Skip to main content

Collaborative European NeuroTrauma Effectiveness Research in TBI

Final Report Summary - CENTER-TBI (Collaborative European NeuroTrauma Effectiveness Research in TBI)

Executive Summary:
Traumatic Brain Injury (TBI) inflicts great personal suffering on victims and families, and leads to huge direct and indirect societal costs. Worldwide TBI affects 50 million people, results in substantial disability, and costs the global economy €325 billion annually (corresponding to about one in every €150 of annual global output). In the European Union and the UK approximately 2.5 million people suffer a TBI each year, of whom 1.5 million are admitted to hospital and 57,000 die. TBI is a complex disease, management of which has not advanced for many decades. Clinical care is not underpinned by strong evidence, and is not individualised. However, emerging diagnostic approaches, research methodologies, and the availability of robust risk adjustment models, could improve matching of patients to therapies (Precision Medicine), comparison of common treatments (through Comparative Effectiveness Research; CER), and more accurate prognostication (of huge value to patients, families, and clinicians).
CENTER-TBI is a large-scale project, aiming to (1) improve characterization of TBI in order to facilitate individualized treatments and (2) identify the most effective clinical care, providing high quality evidence in support of treatment recommendations and guidelines. Leading experts from 47 scientific institutes, worldwide, have worked to generate new knowledge that could improve patient outcomes and reduce the global burden of TBI. CENTER-TBI includes a prospective observational Core Study and a Registry, supported by extensive profiling of participating centres to inform CER analysis. The Core Study collected granular data from over 4500 patients in Europe and Israel and an additional 1200 in Australia and India. Enrolment was in three strata, differentiated by care path: (1) patients discharged home from the emergency room (ER stratum); (2) patients admitted to hospital, but not to the intensive care unit (admission stratum); (3) patients admitted to the Intensive Care Unit (ICU stratum). The Registry collected basic data on all patients presenting with TBI, aiming to assess representativeness of the Core Study and to analyse effects of structural parameters (e.g. organisational) in greater numbers (recruitment: 22,772 patients). CENTER-TBI is part of the International Initiative on TBI Research (InTBIR - http://intbir.nih.gov/).
The Core Study combined emerging techniques (e.g. biomarkers, advanced Magnetic Resonance (MR) imaging, genomics), with innovative approaches to analysis. It created the largest Imaging repositories and Biobank for TBI in the world. We have mapped clinical care (and its variations across participating centres) to outcome after TBI in Europe, and identified disparities in care and substantial variation in management. Outcome variations were, however, lower than in previous studies, suggesting improvements in overall care and systems of care. Best practices were identified, including the demonstration that routine thromboprophylaxis and avoidance of fluid overload in the ICU were associated with better outcomes, and illustrate that strong inferences about key aspects of care can be made. Novel insights were generated regarding multiple aspects of neurotrauma biology, management, and outcome. Examples include: recognition of increasing incidence and poor outcomes of “low impact” TBI in older people who fall; data on the heritability of outcome from TBI; assessment of the incremental benefit of biomarkers and advanced neuroimaging in mapping diagnosis and clinical course; careful analysis of high resolution ICU data to better understand intracranial physiology in intracranial hypertension; recognition that up to 50% of patients with “so called” mild TBI do not experience a full recovery by six months (speaking to the concerns of “concussion” in sport); parcellation of the influence of patient characteristics and injury severity on such outcomes; and understanding the influence of psychological health and cognitive deficits over the spectrum of TBI outcomes. While all these results are highly relevant, we anticipate that from a public health perspective, the greatest benefits can be accrued by improving the follow-up and treatment of patients after mild TBI.
Research results are being widely disseminated to patients, health care professionals and policy makers, and have already resulted in over 200 peer-reviewed manuscripts in the scientific literature. Importantly, though CENTER-TBI officially closed at the end of March 2021 after a project duration of 7 1/2 years, our data, imaging, and biosample repositories continue to be open for the scientific business of improving our understanding and management of TBI. These resources, particularly when combined with those in partner InTBIR studies, will continue to deliver outputs that will help to improve patient outcomes in TBI. Ongoing meta-analyses between CENTER-TBI and its sister study in the US,TRACK-TBI, are already providing novel insights and vital confirmation of results. CENTER-TBI and InTBIR have established productive global networks of researchers and research institutions, who will use the legacy of CENTER- TBI to improve patient care and injury prevention globally for many years to come.
Project Context and Objectives:
Traumatic Brain Injury (TBI) is a major cause of death and disability, causing great personal suffering to victims and relatives as well as huge direct and indirect costs to society. Approximately 2.5 million people in the European Union (EU-28) suffer a TBI, of whom 1.5 million are admitted to hospital and 57,000 die. Worldwide, TBI affects 50 million people and costs the global economy €325 billion annually. This means that one in every €150 annual global output is spent on the costs or consequences of TBI. TBI is a complex disease, but strong evidence in support of treatment recommendations is lacking and clinical management seldom adequately targeted. Conventionally, clinical TBI research has involved reductionist attempts to isolate out single factors for treatment, that do not account for the complexity of TBI and lack generalisability. Modern computational techniques and the availability of robust risk adjustment models facilitate more holistic approaches, such as Comparative Effectiveness Research (CER). CER makes use of differences in treatment and outcome. A specific feature of TBI that favours CER is the large between-centre and between-country differences in management and outcome. CENTER-TBI is a large-scale CER project with the following two overarching Global Aims:
(1) To improve characterization and classification of TBI
(2) To identify the most effective clinical care, providing high quality evidence in support of treatment recommendations and guidelines.

The specific aims are:
1. To collect high quality clinical and epidemiological data with repositories for neuro-imaging, DNA, and serum from patients with TBI (WP 1-6).
2. To refine and improve outcome assessment and develop health utility indices for TBI (WP 10, 11).
3. To develop multidimensional approaches to characterisation and prediction of TBI (WP 7, 8, 9, 10, 12, 15).
4. To define patient profiles which predict efficacy of specific interventions (“Precision Medicine”) (WP 13, 14).
5. To develop performance indicators for quality assurance and quality improvement in TBI care (WP 13).
6. To validate the common data elements (CDEs) for broader use in international settings, and to develop a user-friendly web based data entry instrument and case report form builder (WP 20, 22).
7. To develop an open source database compatible with FITBIR (WP20).
8. To intensify networking activities and international collaborations in TBI (WP 16, 22).
9. To disseminate study results and management recommendations for TBI to health care professionals, policy makers and consumers, aiming to improve health care for TBI at individual and population levels (WP 18, 19).
10. To develop a “knowledge commons” for TBI, integrating CENTER-TBI outputs into systematic reviews (WP18).

The complexity of TBI and research needs
TBI is considered “the most complex disease in our most complex organ”. It is characterized by great heterogeneity in terms of etiology, mechanisms, pathology, severity, and treatment, with widely varying outcomes. Falls and high velocity road traffic incidents cause different types of injury. TBI may consist of diffuse damage, contusional brain damage (bruises) or intracerebral hematoma. Structural abnormalities may or may not be visible on imaging. The clinical severity ranges from minor (minimal complaints, no visible structural damage) to unsurvivable. Conventionally, TBI severity is classified according to the Glasgow Coma Scale (GCS: range 3-15) into mild (GCS 13-15), moderate (GCS 9-12) and severe TBI (GCS<=8). Mild TBI is the most common form of TBI, occurring in around 90% of all cases, but has been least frequently studied. So-called “mild TBI” is however not so mild, and long term complaints are not uncommon. We now also recognize that TBI is not just an acute event, but can trigger a chronic process, with progressive injury over hours, days, weeks, months, and even years. Our past work has shown large differences in outcome between centres with up to a six fold higher risk in “poorer” vs. ”better” centres after adjustment for chance effects and case mix. Whilst basic research has increased our knowledge of the mechanisms involved, improvements in clinical management have not kept pace. Guidelines for the treatment of TBI are available, but the evidence underpinning these recommendations is weak. Moreover, current approaches to the characterization of disease severity and outcome have been unidimensional and not undergone refinement for more than three decades. Recent advances in genomics, advanced neuro-imaging, and biomarker development provide unparalleled opportunities for refinements in clinical characterization, offering more accurate disease phenotyping. Improved disease characterization will aid Precision Medicine, a concept enunciated by the US National Academy of Science. Such improved characterization and stratification allow for more targeted therapies.
Clinical research in TBI is particularly challenging due to disease heterogeneity, and has been further hampered by dispersion of efforts with little collaboration between researchers in acute and post-acute settings, and by research that focuses on isolated disease mechanisms and tests highly specific neuroprotective agents in underpowered clinical trials. Indeed, improvements in TBI care have come not from clinical trials, but rather from observational studies, expert guideline development and meta-analysis of individual patient data. However, the large scale international observational studies on TBI in Europe and the USA that underpin these improvements date back at least 20 years, and do not reflect current clinical care. Rigorously conducted observational studies in large and diverse populations have the potential to better characterize the disease and to reshape future care for patients with TBI. We aimed to address this need through provision of a contemporary observational data set with high quality, prospectively collected highly granular data.

