Trial design

CRYOSTAT-2 was a randomised, parallel-group, unblinded, multicentre, international trial evaluating the effects of early high-dose cryoprecipitate in adult patients with major trauma haemorrhage requiring MHP activation. The trial is registered at ISRCTN, with reference number ISRCTN14998314 (https://doi.org/10.1186/ISRCTN14998314),²⁶ Sections of this chapter are based on the trial protocol published in Transfusion Medicine,²⁷ and the original study protocol document,²⁸ available from the National Institute for Health and Care Research (NIHR) Funding and Awards site.

Approvals

The Medical Research and Ethics Committee and Health Research Authority reviewed the protocol and supporting documents for the CRYOSTAT-2 trial and provided a favourable ethics opinion on 26 May 2017 (Research Ethics Committee reference 17/SC/0164).

Six substantial amendments and eight non-substantial amendments were approved during the project, as detailed in Table 1.

TABLE 1

Ethics approvals

The Confidentiality Advisory Group (CAG) reviewed the application and supporting documents for Section 251 support to process confidential patient information without consent and provided approval on 28 January 2020 (CAG reference 19/CAG/0161).

Participants (inclusion and exclusion)

Patients were eligible for this trial if:

1.

The patient was judged to be an adult, was aged ≥ 16 years in the UK (or according to local guidance) and had sustained severe traumatic injury.

2.

The patient was deemed by the attending clinician to have active haemorrhage.

And required:

3.

Activation of the local MHP for the management of severe blood loss.

And had started or had received:

4.

At least one unit of any blood component.

Patients were not eligible for this trial if they fulfilled one or more of the following criteria:

1. They had been transferred from another hospital.
2. The trauma team leader deemed the injuries incompatible with life.
3. More than 3 hours had elapsed from the time of injury.

Consent

Patients with major trauma haemorrhage have a life-threatening injury. On arrival at the emergency department, patients are usually unconscious, and those who are not are usually in pain, in distress and/or under the influence of medication. Patients were enrolled into the study using an ‘emergency waiver of consent’, whereby the treating senior clinician (the trauma team leader) assessed the patient for eligibility and made the decision whether or not to enter them into the trial. This is an established method of enrolling patients without capacity into emergency medicine trials.

UK law allows patients who need emergency treatment and lack capacity to consent to have a ‘consultee’, who can recommend whether the participant would like to continue in or would object to being in the trial. This consultee can be the treating clinician who is not part of the research team (a professional consultee) or a relative/friend (a personal consultee). This study recommended that a professional consultee be sought in the first instance to allow the participant to move from the emergency waiver to consultee declaration and remain in the study until such time that a personal consultee or the participant could be approached about the study. The aim of the consent process in this trial was always to obtain informed signed consent from participants. The parents/legal guardians of participants who were found to be below 16 years of age after the point of entry were approached as personal consultees. See Appendix 1, Figure 11, for an overview of the consent procedure used in the UK.

Participants enrolled at international sites were subject to the regional or national ethical code of practice for conducting research in incapacitated adults.

There were occasions where a participant died soon after their arrival at hospital, and there had been no opportunity for contact between the research team and a personal consultee. In these situations, it was accepted that approaching the consultee to ask their opinion was likely to be distressing and without benefit. Families of bereaved participants were not contacted by the research team after the participant’s death to obtain personal consultee advice. In these cases, advice from a professional consultee was sought so that data collection could be completed at site. Recruiting hospitals were informed that participation in the trial should be disclosed if a coroners’ court hearing was convened. This process was agreed after receiving advice from the Research Ethics Committee.

The COVID-19 pandemic began during this trial. This made the consent process more complex because of COVID-19-related restrictions in hospitals, particularly when participants were on a COVID-19 ward where no visitors or staff visits were allowed. It was agreed that initial verbal consent could be taken from COVID-19-positive individuals or those on restricted wards in hospital if this was documented appropriately in the medical notes. When it was safe to do so, the participant was approached to sign an informed consent form. If the participant had been discharged, they were contacted at home by telephone or by post. Consent was obtained using an electronic form or via a physical form returned to the hospital. This process was also followed when seeking consultee declarations during this time. The PANDO application³⁰ was used at some participating sites to electronically transfer patient-signed informed consent forms from COVID-19-restricted wards to the research team.

