The conundrum of the definition of haemorrhagic shock: a pragmatic exploration based on a scoping review, experts’ survey and a cohort analysis

A scoping literature review was conducted to identify THS definitions used in the trauma literature and to circumscribe the components of THS [11]. Scoping reviews are form of knowledge synthesis that addresses an exploratory research question aimed at mapping identify knowledge gaps, scope a body of literature or clarify concepts [12, 13]. A search was conducted from January 2010 to November 2018 (included) in MEDLINE (via PubMed) and Google Scholar using a controlled vocabulary (Supplementary material 1) without language restriction. In both databases, the study group investigated the 100 first responses (respectively using best match and by relevance ranking). The clinicaltrials.com database was explored using the same keywords. To broaden this initial search, (1) all the authors (AJ, PSA, PP, SA, CH, TG, SH) were asked to share articles that might include a THS definition and (2) all respective bibliographies were hand-searched using a snowballing method [14]. All published articles that included major trauma patients and provided a definition related to traumatic THS were considered for inclusion. All types of published articles were considered and included: randomized control trials (RCT), observational studies (prospective or retrospective), editorials and protocols. Technological assessments, non-human studies, cost-effectiveness assessments were excluded. In cases of multiple publications of the same study, the most up-to-date or comprehensive publication was included.

Authors (AJ, CH, PP, PSA) identified and extracted from original articles the literal definitions applied to define a THS. They also extracted whether the THS definition was supported by a reference or not.

The survey was developed in a three-step process. First, two members of the working group (AJ, SH) conceived a questionnaire including all components identified from the scoping literature review. Second, all working group members (AJ, PSA, PP, SA, CH, TG, SH) reviewed the questionnaire. Third, a convenience sample of ten international trauma experts for clarity and capacity to provide meaningful answers. After each step, the working group improved the survey according to the provided feedback.

Then, a convenience sample of 64 international experts were selected based on their contribution to major publications about THS (such as guidelines and highly cited publications authors) or identified as opinion leaders (such as conferences speakers) and invited personally to reply to the online survey (www.​surveymonkey.​com). At least three individual e-mail reminders were dispatched to ask for a response. As the survey was based on voluntary participation and information disclosure, the study protocol did not undergo review by an institutional review board. Voluntary participation was considered as consent. The acknowledgments lists all experts who accepted to be named.

The survey consisted of 50 questions organised in three sections: (1) expert profile (country of practice, main clinical area of practice, context (civilian or military), number of personal publications related to the subject), (2) selection of components of THS among clinical & biological components, scores and transfusion-related parameters and (3) conclusion. The ranking of selected components from step 2 ranged from − 3 (I strongly disagree with the importance of this component) to + 3 (I strongly agree with the importance of this component) with 0 being neutral. Components with a median scoring [1;3] were considered as consensual, those scoring between [− 1;1] as neutral and those scoring [− 3;− 1] as non-relevant [15]. For numerical components, experts were requested to provide a threshold on an analogue scale.

Based on the scoping review and on the results from the expert survey, the working group identified consensus and operational definitions of THS. Then, these definitions were applied to a large register of trauma patients, to evaluate how the different definitions affected diagnosis, management and patient outcome. This population was extracted from the Traumabase Registry (www.​traumabase.​eu) from January 2010 (registry initiation) to November 2018 (date of data extraction). The Traumabase registry includes all consecutive patients admitted in all participating level-1 trauma centers for suspected major trauma based on the national Vittel triage criteria [16, 17]. These centers provide the highest level of care for trauma patients in their respective geographical areas. The Traumabase registry samples epidemiological, prehospital, resuscitation, critical care and outcome data. Data collection is performed by trained clinicians and research assistant under supervision of a designated coordinator in each center. The registry database on a secure server, contains numerous consistency and internal validation algorithms. Data management is performed by professional data managers and statisticians from the Research Unit Paris 7. All 14 centers including patients at the time of data extraction participated in the study.

Definitions were translated step by step by two investigators (AJ, SH) to correspond to a mathematical script to classify patients within the data set accordingly (Supplementary material 2).

For each of the five definitions, a THS population was identified and described in terms of:

Quantitative data were described with median and interquartile range [Q1; Q3] and categorical data with counts and percentage (%). A sensitivity analysis examined apart the RCTs. All analysis involved the use of R v4.0.2 (www.​R-project.​org, the R foundation for statistical Computing, Vienna, Austria).

Among the 200 articles screened on MEDLINE and Google Scholar and those highlighted by the snowballing, the scoping review identified 68 studies that involved a definition of THS (44 published articles identified in MEDLINE or Google Scholar and 24 protocols identified in www.​clinicaltrials.​gov) used whether as an inclusion criterion or as an outcome (Supplementary material 3). Among these 68 definitions, 52/68 (76%) were distinct and stand-alone (Supplementary material 3) resulting in only 5 distinct definitions used in more than one article. Among the 44 published articles, 20 (45%) authors provided a reference to the used definition.

