Resuscitation with pre-hospital blood products in adults with trauma-related haemorrhagic shock: the RePHILL RCT

Article history

The research reported in this issue of the journal was funded by the EME programme as project number 14/152/14. The contractual start date was in October 2015. The final report began editorial review in May 2022 and was accepted for publication in December 2022. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the final report document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

Copyright © 2024 Crombie et al. This work was produced by Crombie et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Crombie et al.

Background

For many years, trauma care has been predominantly based upon offering casualties with traumatic injuries basic treatment on scene to then safely allow transfer to hospital for further and definitive care. In the last two decades however, the emphasis in civilian practice has started to change in the direction of delivering more advanced interventions to patients while still in the pre-hospital phase. This shift, intended to provide earlier physiological stability and prevent so-called secondary damage occurring, has been driven in part by lessons learned in conflict by military medical systems, and includes advanced haemorrhage control and blood product-based resuscitation. 1–3

The introduction of blood component resuscitation during military casualty retrieval initially produced encouraging results with reports of reduced mortality amongst recipients receiving red blood cells (RBCs) and pre-thawed plasma. 4–6 However, a subsequent systematic review and meta-analysis of studies investigating outcomes following pre-hospital transfusion across both military and civilian practice found only modest advantages in those with moderate severity of injury. 7 In more recent years, two large randomised trials examining the use of pre-hospital plasma produced differing results – one trial favoured the use of plasma8 and one trial was stopped early due to futility. 9 There is therefore a lack of consistent and reliable data, confounded by the wide variety of systems in which studies have been conducted. In 2020, a review of pre-hospital transfusion of RBCs was unable to demonstrate a survival benefit, again highlighted the absence of randomised controlled trials (RCTs) and recommended further studies incorporating the use of individualised transfusion criteria and the use of plasma. 10

Extrapolating results from military trauma-based studies into civilian practice is not straightforward. The mechanisms and severity of injury sustained in conflict are rarely replicated in civilian practice, the patient demographic of active combatants is comparatively narrow and the medical infrastructure in dedicated field hospitals is different to many civilian emergency departments (EDs).

The provision of blood products as early treatment of major haemorrhage may seem logical, but it is also not without complication. Stored citrated blood can present a significant metabolic burden when administered rapidly and has the potential to cause further disruption to coagulation and myocardial function, especially in trauma. 11 There are also considerations around the provision of blood products including the demand for ‘universal’ blood products, regulatory compliance and secure cold-chain governance to avoid unnecessary wastage of products. 12

The lack of robust evidence surrounding the administration of pre-hospital RBCs or plasma in civilian practice, coupled with the challenges this poses to the transfusion community, make a prospective RCT important if further developments in this area of practice are to be justifiable. 13

We present the results of resuscitation with pre-hospital blood products (PHBP) (RePHILL) – a prospective multi-centre RCT which investigates the hypothesis that the administration of pre-hospital RBCs and lyophilised plasma (LyoPlas) would improve tissue perfusion as measured by the clearance of lactate and/or reduce mortality in patients demonstrating shock secondary to traumatic haemorrhage compared to resuscitation with crystalloid infusion. 14

Study design

We conducted a multi-centred two-arm (1 : 1) open-labelled phase 3 RCT of pre-hospital blood product administration versus standard care for patients following traumatic haemorrhage. Research Ethics Committee approval was granted prior to study initiation (South Central – Oxford C, ref: 15/SC/0691). The trial was registered with the International Standard Randomised Controlled Trial Number (ISRCTN62326938) as well as with the European Union Drug Regulating Authorities Clinical Trials Database (registration: 2015-001401-13) and the Medicines and Healthcare products Regulatory Agency (Clinical Trials Authorisation: 16719/0228/001-0001). The trial was run in accordance with Good Clinical Practice requirements. Reporting is undertaken according to the Consolidated Standards of Reporting Trials (CONSORT) guidance. 15 A trial protocol has been published previously. 14 Figure 1 presents a summary of the trial protocol.

FIGURE 1.

RePHILL trial protocol summary.

Internal pilot

The RePHILL trial included an internal pilot. The purpose of the internal pilot was to assess the trial logistics to determine if it is both feasible and practical to carry on and recruit into the trial. It was intended that the pilot would be run at multiple sites to validate the multi-centre aspects of the trial.

The trial progression criteria were defined as:

• minimum of 25 participants recruited across at least two active sites;

• in participants recruited to the trial intervention arm, at least one unit of RBC and one unit of LyoPlas delivered to at least 80% of participants before reaching hospital;

• at least 90% complete data capture;

• Data Monitoring and Ethics Committee (DMEC) reports no safety concerns, which would prohibit continuation to main trial.

Eligibility criteria

Eligibility was assessed by the Pre-Hospital Emergency Medical (PHEM) team (doctor and/or paramedic). The principal investigator (PI) for intervention delivery sites (IDS) was always a medically qualified doctor, and they were responsible for maintaining oversight of the confirmation of eligibility process.

Clinician training

All clinical staff who participated in the trial completed online Good Clinical Practice and protocol training. At the start of the trial, authorisation for the administration of the interventions was limited to doctors working with the PHEM teams. However, this led to missed recruitments by paramedic-only crews. With agreement from relevant parties, the protocol was amended in 2019 and developed a training programme to ensure the safe administration of blood products by paramedics in the context of the trial, that is non-medical authorisation (NMA).

A bespoke training to enable NMA of blood was developed. The training programme comprised four elements as outlined in Figure 2. The programme was successful delivered to three critical care paramedics who cascaded the training to 14 colleagues.

FIGURE 2.

Non-medical authorisation programme.