It became apparent shortly after the trial began that, for a subset of participants, research staff were at times not able to start or complete the consent process. The reasons for this included:

• patients unexpectedly being discharged (overnight or weekends)
• patients unexpectedly self-discharging (including absconding from authorities)
• participants being rapidly discharged to police custody or returned to the custody of His Majesty’s Prison Service
• participants being rapidly transferred to mental health trusts under section
• participants having been repatriated to a non-participating hospital or residential nursing care home or rehabilitation centre
• participants providing false details or refusing to give any details to hospital staff
• participants being homeless and having no listed general practitioner
• participants in hospital being abusive, aggressive or violent, and whom the research team had been advised not to approach
• participants who would have been extremely distressed if contacted due to the nature of their trauma (e.g. if they had been in an accident in which a loved one had died)
• participants who never regained capacity due to the nature of their injuries (brain injury or coma)
• participants having considerable language difficulties where it was not possible to provide an NHS translator or to ask a family member to translate information about the trial.

To enable follow-up data to be collected for these participants, we applied to the HRA for Section 251 support via the CAG, and this was granted.

The participant or their consultee was free to withdraw their consent or change their opinion about participating in the trial at any time. Participants were withdrawn with or without permission for follow-up data collection. If participants withdrew and did not provide permission for continued data collection, data collected up to the point of withdrawal were retained. Although the participant was not required to give a reason for withdrawing consent, a reasonable effort was be made to establish this reason while fully respecting the participant’s rights.

Randomisation

The allocation sequence was produced by the trial statistician using SAS^® statistical software (SAS Institute Inc, Cary, NC, USA). The allocation sequence had a varying block size that was not disclosed and was stratified by centre. Participants were allocated in a 1 : 1 ratio to the intervention (early cryoprecipitate + standard MHP) or standard care (standard MHP) arm. Allocation cards were prepared and placed in sequentially numbered opaque envelopes, 100% QA checked and sealed with tamperproof tape. Each envelope contained a randomisation number and the allocated treatment and were opened at sites in sequential order. Envelopes were released by the Clinical Trials Unit (CTU) in small batches sufficient to support projected recruitment and replace used envelopes. Envelopes were securely stored in a locked cupboard at each site, access to which was controlled by the research team. In February 2020, an issue was identified with the creation of the allocation sequence, such that there was a tendency for one arm to appear first more often in each block in the early part of the randomisation list, leading to concerns about a small imbalance by the end of the trial. A revised list was created and used when issuing new randomisation envelopes from that point on.

Randomisation took take place in the emergency department or the transfusion laboratory/blood bank, as agreed with the participating hospital. The recruiting staff completed an enrolment log each time an envelope was taken for use, and signed and dated the envelope to confirm that the next available and lowest numbered envelope of the batch had been taken, was unopened and bore no evidence of tampering. The participant’s initials, date of birth and hospital number were also written on the envelope prior to opening (if the initials and date of birth were not known, the unique identifiers used at the participating hospital for unknown patients were used). The correct and sequential use of envelopes was strictly audited by the site research team and CTU staff at site monitoring visits.

If the randomised intervention differed from that which the participant received, a reason for this was requested. These processes allowed for clear and regular auditing of the randomisation process through comparison with the randomisation list held by the statistician. At the end of recruitment, sites returned the randomisation envelopes to the CTU, with participant details obscured.

This was an unblinded trial. As the study control arm was standard care (MHP), it was not possible to blind the intervention of early cryoprecipitate as there was no comparative transfusion of a placebo. Members of the trauma team in emergency department were therefore aware of the treatment arm to which the participant had been randomised. However, the risk to trial integrity was minimised, given that the primary outcome of 28-day mortality was a hard end point.

Trial intervention

The intervention was 3 pools of cryoprecipitate (equivalent to 15 single units of cryoprecipitate or 6 g of fibrinogen supplementation), infused as rapidly as possible, within 90 minutes of arrival at hospital. The intervention was given in addition to the standard MHP.

The comparator was standard MHP only. The standard treatment of major traumatic haemorrhage involves administering RBCs, FFP and platelets. Each participating site followed its local standardised MHP that aligned with the current accepted best practice for transfusion therapy, that is damage control resuscitation, the use of tranexamic acid and the empiric (automatic and unguided) delivery of RBCs, FFP and platelets. The MHP comprises a balanced resuscitation with blood products and was standardised as much as possible across participating sites; however, some degree of variation was inevitable and pragmatic, reflecting actual clinical practice. The transfusion products given vary to some extent from the target ratio according to blood product availability and the participant’s ongoing clinical condition.