CENTER-TBI: Addressing the research needs in TBI
CENTER-TBI brings together leading experts from 47 scientific institutes worldwide. CENTER-TBI is part of the International Initiative on TBI Research (InTBIR - http://intbir.nih.gov/) a collaboration of funding agencies formed in 2011. This initiative provides a platform to encourage international collaborations and application of novel insights and research efforts to improve patient care in TBI. It heralds a shift from the current reductionist approaches to clinical research towards broader approaches requiring multidisciplinary and international collaboration. The basic concept of this project is to exploit the existing heterogeneity in biology, care and outcome of TBI patients to discover underlying pathophysiology, to refine characterisation (paving the way for precision medicine approaches), and to identify effective clinical interventions in comparative effectiveness analyses. Key concepts of our research plan were to include patients of all severity levels, to follow them along their entire disease course across the chain of trauma care from injury scene to longer-term outcome, and to differentiate analyses by care pathway.

CENTER-TBI is an integrative project that optimizes existing knowledge and merges this with new evidence generated from a prospective observational Core Study and a Registry, collecting data on patients with TBI from 20 countries in Europe and Israel. As the study progressed, we have included substantial contributions from Australia, China and India, providing a wider range of clinical practice that allows us to examine greater variations in practice and use such variations to underpin CER analyses. In addition, meta-analysis across studies participating in the InTBIR initiative increases the power of numbers. Extensive profiling of participating centres has been performed to inform CER analysis. In the Core Study, the consortium has collected detailed data in over 4500 patients across all severities of TBI in the EU and Israel and an additional 1,200 in Australia and India. Enrolment was in three strata, differentiated by care path: (1) patients seen in the emergency room and discharged (ER stratum); (2) patients admitted to hospital, but not to the intensive care unit (Admission stratum); (3) patients admitted to the Intensive Care Unit (ICU stratum). Participating centres also maintained a Registry of basic data on all patients presenting with TBI (CENTER-TBI Registry), aiming to assess representativeness of the Core Study and to analyse effects of structural parameters (e.g. organisational) in greater numbers (recruitment: 22,772 patients). The Core Study combines emerging techniques (e.g. biomarkers, advanced Magnetic Resonance (MR) imaging, genomics), with innovative approaches to analysis, including state-of-the-art biostatistics and neuroinformatics. This Precision Medicine approach aimed to allow CENTER-TBI to achieve a step change in the integrated characterization of TBI, and provide clinically relevant constructs that can be used for disease characterisation, pathophysiological inferences, treatment stratification and outcome prognostication. Repositories have been created to allow legacy research with future technologies, benefitting from the extensive and systematic data collection in CENTER-TBI, including long-term outcome. To identify effective medical care (both acute and post-acute), the CENTER-TBI consortium analysed the effects of structure and processes of care at both the organisational (country, region) and at the individual patient level. Treatment of TBI patients varies substantially between centres and countries and depends on trauma organisation, local treatment policies, and physician preferences. In the Quality of Care literature, such characteristics are often differentiated as ‘structure parameters’ (e.g. level 1 or level 2 trauma centre, patient volume) and ‘process parameters’ (e.g. choice of surgical procedures, ICP monitoring and management protocols). It is implausible that all of the systems of care or treatment options offer equal benefit: some may well be better than others. However, substantial patient heterogeneity, coupled with relatively small patient numbers (even in larger centres) means that the link between intervention and outcome is impossible to make in a single centre. CENTER-TBI provides the numbers and methodologies to make such linkages possible. The existing heterogeneity of presentation, differences in management and variability of outcome in TBI provide a compelling argument for rigorous comparative effectiveness research, the outputs of which will provide a rational basis for optimising health care delivery for populations, and clinical management for individual patients.
Research results are integrated with systematic reviews in a process of knowledge transfer and disseminated to patients, health care professionals and policy makers. CENTER-TBI wishes to break with past dogmas and restrictive traditions. As such, the consortium actively seeks global collaborations, includes emergent technologies, involves non-medical scientists, in particular bio-informatics specialists, and seeks collaborative data sharing initiatives. The CENTER-TBI project will contribute towards the overall goals of InTBIR, by identifying more effective and efficient treatment provision, thus improving outcome and reducing costs. The science in the project provides novel information on disease processes, treatment, outcome, and prognosis in TBI, identifying new therapeutic targets and therapies; while the CENTER-TBI repositories ensure opportunities for legacy research. Thus, the project has the potential to improve current health care and its delivery at both population and individual levels, deliver early scientific advances that could improve the care of patients with TBI, and provide a rich investment for future biomedical research.
The project duration was 7 ½ years, including up to 5 years for recruitment and follow-up – revised upwards, due to various delays including those related to COVID-19.
Project Results:
1.3.1 Introduction to structure of this section
CENTER-TBI is a hugely complex project, spanning the entire spectrum of TBI across all severities and along all trajectories of care. The main pillars underpinning our analyses are the Core study (with detailed, highly granular data) with its associated repositories (imaging, biomarkers and genetics), and the Registry, collecting more basic data in larger numbers. In addition, we included systematic reviews of available evidence, introducing the novel concept of Living Systematic reviews, performed extensive profiling of participating study sites to establish their organisational structures and care preferences, and performed in-depth analysis of (differences in) ethical regulations and physician attitudes with a particular focus on patients with acute mental incapacity (as is common after TBI). Here, we provide an integrated summary of the results obtained and translate these into policy and practice recommendations. We have structured this Report to align with the Specific Objectives of CENTER-TBI (see section 1.2) In the final part of this section, we will summarize findings towards attainment of our Global Aims, and present policy and practice recommendations. Citations listed in blue can be found on the CENTER-TBI website (https://www.center-tbi.eu/) and in section 2.1 of this Report.

1.3.2 Setting the Stage
In preparation for the CENTER-TBI studies – and as part of the CENTER-TBI project - we aimed to present an extensive overview of the current knowledge on TBI epidemiology, treatment and research and to perform a detailed characterisation of the organisational structure and treatment preferences of participating centres to the Core study and registry. A major output was the publication of a Commissioned Issue on TBI for the Lancet Neurology, the leading medical journal in the fields of Neurology and Neurosurgery. The manuscript and associated commentaries are available via the link: http://www.thelancet.com/commissions/traumatic-brain-injury. The manuscript presents an up-to-date and comprehensive overview of the science and practice of TBI and identifies gaps in our knowledge. It is now viewed as the main reference resource on TBI and has already been cited over 700 times. The Lancet Neurology Commission was released at the European Parliament on Nov 7, 2017. The occasion was attended by a patient and his Mother, who made a very compelling plea to put the huge public health burden and needs of patients and their relatives, posed by TBI, high on the political and policy agenda. In addition, the Lancet Neurology published four more conventional reviews on specific topics to supplement the Commission: Coagulopathy (Maegele et al 2017), Targeted treatment in the ICU (Stocchetti et al 2017), Chronic and evolving neurological consequences of TBI (Wilson et al 2017), and Paroxysmal sympathetic hyperactivity after acute brain injury (Meyfroidt et al 2018). The evidence base was expanded by 5 Living systematic reviews (a concept pioneered by CENTER-TBI) and 19 conventional systematic reviews (see section 1.3.12).
Extensive provider profiling of CENTER-TBI centres was performed prior to start of data collection and the outputs of this process were published in 11 manuscripts. Additionally, an abbreviated version of the provider profiling performed in Europe was completed by the 45 Chinese centres that participated in the China Registry data collection. We observed both some concordance and substantial variations regarding various aspects of TBI care between Chinese and European centres There were more dedicated neuro-intensive care units in Chinese centres than in Europe (97.8% versus 59.7%) and treatment decisions in the ICU were mainly determined by neurosurgeons (57.8%) in China, while in Europe (neuro) intensivists often took the lead (61.2%). For treatment of refractory intracranial hypertension, a decompressive craniectomy was more frequently seen as general policy in China compared to Europe (89% vs 44.6%).