Informed consent

Prospective informed consent for the participation in the trial was not feasible due to the nature of the injuries and incapacitation that was anticipated to fulfil eligibility criteria. Patients were enrolled in the trial in accordance with Good Clinical Practice, the Declaration of Helsinki, the Clinical Trials Regulations 2004 and the Human Tissue Act 2004. After enrolment, consent for participants to remain in the trial was sought as early as feasible and appropriate when capacity was regained. At the time of enrolment, consent was sought from a professional legal representative shortly after arrival at the receiving hospital. Further consent was sought from a personal or legal representative (such as a close relative or friend). Information regarding the trial was on display in locations likely to be visited by relatives, with a brief summary of the trial and contact details for further information. Some patients may wish to avoid blood product transfusion due to prior held beliefs (such as the Jehovah’s Witness community). In such circumstances, an advance medical directive would have been sought (as per usual practice in emergency resuscitation).

Although an eligibility criterion was that participants should be aged 16 or over, due to the nature of the trial it was not always possible to confirm the age of the participant before recruitment. If a participant was recruited and later found to be under the age of 16, child assent was sought alongside consent from a parent or guardian.

Study setting

The trial was situated within NHS England major trauma network. PHEM teams were recruited to become IDS. At the start of the research, it was anticipated recruitment would take place in six regional air ambulances services. During the set-up of the trial, two air ambulance services withdrew (Dorset and Somerset Air Ambulance and Essex and Herts Air Ambulance) due to loss of equipoise. A further air ambulance service briefly joined the trial (Yorkshire Air Ambulance) but subsequently withdrew prior to recruitment starting due to loss of equipoise.

The following air ambulances participated in the trial:

• East Anglian Air Ambulance, Norwich, UK (www.eaaa.org.uk, accessed 7 February 2022)

• Magpas Air Ambulance, Huntingdon, UK (www.magpas.org.uk, accessed 7 February 2022)

• Midlands Air Ambulance and MERIT, West Midlands Ambulance Service NHS Trust, West Midlands, UK (www.midlandsairambulance.com and https://wmas.nhs.uk, accessed 7 February 2022)

• The Air Ambulance Service, Warwickshire (https://theairambulanceservice.org.uk, accessed 7 February 2022)

The hospital sites which served as receiving hospitals and participating blood banks and pharmacies are listed in Appendix 1.

Randomisation

Interventions

The experimental intervention was a combination of universal donor (blood group O) RBCs and LyoPlas. LyoPlas was used as the plasma product for this trial (License PEI.H.03075.01.1, Germany) in accordance with MHRA approval for import and use as an investigational medicinal product (IMP). The control intervention was 0.9% saline, which was the most commonly delivered non-blood product fluid at the time of trial design. 3 Participants received up to four boluses of the assigned intervention in order to achieve SBP ≥90mmHg or a palpable radial pulse. For the experimental intervention, this was delivered as alternating units of RBC and LyoPlas until a total of 2 RBC and 2 LyoPlas were delivered. For the control intervention, up to four boluses of 250 ml 0.9% saline were delivered. All boluses were administered using fluid warming devices. If any further fluid resuscitation was required in either interventional arm of the trial, this was given in further boluses of 250 ml 0.9% saline and recorded on the PHEM case report form.

Storage and delivery

Lyophilised plasma should be stored between 2 and 25°C and RBC between 2 and 6°C. As LyoPlas is very difficult to reconstitute when cold, it was necessary to store it separately from the RBC. The trial tested and validated two insulated box systems for this purpose – ORCA boxes (Intelsius, Elvington, UK) and Credo boxes (Pelican BioThermal, Leighton Buzzard, UK). The latter boxes were preferred by air ambulances due to their smaller size and weight. The insulated boxes were preconditioned for a minimum of 24 hours in a freezer below –25°C, then removed 30 minutes before use to allow them to reach the required operating temperature before packing. Each transport container also had a temperature data logger, the choice of which was determined by the local blood bank. Sealed treatment boxes were transferred from blood banks to air ambulance/ambulances bases by volunteers from the Blood Bikes charity (www.bloodbikes.org.uk, accessed 9 February 2022). The intervention pathway is shown in Figure 3.

FIGURE 3.

IMP management.

Outcomes

Sample size calculation

There were no prior published analyses using our composite primary outcome. However, there is considerable evidence from observational data for the survival benefit of RBC and plasma in civilian and military studies. Through consultation with those with expertise in pre-hospital trauma resuscitation, we chose an absolute reduction of 10% in the proportion of patients having one of the components of our composite primary outcome as a clinically meaningful effect size. Therefore, using this difference between proportions (two-sided Fisher’s exact test) with 80% power, and a type 1 error rate of 5% (i.e. α = 0.05), we calculated a requirement for 219 participants in each arm of the trial (438 participants in total in our 1 : 1 design). A target of 490 patients was set in order to account for 10% loss to follow-up rate.

The interim analysis for the DMC meeting in May 2018 reported the results on the 192 participants recruited by 20 April 2018. A pooled event rate of 65% experiencing either episode mortality or lactate clearance <20%/h in the two hours post-randomisation was observed in these participants. This observed rate did not correspond with the pooled event rate of 15% assumed in the original sample size calculations. On the DMC’s recommendations, this issue was discussed with the Trial Steering Committee (TSC) in October 2018. The TSC recommended that the power calculations be framed in terms of a relative risk rather than an absolute risk, with the original target sample size of 490 unchanged.

Assuming the pooled event rate remains at 65% and allowing for a 10% loss to follow-up rate, 490 participants will provide 80% power to detect a relative risk ratio of 0.82 (i.e. from 71.7% in the standard care group to 58.3% in the group receiving PHBP) using the method of difference between proportions (two-sided Fisher’s exact test), and a type 1 error rate of 5% (i.e. α = 0.05). This estimated relative risk ratio is consistent with the relative risk ratios of 1.54 and 0.70 reported in two recent pre-hospital RCTs using plasma in one of the treatment arms. 8,9

Statistical analysis

All analyses followed a statistical analysis plan which was finalised prior to database lock (see Report Supplementary Material 1). All analyses were undertaken in SAS v9.4.

Internal pilot

There were delays in starting the trial due to changes to the regulatory requirements, change of chief investigator and difficulty in procuring LyoPlas for use in the trial. The trial opened to recruitment at the Midlands Air Ambulance and West Midlands Ambulance sites on 29 November 2016, with the first patient being recruited on 6 December 2016. By 31 May 2017:

• Twenty-six participants had been randomised across the two active sites.