Early cryoprecipitate (3 pools = 6 g of fibrinogen) was stored in its frozen state and defrosted in accordance with the local standard operating procedures for each blood bank/transfusion laboratory. In the UK, these procedures will be in accordance with the national guidelines.³¹ The cryoprecipitate was made available to the emergency department as quickly as possible. The study team provided trial-specific, yellow-labelled bags for the intervention to differentiate trial cryoprecipitate from any cryoprecipitate given as standard of care. As an alternative, and depending on local approval for use, pre-thawed cryoprecipitate was also permitted during the trial. The early cryoprecipitate was administered as rapidly as possible via an intravenous line and in accordance with local practice. The cryoprecipitate was not mixed with platelets prior to infusion.

There were no restrictions on treatments that could be given during this trial, as long as the treatments were part of standard care and were not administered as part of another trial [unless authorised by the Trial Management Group (TMG) or delegated co-investigator].

Sites

This study was conducted in participating major trauma centres in the UK and the USA. The list of study sites is in Appendix 2, Table 17.

Data collection

Monitoring

CRYOSTAT-2 was assessed as moderate risk; therefore, visits were conducted twice per year at each site. Investigators and their institutions provided direct access to source data for monitoring, auditing and regulatory inspections. We planned to monitor 20% of UK participants through on-site source data verification, but from March 2020, owing to restrictions during the COVID-19 pandemic, as most sites had been visited multiple times monitoring was conducted remotely, with site research teams using a checklist to reconfirm key data. We monitored 30% of UK participants on site. Key data were monitored centrally on a bi-monthly basis as part of the TMG report. This included recruitment, withdrawals, serious adverse events, data queries and completion and consent.

Outcome measures

Adverse events

Data on symptomatic thrombotic events (venous thromboembolism and arterial thrombotic events), death and serious transfusion-related adverse reactions were collected from randomisation up to day 28 or discharge. All serious adverse events were reviewed by the chief investigators or delegated clinical members of the TMG.

Changes to the protocol

Table 2 summarises the changes made to the protocol throughout the lifespan of the trial.

TABLE 2

Changes to the protocol

Sample size

The CRYOSTAT-2 study was designed to detect an absolute mortality difference of 7% from a baseline mortality rate of 26%, that is a reduction to 19%. The baseline mortality was based on the feasibility study CRYOSTAT-1, which reported a 28% mortality rate; the national epidemiological study of trauma transfusion practice, which reported a 39% mortality rate among patients receiving ≥ 10 units of RBC in 24 hours (classified as having ‘massive haemorrhage’); and the PROPPR study of bleeding trauma patients conducted at level 1 trauma centres in North America, which had a baseline mortality rate of 26%.

CRYOSTAT-2 used 90% power to detect a reduction in 28-day all-cause mortality of 7% using a 5% level of significance and a two-tailed test. An initial blinded analysis after the first 300 participants had been recruited and followed up to 28 days was used to reassess sample size requirements and recruitment rates, and at the request of the Data Monitoring Committee (DMC) this was repeated after 750 participants had been recruited.

A group sequential design was used to allow for the DMC reviewing the primary outcome for evidence of harm or benefit (but not futility) after 500 and 1000 participants had been followed up for 28 days. The design used O’Brien–Fleming stopping guidelines²⁹ determined at the study design stage to ensure an overall Type I error of 5% at the end of the trial. O’Brien–Fleming guidelines were used because of the low chance of terminating the trial at early interim analysis, and because of the minimal change to the alpha used at final analysis. The stopping guidelines were used to help guide the DMC’s decision-making alongside other safety data available to the committee. Allowing for the interim analyses in this way, the required sample size to meet specified power requirements was 1530 participants in total. This was initially increased by 2.5% to allow for dropout, but as dropout was higher than anticipated this was later increased to 4.4%, resulting in a total of 1600 participants.

Statistical methods and analysis plan

The Statistical Analysis Plans (SAPs) for 28-day outcomes and for longer-term outcomes are available in Report Supplementary Material 1 and 2. All analyses were on an ‘intention-to-treat’ (ITT) basis and included all randomised participants (including those randomised in error) for whom values of a response variable were obtained, analysed according to randomised group. All analyses were two-sided and the significance level was 5%.