IRB approvals and consent procedures
A basic prerequisite for conducting a clinical study is approval by the country/centre specific Institutional Review Boards (IRB), and implementing procedures for obtaining consent according to national and local regulations. The European Union aims to optimize patient protection and efficacy of health-care related research by harmonizing procedures across Member States. CENTER-TBI, with its broad representation of many European countries, offered a unique opportunity to explore the degree to which such harmonization has been successful. The CENTER-TBI protocol was evaluated in 18 European countries (excluding Israel) by institutional reviews boards (IRBs) of 66 neurotrauma centers. Fourteen IRBs considered CENTER-TBI an observational study, two an interventional study, as the protocol described blood draws and outcome assessments that would not be part of clinical routine. Primary IRB review was conducted centrally in 61% and locally in 39% of countries. Median time till basic approval was 98 (IQR 94-114) days for central review, considered directly applicable to all national centres, and 50 (IQR: 29-102) days for centres only requiring local approval. Basic approval was reached in one (44%), two (33%) or three (23%) review rounds. Additional local IRB approval was required in 55% of the countries with central procedures and increased the time till final approval. Although additional local IRB approval is generally considered more a feasibility check, in practice a full new review was often conducted. The total median duration across centres from submission of the CENTER-TBI protocol until definitive approval was 114 days (IQR 75-224 days) with a range from 1 to 535 days. We conclude that, despite the aim for harmonization, substantial variation remains in IRB procedures across EU Member States, posing challenges to collaborations in research.
Patient informed consent is one of the basic ethical principles in clinical research. A unique feature of research in TBI is that most patients have acute mental incapacity, and cannot provide consent themselves. Several pragmatic alternatives exist, of which proxy consent is most frequently used. However, proxies may be too overwhelmed by emotions to provide a valid consent, and in emergency situations such as severe TBI, there may be insufficient time to consult with proxies, or proxies are unavailable. An option then is to defer consent to a later moment. We found a significant variation in the use of consent procedures between and within EU Member States. Deferred consent was only used in 26% of the neurotrauma centers involved, although considered valid in 82% of the centers and being described as a valid procedure in the EU General Data Protection Regulation and the Clinical Trials Regulation.
Our experience shows that harmonization of informed consent procedures in EU Member States still needs to be improved. Lack of clear directions in European and especially national legislation result in substantial variation in IRB approval of clinical studies, and this may adversely affect the design and conduct of multinational clinical research on TBI, and more in general on all disorders characterized by acute mental incapacity.

1.3.3 Collection of high quality clinical and epidemiological data with repositories for neuro- imaging, DNA, and serum from patients with TBI (WP 1-6).
The CENTER-TBI Core Study and Registry enrolled patients with TBI from Dec 2014 to Dec 2017. Inclusion criteria for the core study were a clinical diagnosis of TBI, presentation <24 hrs after injury, an indication for computerized tomography (CT) scanning, and informed consent obtained according to local and national requirements. Patients were differentiated by care pathway and assigned to the emergency room (ER) stratum (patients discharged from an emergency room), Admission stratum (patients admitted to a hospital ward), or intensive care unit (ICU) stratum (patients admitted to the ICU). The CENTER-TBI Core study was conducted in accordance with all relevant laws of the EU, if directly applicable or of direct effect, and all laws of the country where recruiting sites were located, including, but not limited to, privacy and data protection laws and regulations, the laws and regulations on the use of human materials, and all relevant guidance relating to clinical studies from time to time in force including, but not limited to, the International Council on Harmonisation guideline on Good Clinical Practice (CPMP/ICH/135/95) and the World Medical Association Declaration of Helsinki. The list of sites, ethics committees, approval numbers, and approval dates is available online (https://www.center-tbi.eu/project/ethical-approval). The Registry collected administrative data not requiring consent and covered a site-specific, convenience-based period during the recruitment period of the core study. A total of 65 sites from 20 countries participated). The CENTER-TBI initiative attracted global interest and included substantial contributions from Australia, China and India).
CENTER-TBI Core Study: Descriptive analysis:
Data from 4509 patients enrolled in Europe and Israel were analysed. Of these, 848 (19%) patients were in the ER stratum, 1523 (34%) in the Admission stratum, and 2138 (47%) in the ICU stratum. The relative distribution across strata to a large extent reflects logistic considerations with regard to enrolment at sites. The strong drive in busy Emergency Rooms for a fast turnover of patients proved challenging for recruiting patients to the ER stratum. A more generalizable picture of care pathways and strata is provided in the Registry. Results of descriptive analyses have been published (Steyerberg et al 2019; Huijben et al 2020). We summarize some of the main findings:
• The median age was 50 years [IQR 30–66], substantially higher than in previous studies, and 1254 [28%] patients were aged >65 years. The high percentage of older patients who suffer TBI is highly relevant as up till now most clinical trials excluded patients over the age of 65. Older patients have therefore been disenfranchised from clinical trials, and as a consequence little evidence exists to support their treatment and guidelines are not applicable to this age group.
• A total of 462 (11%) patients had serious comorbidities, illustrating that TBI is no longer a disease of previously healthy young people. The presence of co-morbidities can adversely affect the disease course.
• A total of 772 (18%) patients were taking anticoagulant or antiplatelet medication. Such medication may lead to rapid progression of haemorrhagic lesions (see also section 1.3.6)
• An incidental fall was the most common cause of injury in the ER (51%) and admission strata (51%), but not in the ICU stratum (41%), where road traffic incidents were the main injury cause (45%).
• Alcohol was contributory in 1054 (25%) patients, but varied by cause of injury (17% in road traffic incidents, 28% in incidental falls, and 64% in violence-related TBI). These data illustrate the success of traffic-related alcohol prevention campaigns, but highlight a need for targeted prevention campaigns to reduce the number of TBIs due to falls, particularly in the older people.
• Major extracranial injuries (abbreviated injury score ≥3) were reported in 422 (28%) patients in the admission stratum and in 1174 (55%) in the ICU stratum. The body region most commonly injured was thorax and chest (n=742 [35%]), and concomitant serious spinal injuries occurred in 374 (18%) patients. The co-occurrence of TBI with injuries to other parts of the body emphasizes the need for a multidisciplinary approach to treatment.
• Substantial inter-country differences existed in care pathways and practice, but not in outcome.
• 6 month mortality was 1.3 % (9/694) in the ER stratum, 5.5% in the Admission (70/1264) and 21.3% in the ICU stratum (394/1846). Incomplete recovery (defined as a 6 month GOSE <8) was found in 30% of patients in the ER, 53% in the Admission and in 84% in the ICU stratum. The high percentage of incomplete recovery in the ER and Adm strata, in which most patients had mild TBI, illustrates that “mild TBI is not so mild”.
• In patients with moderate to severe TBI mortality was lower than predicted from the IMPACT prognostic model (observed to expected ratio 0.70 [0.62– 0.76]) but unfavourable outcome (defined as a GOSE<5), was not (1.06 [95% CI 0.97– 1.14]). These data suggest that treatment has improved with fewer deaths, but at a cost of more survivors with disability.