• 100% of patients randomised to the PHBP arm have received at least one unit of RBC and one unit of LyoPlas.

• A review of 22 completed case report forms confirmed 100% data completion for the primary outcome and 93% data for the secondary outcome.

• The TSC and DMC reviewed trial progress and recommended to the funder that the trial be continued.

The funder confirmed successful completion of the internal pilot and approved progression to the main trial.

Recruitment

Following successful completion of the pilot, the trial extended to the three remaining additional sites. Participant recruitment continued through to 1 January 2021. The trial was closed on 2 January 2021 prior to achieving the intended sample size due to the ongoing impact of the COVID-19 pandemic. The decision to close the trial was made by the TSC and sponsor, without any knowledge of the data or results of interim analyses. During the 49 months of trial recruitment, at least 580 patients were assessed for eligibility, with 432 (74%) deemed eligible and recruited into the trial (see Figure 4). The flow of participants and reasons for non-recruitment are summarised in Figure 5. Recruitment was approximately equal between both trial arms with 209 (46%) participants allocated to receive RBC/LyoPlas and 223 (54%) allocated to receive 0.9% saline.

FIGURE 4.

Recruitment. Thick red line denotes observed recruitment. Aqua line denotes the revised target recruitment still based on a total of 490. Dashed blue line denotes projected recruitment based on rate observed up to the previous DMEC meeting in January 2020. Lighter aqua shaded region denotes period during which no LyoPlas was available to be supplied to the trial. Dashed black line denotes the scheduled end of trial recruitment. Darker aqua shaded region denotes period during which recruitment was paused due to COVID-19. A further breakdown of recruitment (by site) is provided in Appendix 2.

FIGURE 5.

CONSORT flow diagram. ^abased on screening lists provided by each IDS. IO = intraosseous, TCA = traumatic cardiac arrest. ^ballocated interventions are, unless clinically justified: the administration of at least one unit of RBC and one unit of LyoPlas in the PHBP/LyoPlas arm of the trial, and the administration of at least one bolus of fluid in the 0.9% saline arm of the trial. ^creasons for participants not receiving any units of allocated intervention (with no clinical justification) were: 9 due to equipment absence/failure (e.g. of giving sets or lactate monitors), 1 due to complex scene conditions, 1 due to decision to stop resuscitation, 1 due to non-trial saline already being administered to patient and 5 gave no reason.

All participants were followed up until trial exit, with data collection ending at the first occurrence of: withdrawal, acute care discharge, death or at 30 days follow‐up. Episode mortality data was collected up to discharge from the acute care setting, which may be >30 days. The median duration of study follow-up was 8 days [interquartile range (IQR) to 34] across all 432 participants. The median follow-ups for each treatment group were 9 days (IQR 1 to 34) for participants in the RBC/LyoPlas group and 7 days (IQR 0 to 31) for participants in the 0.9% saline group.

The allocation of participants between treatment arms exhibited a slight imbalance, with 14 more participants allocated to receive 0.9% saline than to receive RBC/LyoPlas. Across all 8188 treatment boxes issued by the blood banks, 4094 contained 0.9% saline and 4094 contained RBC/LyoPlas. Treatment boxes issued by blood banks were split 50:50 between RBC/LyoPlas and 0.9% saline in each IDS. Although the proportion of issued treatment boxes used by participants did vary by IDS, from 2.8% to 11.5%, there was no systematic imbalance by study arm. The pre-hospital teams were unaware of the treatment allocation prior to enrolling a participant, so we assume that the imbalance in the number of participants in each arm is down to random chance.

Baseline demographic and clinical characteristics

Baseline demographic and clinical characteristics by treatment arm are presented in Table 4. Participants were well-balanced across baseline characteristics, with most participants being male (82%), of white ethnicity (62%) and with a median age of 38 years (IQR 26 to 58). Participants could record more than one injury mechanism and the most commonly recorded was road traffic collision (62%), with stabbing (16%) and falls (14%) being the other main mechanisms of injury. Other mechanisms of injury were recorded by 9% of participants and comprised 13 laceration injuries, 6 pedestrian incidents with trains, 4 agricultural incidents, 4 industrial accidents and 5 other injuries.

Injury characteristics were similar across treatment arms. The presence of a concurrent brain injury was only collected in the last 16 months of the trial and, during that period, around half of the participants had a concurrent brain injury (48%). Participants could experience both compressible and non-compressible haemorrhages, with non-compressible haemorrhages being the most commonly recorded (83%). Traumatic cardiac arrest, defined as those with a heart rate of 0 and blood pressure of 0, was experienced by 13% of participants providing on-scene heart rate and blood pressure measurements. Blunt force trauma injuries were recorded in 79% of participants, with penetrating trauma injuries recorded in 22% of participants. Acute excessive consumption or alcoholism was suspected in 14% of participants, and the presence of other illicit substances was suspected in 6% of participants. On-scene vital signs were very similar between both groups, with participants exhibiting a mean [(standard deviation (SD)] blood pressure of 73 (18)/46 (15) mmHg, a mean (SD) heart rate of 112 (32) bpm, a mean (SD) respiratory rate of 23.8 (10.1) per minute, a mean (SD) oxygen saturation of 92% (9%) and a median (IQR) GCS of 7 (3, 14). Capillary lactate concentration was very similar across both groups, with a mean (SD) of 9.15 (4.69) mmol/L across all participants.

Both measures of injury severity, injury severity score (ISS) and new injury severity score (NISS), are recorded only on those participants that were Trauma Audit Research Network (TARN) eligible. This excludes participants who died prior to arrival at the ED, so cannot strictly be defined as a baseline characteristic. For the 300 participants providing injury severity scores, the median ISS was 36 (IQR 25 to 50) and median NISS was 43 (IQR 34 to 57). The pre-hospital teams enrolling RePHILL patients attended scene via road ambulance for 62% of participants. They arrived on scene at a median of 26 minutes (IQR 19 to 37) after the time of injury, approximated via the time of the 999 call, and administered the first bolus of trial intervention at a median of 22 minutes (IQR 15 to 34) later.

