Predictive value of S-100β protein for prognosis in patients with moderate and severe traumatic brain injury: systematic review and meta-analysisBMJ 2013; 346 doi: http://dx.doi.org/10.1136/bmj.f1757 (Published 04 April 2013) Cite this as: BMJ 2013;346:f1757
- Eric Mercier, MSc candidate1,
- Amélie Boutin, PhD candidate1,
- François Lauzier, assistant professor123,
- Dean A Fergusson, associate professor4,
- Jean-François Simard, research assistant1,
- Ryan Zarychanski, assistant professor5,
- Lynne Moore, assistant professor16,
- Lauralyn A McIntyre, assistant professor47,
- Patrick Archambault, assistant professor8,
- François Lamontagne, assistant professor9,
- France Légaré, professor810,
- Edward Randell, associate professor11,
- Linda Nadeau, assistant professor12,
- François Rousseau, professor1012,
- Alexis F Turgeon, assistant professor12
- 1Centre de Recherche du Centre Hospitalier Universitaire (CHU) de Québec (Hôpital de l’Enfant-Jésus), Traumatologie - Urgence - Soins Intensifs (Trauma - Emergency - Critical Care Medicine), Université Laval, Québec City, QC, Canada
- 2Department of Anesthesiology, Division of Critical Care, Université Laval, Québec City, QC, Canada
- 3Department of Medicine, Université Laval, Québec City, QC, Canada
- 4Clinical Epidemiology Unit, Ottawa Hospital Research Institute, Ottawa, ON, Canada
- 5Department of Internal Medicine, Section of Critical Care Medicine, University of Manitoba, Winnipeg, MB, Canada
- 6Department of Social and Preventive Medicine, Université Laval, Québec, QC, Canada
- 7Department of Medicine, Division of Critical Care, University of Ottawa, Ottawa, ON Canada
- 8Department of Family and Emergency Medicine, Université Laval, Québec, QC, Canada
- 9Centre de Recherche Clinique Étienne-Le Bel du CHUS, Université de Sherbrooke, Sherbrooke, QC, Canada
- 10Centre de Recherche du CHU de Québec, Knowledge Transfer and Health Technology Assessment, Université Laval, Québec City, QC, Canada
- 11Department of Laboratory Medicine, Memorial University, St John’s, NF, Canada
- 12Department of Molecular Biology, Medical Biochemistry and Pathology, Université Laval, Québec City, QC, Canada
- Correspondence to: A F Turgeon, Centre de Recherche du CHU de Québec (Hôpital de l’Enfant-Jésus), Traumatologie - Urgence - Soins Intensifs (Trauma - Emergency - Critical Care Medicine), 1401, 18e rue, local H-012a, QC, Canada G1J 1Z4
- Accepted 12 February 2013
Objectives To determine the ability and accuracy of the S-100β protein in predicting prognosis after a moderate or severe traumatic brain injury.
Design Systematic review and meta-analysis of randomised controlled trials and observational studies.
Data sources Medline, Embase, Cochrane Central Register of Controlled Trials, BIOSIS (from their inception to April 2012), conference abstracts, bibliographies of eligible articles, and relevant narrative reviews.
Study selection Two reviewers independently reviewed citations and selected eligible studies, defined as cohort studies or randomised control trials including patients with moderate or severe traumatic brain injury and evaluating the prognostic value of S-100β protein. Outcomes evaluated were mortality, score on the Glasgow outcome scale, or brain death.
Data extraction Two independent reviewers extracted data using a standardised form and evaluated the methodological quality of included studies. Pooled results were presented with geometric means ratios and analysed with random effect models. Prespecified sensitivity analyses were performed to explain heterogeneity.
Results The search strategy yielded 9228 citations. Two randomised controlled trials and 39 cohort studies were considered eligible (1862 patients). Most studies (n=23) considered Glasgow outcome score ≤3 as an unfavourable outcome. All studies reported at least one measurement of S-100β within 24 hours after traumatic brain injury. There was a significant positive association between S-100β protein concentrations and mortality (12 studies: geometric mean ratio 2.55, 95% confidence interval 2.02 to 3.21, I2=56%) and score ≤3 (18 studies: 2.62, 2.01 to 3.42, I2=79%). Sensitivity analysis based on sampling time, sampling type, blinding of outcome assessors, and timing of outcome assessment yielded similar results. Thresholds for serum S-100β protein values with 100% specificity ranged from 1.38 to 10.50 µg/L for mortality (six studies) and from 2.16 to 14.00 µg/L for unfavourable neurological prognosis as defined by the Glasgow outcome score.