The most recurrent definition of THS (n = 12/68) was the presence of “a systolic blood pressure (SBP) less than or equal to 70 mmHg or between 71 and 90 mmHg if the heart rate is greater than or equal to 108 beats per min”. Cross-referencing several studies applying this definition identified randomized controlled trial by Bulger et al. on Out-of-hospital Hypertonic Resuscitation After Traumatic Hypovolemic Shock as the initial source for this definition [22]. SBP was involved in 48/68 (71%) definitions, with a systolic pressure threshold ranging from 70 to 120 mmHg. The most frequently used threshold for SBP was 90 mmHg (n = 41/68, 60%). Red blood cell (RBC) transfusion was involved in 17/68 definitions (25%), with very heterogeneous volume and time thresholds. Serum lactate concentration was involved in 8/68 different definitions (12%) with a threshold ranging from 2 to 5 mmol/L. 5/68 definitions (7%) applied the Advanced Trauma Life Support (ATLS) classification of hypovolemic shock (with threshold ranging from grade 1 to 3 when a threshold was specified).

The responses to the survey were obtained between November 2018 and February 2019. Among the 64 experts invited to take part in the survey, 29 (45% response rate) from 10 different countries responded (France, n = 8; United Kingdom, n = 4; United states of America, n = 4; Germany, n = 4; Austria, n = 2; Switzerland, n = 2; Norway, n = 2; Israel, n = 1; Algeria, n = 1; Belgium, n = 1). These experts shared miscellaneous background with 8/29 (28%) practising mainly in prehospital care, 13/29 (45%) in intensive care, 11/29 (38%) in operating the, 7 (24%) in the emergency department, 6/29 (20%) in surgery and 3/29 (10%) in a laboratory (with some experts reporting multiple categories). Of these 21/29 (72%) experts were affiliated to civilian care; 2/29 (7%) to military care only and 6/29 (21%) were affiliated both to civilian and military care. The survey participants had a high level in expertise in trauma care with 11/29 (38%) reporting more than 20 publications as first author on trauma, 8/29 (28%) that report between 10 and 20 publications, 7/29 (24%) that report 5 to 10 publications and 3/29 (10%) that report less than five publications.

The survey findings are summarized in Fig. 1 and Supplementary material 4 provides the median values and thresholds proposed by the experts.

The survey identified four consensual components of a THS definition (mean score [1–3]):

The survey identified 16 components with a lower level of consensus (mean score [0–1]): response to fluid challenge, variation in hemoglobin, hemoglobin at hospital, mean blood pressure, respiratory rate, intervention needed to stop the bleeding, pulse pressure, use of vasoactive drugs, hemoglobin on scene, viscoelastic tests, capillary refill time, fibrinogen, blood lost volume, mottling, treatment response, INR and five non-consensus components (Injury mechanism, platelet, injury location, Prothrombin Time and Activated Partial Thromboplastin Time) were also identified.

Based on the review and survey results, the study group purposively selected five definitions of THS (Box 1):

These five definitions were applied in a cohort of 21,807 trauma patients from the Traumabase registry. Five cohorts of THS were identified according to the chosen definition: PROPPR (n = 881, 4.0%), PROMMTT (n = 2510, 11.5%), Traumabase (n = 1765, 8.1%), Expert-based definition (n = 2738, 12.6%) and the “Most Frequently Cited” definition (n = 3527, 16.2%).

These five THS populations were compared for baseline characteristics (Table 1) and outcomes (Table 2) resulting in important differences across groups, including the proportion of massive transfusion (ranging from 19.5% with the “most frequently cited” definition to 43.0% with the Traumabase definition), the proportion of patients requiring surgery within the first 24 h (ranging from 54.6% with the “most cited definition” to 75.0% with the Traumabase definition), in-hospital mortality (ranging from 27.7% with the experts definition to 37.1% with PROPPR definition), in-hospital length of stay (ranging from 10 days with the most cited definition to 17 days with the Traumabase definition).

The objective of this study was not to provide a definitive definition of THS, but provide substance to advocate for a concerted, international effort to define haemorrhagic shock and to inform and prepare this process. The three-step exercise identified five candidate definitions and demonstrated when applied to a large trauma data set how they generate considerably divergent study cohorts in terms of need for resource use and patient outcome including mortality.

Comparing the effect of the five candidate definitions, the groups appeared similar in terms of basic characteristics such as age, sex, mechanism, proportion of head injury as important confounder for overall mortality and AIS composition. All five candidate definitions seemed able to target groups of severe patients with SAPS 2 ranging from 32 (expert) to 55 (PROPPR), ISS ranging from 25 (Most Cited) to 32 (PROPPR), organ failure and high predicted and observed mortality for all groups. All definitions selected patients with severe physiological derangement with a Shock Index of 1 and above (1.3 for PROPPR), haemoglobin < 12 g/dl, Lactate > 3 mmol/l and BE > 7. The two definitions by far generating the most intense resource use and blood product consumption use on admission and at 24 h are consistently PROPPR and Traumabase. There indicators correlate well in these groups with marker of physiological derangement and patient outcome. The five candidate definitions were more or less selective, PROPPR being the most selective and the Most Cited recruiting the largest group but with low severity.