CENTER-TBI Registry:
Fifty-six study centres from 17 European Countries and Israel enrolled 22782 patients to the CENTER TBI Registry. A total of 21681 TBI patients - with clinical care pathway and known injury mechanism data - were included for analysis - a median of 247 (IQR 63–473) from each centre. Patients enrolled to the CENTER-TBI registry had a median age of 55 years (IQR 32-75 years), and a 61% male preponderance; 55% had pre-existing medical conditions, 12%(2578) and 11·4%(2466) were taking pre injury anticoagulants or antiplatelet therapy respectively. 82%(17702) presented to the study hospital Emergency Department with mild TBI (Glasgow Coma Scale (GCS) 13-15. Patients presenting directly had CT brain imaging conducted a median of 68 minutes after ER arrival (IQR: 34-141). 31·1% (6746) had abnormalities on CT imaging. The majority (57·1%) were admitted to hospital, 19·2% to intensive care on average two and a half hours after ED arrival. Our analyses focussed on 1) generalizability of Core data and 2) impact of energy transfer mechanisms. Overall, there were differences in patient characteristics between the Core study and the registry, caused by exclusion of patients with pre-existing neurological disorders (including dementia) in the Core study, and differential recruitment to strata with relatively more patients enrolled into the ICU stratum in the Core study. When, analysed by stratum, however, patients in the Core study broadly resembled those in the Registry. These data and comparisons to external registries, such as the UK Trauma Audit and Research Network (TARN), indicate that the CENTER TBI recruiting centres had used an appropriate purposive sampling strategy to recruit patients reflecting the generality of TBI presenting to each centre. Our interest for the impact of energy transfer mechanisms was ignited by the observed high number of older people injured by falls, whilst patients injured by high energy transfer (e.g. road traffic collisions) are prioritised by Emergency Medical Service trauma triage tools as they are considered to have more severe injury. We found that 40% (8622/21,681) of patients in the Registry were injured by low energy falls. These patients have similar rates of CT abnormalities and in-hospital mortality as those injured by other mechanisms, but are 50% less likely to receive ICU care or emergency interventions. This indicates that high-energy injury characteristics should be de-emphasized for injury scene and ED triage of older people with TBI.

CENTER-TBI China Registry:
The CENTER China Registry collected data on patients with TBI admitted to hospitals across China in the same period and according to a similar format as the CENTER-TBI Registry. This intrinsic feature of "twin" studies illustrates the benefits of standardized data collection according to a common format, and highlights the relevance of understanding the heterogeneous nature of TBI and its treatments in different continents. Data of 13138 patients from 52 hospitals in 22 provinces of China were analysed (Gao et al 2020). Most patients were male (9782 [74%]), with a median age of 48 (IQR: 33-61), and 2217 (17%) > 65 years of age. Road traffic incidents were the major cause of TBI (6548 [50%]). Injuries causing TBI most commonly occurred between 9 am and 11 pm and peaked at 10 am (n = 1165; 8.9%). A total of 3882 patients (30%) were transferred from another hospital to the study centre, with substantial variations in secondary referral rates across provinces. ICP monitoring, external ventricular drainage (EVD), craniotomy and decompressive craniectomy were performed in 1509 (11%), 774 (5·9%), 2679 (20%) and 2170 (17%) patients respectively with substantial variation occurring between provinces and centres. Between centre variations were particularly large for ICP monitoring (MOR: 7.64 CI: 4.77-12.98) and for the use of external ventricular drainage (MOR: 9.37 (CI: 4.77 – 18.63). Overall hospital mortality was 4.8% (637), and in severe TBI 19.7% (552). The observed mortality was lower than expected according to the CRASH basic model (O/E ratio 0·49, 95% CI 0·45-0·53). Substantial variation existed between centres (MOR: 2.0 (1.55-2.42)) which is larger than observed in the CENTER-TBI data from Europe. Comparison with the CENTER-TBI data from Europe show that in China, TBI remains a problem primarily of young and middle-aged adults, leading to huge losses in health and labour capacity. We anticipate that the changing demographics (ageing) of the population in China combined with further improvements in road traffic safety will lead to an increase in domestic injuries as cause of TBI in the near future, in particular in older people, thus following a trend observed in high income countries. Between centre differences in treatment and in outcome were larger in China compared to Europe. Whilst the observed differences between provinces and centres offer potential to evaluate the performance of organization and professional behaviour at the level of institutions, they also indicate the need for initiatives to improve health care policy for TBI to take local aspects into consideration and to tailor trauma systems to better fit the situation in different areas. The results of the CENTER China Registry with comparable mortality rates to European data highlight the huge potential that collaborations with China may offer to advance the care for patients with TBI.

CENTER-TBI repositories:
In the context of the Core Study, we have established Repositories for Imaging studies (CT and MR Images), for blood samples and for DNA. These are the largest in the world in TBI, offering opportunities for legacy research after the formal end of CENTER-TBI. The Neuro-imaging Repository is maintained by Icometrix (Leuven, Belgium), and contains a total of 8545 CT images, 630 early MR scans and 719 MR scans obtained at follow-up). The results of standardized qualitative and quantitative reporting of radiological characteristics have informed many of the analyses of CENTER-TBI.
The CENTER-TBI Biobank, maintained in Pecs, Hungary, has been populated with 60187 aliquots from 8026 sample collections of 3803 patients. Serum samples were divided across 8 aliquots each and stored at -800C. Three of these aliquots have been used for biomarker assays (S100B and NSE in Pecs and GFAP, UCHL1, total tau and NFL at the McKnight Brain Institute, University of Florida, USA). Collaborations with external academic and commercial Parties have been established to facilitate extended analyses (e.g. metabolomics and lipidomics) and to inform the design of clinical trials. In this context, aliquots have been provided to ABCDx SA; Geneva; Switzerland, University of Örebro; Örebro; Sweden, NanoDx Inc. (former BioDirection); Southborough, MA, USA, and the University of Edinburgh; Edinburgh; UK. A substantial number of aliquots (both pristine and smaller left-over aliquots), however, remain in the Biobank, and are available for novel research initiatives. A total of 3695 whole blood samples were transferred to Cambridge for banking and DNA extraction. Leveraging with samples from previous EU and UK funded studies, we were able to add an additional ~700 samples. After exclusion of patients based on failed DNA extraction, non-European ancestry (so imputation not possible), incomplete outcome data, and missing covariates, genotyping data are available from 3187 patients (73%) for association analysis. Genotyping was performed at the Finnish Institute for Molecular Medicine (FIMM). Aiming for more robust analyses in larger numbers, these data will be combined with a similar exercise in TRACK-TBI (our sister study in the USA) using identical phenotyping and genotyping, and with an additional 409 patients recruited at the Massachusetts General Hospital (Massachusetts General Partners; MGB) in Boston. This has resulted in a total cohort of 4710 TBI patients with European ancestry, and a further 558 non-European ancestry patients –providing a total cohort of 5,268 datasets for genetic association analysis, yielding adequate power to detect OR of ~1.5 for genetic effects in a GWAS. This allowed us to report the first GWAS/TWAS of TBI outcome, utilizing the largest sample for any genetic association study of TBI to date (see section 1.3.5).

1.3.4 Refining and improving outcome assessment and developing health utility indices for TBI (WP10, 11).
Traumatic Brain Injury should not be considered as “an event”, but as a “process”, resulting in a large number of survivors with with functional, cognitive, emotional and physical consequences. These impairments occur in all grades of severity, and their broad range implies a need for multidimensional approaches to outcome assessment. Conventionally, studies have assessed outcome according to the Glasgow Outcome Scale – Extended (GOSE) that ranges from 1 (death) to 8 (Upper good recovery). In CENTER-TBI, we sought to obtain a comprehensive assessment of outcome, including clinician-reported outcomes, patient-reported outcomes, and performance-based physical and cognitive outcomes. We collected data on functional outcome (GOSE), Health related Quality of Life (generic: SF-36v2, SF-12v2, and disease-specific: Qolibri and Qolibri-OS), anxiety and depression (GAD-7 and PHQ-9), posttraumatic stress disorder symptoms (PCL-5), post-concussion symptoms (RPQ) and cognitive performance (Cambridge Neuropsychological Test Automated Battery (CANTAB), Rey Auditory Verbal Learning Test (RAVLT) and Trail Making Parts A & B). Our main aim was the selection of the most sensitive instruments to inform multidimensional approaches to outcome assessment after TBI, including the refinement of procedures for administration and interpretation of the GOSE, and exploration of inter-dependencies between outcomes.