Primary outcomes

The composite primary outcome of episode mortality and/or failure to clear lactate occurred in 128/199 (64%) of participants in the RBC/LyoPlas group, and in 136/210 (65%) of participants in the saline 0.9% group (see Table 5). After adjusting for IDS, allocation to treatment with RBC/LyoPlas was observed across the 432 study participants to have a 1% higher relative risk of episode mortality and/or failure to clear lactate than allocation to treatment with 0.9% saline with a 95% two-sided compatibility interval of 0.88 to 1.17, indicating a moderate range of plausible true treatment effects. The degree of evidence against the null hypothesis that the treatments are interchangeable is p = 0.86. After adjusting for IDS, allocation to treatment with RBC/LyoPlas was observed across the 432 study participants to have a 0.025% lower absolute risk of episode mortality and/or failure to clear lactate than allocation to treatment with 0.9% saline with a 95% two-sided compatibility interval of −9% to 9%, indicating a moderate range of plausible true treatment effects. The degree of evidence against the null hypothesis that the treatments are interchangeable is p = 0.996.

The qualifying events for the primary outcome in each treatment group are given in Table 7. Due to participant drop-out and missing data, 23 participants did not provide primary outcome data. For the 199 participants providing a primary outcome in the RBC/LyoPlas group, the most common outcome was survival and clearance of lactate (36%), 29% of participants experienced episode mortality and failed to clear their lactate, 20% survived but failed to clear their lactate and 15% cleared their lactate but experienced episode mortality. For the 210 participants providing a primary outcome in the 0.9% saline group, the most common outcome was experiencing episode mortality and failure to clear their lactate (36%), 35% of participants survived and cleared their lactate, 18% survived but failed to clear their lactate and 11% cleared their lactate but experienced episode mortality.

The individual components of the primary outcome were analysed separately (see Table 8). Episode mortality occurred in 88/203 (43%) of participants in the RBC/LyoPlas group, and in 99/218 (45%) of participants in the saline 0.9% group (see Table 8). Analyses, adjusted for IDS, yielded an estimated risk ratio of 0.97 (95% CI 0.78 to 1.20; p-value: 0.75) and an estimated risk difference of −3% (95% CI: −12% to 7%; p-value: 0.57). Failure to clear lactate occurred in 98/196 (50%) of participants in the RBC/LyoPlas group, and in 113/206 (55%) of participants in the saline 0.9% group (see Table 8). Analyses, adjusted for IDS, yielded an estimated risk ratio of 0.94 (95% CI 0.78 to 1.13; p-value: 0.52) and an estimated risk difference of −5% (95% CI −14% to 5%; p-value: 0.33). The estimates of treatment effects for each individual component produced relatively wide CIs including both benefits and harms. The degree of evidence against the null hypothesis that the treatments are interchangeable ranged from 0.33 to 0.75 across these analyses.

Table 6 summarises which of the primary outcome components were met by the 432 participants: the 409 who recorded a primary outcome, and the 23 who did not provide information to determine a primary outcome.

Secondary outcomes

The secondary outcomes are summarised in Tables 9–13. Selected secondary outcomes are summarised in Table 9. The volume of post intervention fluids was similar between treatment groups [adjusted mean difference −34 ml (95% CI –101 to 32)]. There was little evidence of any meaningful difference in the time taken to arrive at ED from either the time of injury [recorded as the time of the 999 call, adjusted mean difference 0.60 minutes (95% CI –6.14 to 7.35)] or from the time of randomisation [recorded as the time the first treatment box was opened, adjusted mean difference 3.03 minutes (95% CI –1.40 to 7.46)]. Vital signs recorded at ED arrival were similar across treatment groups [adjusted mean differences for heart rate −0.80 (95% CI –5.83 to 4.23) beats per minute, SBP −1.19 (95% CI –8.19 to 5.82) mmHg, diastolic blood pressure 2.26 (95% CI –3.77 to 8.29) mmHg, respiratory rate 0.59 (95% CI –0.79 to 1.97) per minute and oxygen saturation 0.48 (95% CI –0.86 to 1.82)]. Laboratory results showed that lactate concentration on ED arrival was similar across groups [adjusted mean difference −0.08 (95% CI −0.97 to 0.82) mmol/L]. Although INR was only collected on 158 participants, the proportion with an INR >1.5 was similar across treatment groups [adjusted risk ratio 0.91 (95% CI 0.44 to 1.90)]. Haemoglobin concentration on ED arrival was significantly higher in the RBC/LyoPlas group compared to the 0.9% saline group [adjusted mean difference 15 (95% CI 10 to 19) g/L]. A post-hoc analysis found that blood product use from admission to hospital through to 24 hours later was higher in the RBC/LyoPlas group than in the 0.9% saline group [adjusted mean difference of 1.80 (95% CI 0.58 to 3.01) units of RBCs and 1.54 (95% CI 0.57 to 2.50) units of plasma]. Measures of mortality within 3 hours of injury and within 30 days of mortality were both slightly lower in the RBC/LyoPlas group. Analyses, adjusted for IDS, yielded an estimated risk ratio of 0.75 (95% CI 0.50 to 1.13) and an estimated risk difference of −7% (95% CI –15% to 1%) for 3-hour mortality. For 30-day mortality, adjusted analyses produced an estimated risk ratio of 0.94 (95% CI 0.76 to 1.17) and an estimated risk difference of −4% (95% CI –13% to 6%).