The authors struggled to find a reference for the Most cited definition; none of the published articles that applied it provided an original reference and the definitions seems not derived on any clinical cohort. It can probably be considered as outdated and be replaced by studies based on more rigorous and prospective approaches [27].

Figure 2 illustrates the time dependency of each definition related to the patient pathway and how this relationship to time may influence the identification of patients in shock to prompt expedient treatment and/or the inclusion into a study protocol or observational cohort. The Most Cited can be applied on scene, all other definitions at the earliest in the resuscitation room or require a few hours to elapse. All take into account the fact that the early hours are essential.

To our knowledge, only one recent work addressed the challenge to derive a consensual definition for massive transfusion, a proxy for hemorrhagic shock, through an international expert Delphi process [2]. The experts voted for the following definition: 4 or more units of multiple blood components within 2 h of injury. This definition was unfortunately not available when the present work was under way. The definition is applicable within two hours of injury and resource based using all types of blood products but entails no physiological parameters. Excluding physiological data may omit patients initially in shock but stabilised through expedient medical and/or surgical intervention or die from shock on admission. A definition requiring a certain amount of blood products may discount patients that stabilised with smaller amounts, a recently observed tendency [28]. We applaud the work of Wong et al. and share their approach as the way forward to find a consensual definition. However, as the authors state themselves, the composition of their expert group is over-representative of North America and the UK and more suitable for systems with similar transfusion practice. Neither the definition used by Wong et al., nor any of the five candidate definitions explored in the present study includes an element of haemorrhage control resource need other than blood products. An aspect that may require further exploration. The international trauma community needs to take into account definitions for resource poor settings.

Clinical research relies on a concise delineation of inclusion or outcomes criteria [7]. The use of poorly delineated definitions is likely to affect results by reducing reproducibility either in randomised controlled trials or in epidemiologic studies, with a risk to waste precious research resource and divert patient recruitment [29]. The present results demonstrate the existing heterogeneity in the available literature and explain the challenge to perform meaningful systematic reviews [30‐32] and meta-analysis [33]. In consequence, a consensual international definition of THS is necessary.

In the present study, authors were careful to avoid recommending a single definition. But the results inform the inevitable and necessary future international consensus with precious insight. In agreement with the definition provided by Wong et al., a future THS definition should focus on the first hours of shock management to acknowledge the time-dependant and dynamic nature of THS and integrate consumption of all blood type of blood products. Such definition would also need to reflect that the later the THS is defined, more information (physiological, clinical, therapeutic) is available to make the diagnosis more specific. THS can indeed present with various presentations depending on the timeline. For example, immediately after the trauma a patient can present a significant bleeding from extremity injury with normal hemodynamic parameters and still functional physiological compensation. Despite hemorrhage control after application of a tourniquet, the same patient might then become hypotensive because of strained physiological compensation. A blood transfusion might thus be required to restore hemodynamic. In addition, a composite definition should include physiological parameters to not omit patients that stabilise or die early and expressed by the expert consensus. The inclusion of haemorrhage control interventions should be considered [27]. As demonstrated by the registry analysis in this work, the application of a specific definition has measurable impact on the composition of either prospective or retrospective study cohorts, patient outcome, resource use, the detection and diagnosis and in consequence the treatment of THS. A very inclusive definition that assures large recruitment, may not be ideal to identify bleeding patients that will require haemorrhage control and vice versa. Ideally, all these aspects require to be taken into account to find an international consensus. The example of the five candidate definitions show that several future candidate definitions need to be tested in large international cohorts to reach a final consensus.

This work also presents several limitations. First, our review is not a systematic review. However, scoping reviews are now endorsed by the PRISMA collaboration group [11] and the purpose of the study matches with a scoping review extent (clarify definition, explore breadth and map evidence) [34‐36]. Second, a single registry was used to explore the impact of these different definitions. External validation of the finding on a separate, ideally international, data set appears is mandatory. The authors suggest to organise an international consensus conference including resource settings and test several candidate definitions in a large international dataset. Finally, this study did not aim to recommend a single definition of THS. The objective was to provide useful methodological and conceptual insight to inform an international consensus process. Third and last, the authors acknowledge that alternative definitions could have been included in the registry-based study (a single SBP threshold, the shock index or the ATLS definition). The authors, however, believe that providing more definitions carries the risk to generate more noise without providing any gain in the main message. Namely, that THS definitions are heterogenous in the literature and how different definitions have a different impact on patient outcomes. Similarly, other THS-related parameters could have been considered in the survey such as those able to explain individual sensitivity to haemorrhage like age, comorbidities or concomitant TBI. We believe that these parameters could be of interest in further studies.