The primary outcome, all-cause mortality at 28 days from arrival at hospital, was determined as the proportion of participants in each treatment arm who died within 28 days. The odds ratio (OR) for death within 28 days [with 95% confidence interval (CI) and p-value for the treatment arm term] was presented, and adjusted for centre using a marginal model, and this was the primary analysis of the outcome. This was supplemented by logistic regression analysis so that account could be taken of any prespecified factors that might have differed between the treatment arms and had a statistically significant association with 28-day mortality. Mortality rates according to timing of cryoprecipitate administration were also analysed, using categories of ≤ 45 minutes, 46–60 minutes, 61–90 minutes and > 90 minutes from arrival at hospital. Mortality rates among those who did or not receive cryoprecipitate in the cryoprecipitate arm were also analysed.

As no placebo was used in the study, a per-protocol analysis that excluded patients who did not receive trial treatment would differentially exclude those in the intervention arm. Per-protocol analysis therefore focused on the cohort of patients who could have benefitted from the intervention and excluded only protocol deviations unrelated to the details of cryoprecipitate administration, randomisations in error, those who died within 90 minutes of admission and those who did not receive any blood products after arrival at hospital (an indication that they had already stopped bleeding). A per-protocol analysis was conducted for all mortality end points up to 28 days (6 hours, 24 hours and 28 days), transfusion requirements, hospital stay and thrombotic events.

Multiple imputation based on full conditional specification was used to impute values of potential risk adjustment factors. The set of variables used in the multiple imputation model was specified in the SAP and included the primary outcome variable. Primary and secondary outcome measures were not imputed, and these were treated as missing data. The methods used for the analysis of secondary outcomes in the form of proportions were similar to those described for the primary outcome. Survival times and rates were estimated using the Kaplan–Meier method and compared using Cox proportional hazards regression. Transfusion requirements were summarised as the median and interquartile range (IQR) of the number of units administered from injury to 24 hours post arrival at hospital, and the mean products transfused per participant per hour over the first 24 hours compared between the arms. Hospital stay, critical care stay and ventilator-days were estimated using a competing risks analysis,^32,33 with discharge/extubation as the event and death as the competing risk.

The number of symptomatic thrombotic events up to day 28 were presented overall and by treatment arm. In particular, the number of venous thromboembolisms (pulmonary embolism, deep-vein thrombosis) and arterial thrombotic events (myocardial infarction, stroke) were calculated.

Economic evaluation methods

An economic analysis model was developed to analyse the cost-effectiveness of early cryoprecipitate plus standard of care versus standard of care alone from an NHS and Personal Social Services perspective. Full details of the methods and results are provided in Chapter 4.

Patient and public involvement

The principles set out by the INVOLVE advisory group (active until 2020) were used to guide our approach to patient and public involvement and experience (PPIE) engagement. The research team also built on their pre-existing patient and public involvement (PPI) links that had been set up prior to the start of the feasibility study, CRYOSTAT-1, in 2012. The initial focus of CRYOSTAT-2 PPIE was to engage with patient and public stakeholders to inform the design of the study. Advertising documents (posters/e-mails/website adverts) were sent out both to patients who had previously been injured and to all the lay members of one participating site's trust mailing list to ask for interested volunteers to join a meeting to discuss the study. In total, 2 face-to-face meetings were held (2014, 2016) and 1 survey was distributed to 50 members of the public, asking for views about 2 important trial design questions: what they felt would be the most relevant outcome to test in the proposed CRYOSTAT-2 study (e.g. survival, functional status, or other) and what their views were about consenting participants to a study for which most eligible patients lacked capacity.

Following the second meeting, which specifically focused on consent, in 2016 a dedicated PPI group Patient/Public Advisors for Injury Research (PAIR) was formed to support CRYOSTAT-2 trial development. This group provided ongoing support to the CRYOSTAT-2 trial in the following ways: (1) most importantly, they helped the study team shape the trial design with regard to the primary end point and most acceptable method of gaining consent; (2) the group nominated a member to sit on the Trial Steering Committee (TSC) and this member actively contributed to each TSC meeting throughout the study (a second, more experienced PPI representative was also invited to sit on the TSC); (3) the PAIR group collectively co-wrote the plain language summary for the application to the NIHR Health Technology Assessment (HTA) programme as well as the CRYOSTAT-2 trial protocol and will continue to help with the dissemination of information; (4) the group helped to draft and subsequently edited the patient-facing trial documents such as the patient/patient representative information leaflets and consent/assent forms; and (5) the TSC members helped to develop the NIHR HTA funding application as an integral part of the research team. We will ask the PAIR group for assistance in disseminating the results of this study more widely through social media and patient forums.