Pre-hospital fluid (type and volume) and vital signs are summarised in Table 10. The proportion of participants receiving fluids given prior to intervention, and the volumes of fluids received, were similar between both groups [adjusted risk ratio 0.95 (95% CI 0.83 to 1.07) and adjusted mean difference −17 (−108 to 74) ml]. The proportion of participants receiving fluids given after intervention, and the volumes of fluids received, were also similar between both groups [adjusted risk ratio 0.84 (95% CI 0.58 to 1.21) and adjusted mean difference −34 (−101 to 32) ml]. Vital signs were recorded at regular intervals from arrival on scene to 24 hours after arrival at hospital. Vital signs are only summarised for participants who were still alive at the time of assessments (i.e. no values of 0 are imputed for participants known to have died at a previous time point). All vital signs were broadly similar across both treatment groups at each time point. Mean heart rate values appeared to stabilise at around 86–90 beats per minute two hours after participants arrived in hospital. SBP increased from a mean of 73 mmHg on scene to around 110–117 mmHg at all in-hospital assessments. Diastolic blood pressure increased from a mean of 46 mmHg on scene to a mean of 73 mmHg when arriving at hospital, before reducing to a mean of 62 mmHg 12 hours after arriving at hospital. Respiratory rate was slightly elevated on scene with a mean of 23 per minute, but this reduced to an average rate between 18 and 20 per minute after arrival at hospital. Oxygen saturation was relatively low on scene with a mean of 92%, but this increased to a mean saturation greater than 97% once participants arrived in hospital. Lactate concentration was similar across both arms at each time point with values decreasing from arrival at hospital to 2 hours after arrival at hospital. At 2 and 6 hours after arrival at hospital, the proportion of participants with an INR >1.5 was similar across treatment groups, however given the low incidence of participants with an INR >1.5 there is considerable uncertainty associated with these estimates. Calcium concentration on ED arrival was similar across treatment groups [adjusted mean difference −0.03 (95% CI −0.12 to 0.05) mmol/L].

Total blood product receipt was recorded at 6, 12 and 24 hours after arrival at hospital and is summarised in Table 12. The mean number of units of RBCs used at each time point was slightly higher in participants in the RBC/LyoPlas group compared to the 0.9% saline group. Similarly, the mean number of units of plasma used at each time point was also slightly higher in the RBC/LyoPlas group compared to the 0.9% saline group. Participants on the RBC/LyoPlas arm received a higher volume of crystalloid than participants on the 0.9% saline arm [adjusted mean difference at 12 hours after hospital arrival 628 (95% CI 211 to 1034) ml, adjusted mean difference at 24 hours after hospital arrival 708 (95% CI 180 to 1236) ml]. The number of bags of cryoprecipitate and platelets was similar across all three time points. The volume of colloid was lower in participants in the RBC/LyoPlas group compared to the 0.9% saline group, but the high variability of colloid volume means there is considerable uncertainty associated with these adjusted mean differences.

Further secondary outcomes are summarised in Table 13. Acute respiratory distress syndrome (ARDS) was recorded in 9/142 (6%) participants in the RBC/LyoPlas group and 3/129 (2%) in the saline 0.9% group, adjusted relative risk 2.71 (95% CI 0.75 to 9.81). The rate of transfusion-related complications in the first 24 hours after arrival in hospital was similar across the two treatment groups [11/148 (7%) in the RBC/LyoPlas group and 9/137 (7%) in the 0.9% saline group, adjusted risk ratio 1.05 (95% CI 0.46 to 2.42)]. The number of organ failure-free days experienced by participants was similar across treatment groups [mean 12.19 days (SD 13.0) in the RBC/LyoPlas group and 12.1 days (SD 13.1) in the 0.9% saline group, adjusted mean difference 0.86 (95% CI −1.64 to 3.36) days].

Laboratory results, ROTEM and Multiplate are summarised in Table 11. Coagulation measured viscoelastically by rotational thromboelastometry (ROTEM^©) and platelet function using multiple electrode impedance aggregometry (MultiPlate) data was only collected on participants from selected receiving hospitals. The measurements taken for both sets of EXTEM and FIBTEM tests were similar across treatment group, suggesting that participants in each treatment arm exhibited equivalent coagulation measured viscoelastically profiles. Multiplate data were only collected on 32 participants, and the uncertainty associated with these small sample sizes makes it hard to reach any meaningful conclusions other than that there was an absence of evidence of any meaningful difference between treatment arms.

Exploratory outcomes are summarised in Table 14. Both the length of stay in ITU and in hospital, recorded from admission to hospital up to a maximum of 30 days, were similar between treatment groups [adjusted mean differences of −0.40 days (95% CI −2.84 to 2.04) and −1.67 days (95% CI −4.31 to 0.97) respectively]. The occurrence of organ failure on at least one day during hospital stay was assessed for each organ system. Around two-thirds of participants experienced organ failure on at least one day in their respiratory, neurological or cardiovascular systems. One in four participants experienced organ failure on at least one day in their renal system, one in eight participants experienced organ failure on at least one day in their coagulation system, and one in 13 experienced organ failure on at least one day in their liver system. Of the 406 participants providing a response, nearly all (97%) received an initial dose of TXA, of the 307 participants providing a response, over two-thirds (69%) received a second dose. Rates of surgery in the first 24 hours following admission to hospital were similar in both treatment groups, with the highest rates reported up to 6 hours after arrival at hospital.

Exploratory Bayesian analyses

We decided a priori to include a Bayesian analysis of the primary outcome and its individual components. The rationale for doing so was to directly estimate the probability of a clinically meaningful treatment effect, which is a measurement of direct interest to clinicians. 18

Following the DMEC and TSC meetings in May and October 2018, the power calculations were reframed in terms of relative risk rather than absolute risk while maintaining the original target sample size of 490. Based on the observed pooled event rate in May 2018 (65%), and allowing for a 10% loss to follow-up rate, 490 participants would provide 80% power to detect a relative risk ratio of 0.82. This effect size was used to inform the sceptical and information prior distributions in the exploratory Bayesian analyses.

Bayesian models were fitted using three different prior distributions: a non-informative prior, a sceptical prior such that the probability of observing a treatment effect at least as large as a relative risk ratio of 0.82 is <5% and an informative prior reflecting current knowledge. For each set of the priors, Table 15 provides summary statistics (the mean value, and upper and lower 2.5% quantiles) of the primary outcome event rates in each treatment group. Posterior probabilities of primary outcome rates by treatment group were estimated for Bayesian analyses using each of the three priors, encompassing varying assumptions of benefit from RBC/LyoPlas.