Linguistic validation and psychometric evaluation
A major challenge was to ensure applicability and comparability of outcome instruments in an international setting. Many of instruments were only available in English or, at best, in a limited number of languages. We undertook linguistic validation and psychometric evaluation. Linguistic validation is challenging as it needs to address semantic, syntactic, cultural and conceptual differences, while maintaining the content of each instrument across languages. In total, 237 translations and 211 linguistic validations were carried out in 20 languages. Psychometric analyses showed that reliability of all instruments was satisfactory to excellent, and that the instruments were comparable with each other and to the original versions. Validity analyses demonstrated that correlations between measures were consistent across languages. Translations of the outcome instruments are a major output of CENTER-TBI and provide a solid basis for multinational TBI research and practice. They are available on the CENTER-TBI website (https://www.center-tbi.eu/project/validated-translations-outcome-instruments).

Approaches to GOS/GOSE administration and rating
The GOS(E) is an ordinal scale, that is commonly administered in a structured format, addressing 7 areas of functioning. It may be administered by personal interview, telephone interview or by postal or web-based questionnaire. Assignment of the most appropriate rating (1-8) can be performed centrally, by an automated algorithm or by the Investigator. We explored agreement between interview- and questionnaire-based assessments. Overall, both methods agreed well. However, some differences were noted: Compared to questionnaires, interviewers recorded more problems with work, fewer limitations in social and leisure activities, and more symptoms. Interviewers also sometimes applied judgement when assigning an overall rating, particularly for cases with potentially unfavourable outcomes. However, associations with prognostic factors and patient reported outcomes were very similar in strength for interviews and questionnaires. The findings support the utility of questionnaires in studies where this form of contact can offer practical advantages over interviews. In CENTER-TBI, central assignment of GOS(E) rating was preferred for a composite GOSE. which combined ratings from interviews and questionnaires.
Discussions with our US colleagues from TRACK-TBI revealed that different approaches to assignment of the GOSE rating exist across the Atlantic. In Europe, the intent is to capture the overall consequences of injury for function, including possible effects of extracranial injuries, whilst in the US, the primary aim has been to capture TBI-related disability and exclude effects of extracranial injuries. Clearly, this may lead to substantial differences in reported outcome outcomes between studies. Although we recognize that in practice it may be difficult to disentangle effects of systemic injuries from those of brain injury, and that attempting to do so risks introducing an element of subjectivity, both approaches may have merits. We suggest that future studies should differentiate between “GOSE-All” (including polytrauma and any side-effects of an intervention) and GOSE-TBI, in which disability that is clearly unrelated to brain injury is discounted. A set of guidelines for use of the GOSE was developed in collaboration with colleagues in TRACK-TBI and thus represents a consensus for Europe and North America. The manual covers administration of the GOSE and common issues that arise; it has been published in an open access format to maximise dissemination (Wilson et al., 2021. https://doi.org/10.1089/neu.2020.7527).
Any study, and in particular studies on TBI, can suffer from loss to follow-up. In general terms, imputation is preferred over a complete case analysis. Previous studies, including most clinical trials, have imputed missing values according to the last observation carried forward (LOCF) approach. This approach, however, insufficiently takes the natural recovery trajectory into consideration. We explored alternative approaches to imputation that take multiple time points into account. We found a multi-state model to interpolate missing outcomes based on available observations, both before and after the pre-defined time window for assessment, to function best with performance superior to the LOCF approach (Kunzmann et al 2019).

Outcomes after TBI
Wide-ranging analyses have been performed on the data available. At 6 months a GOSE rating was available for 84% of patients, and for survivors completion rates ranged from 55% to 57% for patient-reported outcomes and from 37% to 46% for cognitive tests.
We conducted extensive methodological analyses to examine the ability of assessments to identify differences after TBI, and compare their relative sensitivities. The GOSE (recovery) displayed the highest ability to capture changes in subgroup analyses over all time points, followed by the QOLIBRI and QOLIBRI-OS (TBI-specific HRQoL) and then by the SF-36v2 and SF-12v2 (generic HRQoL). Psychological outcome measures (anxiety: GAD-7, depression: PHQ-9, posttraumatic stress disorder: PCL-5, and post-concussion symptoms: RPQ), the QOLIBRI and the QOLIBRI-OS as well as the mental component of the SF-12v2 and SF-36v2 were found to differentiate well between individuals with premorbid psychological problems, especially at later time points (i.e. 12 months after TBI). QOLIBRI and QOLIBRI-OS were found to be most sensitive in capturing differences between patient subgroups. The PHQ-9 and then the RPQ are able to distinguish functional recovery states as measured by the GOSE at three time points (3, 6, and 12 months after TBI). The PCL-5 and SF-12v2 mental component score (MCS) appeared more sensitive at 12 months after TBI.
Six-month cognitive outcomes were collected in 1554 patients, making this one of the largest studies of cognition after TBI conducted to date. We related cognitive functioning to GOSE scores and found that processing speed is the domain most strongly related to function in daily life (Wilson et al 2021). Deficits in cognitive performance were particularly evident in patients who were dependent (GOSE 3 or 4) or unable to participate in one or more major life activities (GOSE 5).
At higher levels of function (GOSE 6 to 8), cognitive performance was similar across categories. We also found that cognitive function was associated with symptoms of post-traumatic stress disorder assessed by the PCL-5. A total of 13% (153/1134) of patients screened positive for probable PTSD. Low performance on cognitive tests assessing attention, cognitive flexibility and verbal long-term memory were associated with probable PTSD following TBI.
A particular focus in our studies was on patients with a mild TBI (GCS 13-15). We found that 50% (1239/2464) had a GOSE below 8 at 6 months, demonstrating that “mild TBI is not so mild”. Around 25% had SF12v2 summary scores below threshold for impairment (scores <40) and 26% had RPQ scores ≥ 16, indicating significant postconcussion symptoms. In patients with mild TBI who were discharged home from the ER, the respective rates were 29% (GOSE), 21-23% (SF12v2) and 21% (RPQ).
These high rates of impairments are of particular concern as in current clinical practice most patients discharged from the ER are not routinely scheduled for any follow-up. We explored in greater detail the presence of post-concussion symptoms and other outcomes in patients with complicated (with abnormalities on CT scan) versus uncomplicated (normal CT scan) mild TBI and their evolution between 3 and 6 months. A higher percentage of patients after complicated mTBI were classified as having significant symptoms at three (complicated: 46% vs. uncomplicated: 35%) and six months (complicated: 43% vs. uncomplicated 34%). However, after adjusting for baseline covariates, the difference between complicated and uncomplicated mTBI at three months appeared minimal: odds ratio: 1.28 (95%CI: 0.98 - 1.70). We note that post-concussion symptoms are non-specific and are reported in a substantial proportion of the general population (up to 18% for a rating score of 3, and up to 45% for a rating score of 2). We further note a large similarity in symptoms captured by the RPQ and symptoms reported in long-term COVID, and indeed in other survivors of critical illness. These similarities suggest that some part of the cognitive and psychological features seen in TBI may be a consequence of the host response, rather than the primary injury. Patients after complicated mTBI had significantly lower GOSE scores, and reported lower TBI-specific and generic HRQoL compared to those after uncomplicated mTBI. Both groups showed a tendency to improve from three to six months after TBI. Overall, impairment rates after mTBI were much more strongly related to stratum than to presence of CT abnormalities, suggesting that selection by stratum may be more appropriate than targeting complicated mTBI for achieving an enriched population in clinical trials. Our data, showing high impairment rates after mTBI, highlight the need to take account of the short and long-term impact on outcome for patients after mTBI, and to provide structured screening and individualized tailored therapy when such impairments are detected.

