The acute management of trauma hemorrhage: a systematic review of randomized controlled trials

Trauma is one of the world's leading causes of death and disability. Around 40% of deaths are due to bleeding or its consequences, establishing hemorrhage as the most common cause of preventable death in this clinical group [1‐3]. The relationship between trauma hemorrhage and poor outcomes has been well recognized for over 30 years [4], and applies globally [5, 6], in both civilian and military settings [7]. However, outcomes from severe hemorrhage remain poor, with mortality rates approaching 50% for patients who require massive blood transfusion or who develop a significant coagulopathy [8, 9]. Management of trauma hemorrhage depends on a multifactorial approach of timely surgical intervention, fluid resuscitation and blood transfusion therapy [10].

Advances have taken place in our understanding of the pathophysiology of trauma induced coagulopathy [11, 12], in the availability of rapid diagnostic modalities [13], and the introduction of hemostatic resuscitation strategies [14]. Conversely, evidence reviews have shown that some accepted therapies such as blood or plasma transfusion may be ineffective or associated with worse outcomes [15, 16].

Existing reviews have focused on individual interventions, such as transfusion ratios [16‐19], blood substitutes [20], or pharmaceutical agents [21, 22]. Our objective was to conduct a systematic review of the wider trial literature for all randomized controlled trials (RCTs) relevant to the early management of trauma patients with bleeding. We specifically aimed to appraise the methodology of the trials and to assess a broad range of outcomes focusing on bleeding and transfusion requirements, correction of coagulopathy and mortality.

The search strategy identified 11,856 citations. A total of 120 citations were relevant and reviewed at full text. After exclusions (Figure 1) [24], 35 completed RCTs were eligible for analysis (Additional file 2) [25‐63]. Four trials are ongoing [64‐67] (Table 1) and three have been terminated [68‐70] (Table 2). Trials ranged in size from 32 to 20,211 participants and the majority (n = 23) were single centre studies. Thirty-four trials were of parallel group design and one a crossover trial [49]. Nine studies examined a pre-hospital intervention [32, 34, 41‐44, 47‐49], one study used an intervention in both pre-hospital and hospital settings [31] and the remaining interventions were administered in-hospital [25, 26, 29, 30, 33, 35‐40, 45, 46, 50‐57, 61‐63].

The majority of trials (n = 31) recruited trauma patients exclusively, but four studies included non-trauma patients comprising between 4 and 25% of participants [25, 32, 45, 46], totalling 81 patients. All 35 studies included civilian patients only. Six trials only recruited participants with penetrating injuries [29, 34, 35, 41, 48, 57] and one only blunt injury [57]. The 22 studies that included both types of injury had a mean penetrating injury rate of 37% (range 1 to 89%). Twenty-five studies provided data on injury severity scores (ISS) of participants. The mean ISS for studies reporting ISS was 24, range 15 to 33. The inclusion criteria for participants varied. Three studies used a systolic blood pressure (SBP) below 80 mmHg [38, 37, 46], 15 RCTs used 90 mmHg [29, 31, 33, 34, 39‐41, 43, 44, 48, 51, 52, 56, 61, 63] and 3 studies used 100 mmHg [30, 42, 53]. Only one RCT used base deficit as an inclusion criterion [61]. Seventeen studies provided data on the percentage of participants receiving blood transfusions (overall mean 74%, range: 5 to 100%) [25, 26, 31, 33, 35, 36, 38, 39, 42, 46, 55‐57, 61, 62]. Enrolled patients receiving massive transfusion (over 10 units of RBC in 24 hours) varied from 6 to 100% (mean 30%) [25, 36, 37, 42, 53, 57, 61].

Methodological quality is summarized in Additional file 3 and Figure 2[71]. Only 12 studies described adequate sequence generation methods. Allocation concealment was detailed in 23 studies and adequate in 13. Twenty-one trials did not report blinding, 14 reported blinding of either participants or personnel and 4 of these also reported blinding of the outcome assessor. Most studies (n = 26) had no loss of patients, and five had less than 10% loss to follow-up. Only one study used good methodological practices in all areas examined [56]. There was no trend to improvement in methodological quality over time.