Sensitivity analyses

Sensitivity analyses were undertaken for the primary outcome and each component of the primary outcome: episode mortality and failure to clear lactate. For each sensitivity analysis, the adjusted risk ratio and adjusted risk differences are reported.

Subgroup analyses

Subgroup analyses were conducted for the composite primary outcome and for episode mortality.

Safety outcomes

The rates of complications and adverse events by treatment group are reported in Table 28. There was only one reported transfusion-related acute lung injury, and rates of thromboembolism were similar across the treatment groups. There was the suspicion or clinical evidence of infection in 65% of participants who completed at least one in-hospital daily assessment. This rate was similar across both treatment groups. There were only two serious adverse events recorded in the RePHILL trial. There were no required dose reductions during the trial, and no participants needed to have their trial treatment discontinued for drug-related toxicity. None of the deaths in the RePHILL study were related to the trial treatments.

Blood product wastage

RBC and LyoPlas wastage was measured at two of the trial blood banks. In total, 2496 RBC units were issued, 171 units were administered to patients, 2325 units were returned to blood bank stock and 53 units required disposal. This equates to a wastage rate of 2.12%. Amongst 2496 units of LyoPlas that were issued, 31 were wasted, equating to a rate of 1.24%.

Good clinical practice of compliance statement

The trial was run in accordance with the requirements of Good Clinical Practice. No serious breaches occurred during the conduct of the trial.

Equality, diversity and inclusion

Given the unpredictable nature of life-threatening haemorrhagic shock and the need for emergency treatment, it was not possible for us to take active steps to enrich the diversity of the population of patients recruited. However, the population enrolled broadly reflected the ethnicity characteristics of England and Wales (www.ethnicity-facts-figures.service.gov.uk/uk-population-by-ethnicity/national-and-regional-populations/population-of-england-and-wales/latest, accessed 6 July 2022). The study protocol excluded people with a known objection to blood or blood product transfusion in order to respect the individual’s rights to accept or refuse treatment. We also excluded women who were known or suspected to be pregnant due to the potential greater transfusion risks in this population. Finally, we excluded prisoners due to the perceived difficulty of obtaining follow-up data for this population. In subsequent research, we have shown that it is feasible to obtain survival outcomes data for this group of patients through the Office for National Statistics. 31

Strengths and limitations

RePHILL is the first RCT to evaluate the clinical effectiveness of pre-hospital blood transfusion. The trial overcame substantial regulatory, supply and logistical challenges. It is the first to engage multiple air ambulance charities in a randomised clinical trial of an investigational medicinal product. The case mix of participants enrolled were typical of UK civilian major trauma. The injury severity scores confirm that the group of participants enrolled had, as anticipated, severe injuries. These findings support the generalisability of the research findings to the UK civilian trauma population.

Alongside these strengths, the trial also has limitations. First, we recruited only 88% of our planned sample size due to the impact of COVID-19 and the withdrawal of several air ambulance partners. Given the similarity of the primary and secondary outcomes between the intervention and control group, although it is possible that not recruiting the full sample size may have led to a type 2 error, we consider it unlikely that it would have led to a substantively different finding. This position is further supported by simulations examining the effect of large increases in sample sizes which would not have materially altered the findings for the primary outcome. Whether a larger trial would have influenced the secondary outcomes, remains to be determined in future studies.

COVID-19 had a global impact on clinical trials leading to thousands of clinical trials closing prior to reaching the intended sample size. 32 The CONSORT and SPIRIT Extension for RCTs Revised in Extenuating Circumstances (CONSERVE) statement provides guidance to help improve the transparency, quality and completeness of reporting. 33 Adopting that framework in the context of RePHILL can be summarised as:

• Extenuating circumstance: (1) COVID-19 led to an inability to maintain safe oversight for the trial due to a combination sickness and redeployment of clinical and research staff. (2) National lockdowns reduced the incidence of major trauma.

• Impact: (1) Sponsor paused trial recruitment (March 2020 through to July 2020). (2) Fewer cases of major trauma leading to lower recruitment rates. Together these meant that the study fell short of reaching the intended sample size.

• Mitigations: Recruitment window extended; addition of Bayesian analysis.

• These are important modifications as they reduced the statistical power of the study.

Although multi-centre RCTs are considered the gold standard of evidence, it takes considerable time to design and recruit sufficient numbers. Other pre-hospital transfusion trials have faced challenges – with the Control of Major Bleeding After Trauma – COMBAT stopped prematurely due to futility9 and the Pre-Hospital Use of Plasma for Traumatic Hemorrhage – PUPTH due to insufficient recruitment. 34 The withdrawal of several air ambulance services prior to the start of the trial also adversely affected delivering the trial on time and target. The clinical and commercial pressures for early adoption (www.londonsairambulance.org.uk/news-and-stories/charity/uks-first-air-ambulance-to-carry-blood-on-board, accessed 25 February 2022)35,36 and loss of equipoise37 meant that early evaluations of pre-hospital blood were limited to observational studies38,39 and reduced the pool willing to participate in a randomised evaluation. Future randomised studies will need to carefully consider the pressures for early adoption inherent in this sector and work carefully with air ambulance partners to agree future research priorities.

Research in trauma-related haemorrhagic shock is limited by the absence of a core outcome set. A systematic review examining outcome measures used in clinical research evaluating pre-hospital blood component transfusion in traumatically injured bleeding patients identified over 212 outcome measures across 34 studies,40 highlighting significant heterogeneity in outcomes across previous trials. A composite primary outcome reflecting the efficacy of initial resuscitation (lactate clearance) and overall effectiveness (death) was chosen after careful consideration by the trial investigators, funders and patient and public involvement groups. Despite the empirical attractiveness to this composite outcome, the direction of effects observed in RePHILL for mortality −3% (−12% to 7%) and improved lactate clearance −5% (−14% to 5%) do not appear to have been additive when combined as a composite outcome [i.e. negligible difference seen in the composite outcome −0.025% (−9% to 9%)]. This could be explained by the most seriously injured patients, at highest risk of early death, would also have high lactate loads. An effective treatment that averted early death could be mitigated by producing a survivor with impaired lactate clearance. This suggests investigators should be cautious about its use in future studies.