Towards a multidimensional approach to outcome assessment
There is an increasing awareness of the importance of multi-dimensional outcome assessment in TBI, but a lack of practical advice on implementation. We propose that functional outcome can be used as a framework for guiding the application of patient-specific assessments. This is a ‘sliding’ approach, which uses severity of disability on the GOSE as a guide to the suitability of assessments. Support for this approach comes from examining relationships between global functional outcome and other assessments based on cross-sectional data collected from 2573 patients. Outcome completion rates were 80% or above for the entire sample, but substantially lower among patients with severe disability. Impairments of mental health and health-related quality of life were common in all groups except in patients with upper good recovery. Broadly, it appears that assessment might usefully be tailored to at least three levels of recovery (severe disability, moderate disability and lower good recovery) reflecting severity of disability and impairment. Upper good recovery may represent a fourth level, where more sensitive assessments are needed.
Besides a “sliding approach”, in the search for multidimensional descriptions of outcome it is logical to consider combining assessments. For example, severe disability can be subdivided based on the extent of impairment. Similarly, upper levels of outcome could be subdivided by the presence of impaired cognition, health-related quality of life, or mental health. For mild TBI, a multidimensional dichotomized outcome can be created contrasting “complete recovery” and “incomplete recovery”, in which the latter is defined as GOSE <8 or impairment on another outcome domain. In 2464 patients with mild TBI and complete GOSE assessment, we found that 11% would be rated as having an unfavourable outcome according to the conventional dichotomization of the GOSE (GOS<5 vs GOSE>=5), but that 51% had a GOSE<8. This confirms that dichotomizing the GOSE at a level of <8 versus 8 is more appropriate for mild TBI. We further explored the concept of “incomplete recovery” based on assessment with multiple instruments in a cohort of 1612 patients with mild TBI in whom measures were available. We found that in most cases, a GOSE <8 was the driving impairment.
However, a substantial number of patients had impairments in other domains. Creating a composite outcome, considering in addition to the GOSE, the SF12v2 MCS and PCS, the Qolibri and RPQ scores, we found that 63% of patients would be considered as having incomplete recovery defined by impairment on one or more of the instruments and 40% when defined by impairments on 2 or more instruments. These approaches comprise multidimensional outcome tools that have the potential to better capture the consequences of mild TBI.We suggest that the “sliding approach” is appropriate to clinical practice for targeting additional assessments after TBI and that the concept of “incomplete recovery” be considered as endpoint for clinical studies on mild TBI.
Health Utility
We performed a web-based survey, which was completed by 13,623 respondents from the Netherlands, United Kingdom and Italy (Voormolen et al, 2020). We derived a value set for the QOLIBRI-OS which allows calculation of utility scores for TBI health states. We also calculated disability weights by Glasgow Outcome Scale Extended (GOSE) severity level derived from health-related quality of life (HRQoL) data of 2215 TBI patients. The utility scores and disability weights are needed for economic evaluations, and for the calculation of summary measures of population health, which may be used to inform decision-makers on the best interventions and strategies for TBI patients. In a case study, we explored this approach to compare the cost-effectiveness of two treatments used to reverse brain swelling after traumatic brain injury: “hypertonic saline” (HTS) and “mannitol” and demonstrated that the statistical power to detect treatments effects is higher when the incremental cost effectiveness ratio (ICER) is based on QALY rather than "good recovery" versus “non-optimal recovery”. The development of disease-specific health utility indices for TBI now enables cost-effectiveness analyses and policy making.

1.3.5 Multidimensional approaches to characterisation and prediction of TBI (WP 7, 8, 9, 10, 12, 15).
Improved characterisation and classification of TBI is essential to developing individualized treatment approaches in the context of Precision Medicine. Although characterisation and classification of TBI are multidimensional concepts, most previous approaches have focused on a single dimension, e.g. clinical severity as measured by the Glasgow Coma Scale (GCS) or the presence of structural damage assessed by computerized tomography (CT scanning), and often reduced these further to over-simplified constructs. Novel approaches (genetic risk stratification, advanced imaging, and emerging biomarkers) could offer substantial gains in Precision Medicine and therapy stratification. Prediction models, which combine several characteristics to predict outcome, have many applications. They can aid precision medicine by identifying patients at higher risk, allowing selection for more aggressive therapies; provide more objective information regarding outcome expectations to patients and their relatives; support timely clinical decision-making; aid stratification of patients in randomized controlled trials (RCTs); and provide a basis for benchmarking quality of care. We present a summary of the CENTER-TBI results in three parts: epidemiological insights, characterization and prediction.
Epidemiological insights
Aiming for an integrated insight into the current epidemiology of TBI, we compared findings from the CENTER-TBI Core study and Registry to those of our living systematic review (for more on LSRs, see section 1.3.12). On comparison between the LSR and Core study, the median ages were very similar (50 and 51 years). The proportion of males was somewhat higher in the systematic review (72% vs 67%). In both cases, falls (46% in CENTER-TBI and 41% in the systematic review) and traffic accidents (38% and 39%) were the predominant causes of injury. However, studies in the LSR and the Core Study were biased towards in-hospital patients. The CENTER Registry provides a more general picture of TBI seen in hospitals (including patients discharged from the ER). Here, the higher age and greater incidence of falls as cause of injury were more apparent. We conclude that TBI is moving towards older ages with incidental falls becoming the predominant cause of injury. Age is recognised as a strong prognostic factor in TBI outcome, but is associated with comorbidities, treatment of these comorbidities, and frailty (see section 0). Separation of these complex relationships is critical since some aspects (e.g. therapeutic anticoagulation; see later) are correctable causes of poor outcome which demand personalised therapy. Detailed data of the Core Study showed that alcohol abuse was observed in 28% of incidental falls, and in 9% sedatives or sleeping pills were used before the accident. Alcohol-related road accidents were recorded in 17% of cases and abuse of cannabis commonly seen in violence-related injuries (15%).

Characterisation of TBI
In our aim towards innovation, we focused on genetics, (advanced) neuro-imaging, biomarkers and coagulopathy.
Genetic analyses and heritability
We conducted the first genome-wide and transcriptome-wide association (GWAS/TWAS) study of TBI outcome, utilizing the largest sample (n=5,268) for any genetic association study of TBI to date. To this purpose, the GAIN consortium (Genetic Associations in Neurotrauma) was established, including patients from TRACK-TBI, previous EU and UK funded studies and a cohort from the Massachusetts General Hospital in Boston (USA). The estimated heritability of TBI outcome was 28 (+14)%. GWAS revealed no hits with genome-wide significance (p < 5 x 10−8), but identified 84 variants in 12 independent loci which met a lower pre-specified sub-genomic statistical threshold (p < 10−5) for association with TBI outcome. An exploratory analysis of past published candidate variants only revealed a single variant that reached significance after correction for the number of candidate gene associations studied (rs1800450 in the MBL2 gene). Notably, APOE genotype was not significantly associated with outcome – in keeping with the relatively modest effects we demonstrated in our LSR on Apolipoprotein E4 polymorphisms, despite the many past positive studies. These findings indicate the need for extreme caution when interpreting results from previous candidate genetic association studies that have often been underpowered and subject to publication bias. . The overall heritability estimate we found is consistent with the hypothesis that common genetic variation significantly contributes to inter-individual variability in host response and TBI outcome, a finding which will be refined in subsequent studies.

