on Harmonisation-Good Clinical Practice Guidelines.32
Rationale for protocol change
During the CRASH-3 trial, new research emerged suggesting that TXA is likely to be most effective in the first few hours after injury and less effective when given later.18 Trauma triggers the early release of tPA, the enzyme that converts plasminogen to the fibrinolytic enzyme plasmin, resulting in increased clot breakdown and bleeding.33,34 tPA levels peak about 30 minutes after injury and plasmin peaks at 1 hour.34
By inhibiting early fibrinolysis, TXA prevents coagulopathic bleeding;35 however, the effects appear to be short lived. Around 2 hours after injury, plasminogen activator inhibitor (PAI-1) levels increase, reaching a peak at 3 hours.34 Plasminogen activator inhibitor inhibits fibrinolysis, resulting in ‘fibrinolytic shutdown’.36 This might explain why the benefits of TXA in polytrauma patients appear to be limited to the first 3 hours.18 As recent research shows that the coagulopathy after TBI is similar to that in poly-trauma, a similar time-dependent effect might be expected after TBI.37,38 If the pathophysiological mechanisms affected by TXA are most relevant in the early hours after injury, the effect of TXA in this early period is the outcome of greatest importance. Nevertheless, intracranial bleeding can continue for up to 24 hours after injury and, therefore, examination of the effects of TXA within and beyond 3 hours remains an important scientific objective that will be addressed in preplanned subgroup analyses.
Sample size
Prior to implementing the amendment on limiting recruitment to within 3 hours of injury, 3535 participants had been recruited. It was originally estimated that a trial with about 10,000 patients would have 90% power (two-sided alpha of 1%) to detect a 15% relative reduction (20% to 17%) in mortality. We increased the sample size to 13,000 to get enough patients (about 10,000 as per the original sample size calculation) within 3 hours of injury to confirm or refute an early benefit. With 10,000 patients, the study would also have > 90% power to detect a difference in mean DRS score of 1.0 [assuming a standard deviation (SD) of DRS score of 9.0]. Experience from the CRASH-1 and CRASH-2 trials suggests that the anticipated rates of loss to follow-up (< 1%) would not have an important impact on study power.17,39
Statistical methods and analysis plan
The statistical analysis plan was published before unblinding (see www.journalslibrary.nihr.ac.uk/programmes/hta/1419001/#/documentation; accessed November 2020).19 The plan gave our reasons for limiting recruitment to within 3 hours of injury and stated that outcomes for patients treated after 3 hours of injury would be presented separately. All analyses were on an ‘intention-to-treat’ basis. For each binary outcome, we calculated RRs and 95% CIs. We conducted a complete-case analysis with no imputation for missing data. The safety of participants was overseen by an independent Data Monitoring Committee, which reviewed four unblinded interim analyses.
Subgroup analyses
In order to test the hypothesis that TXA is most effective when given soon after injury, a subgroup analysis was conducted of the effect of TXA according to the time interval between injury and TXA treatment (≤ 1 hour, > 1 to ≤ 3 hours, > 3 hours). We prespecified that this analysis would include patients treated within and beyond 3 hours of injury. As TBI severity, SBP and age could confound the impact of time to treatment on treatment effectiveness, we planned to control for these variables in a multivariable model. We expected that any beneficial effect of TXA would vary by time to treatment, with earlier treatment being most effective. We examined this hypothesis in a subgroup analysis of the effect of TXA according to the estimated time interval between injury and treatment (≤ 1 hour, > 1 to ≤ 3 hours, > 3 hours).
The effects of TXA on the primary outcome were also stratified by severity of head injury and age. Severity of head injury was assessed using the baseline GCS score, mild to moderate (GCS score of 9–15) or severe (GCS score of 3–8), and by pupil reactivity. In addition, we assessed the impact of severity in a regression analysis that included continuous terms for GCS and its square.
Traumatic brain injury patients who have a GCS score of 3 and bilateral unreactive pupils have a very poor prognosis, with a mortality risk of about 75%. The inclusion in the CRASH-3 trial of such severely injured patients, who may have little potential to benefit from the trial treatment, would bias the treatment effect towards the null. We therefore prespecified a sensitivity analysis that excluded patients with a GCS score of 3 and bilateral unreactive pupils.
As fibrinolytic activation after TBI may increase with age, we examined the effect of TXA on head injury death stratified by age: younger (≤ 30 years), middle (31–60 years) and older (> 60 years). For subgroup analyses, we report p-values for the test for heterogeneity.
Economic evaluation methods
An economic model was developed to analyse the cost-effectiveness of TXA treatment versus no treatment for patients with TBI. The analysis was performed in line with National Institute for Health and Care Excellence (NICE) guidance for economic evaluations, comparing the incremental costs and outcomes associated with providing TXA, over a lifetime time horizon, from the perspective of the UK NHS.40 Full details of the methods and results are provided in Chapter 5.
Patient and public involvement
The CRASH-3 trial included patient and public involvement (PPI) to achieve the following objectives, namely to:
- gain a lay perspective on PPI involvement in the design and management of emergency care clinical trials
- identify an appropriate consent procedure for entering critically ill trauma patients into emergency clinical trials, which could be used for CRASH-3
- ensure that we collect outcomes that are of primary concern to patients and their families after TBI
- ensure that patient-facing documents for the trial were appropriate and clear
- provide a lay perspective on the management of the trial and interpretation of the results
- assist in developing and implementing the results dissemination strategy, and to help with presenting the trial results in a public-friendly format.