Given the nature of the intervention and the setting for administration, it was not possible to mask treatment allocation to those delivering the intervention. It is possible that this may have led to performance bias during both the pre-hospital and ED treatment of enrolled patients. The key trial outcomes were objective measurements and thus it was unlikely that they were influenced by knowledge of treatment assignment. More subjective measures (e.g. transfusion reactions, adverse events) may have been more susceptible to detection bias.

Future research

There remains an urgent need to develop evidence-based interventions to improve outcomes from haemorrhagic shock secondary to acute traumatic injuries.

A key challenge is to identify the population of patients most likely to benefit from pre-hospital interventions. 41 The RePHILL trial recruited a population of severely injured patients characterised by high injury severity scores and overall high episode mortality. 42 By contrast, the COMBAT trial recruited less severely injured patients (and had a short time between intervention and hospital arrival) and similarly did not find evidence of benefit from up to two units of thawed plasma. 9 The PAMPer trial,8 the only trial to find a benefit from pre-hospital transfusion (of up to two units of plasma) recruited a group of moderately injured patients who were on average 42 minutes from away from definitive hospital-delivered treatments. This suggests there may be an optimal set of inclusion/exclusion criteria for recruitment to future trials which avoids recruiting those unlikely to benefit from having either catastrophic injury (e.g. traumatic cardiac arrest) or least severe injuries who are likely to survive irrespective of pre-hospital transfusions. However, there remains a key challenge to identify this population of patients at the roadside. Whilst some post-hoc analyses (e.g. longer pre-hospital transport time, CT evidence of brain injury) suggest there may be subgroups with better outcomes, none of the a priori subgroup analyses across the three trials, which can be assessed at the point of injury, demonstrated evidence of benefit. Further research is required to identify the characteristics of patients most likely to benefit from pre-hospital transfusion.

The discordant findings from the four prior randomised transfusion trials highlight the need for further research to identify the optimal intervention(s) in trauma-related haemorrhagic shock. 8,9,19,42,43 Consideration should be given to both the components transfused, and the sequence and dose. Such research should be informed by the findings (once published) of the Pre-hospital Plasma or Red Blood Cell Transfusion Strategy in Major Bleeding (PRIEST; www.clinicaltrials.gov/ct2/show/NCT04879485, accessed 28 September 2022). 44 The role of whole blood, which reduces the need for transfusing individual components, showed promise in a single-centre pilot randomised trial45 and is being explored in the Type O Whole Blood and Assessment of Age During Pre-hospital Resuscitation Trial (TOWAR; https://clinicaltrials.gov/ct2/show/NCT04684719, accessed 28 September 2022)46 and Study of Whole Blood In Frontline Trauma (SWIFT; www.nhsbt.nhs.uk/clinical-trials-unit/current-trials-and-studies/swift/trial, accessed 28 September 2022). 47 Noting that early treatment is critical for best outcomes in traumatic haemorrhage,48,49 future trials will need to consider how to enable interventions to be delivered earlier in the patient pathway.

A systematic review of outcome measures used in clinical research evaluating pre-hospital blood component transfusion patients with trauma-related haemorrhagic shock identified substantial heterogeneity in the outcomes reported. Many outcomes focused on evidence of clinical effectiveness with limited insights provided into safety and complications. 40 The absence of a core outcome set for pre-hospital transfusion trials presents a significant challenge for researchers seeking to identify the optimal outcomes for individual trials and to facilitate meta-analyses between trials. The Core Outcomes Set for Cardiac Arrest (COSCA) includes survival to hospital discharge (or 30 days), survival with a favourable neurological outcome and health-related quality of life. 50 While survival to and beyond hospital discharge may be desirable for large effectiveness trials, any treatment effect from pre-hospital transfusion will likely be diluted due to severity of underlying injuries and treatments administered after hospital arrival. Future trials that wish to demonstrate evidence of efficacy may need to focus on outcomes earlier in the patient journey such as survival to hospital admission or within the first 24 hours of injury. Further research and consensus work is needed to define the optimal outcomes set for efficacy and effectiveness trials, what constitutes the minimal clinically important differences for these outcomes and what the optimal experimental designs are to answer the specific research questions. 18

Contributions of authors

Nicholas Crombie (https://orcid.org/0000-0003-3740-5530) (Consultant Trauma Anaesthetist) conceived, designed and obtained funding for the study, supervised the research, wrote the original draft and reviewed and approved the draft.

Heidi A Doughty (https://orcid.org/0000-0002-5444-6222) (Consultant in Transfusion Medicine) conceived, designed and obtained funding for the study, supervised blood and LyoPlas supplies, wrote the original draft and reviewed and approved the draft.

Jonathan RB Bishop (https://orcid.org/0000-0003-1789-5886) (Senior Medical Statistician) conceived, designed and obtained funding for the study, was responsible for data curation, formal analysis, data verification and data visualisation, wrote the original draft and reviewed and approved the draft.

Amisha Desai (https://orcid.org/0000-0002-7305-8401) (Advanced Clinical Pharmacist) supervised investigational medicinal product management, and reviewed and approved the draft.

Emily F Dixon (https://orcid.org/0000-0003-3660-5354) (Trial Manager) co-ordinated the study and reviewed and approved the draft.

James M Hancox (https://orcid.org/0000-0002-5797-0898) (Critical Care Paramedic) supervised patient recruitment in collaboration with sites and reviewed and approved the draft.

Mike J Herbert (Blood Transfusion Discipline Lead) supervised blood and LyoPlas supplies and reviewed and approved the draft.

Caroline Leech (https://orcid.org/0000-0002-5301-9798) (Consultant in Emergency Medicine and Pre-hospital Emergency Medicine) managed resources at sites and supervised patient recruitment and reviewed and approved the draft.