Neuro-imaging
We established the largest neuro-imaging repository in TBI to date, including 8545 CT images, 630 early MR scans and 719 MR scans obtained at follow-up (see section 0). We developed tools for automated segmentation and lesion detection of CT images (icobrain tbi: https://icometrix.com/services/icobrain-tbi and BLAST-CT: http://deepmedic2.doc.ic.ac.uk:8080/). We demonstrated, that compared to the ABC/2 method, automated segmentations are accurate and have a great potential to expedite the interpretation of large numbers of scans. We showed that central reporting of Neuro-images should be preferred over investigator-based assessments in the context of multicentre clinical studies. For harmonisation of advanced MR imaging, we conducted phantom studies and studies on healthy volunteers.
In collaboration with TRACK-TBI we performed a hierarchical clustering analysis of CT features in mild TBI, and identified 3 major clusters of CT features: 1) Contusion/subarachnoid hemorrhage/subdural hematoma; 2) intraventricular, hemorrhage/petechial hemorrhage; and 3) epidural hematoma. In particular Cluster 1 was predictive of both incomplete recovery (Glasgow Outcome Scale- Extended (GOSE) score <8) and more severe impairment (GOSE<5) out to 12 months post-mTBI. Our MR studies showed that 30% of patients with mild TBI and a normal CT scan on presentation demonstrated structural abnormalities on MR imaging.
We found that Qualitative MRI added to the prognostic accuracy of outcome prediction beyond conventional clinical and CT based variables, and quantitative volumetric MRI and DTI metrics provided further added value beyond that afforded by clinical reporting of MR images. These findings suggest that advanced MRI reveals potential neuroanatomical substrates of mTBI in white matter and is most strongly associated with odds of recovery if performed within 72 hrs.
The neuroanatomical substrates of persistent late symptoms are particularly difficult to explain in mild TBI, where CT images are often normal, and even structural MRI may show only minimal changes. In this context, we explored functional MRI, and found that individuals with mTBI showed greater global connectivity (p=0.013) yet reduced functional complexity (p=0.027) compared to healthy age-matched controls obtained from the CamCan study (https://www.cam-can.org/). These changes in brain connectivity in mTBI may be a compensatory adaptation to reduced complexity (and hence efficiency) of information transfer.
We further explored the added value of advanced MR imaging in patients with very severe disorders of consciousness after TBI. A model including age and deep white matter diffusion metrics (fractional anisotropy and mean diffusivity) obtained an AUC of 0.93 on the MRI-COMA training dataset. On the validation dataset, the model successfully (specificity above 95%) identified one in two patients who had an unfavourable outcome at one year post TBI, and two-thirds of the patients who experienced a favourable outcome. These results imply that advanced MR imaging can potentially support decision-making at the individual level within the framework of a multimodal evaluation while the patient is still in the ICU.
Biomarkers
The main focus of our work was on the (added) value of biomarkers for triaging patients with mTBI for CT scanning. We analyzed 6 biomarkers ((S100B, NSE, GFAP, UCHL1, total tau and NFL) in serum samples obtained within 24 hours of injury (2867 patients overall of whom 1951 with mTBI), and related these to the presence of traumatic intracranial abnormalities on the first CT scan (Czeiter et al 2020). All biomarkers scaled with clinical severity and stratum, and with presence of CT abnormalities. GFAP achieved the highest discrimination for predicting CT abnormalities (AUC 0.89 [95%CI: 0.87-0.90]) with a 99% likelihood of better discriminating CT-positive patients than clinical characteristics used in contemporary decision rules. Results were consistent across strata, and injury severity. In patients with mild TBI, GFAP also showed incremental diagnostic value: discrimination increased from 0.84 [95%CI: 0.83-0.86] to 0.89 [95%CI: 0.87-0.90] when GFAP was included. GFAP performed better than S100B, a marker currently included in Scandinavian Guidelines for triaging patients with mild TBI for CT scanning. Combinations of biomarkers did not improve discrimination compared to GFAP alone. We further explored the added value of biomarkers compared to four clinical decision rules ((CCHR=Canadian CT Head Rule; CHIP=CT in Head Injury Patients; NICE= National Institute for Health and Care Excellence; NOC= New Orleans Criteria) in 1889 patients with mTBI. We found that GFAP not only provided added value, but also outperformed all clinical decision rules in diagnostic accuracy. GFAP alone had a higher discriminative ability (AUC) for detecting intracranial abnormalities than the components of the rules, with a relatively small increase when the components of the rule were added to GFAP. Our results support the development of novel CT decision rules, combining serum GFAP with clinical characteristics, for triaging patients with mild TBI for CT scanning. Further validation studies are required to determine if GFAP may even replace existing CDRs.
In a collaboration with Orebro University, we undertook a comprehensive metabolomics study, and the first substantive lipidomic analysis, in a cohort of 716 TBI patients and 229 non-TBI controls (orthopaedic, internal medicine, and neurological patients). We identified metabolites specifically associated with TBI severity and outcomes. Choline phospholipids (lysophosphatidylcholines, ether phosphatidylcholines and sphingomyelins) were inversely associated with TBI severity and were among the strongest predictors of outcome. These data show that metabolite-based signatures hold promise for improving current clinical or protein-based outcome prediction models. The observed metabolic patterns likely reflect different pathophysiological mechanisms, including protective changes of systemic lipid metabolism aiming to maintain lipid homeostasis in the brain.

Coagulopathy
One in five patients (19.6%) with isolated TBI displayed laboratory signs of coagulopathy based upon conventional coagulation parameters upon emergency department arrival (Böhm et al 2020). Patients on pre-injury anticoagulant and/or antiplatelet therapy had a two-fold exacerbated coagulation profile compared to those without. In patients without pre-injury anticoagulant and/or antiplatelet therapy, conventional coagulation parameters deteriorated with increasing TBI severity. Patients with isolated TBI that were on pre-injury anticoagulant and/or antiplatelet (APAC) therapy had a three-fold higher mortality and a higher frequency of unfavourable outcome at six months (GOSE 1-4) compared to those without (51.9% vs 23.5%). The higher mortality in patients on pre-injury APAC use was confirmed in the CENTER China registry and found to be determined by use of anticoagulants. On more detailed analysis of sequential CT scans (Mathieu et al 2020), we found that lesion progression was significantly higher in the APAC group for extra-axial (3.1 vs. 1.3 mL, p = 0.01) but not intraparenchymal (3.8 vs. 4.6 mL, p = 0.65) intraventricular (0.2 vs. 0.0 mL, p = 0.79) or total intracranial haemorrhage (ICH; 7.0 vs. 6.0 mL, p = 0.08). Extended coagulation profiling in patients with isolated TBI and raised INR displayed deterioration within the thrombin regulating process with increased fibrinolysis and dysregulation of fibrinolysis regulating mechanisms. Increasing endothelial dysfunction (VE-Cadherin) and damage (Syndecan-1, EDMP) indicated presence of endothelial damage in these patients.

Towards a new multidimensional disease classification for TBI
We performed a hierarchical clustering analysis and identified three main characteristics by which TBI can be described: GCS, trauma mechanism, and major extracranial injury (Gravesteijn et al 2019).

Prediction
Many prognostic models exist for moderate and severe TBI (IMPACT1 and CRASH2 models), but not for mTBI. None of the existing prognostic models for early prediction of GOSE in mTBI have had both good calibration and discrimination. We externally validated the IMPACT and CRASH models in relevant subcohorts of CENTER-TBI (Dijkland et al 2020). For the IMPACT models, discrimination was good, with AUCs ranging between 0.77-0.85 in 1173 patients and between 0.80-0.88 in the broader CRASH selection (n=1742). For the CRASH models, AUCs ranged between 0.82-0.88 in 1742 patients and between 0.66-0.80 in the stricter IMPACT selection (n=1173). Calibration was only moderate with a lower-than-expected mortality. This indicates a need to update the models. We explored the added value of biomarkers obtained within 24 hours of injury and found that all biomarkers provided incremental value with the greatest effect for the combination of UCHL1 with GFAP (increase in AUC from 0.876 to 0.909 for mortality and from 0.850 to 0.876 for unfavourable outcome. We examined the added value of common machine learning (ML) algorithms over logistic regression, LASSO and ridge regression for prediction of outcome after moderate and severe TBI (Gravesteijn, Nieboer et al. 2020). These advanced data analytical techniques did not succeed in better characterization of the complex prognostic patterns in TBI. We further explored prediction of other outcomes than the GOSE in 2666 adult patients who had completed the HRQoL questionnaires at six months after injury. We found that medical and injury related characteristics were of greatest importance for the prediction of PCS, whereas patient related characteristics were more important for MCS and the QOLIBRI following TBI. However, the proportion of variance explained (R2) was relatively low (19% for the physical component score of SF12, 9% for mental component score and 13% for the QOLIBRI). This could be improved substantially in mTBI by including HRQoL assessments at 2 to 3 weeks after injury. This increased the R2 to 37% for the PCS, to 36%, for the MCS and to 48% for Qolibri). For patients with mTBI, we developed a new model for early prediction of ordinal GOSE (1-8) based on readily available admission characteristics. The core clinical model included age, sex, psychiatric history, preinjury health, Glasgow Coma Score, and Injury Severity Score. The model had an AUC of 0.70 after correcting for optimism.