Conclusions After moderate or severe traumatic brain injury, serum S-100β protein concentrations are significantly associated with unfavourable prognosis in the short, mid, or long term. Optimal thresholds for discrimination remain unclear. Measuring the S-100β protein could be useful in evaluating the severity of traumatic brain injury and in the determination of long term prognosis in patients with moderate and severe injury.
Early determination of prognosis after traumatic brain injury is a priority for relatives and physicians involved in the care of these patients.1 2 Despite recent improvement in management of patients with traumatic brain injury in intensive care and the development of guidelines to standardise care,3 4 mortality and morbidity in these patients remain high.5 6 7 About 30% of patients admitted after severe traumatic brain injury will die, and 50% will be moderately disabled.7 8 In a recent multicentre cohort study, we observed variable mortality rates across Canadian trauma centres, despite comparable severity of injury, and considerable variation in the incidence of withdrawal of life sustaining treatments.9 As many of these patients are young with no previous comorbidity, the decision to withdraw life sustaining treatments is based mainly on prognostic evaluation. Current prognostic indicators and models, however, are limited by their lack of sufficient discriminative capacity to inform clinical decision making.10 11 12 New prognostic information beyond the clinical examination, patient demographics, and radiological imaging from admission is needed to allow early prediction of short, mid, and long term outcome of patients with moderate and severe traumatic brain injury.13
Over the past 20 years, biochemical markers of brain damage have been increasingly studied as potential tools for prognostic evaluation.13 14 15 16 17 Concentrations of S-100β protein, the β subunit of a calcium binding protein present mainly in glial and Schwann cells,18 increase in human blood and cerebrospinal fluid after a wide range of diseases or conditions leading to brain damage.19 20 21 22 23 24 25 26 27 28 Increased concentrations in blood and cerebrospinal fluid have been reported in patients with traumatic brain injury.29 Despite growing evidence suggesting a potential clinical role for S-100β as a biomarker, its association with short, mid, and long term prognosis remains unclear in patients with traumatic brain injury. There are also concerns that extracerebral injuries could contribute to increases in concentrations. Measurements of S-100β protein, or other biomarkers, are not widely used in clinical practice and are not considered standard of care.14 15 We therefore conducted a systematic review to evaluate the prognostic value of the S-100β protein after moderate or severe traumatic brain injury.
Materials and methods
We searched Medline, Embase, Cochrane Central Register of Controlled Trials (Central), and BIOSIS from their inception to April 2012 for relevant studies. For Medline and Embase, we used validated combinations of terms for prognostic studies to achieve optimal search sensitivity and specificity.30 31 Broad text and MeSH or EMTREE terms for biomarkers were used to maximise sensitivity. Our search strategy was designed to identify a wide range of biomarkers to increase sensitivity. The full search strategy for Medline is provided in appendix 1. We screened abstracts from relevant meetings (American Association of Neurological Surgeons, European Association of Neurosurgical Societies, Société de Neurochirurgie de Langue Française, Congress of Neurological Surgeons, Critical Care Canada Forum, International Trauma Anesthesia and Critical Care Society, World Federation of Societies of Intensive and Critical Care Medicine, Society of Critical Care Medicine, European Society of Intensive Care Medicine, International Symposium on Intensive Care Medicine, American association for the surgeons of trauma) and reference lists of selected articles and relevant narrative reviews.
Search results were combined and duplicates were excluded with EndNote (version X5, Thomson Reuters, 2011). Two reviewers (EM and JFS or AB) independently reviewed all citations and selected eligible studies. A third author (AFT) was consulted in case of disagreement.
We included cohort studies and randomised controlled trials that determined S-100β protein concentrations in patients with moderate and/or severe traumatic brain injury as defined by a Glasgow coma score <13.32 Included studies had to report at least one outcome of interest (mortality, Glasgow outcome score,33 or brain stem death) and had to report S-100β protein concentrations in cerebrospinal fluid, venous blood, arterial blood, and/or urine. One quantitative measurement of S-100β protein in the emergency room or the intensive care unit, along with at least one follow-up outcome measure after discharge from intensive care, was also required for inclusion. Studies with one or no patient having a favourable or an unfavourable outcome were excluded as no standard deviation could be computed. We included prospective and retrospective outcome assessments and avoided language based study exclusions. We excluded studies limited exclusively to children (aged <18) and studies in which less than half of included patients had moderate or severe traumatic brain injury, unless we could extract the data related to patients with moderate or severe traumatic brain injury.