We included PPI groups to input to different stages of the trial. This included people who are at high risk of TBI, charitable organisations that support victims of trauma (RoadPeace) and people who have suffered TBI (Headway, Nottingham, UK).
Prior to working with our group, we carried out formative research to help guide PPI activities.
Formative research
Method
A qualitative study was conducted to elicit views on how best to involve patients and the public in the design, conduct and reporting of clinical trials involving people in emergency situations, gathering perspectives on which areas of the research programme they believed public contribution would be most appropriate. Approaches to designing a consent process to enter patients into emergency clinical trials were also explored.
Three focus group discussions were conducted, one with young people involved in an amateur boxing club, the other with a group of older men belonging to a social club and the third with a group of older women who were involved in a continuing education project and crafts-based activities. In total, 19 people took part (12 men and 7 women).
The sessions included a PowerPoint® (Microsoft Corporation, Redmond, WA, USA) presentation detailing why clinical trials are conducted in emergency medicine, how they are conducted and the key principles, including issues of consent, randomisation and the use of placebos. This was followed by three exercises using group work and discussion techniques.
Two key areas of inquiry emerged from these discussions: public involvement in the design and management of clinical trials and decisions about entering patients into clinical trials in an emergency.
Involvement in clinical trial design and management
Participants were highly supportive of clinical medical research, seeing it as essential for the progress of medical science. They also had a sense that the public should be consulted in principle. However, they struggled to identify how they might usefully contribute to the design and management of clinical trials in practice, seeing this as the province of highly skilled and qualified experts. Although there were individuals who could envisage a role for themselves with appropriate information and preparation, it was important to acknowledge that others felt that they had neither the inclination nor the aptitude to become involved, trusting in the expertise and competence of clinical researchers. Participants did have strong opinions in one area: that decision-making about the outcomes of clinical research must take account of quality-of-life issues and not be confined to treatment efficacy or safety, which they saw patients and the public as being well placed to comment on.
Consent process for involving patients in clinical trials in an emergency
Initially, a minority opposed entering patients into trials without their consent but these views tended be modified as participants considered the comments of others about the incapacitation of patients, the time-critical nature of emergency medicine and the necessity of clinical trials for medical progress.
Overall, among all groups, there was a very high regard for the medical profession and a strong faith in the skills and competence of medics, as well as the belief that clinicians would always act in the best interest of the patient. This was reflected in a sense that clinicians should be allowed to exercise their clinical judgement without undue burden to seek consent from next of kin when patients could not consent for themselves. However, moderating this perspective for some was a belief in the principle that, where practicable, next of kin should be consulted. Others argued that this might place a heavy burden of responsibility on families, and that the clinician’s greater expertise may in fact render better decisions.
Interestingly, when the participants were asked what they would want for themselves, all the participants expressed a desire for the clinician (or their family) to enter them into the trial.
Patient and public involvement group
The PPI group was responsible for providing input on the development of quality-of-life outcome measures to be used in the trial. They provided feedback from individuals with TBI and their caregivers on items of primary concern to patients after TBI
The PPI group reviewed drafts of the patient representative and patient information sheets, and consent forms.
A member of the PPI group from RoadPeace provided a lay perspective on the management of the trial as part of the TSC. RoadPeace is the national charity for road crash victims in the UK. Road traffic collisions are responsible for the majority of cases of TBI globally. RoadPeace supports survivors and their families and works to prevent serious injury and deaths from road crashes.
RoadPeace provided input in the CRASH-3 dissemination strategy (see Chapter 7). RoadPeace was involved in interpreting the data as part of the writing committee responsible for the main result publication. Both RoadPeace and Headway (National Head Injuries Association) provided help in dissemination of the results. A film to report the main trial results was led by a member of Headway (https://crash3.lshtm.ac.uk/blog/crash-3-trial-results/; accessed 23 February 2020).
Outcome of patient and public involvement
Patient and public involvement contributed to the success of the trial. The consent process that was developed with PPI groups was used in all countries that took part in the trial. The main structure and content of the brief information sheet, participant/legal representative information sheets and consent forms were utilised globally. They were accepted by all ethics committees and regulatory agencies with only local modifications needed.
The outcome measure developed with the PPI group included the following domains, which were considered to be important to TBI patients and their families: (a) walking, (b) washing/dressing, (c) pain/discomfort, (d) anxiety/depression, (e) agitation/aggression and (f) fatigue. A three-point scale response for each domain was used (none, moderate, extreme).
Role of funding source
The run-in phase (the first 500 patients) was funded by the JP Moulton Charitable Trust. The main phase was funded jointly by the National Institute for Health Research Health Technology Assessment (HTA) (project number 14/190/01) and Joint Global Health Trials [Medical Research Council (MRC), Department for International Development, Wellcome Trust] (project number MRM0092111). Dr Paul Atkinson, Saint John Regional Hospital, Canada, received a CA$10,000 grant from the New Brunswick Trauma Program to support the trial in Canada. The funders of the study had no role in study design, data collection, data analysis, data interpretation or writing the report. The corresponding author/writing committee had full access to all the data in the study and had final responsibility for the decision to submit for publication.