Background
Uncontrolled bleeding is the most common preventable cause of death in major trauma [
1]; it affects up to 40% of patients with severe injury. Trauma haemorrhage is exacerbated by a complex interplay of clotting abnormalities and may result in a trauma-induced coagulopathy (TIC). Activation of protein C is central to TIC [
2] and associated with increased fibrinolysis [
3] and loss of fibrinogen [
4]. Hypofibrinogenaemia occurs early after injury and is an independent predictor of death [
5]. Augmentation of blood fibrinogen levels necessitates the use of a concentrated form of fibrinogen supplementation (e.g., cryoprecipitate or fibrinogen concentrate [FgC]) because fresh frozen plasma (FFP) alone is ineffective [
4,
6].
There is continued debate about the comparative effectiveness of the two concentrated fibrinogen treatments [
7,
8], despite two recent feasibility randomised controlled trials (RCTs) in trauma haemorrhage (CRYOSTAT 1 [
9] and FiiRST [
10]). The aim of this paper is to report the findings of a second feasibility study in adult trauma haemorrhage using FgC. The primary objectives of the E-FIT 1 study were to determine whether it was possible to deliver FgC therapy early (within 45 minutes) to adult trauma patients and the proportion of participants whose fibrinogen levels were maintained ≥ 2 g/L during active haemorrhage.
Methods
Study design E-FIT 1 trial
The E-FIT 1 study was a multi-centre, double-blind, placebo-controlled RCT conducted in five UK major trauma centres. The study is registered with
www.controlled-trials.com (ISRCTN67540073).
Eligibility criteria and randomisation
Trauma patients were eligible if they were adults (judged to be aged 16 years or older), were actively bleeding and in haemorrhagic shock and therefore required activation of the major haemorrhage protocol (MHP) or had already received a transfusion of emergency red blood cells (RBC). Exclusion criteria included patient transferred from another hospital, the trauma team leader deemed the injury incompatible with life, more than 3 hours had elapsed from time of injury, pregnant women and severe isolated or unsalvageable head injury. Women of childbearing age had a point-of-care blood test performed to rule out pregnancy (Abbott, Princeton, NJ, USA).
Participants were block randomised in a 1:1 ratio of placebo to active arm. An independent statistician produced the allocation sequence using a computer-generated random sequence, and randomisation lists were produced centrally. Allocation concealment was maintained by the labelling of study packs prior to release to sites. Identical study packs containing either FgC or placebo were sequentially labelled, and participants were allocated to the next available study pack. All research site staff, participants and trial management staff were blinded to study allocation.
Consent
An emergency waiver with independent agreement process was used. Written informed consent from the participant was sought as soon as practically possible after study entry for continuation in the trial. If the participant did not regain capacity, agreement was sought from the participant’s next of kin or other appropriate representative. The protocol and consent process was approved by the NHS National Research Ethics Service (NRES) Committee South Central Oxford C Ethics Committee (15/S3/0316) and the Medicines and Healthcare products Regulatory Agency (MHRA) (25224/0003/001-0001).
Study intervention
Participants were randomised equally to treatment or placebo. All randomised participants received MHP, and the study intervention was started as soon as possible and within 45 minutes of hospital arrival. Typically an MHP constituted two transfusion packs—pack 1 followed by repeated use of pack 2—until bleeding was controlled. Pack 1 included 4 RBC and 4 FFP; pack 2 included 4 RBC, 4 FFP, 10 U of cryoprecipitate (approximately 300 ml, 4 g of fibrinogen) and 1 pool platelets. (Specifications for blood components are found in the UK ‘red book’ [
11]). Standard laboratory clotting tests were taken throughout active bleeding, and a fibrinogen level of < 1.5 g/L was the trigger for additional cryoprecipitate, when necessary.
The dose of FgC was chosen using data from the CRYOSTAT 1 trial [
9]. In this trial, two pools of UK cryoprecipitate, containing approximately 4 g of fibrinogen, constituted the study intervention, and fibrinogen levels were maintained above 1.8 g/L during active bleeding. Modelling data from a multi-centre European study have demonstrated that a fibrinogen level of 2.3 g/L is associated with the lowest mortality rates [
12]. A dose of 6 g of FgC was therefore chosen because 2 g of fibrinogen raises the blood level in patients with major bleeding by approximately 0.5 g/L [
13].
In the treatment arm, an infusion of 6 g of FgC (RiaSTAP; CSL Behring, King of Prussia, PA, USA) was administered as soon as possible, and an equivalent volume (300 ml) 0.9% saline was administered in the placebo arm. FgC and placebo were blinded interventions. Study intervention packs were held in the emergency department (ED). Blinding was maintained by research staff following a validated protocol [
14,
15]. Reconstitution of the study intervention was completed at the patient’s bedside by research staff. The intervention drug was drawn into black syringes and infused as an intravenous bolus over 5 minutes. Participants were deemed to have received the study intervention if at least five whole syringes were infused.