Simon J Lewis (https://orcid.org/0000-0002-2350-1436) (Consultant in Emergency Medicine and Pre-hospital Emergency Medicine) managed resources at sites and supervised patient recruitment and reviewed and approved the draft.

Mark R Nash (https://orcid.org/0000-0002-3793-4937) (Consultant Anaesthetist) managed resources at sites and supervised patient recruitment and reviewed and approved the draft.

David N Naumann (https://orcid.org/0000-0003-2243-2325) (Military Surgeon, Trauma Fellow) conceived, designed and obtained funding for the study and wrote the original draft.

Karen Piper (https://orcid.org/0000-0001-5087-7052) (Programme Manager: NIHR Surgical Reconstruction and Microbiology Research Centre) who played a substantial role in the conception and design of the study and obtained funding for the study.

Gemma Slinn (https://orcid.org/0000-0003-1096-4148) (Trials Management Team Leader) managed the set-up and oversaw the operational delivery of the study at Birmingham Clinical Trials Unit, and reviewed and approved the draft.

Hazel Smith (https://orcid.org/0000-0002-7652-7699) (Research Paramedic) supervised patient recruitment in collaboration with sites, wrote the original draft, and reviewed and approved the draft.

Iain M Smith (https://orcid.org/0000-0001-5941-410X) (Specialty Registrar in General Surgery) conceived, designed and obtained funding for the study and reviewed and approved the draft.

Rebekah K Wale (https://orcid.org/0000-0003-1343-7538) (Senior Trial Manager) co-ordinated the study and reviewed and approved the draft.

Alistair Wilson (https://orcid.org/0000-0001-8042-1654) (Consultant Trauma Surgeon) managed resources at sites and supervised patient recruitment and reviewed and approved the draft.

Aisling Crombie (https://orcid.org/0000-0003-4739-9613) (Deputy Director of Nursing and Quality) played a substantial role in the conception and design of the study and obtained funding for the study.

Mark Midwinter (https://orcid.org/0000-0003-1836-7137) (Professor of Clinical Anatomy, General and Trauma Surgeon) played a substantial role in the conception and design of the study and obtained funding for the study.

Natalie Ives (https://orcid.org/0000-0002-1664-7541) (Reader in Clinical Trials) conceived, designed and obtained funding for the study, was responsible for data curation, formal analysis and data visualisation, supervised the research, wrote the original draft and reviewed and approved the draft.

Gavin D Perkins (https://orcid.org/0000-0003-3027-7548) (Professor in Critical Care Medicine) conceived, designed and obtained funding for the study, supervised the research, wrote the original draft and reviewed and approved the draft.

We thank the patients and their relatives who supported the trial, principal investigators and other research staff at ambulance services, receiving hospitals, air ambulance charities, participating blood banks and pharmacy staff, the blood bikers, Birmingham Clinical Trials Unit and Sponsors office, the Trial Steering Committee, and the Data Monitoring and Ethics Committee. This project is funded by the Efficacy and Mechanism Evaluation Programme (study reference 14/152/14) of the National Institute for Health and Care Research (NIHR).

Ethics statement

The study was approved by South Central Research Ethics Committee (15/SC/0691) on 15 December 2015.

Data-sharing statement

Requests for access to data from the RePHILL trial should be addressed to the corresponding author at rephill@trials.bham.ac.uk. The individual participant data collected during the trial (including the data dictionary) will be available, after de-identification, with no end date. All proposals requesting data access will need to specify how the data will be used, and all proposals will need the approval of the trial management group before data release.

Disclaimers

This report presents independent research. 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, the MRC, the EME programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the EME programme or the Department of Health and Social Care.

Trial Management Group

Prof. Gavin Perkins, Dr Nicholas Crombie, Iain Smith, Dr Heidi Doughty, Dr David Naumann, Hazel Smith, Dr Margaret Grant, Gemma Slinn, Dr Rebekah Wale, Emily Dixon, Dr Karen Piper, Deborah Papoola, Amisha Desai, Natalie Ives, Dr Jon Bishop

Data Monitoring Committee

Prof. John Nicholl (Chair), Dr Jan Jansen, Prof. Fiona Lecky

Trial Steering Committee

Prof. Ian Roberts (Chair), Prof. John Holcomb, Dr Simon Stanworth, Prof. Jason Smith, Prof. Timothy Coats, Andrew Cox, Timothy Marshall

Blood Banks and Pharmacies

Mike Herbert, Mindy Sahota (New Cross Hospital); Camran Khan, Zeeshan Parvez (Worcester Acute Hospital); Tina Taylor, Julie Nortcote, Mojid Khan (University Hospitals Coventry and Warwickshire); Claire Newsam, Katherine Philpott, Lynne Whitehead (Addenbrooke’s Hospital); Debbie Asher, Gail Healey (Norfolk and Norwich University Hospital)

Blood Bikers

Midland Freewheelers, Warwickshire and Solihull Blood Bikers

Air Ambulance Sites (IDS)

Alistair Wilson (PI), Pam Chrispin (PI), Richard Hinson (East Anglian Air Ambulance); Caroline Leech (PI), Mark Beasley, Sam Cooper (The Air Ambulance Service); Mark Nash (PI) (West Midlands Ambulance Service); Simon Lewis (PI), Oliver Robinson, Rod Mackenzie (Magpas Air Ambulance)

Receiving Hospital Sites

David Yeo (PI; Queen Elizabeth Hospital); Caroline Leech (PI; University Hospitals Coventry and Warwickshire); Tom James (University Hospitals of North Midlands); Jason Kendall (PI), Johannes von Vopelius-Feldt (PI) (Southmead Hospital Bristol); Christopher Gough (PI) (Queen’s Medical Centre Nottingham); Gary Mills (PI; Northern General Hospital Sheffield); Aquib Hafeez (PI; Oxford University Hospitals); Sarah Hazleman (PI; Addenbrooke’s Hospital); Francoise Sheppard (PI; Norfolk and Norwich University Hospital); Ben Bloom (PI; Royal London Hospital)