Two reviewers (EM and JFS or AB) independently collected data using a standardised data abstraction form. We abstracted information related to study design, patient characteristics (age, sex, severity of injury, blunt or penetrating injury, type of lesions, mechanism of injury, Marshall score for computed tomography, clinical pupillary reaction, hypotension, hypoxaemia, intracranial pressure, and mechanical ventilation), treatments (operative and pharmacological), laboratory aspects of S-100β protein testing (type of assay used, time period of sampling, and sampling type), and clinical outcomes (outcome type and timing of assessment). In instances of duplicate reporting, we used data from the study that included the largest number of patients or, when available, individual patient data from each study. We contacted authors for clarification on study sample or for missing data.
If multiple measurements of S-100β were carried out, we used the first measurement after the injury for analysis. If outcomes were assessed at multiple time points, we used the measurement furthest from injury for analysis. When the Glasgow outcome scale was dichotomised by the authors, we retained their definition of unfavourable outcome. When the entire spectrum of the Glasgow outcome score was provided, we defined an unfavourable outcome as a score ≤3.
Methodological quality and risk of bias of included studies
We developed a modified version of the QUADAS-2 assessment tool34 (appendix 2) to evaluate the risk of bias in prognostic studies. We also used the criteria for reporting observational studies proposed in the STROBE statement35 to complete the methodological evaluation of the included studies (appendix 3).
The distribution of S-100β concentrations were right skewed and we therefore log transformed them to yield a normal distribution, assessed with Shapiro-Wilk and Kolmogorov-Smirnov normality tests. A log normal distribution facilitated the analysis and presentation of outcomes between groups with geometric means ratios, for which the null value is one.36 Therefore, a ratio greater than one indicates that mean concentrations are higher in the group with an unfavourable prognosis compared with the group with a favourable prognosis.36
Analyses were performed with random effects models. The presence of potential heterogeneity was assessed with the I2 statistic.37 Sensitivity analyses based on a priori hypotheses (time period of evaluation, sampling time, sampling type, severity of traumatic brain injury, isolated traumatic brain injury, biochemical technique, blinding of outcome assessment) were performed to investigate expected or measured heterogeneity. When individual patient data were available, we computed receiver operating characteristics curves for each study and used a bivariate random effects regression model38 to pool the sensitivity and specificity of intervals of S-100β threshold values for mortality and Glasgow outcome score. We also computed discrimination threshold values for 100% specificity for each of these studies.
In some studies, it was unclear whether the authors presented standard deviations or standard errors. In these cases, to prevent an incorrect rejection of the null hypothesis (type I error), we assumed the reported statistics to be standard errors. Analyses were conducted with Review Manager version 5.0 (Cochrane Collaboration, Copenhagen, Denmark) and SAS version 9.2 (SAS Institute, Cary, NC). For all tests and confidence intervals we used a two tailed type I error rate of 5%. The reporting of this systematic review complies with the PRISMA statement.39 Publication bias was evaluated through visual inspection of funnel plots.
Quality of the evidence
The quality of the evidence for the three main outcomes was determined with the GRADE approach40 with the GRADEpro software (version 3.2 for Windows. Jan Brozek, Andrew Oxman, Holger Schünemann, 2008).
Study identification and selection
Our search strategy retrieved 9228 citations after removal of duplicates. After screening and the application of our inclusion and exclusion criteria, we included 41 studies41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 published between 1987 and April 2012 (1862 patients) (fig 1⇓). With the exception of one conference proceeding44 and one pilot study reported as a table in the final study publication,75 all included studies were published peer reviewed manuscripts.
Thirty nine studies were observational cohort studies and two were randomised controlled trials68 72 (table 1⇓). Three studies were published in languages other than English: Chinese,61 Japanese,65 and Czech.70 Each study evaluated between four and 149 patients with moderate and severe traumatic brain injury. Only one study reported including penetrating trauma, which represented 6.7% of its sample.53 The main outcome measures presented were the Glasgow outcome score (30 studies), mortality (18 studies), and brain stem death (two studies). Site of S-100β protein sampling was venous (31 studies), arterial (10 studies), or cerebrospinal fluid (five studies). Eighteen studies presented data from two or more samples at different time points after traumatic brain injury. Ten studies presented individual patient data.42 45 49 51 53 57 66 73 74 75 Individual data for three studies by the same group42 49 57 were combined in the meta-analysis as they presented data from the same patients. All analyses were performed with serum (arterial or venous) S-100β protein concentrations.