Outcomes
The primary outcome was the feasibility of administering the study intervention within 45 minutes of admission, defined by the proportion of all participants randomised who started their infusion within that time. For this trial to be successful, at least 90% of the participants were required to achieve this target. An additional primary outcome was to determine the proportion of participants whose fibrinogen level remained at 2 g/L or above during active haemorrhage.
Secondary outcome measures were clinical and laboratory measures of efficacy and safety. Clinical outcomes included mortality at 3, 6 and 24 hours, and 28 days from admission; transfusion requirements, in numbers of units, at 3, 6 and 24 hours; duration of organ support; in-patient stay, including the intensive care unit/high-dependency unit; and quality of life. Safety was measured by symptomatic thrombotic events and arterial (e.g., myocardial infarction, stroke) and venous (e.g., pulmonary embolism, deep venous thrombosis) events during hospital stay. Laboratory measures included Clauss fibrinogen level at 2 hours after admission and at day 7 from admission. Standard MHP protocols in the UK recommend frequent blood samples every 30–60 minutes during trauma haemorrhage [
16] and were taken as standard measurements.
Sample and data collection
Blood samples were drawn immediately upon admission to the resuscitation room. Clauss fibrinogen measures were analysed in the hospital coagulation laboratories, according to standard operating procedures. Patient characteristics, mechanism and severity of injury, and admission physiology were collected and scored using the Abbreviated Injury Scale, Injury Severity Score and Glasgow Coma Scale. Tranexamic acid and other haemostatic drug administrations were recorded. Organ support was defined using the composite time to complete organ failure resolution score [
17]. Organ failure was defined using the Sequential Organ Failure Assessment score. Data on timing of transfusions, clear fluids and mortality were collected in the first 24 hours. Measures to mitigate venous thromboembolic risk were recorded weekly to day 28. Symptomatic thromboembolic disease was categorised according to standard clinical and/or radiological measures. Quality of life (EQ-5D-5L) questionnaires were completed upon discharge or day 28, whichever was the sooner. The EQ-5D-5L descriptive systems index value was calculated, allowing the five health dimensions to be converted into a single numeric measure [
18].
Sample size and data analysis
If the proportion of participants who received the study intervention within 45 minutes from admission was 90%, a sample size of 40 would yield a 95% CI for this estimate of between 76% and 97%. To allow for 20% drop-out, the final sample size was chosen to be 48 participants in total, with 24 participants per arm. All analyses were performed according to the intention-to-treat principle and included all randomised participants.
Clinical and laboratory measures were compared using Fisher’s exact test or the Mann-Whitney test for categorical or continuous data, as appropriate. Normal linear regression was used to assess whether fibrinogen at 2 hours from admission was different between the two arms, adjusting for values at admission. Normal linear regression, unadjusted for any other factors, was used to assess whether fibrinogen at 7 days from admission was different between arms. Residual plots from each normal linear regression were examined for evidence of non-linearity, skew or non-constant variance. Log-transformed fibrinogen would be used in the regression model(s) if any evidence was found. All-cause mortality was estimated using the Kaplan-Meier method and compared using the log-rank test. Duration of organ support and hospital stay were estimated using competing risks methods and compared using Gray’s test. Death prior to the event of interest was considered the competing risk.
Sensitivity analyses were conducted for the primary outcomes, treating cases with missing time of administration of the study intervention as administration within 45 minutes (best-case scenario) or as administration beyond 45 minutes (worst-case scenario) and treating cases with missing fibrinogen at 2 hours from admission as achieving (best-case scenario) or not achieving (worst-case scenario) a level of at least 2 g/L. Except where specified, all analyses were unadjusted, and there was no adjustment for multiple testing. All statistical tests were two-sided. All analyses were undertaken using SAS/STAT software version 9.4 (SAS Institute Inc., Cary, NC, USA).
An independent data monitoring committee monitored all safety events throughout the study. All serious adverse events were evaluated and classified independently by the co-chief investigators, and any disagreements were resolved by consensus. The study manuscript was produced according to Consolidated Standards of Reporting Trials (CONSORT) recommendations for the reporting of randomised clinical trials [
19].
Discussion
This is the second RCT evaluating administration of FgC in trauma. Despite median administration times approximating 40 minutes, the range varied widely (10 to 82 minutes), and only 69% of participants received study intervention within 45 minutes of admission. There was some evidence of a difference in the proportion of participants with fibrinogen levels ≥ 2 g/L between arms (
p = 0.10). However, the FgC group had a fibrinogen level of 1.6 g/L on admission (0.7 g/L lower than the placebo arm), which likely reflects the trends towards higher injury severity and greater degree of shock in the intervention arm. Importantly, the fibrinogen level in the FgC arm rose by a mean of 0.9 g/L (SD, 0.5) compared with a reduction of 0.2 g/L (SD, 0.5) in the placebo arm, and average fibrinogen level at 2 hours from admission was significantly higher in the FgC arm (
p < 0.0001). Fibrinogen levels did not remain elevated by 7 days, a finding also shown in the two other feasibility RCTs [
9,
10], suggesting no long-term effect of fibrinogen replacement.