The 35 RCTs identified might be expected to provide a strong evidence base for a single clinical condition. However, the multifactorial nature of trauma hemorrhage, the multiplicity of interventions, issues with trial design, difficulties with the conduct of trauma trials and lack of a coordinated approach mean that only limited conclusions can be drawn. The largest sub group of included RCTs evaluated different strategies for using fluids during resuscitation, but did not consistently identify improvements in outcomes. The RCTs evaluating hemoglobin substitutes reported a reduction in RBC requirements but safety remains a concern [20]. Very few studies were identified evaluating the clinical effectiveness of RBC or blood component therapy. Only two studies were identified which evaluated surgical or mechanical interventions, which is surprising given the interest in damage control surgery [72] and angio-embolization [73]. Tranexamic acid was the only pharmaceutical agent that improved mortality [56].

Two studies reported bleeding endpoints using time taken to achieve hemorrhage control as their endpoint [37, 52], all other studies reported surrogate outcomes. Transfusion requirement was commonly used as a surrogate outcome for bleeding, but its use introduces issues with variations in transfusion practice, differences in product type and availability, and survivor bias [74]. Although transfusion for trauma hemorrhage is usually completed within a few hours of injury [75], a large proportion of the transfusion data was reported over a much longer timeframe. Differentiation between early and late transfusion use is an important distinction in understanding the effects of interventions for acute bleeding.

There was no demonstrable association between survival and transfusion requirements, despite evidence from observational studies [76, 77]. None of the nine trials reporting a reduction in RBC use had an associated survival improvement [29‐31, 33, 39, 45, 51, 57, 61]. Conversely, other studies reported survival benefits but did not observe differences in transfusion use [46, 48, 56]. No study used correction of coagulopathy as a defined endpoint. Newer methods of assessing hemostasis such as thromboelastography were not used and a definition of coagulopathy was variable and provided by a limited number of trials [32, 58, 60].

Many of the included trials were poorly designed or conducted, underpowered or recruited small numbers of participants. Recruitment to trauma RCTs can be difficult, not least because of the challenges of enrolling incapacitated patients where informed consent is impossible, although some countries now have recognized processes for emergency consenting. Low patient numbers affect study power and increase the risk of bias, since baseline imbalances between patient groups is likely to occur even if randomization has been rigorous [78]. Only five studies were powered to provide mortality results, and it is likely that the improvement in mortality suggested by the sample size calculations (ranging between 6 and 35%) was over optimistic in many studies [79]. In contrast the CRASH-2 study tested the hypothesis that tranexamic acid would provide a 2% survival benefit which projected a sample size of 20,000 participants [56].

There are limitations to this review. A quantitative analysis was not possible because of the heterogeneity between studies. For example, the inclusion criteria for patients varied widely, such as SBP values for shock. This increases the risk of missing low levels of benefit or harm, which were not large enough to be statistically relevant in any single RCT. The heterogeneity also highlights the importance of working towards uniformity in clinical trials. Attempts were made to identify all relevant RCTs including those in the non-English literature, but some studies may have been missed. Our literature search spanned 60 years, a time frame which has seen trauma care alter significantly. The included RCTs are all from civilian settings, and, therefore, RCT data do not exist to evaluate changes in military practice, although the recent changes in transfusion support for trauma patients have been driven by military data. There were no eligible RCTs examining, for example, the role of tourniquets and, therefore, this area has not been addressed in our review, although RCTs may not be indicated for every intervention.

The acute management of trauma hemorrhage has been evaluated in a large number of trials but these have not in the main produced results that have changed management or improved outcomes. This systematic review set out to examine RCTs, as the most robust form of study design and in so doing observational data have not been identified and appraised. However, it demonstrates that the difficulties associated with recruitment, design and conduct of trauma trials can be overcome to produce better quality RCTs. As our understanding of the pathophysiology of trauma hemorrhage grows, a coordinated strategy is required for this globally important condition.