Fifteen of the 41 included studies could not be incorporated in the meta-analysis: nine presented the peak concentrations of serial samples of S-100β protein plasma or cerebrospinal fluid43 47 48 50 60 71 72 or the mean value52 79; two did not report measures of dispersion44 61; two presented data on four or five patients, with one patient having a favourable outcome in each case41 56; and one reported only the threshold value for a 100% specificity for unfavourable prognosis.55 Finally, two studies reported brain stem deaths,58 79 but one reported the mean value of serial samples, which precluded the application of meta-analysis for this specific outcome.
Methodological quality of included studies
Seventeen studies examined the risk of bias, six studies presented a flow diagram of participants, 14 studies adequately described their study population (including missing data and patients lost during follow-up), and 13 studies presented their funding sources. Figure 2 and appendix 3 present a more complete evaluation of the methodological quality and risk of bias⇓.
We observed significant positive associations between serum concentrations of S-100β protein and outcome. Increased concentrations correlated with increased mortality (12 studies: geometric mean ratio 2.55, 95% confidence interval 2.02 to 3.21; I2 56%; fig 3⇓), a Glasgow outcome score ≤3 (18 studies: 2.62, 2.01 to 3.42; I2 79%; fig 4⇓), and brain stem death (one study: 2.9, 2.3 to 3.5). The results were consistent in all sensitivity analyses and were not influenced by the presence of associated traumatic injuries in other parts of the body (tables 2⇓ and 3⇓). In mortality subgroup analyses, heterogeneity was lowered according to testing method and timing of outcome assessment. In the studies that we excluded because of lack of information on measures of dispersion,44 61 80 we observed a significant and consistently positive association between serum concentrations of S-100β protein and mortality. In eight studies excluded from the meta-analysis because they reported only the peak or the mean values over serial samples, authors reported a significant association (P<0.05) between serum concentrations and mortality,48 72 Glasgow outcome score ≤3,43 47 48 50 60 71 and brain stem death.79 Three studies reported a significant association between S-100β protein concentrations in cerebrospinal fluid and a Glasgow outcome score ≤3.51 59 62 We also analysed the data using the Taylor series method,82 83 and this did not substantially change the results (data available from authors). These analyses, however, yielded to a quasi-absence of statistical heterogeneity in all analyses.
For mortality (six studies), serum threshold values of 2.5-3.0 µg/L yielded a mean specificity of 91% (95% confidence interval 84% to 95%) and a sensitivity of 39% (24% to 57%), while concentrations >3.0 µg/L yielded a mean specificity of 97% (95% to 98%) (see appendix 4). When we considered each study individually, the respective serum thresholds to attain 100% specificity for prognosis of death, meaning that all surviving patients are correctly identified by the test (no false positive overdetection of prognosis of death), ranged from 1.38 µg/L to 10.50 µg/L, with an associated sensitivity ranging from 14% to 60% (fig 5⇓).
Similarly, for unfavourable neurological prognosis (Glasgow outcome score ≤3) (five studies), threshold values of 2.5 µg/L to 3.0 yielded a specificity of 94% (95% confidence interval 85% to 98%) and a sensitivity of 38% (15% to 67%) and values >3.0 µg/L yielded a specificity of 96% (91% to 98%) (appendix 4). Again, when we considered each study individually, threshold values for 100% specificity for unfavourable neurological prognosis ranged from 2.16 µg/L to 14.0 µg/L, with an associated sensitivity ranging from 9% to 50% (fig 6⇓).
Publication bias and quality of evidence
Visual evaluation of funnel plots did not indicate any publication bias (see appendix 5). The quality of the evidence for mortality and for unfavourable neurological outcome (Glasgow outcome score ≤3) was moderate (table 4⇓).
This meta-analysis identified a significant association between S-100β protein serum concentrations and short (less than three months), mid (three to six months) or long term (six months and above) prognosis in patients with moderate or severe traumatic brain injury. The concentrations were significantly correlated with unfavourable prognosis, as defined by mortality or Glasgow outcome score ≤3, irrespective of concomitant traumatic injuries. Serum thresholds values ranging from 1.38 µg/L to 10.5 µg/L and from 2.16 µg/L to 14.0 µg/L were associated with 100% specificity for mortality and a Glasgow outcome score ≤3, respectively. Our findings are highly relevant to the care of critically ill patients with traumatic brain injury, especially as to help informed decision with respect to the determination of prognosis.