Transfusion requirements were not different between arms; however, it was notable that there was a trend towards increased cryoprecipitate use in the FgC arm. Transfusion needs were greater for all blood components in the FgC arm, again a difference likely reflecting the higher injury burden and shock, the empiric nature of transfusion administration, and the lower admission level of fibrinogen. The proportion of patients achieving haemostasis at 3 hours, however, was identical in both arms, suggesting that duration of bleeding does not fully explain the need for greater cryoprecipitate use in the FgC arm (data not shown).
No safety signal for thrombotic events was detected in this study. Venous thromboembolism was reported only in the FgC arm and was under 10%, a rate similar to that in a recent FgC trauma study in which researchers also used 6 g of FgC as their intervention [
10]. More deaths were recorded in the FgC arm (
n = 8 vs.
n = 3), with no difference in deaths resulting from bleeding between arms, although the study was not powered for mortality. Differences in death rates may represent the variability in baseline characteristics between groups (i.e., higher injury severity and worse organ failure in the FgC arm).
Rapid reconstitution is often cited as an important benefit of FgC. Our study, and that of Nascimento et al. [
10], reported similar median FgC reconstitution times: 23 minutes (SD, 9) (E-FIT1) and 26 minutes (SD, 5) (FiiRST). In a non-RCT setting, dissolution of fibrinogen for in vitro testing has been reported within 30 seconds [
20], although it most commonly takes 10 minutes [
21]. The longer reconstitution times in our study were due to the need to dissolve the fibrinogen whilst maintaining allocation concealment. The study vials were kept in cardboard tamper-proof boxes, meaning that dissolution of fibrinogen powder (or placebo) could not be visually assessed and that each vial needed to ‘rest’ for 3 minutes prior to being drawn into a syringe [
14]. In an unblinded RCT, time to factor concentrate administration was shorter at 10 minutes (IQR, 10–16) [
22]. A future alternative source of concentrated fibrinogen may be extended shelf-life cryoprecipitate [
23], which could be pre-thawed and held in the ED, avoiding reconstitution times, but this has yet to be tested in an RCT setting.
The inclusion criteria for this trial did not use fibrinogen level. Goal-directed transfusion therapy using viscoelastic haemostatic assays (VHAs) (thromboelastography or rotational thromboelastometry [ROTEM]) is standard practice in many trauma centres in Europe and the United States [
22,
24,
25], and it is advocated for guiding transfusion therapy. In a retrospective European study, VHA-guided therapy has led to a 50% reduction in the incidence of massive transfusions for trauma compared with that predicted by the trauma-associated severe haemorrhage score, with an associated reduction in mortality from 33% to 22% [
26]. However, delays to first results are incurred (on average, 15–20 min [
21,
22]), and empiric, immediate transfusion therapy for patients with uncontrolled haemorrhage is a standard of care [
27]. To date, in trauma haemorrhage, there are no evidence-based thresholds to guide fibrinogen treatment, only expert consensus [
28]. A small RCT, which was terminated early, used dual ROTEM measures to guide transfusion (FIBTEM A10 > 8 mm, EXTEM CT < 78 seconds), and these data may support the use of ROTEM to direct transfusion therapy [
22] but need validation in a larger study. A large European study is due to be completed in 2018 (iTACTIC; ClinicalTrials.gov, NCT02593877) which will compare the efficacy of VHA and standard clotting tests for transfusion in trauma.
Our study was designed to include participants with the clinical phenotype of severe bleeding, not those exclusively with a low fibrinogen level. We did not use VHA or Clauss fibrinogen for two reasons: (1) to minimise delay to randomisation (VHA testing incurs 10–15-min delays) when speed of delivery of transfusion is known to be important [
27,
29,
30] and (2) to include participants with the broader clinical entity of bleeding. The recent RETIC trial used FIBTEM A10 for eligibility, and randomisation took approximately 35–38 minutes, compared with 15–16 minutes in our study [
22].
This is one of the first published multi-centre UK trauma transfusion studies, and it provides evidence that the delivery and conduct of RCTs in the challenging environment of the ED is possible. Prior to study start it was predicted that the recruitment would take 18 months. Recruitment was completed within 10 months, facilitated both by the research infrastructure and by the ability of research teams to recruit outside core hours. Rapid recruitment has recently been seen in another transfusion trauma study (FEISTY; ClinicalTrials.gov, NCT02745041) showing an encouraging appetite for the collection of high-quality collaborative study data.
This study has several limitations. It is a small feasibility study, and large differences were seen between treatment groups, in particular shock parameters and fibrinogen levels, which will alter treatment effects. Treatments for trauma haemorrhage are more effective when given earlier in the time course of major bleeding, as demonstrated by the CRASH-2 data [
29], and it is possible that a time delay incurred in the delivery of FgC in this study may have attenuated potential clinical benefits. This study was designed to test the ability of the research teams to administer FgC rapidly, and secondary endpoints should be viewed cautiously because the study was not powered for evaluation of clinical outcomes.