Strengths and weakness of study
There are limitations of our systematic review. Firstly, there was considerable heterogeneity for all outcomes of interest. Heterogeneity among studies that assessed mortality was explained by the testing method used and by the time period over which outcome was evaluated. Sensitivity analyses including the type of assay used, the timing of sampling, the sampling type, isolated versus multiple trauma, and the timing of outcome evaluation after traumatic brain injury, however, did not fully explain the observed heterogeneity for the Glasgow outcome score. Secondly, the use of the first measurement of S-100β in our meta-analysis when more than one sample was collected could have generated more conservative estimates as samples obtained between 12 and 24 hours after admission showed a stronger association with outcome measures, which could reflect the impact of secondary neurological injuries like hypoxaemia, hypotension, and increased intracranial pressure. Thirdly, though we carried out our systematic review according to high methodological standards,39 the results of the meta-analysis are limited by the quality of studies included. For example, only 16 studies reported outcome assessment that was blinded from S-100β protein concentrations, which implies a high risk of bias. Moreover, we cannot exclude potential publication bias.
Fourthly, we could not perform sensitivity analyses related to age, pupillary reactivity, or the motor component of the Glasgow coma score, which are known indicators of prognosis in such patients, because of the variable presentations or absence of these data in included studies. Finally, the different chemical assays used could have affected the accuracy and precision of the measured thresholds of S-100β protein concentrations. Although the sensitivity analyses did not show any major impact on the results, some of the assays were used in only a few studies, thus precluding a robust interpretation of their impact. Finally, the S-100β protein concentrations could potentially be affected by previous neurological diseases84 or high serum alcohol concentrations.85 Data on those variables were rarely available and precluded any sensitivity analysis. While these variables could potentially have an impact in mild traumatic brain injury, however, this is unlikely to be considerable for moderate and severe injuries considering the importance of the traumatic brain injury.
Our systematic review had important strengths. We conducted a thorough systematic search, including different databases, and used a comprehensive analytical approach that allowed the inclusion of studies presenting not only means and standard deviations, but also centiles such as medians, thus improving the exhaustiveness of the results. Our rigorous methods were based on guidelines for conducting and reporting systematic reviews.
Comparison with previous knowledge
Previous narrative reviews published to date have outlined the potential of S-100β protein concentrations for predicting outcome after moderate or severe traumatic brain injury, but none of these used systematic review methods or incorporated meta-analyses.14 15 17 86 87 88 89 90 91 92 93 94 95 The results of our study are consistent with those from two previous systematic reviews conducted in patients with stroke or cardiac arrest.96 97 The first review found an association between S-100β protein concentrations and prognostic features (infarct volume and stroke severity),97 while the second review showed that S-100β protein might be a better outcome predictor than the neurone specific enolase after cardiac arrest.96 Our results are also consistent with a large observational study performed in unselected neurocritically ill patients that found that S-100β was associated with neurological deterioration or complications.98
The presence of extracerebral sources of S-100β protein could lead to an overestimation of the severity of the brain lesion in the early phase after traumatic brain injury in patients with multiple injuries.16 99 100 101 102 Only four studies included in our meta-analysis62 68 70 81 specified not enrolling patients with associated multiple trauma. The association between S-100β protein concentrations and prognosis, however, was consistent irrespective of other injuries. This result is concordant with the observations that S-100β protein concentrations are more specific to the brain than to any other organ. Given that 80-90% of the total amount of S-100β is found in cerebral tissue,93 and that serum concentrations of S-100β protein have been correlated with the extent of brain damage in traumatic brain injury on computed tomography46 and in patients with ischaemic stroke,103 the attributable concentrations and influence of extracerebral sources of S-100β is thus likely to be minimal.100 One excluded study previously proposed such an approach,104 but we could not evaluate this hypothesis as no study that included isolated head trauma reported individual patient data. Furthermore, we could not explore potential confounding from severity of extracerebral injuries as data were not reported by outcome groups.
The discriminative capacity of the S-100β protein in the prediction of mortality and neurological outcome in patients with moderate and severe traumatic brain injury provides a glimpse at its potential usefulness as part of a shared decision making process. Indeed, medical teams and relatives faced with decisions about level of care are often left with little information on probabilistic expectations regarding the prognosis in these patients. The high specificity observed at thresholds over 2.5 µg/L makes the S-100β protein a candidate variable to include—in combination with other prognostic indicators such as data from the clinical examination, imaging, and electrophysiological tests—in a prognostic model to help in a shared decision making process. Such a model could better inform clinical teams and relatives on expected clinically important outcomes and optimise the provision of healthcare. On the other hand, the high sensitivity of the S-100β protein to rule out a clinically important brain injury could be useful to provide guidance for the decision whether to perform additional diagnostic assessment such as imaging in patients with traumatic brain injury. As part of a decision aid, the S-100β protein concentration could serve to rule out important traumatic brain injury and avoid exposing patients to unnecessary radiation from imaging, allow better triage and use of resources, and thus be a potentially cost effective measure.
Many questions remain unanswered, such as the optimal biochemical method, timing of sampling, and prognostic threshold. Different assays and timing of sampling might call for different thresholds. With the current level of evidence, we cannot comment on the optimal parameters for prognostic evaluation. Further research is needed to explore combination of variables known to be associated with clinical outcomes of traumatic brain injuries to develop a prognostic model with a high discriminative capacity.
We observed a significant association between serum concentrations of S-100β protein and unfavourable prognosis as defined by mortality, Glasgow outcome score ≤3, and brain stem death. The optimal discrimination threshold values for S-100β protein and the optimal sampling time remain uncertain as there were important variations between studies. The measure of S-100β protein concentrations could potentially play a role as part of a decision aid in the prognostic evaluation of patients with traumatic brain injury as well as to potentially rule out important traumatic brain injury. Further efforts should focus on standardising testing methods and further research on identifying optimal threshold values and sampling time for prognosis determination and on combining S-100β protein concentrations with other prognostic indicators to improve the accuracy of prognostic models and help guiding level of care decisions in a shared decision making process.
What is already known on this topic
Many indicators have been independently associated with prognosis after traumatic brain injury, but they are of limited clinical use when considered separately and current prognostic models do not have sufficient discriminative capacity to inform clinical decision making
S-100β protein concentrations have been shown to increase in blood and cerebrospinal fluid after a wide range of diseases or conditions leading to brain damage
What this study adds
S-100β protein serum concentrations correlate significantly with unfavourable prognosis in patients with moderate or severe traumatic brain injury, as defined by mortality, Glasgow outcome score ≤3, or brain stem death, with or without concomitant traumatic injuries
The association between serum concentrations of S-100β protein and prognosis was observed at discharge from intensive care and at one, three, and six months.
Serum threshold values ranging from 1.38 µg/L to 10.50 µg/L and from 2.16 µg/L to 14.00 µg/L were associated with 100% specificity for mortality and a Glasgow outcome score ≤3, respectively
Cite this as: BMJ 2013;346:f1757
We thank Lucie Côté from the Library of the CHU de Québec, Hôpital de l’Enfant-Jésus, for her help with the retrieval of study publications and Marie-Joëlle Poitras-Pariseau, Information Specialist at the Library of the Université Laval, for her help in the design of the search strategy.
Contributors: E Mercier, F Lauzier, R Zarychanski, L Moore, PA Archambault, F Lamontagne, F Légaré, E Randell, F Rousseau, DA Fergusson and AF Turgeon contributed to the conception and design of the study. EM, J-FS, AB, and AFT determined eligibility of search results and extracted data from included studies. EM, J-FS, AB, and AFT performed and reviewed the analyses. EM, AB, and AFT drafted the manuscript. All authors participated to the interpretation of the data, the critical review of the manuscript for important intellectual content and approved the final version. EM and AFT are guarantors.
Funding: This work was funded by the Fonds de la Recherche du Québec - Santé (FRQ-S) (Traumatology Research Consortium, grant No 23698), the Canadian Anesthesia Research Foundation (CARF), the Foundation of the Association des Anesthésiologistes du Québec (AAQ Foundation) and the Regroupement en Soins Critiques of the FRQ-S Respiratory Health Network. EM was supported by a research training grant (MSc) from the FRQ-S during the conduction of the study (No 23819). AFT, FL, and PL are recipients of a research career award from the FRQ-S. AFT and FL are supported by the Traumatology Research Consortium of the FRQ-S. LM, LAM, and DAF are recipients of new investigator awards from the Canadian Institutes of Health Research (CIHR). RZ is a recipient of an RCT mentorship award from the CIHR. Funding agencies had no role in any part of conduct of the study or preparation of the manuscript.
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: no support from any organisation for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.
Ethical approval: Not required.
Data sharing: No additional data available.
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/.