1.3.6 Patient profiles which predict efficacy of specific interventions (WP 13, 14)
In this section we consider patient profiles by care pathway, gender effects, disparities in care, efficacy of specific interventions and complications with a focus on comparative effectiveness research (CER) in the ICU population. CER provides a promising framework to identify best practices and improve outcome after TBI. CER is the generation and synthesis of evidence that compares the benefits and harms of alternative methods to prevent, diagnose, treat, and monitor a clinical condition or to improve the delivery of care. A basic concept of CER is to study differences in care and outcome in observational studies, thus turning natural variability into an asset. Natural links exist between CER and individualized approaches, as CER aims to identify the best treatment for an individual patient, with a specific type of injury, severity, comorbidities and other aspects that determine optimal treatment.

Patient profiles by care pathway
A unique feature of CENTER-TBI was the differentiation of patients by stratum according to care pathway (ER, Admission and ICU). Overall, 28% of patients in the Core study were over 65 years of age, but this was significantly higher in the Admission stratum (32%) compared to the ER and ICU strata (25 and 26% respectively). This high percentage of older patients with TBI has direct implications for clinical care and research. Most clinical trials to date have excluded patients over 65, and as a consequence evidence to underpin treatment for TBI in older patients is lacking. A clear need exists for research in older patients with TBI. Patients in the ICU stratum were more often injured in traffic accidents (45%) compared to the ER and ICU strata (32 and 33%), whilst falls were the most common cause of injury in the ER (51%) and Adm strata (51%). We explored changing care pathways and between centre practice variations in a total of 2138 patients admitted to the ICU in Europe, and found that 36% of patients were classified as having a mild TBI (Glasgow Coma Scale; GCS 13–15). Some of these admissions are motivated by the presence of serious extracranial injuries, significant comorbidities (especially in older patients), by secondary deterioration, or by a substantial risk of deterioration due to possible progression of traumatic intracranial lesions. However, it appears that some ICU admissions in some centres may be driven by lack of resources on other wards, or by local clinical culture. Addressing these drivers of inappropriate ICU admission could result in more efficient use of ICU resources. We noted substantial between-centre variations in the use of intracranial pressure (ICP) monitoring and aggressive treatments (Huijben et al 2020), with Median Odds Ratios (MOR) of 2.5 – 2.9 respectively, but these did not translate into differences in 6-month outcome (MOR: 1.2). This variation in outcome between centres was much lower than observed in previous studies, and posed challenges to CER analyses of specific interventions. However, this observation provides important evidence that treatment standards have improved over time - consistent with the lower than expected mortality described in section 0. It further implies that high quality intensive care is likely more important than specific treatment approaches.

Gender, age, and comorbidities – interacting effects
Our data disprove the traditional perception that TBI is a disease of young, otherwise healthy adult males. Our patients were older and commonly had comorbidities (11% with severe systemic disease and a further 32% with mild systemic disease). While male preponderance persisted (67% overall; 73% in the ICU stratum), this was less than in previous studies, and was lost in patients over the age of 75 years. Debate exists if outcome may differ by gender. We explored sex and gender differences in care pathways, treatment characteristics and functional, health-related quality of life and mental health outcomes after mild, moderate and severe TBI (Mikolić et al 2021). We included 2862 adults (36% women) with mild TBI and 1333 adults (26% women) with moderate/severe TBI. We found no substantial differences between men and women in treatment characteristics and care pathways, but women with mild TBI had poorer 6-month outcomes across different domains of functioning. Following mTBI, Women under age 45 and above age 65 years showed worse 6-month outcomes compared to men of the same age. We used natural effects models to decompose the total effect of sex/gender on outcomes into indirect effects that passed through the specified mediators (socio-demographic variables and injury-related characteristics) and residual direct effects. We found that outcome differences were not clearly mediated by sociodemographic variables or by care pathways. We conclude that other features underlie observed sex differences in outcomes after mTBI.

Disparities and deficiencies in care
An earlier section detailed the disenfranchisement of patients over 65 years of age in past clinical trials, and our data show significant disparities of care in older people: In the CENTER-TBI Registry we found that 40% of patients with TBI are injured by low energy falls and these mostly occur in older patients. These patients have similar rates of CT brain scan abnormalities and in-hospital mortality as those injured by other mechanisms, but are 50% less likely to receive critical care or emergency interventions. This indicates that high energy transfer should no longer dominate injury scene and emergency department TBI triage of injured older people. Our provider profiling survey on end-of-life practices further showed that age was reported to influence the decision-making process in 81% of the centres (van Veen et al 2020). However, in the Baltic States and Eastern Europe, age did not play a role in 60%, and 50% of the centres respectively. Although strong evidence exists that outcome is poorer with increasing age, we caution against the dangers of self-fulfilling prophecies.
We found further evidence for deficiencies in care with regard to discharge policy from the ER and rehabilitation needs. Our provider profiling data showed that 90% of centres do not routinely schedule a follow-up appointment for TBI patients discharged home from the ED, and around 50% on discharge from the ward (Foks et al 2017). In the Core data we found that only 26% of patients discharged from the ER received written information and 6% a follow-up appointment in hospital. Yet, we find that 30% of patients discharged from the ER do not attain a full recovery by 6 months. This indicates a need for more stringent follow-up and support after mild TBI.
Clinical outcome at 6 months was classified as moderate to severe disability in1206 of the individuals recruited to CENTER-TBI. Of these, 90% reported rehabilitation needs (Andelic et al 2021), but only 30% received in-patient rehabilitation and 15% out-patient rehabilitation. Physiotherapy was the most frequently provided modality, but cognitive therapy and psychological counselling were provided in only approximately one third of patients reporting impairments in these domains and who may have benefited. These results were confirmed on analysis across the entire CENTER-TBI cohort, showing that in the year following TBI, only 31.4% of patients received rehabilitation services. Significant negative predictors for receipt of rehabilitation were preinjury unemployment (OR = 0.80) living in Central or Eastern Europe (OR = 0.42) admission to hospital ward (OR = 0.47; reference: admission to intensive care unit), or direct discharge from emergency room (OR = 0.24). We conclude that rehabilitation referral is not only driven by clinical needs, but also by demographic and organizational factors, raising issues related to inequality in access to appropriate rehabilitation care throughout Europe. Based on these findings, there is an urgent need to implement national and international guidelines and strategies for access to rehabilitation after TBI.

Efficacy of specific interventions
Airway management: Intubation and tracheostomy
We explored the use and benefits of intubation and of performing an early tracheostomy in ventilated patients (Gravesteijn et al 2020; Robba et al 2020). Intubation was performed in 890/3736 (24%) patients at the accident scene and in a further 460/2930 (16%) on arrival to hospital in the ED. Substantial variation in intubation practices existed between countries. Overall, prehospital intubation had no adjusted overall effect on functional outcome (OR:1.01,Intubation practice variation between countries 95%CI:0.79–1.28 p=0.96) but prehospital intubation was associated with better functional outcome in patients with higher AIS scores in thorax and abdominal (p=0.009 and p=0.02 respectively).
In-hospital intubation showed a non-significant beneficial effect on outcome (OR:0.86 95%CI:0.65–1.13 p=0.28) but on subgroup analysis of patients with GCS scores of 10 or lower the effect was significant (p=0.01). These results suggest that major extracranial injury should drive the decision to intubate in prehospital setting, and that indications for intubation in-hospital should be broadened to include also patients with a GCS of 9 or 10, rather than <8 as is commonly advised. We explored the benefit of early tracheostomy in patients with an ICU stay >72h and found considerable heterogeneity between countries in tracheostomy frequency (7.9–50.2%) and timing (early: 0-17.6%). Tracheostomy in the first week was associated with a better neurological outcome and reduced length of stay in hospital and ICU. We need definitive trials to assess the benefit of early tracheostomy suggested by these findings.

Intracranial pressure monitoring
A total of 921 of the 2138 patients admitted to the ICU (43%) received an ICP monitor. However, 370 out of the 961 severe TBI patients (38.5%) who met authoritative Guideline criteria for ICP monitoring (eg GCS <=8) did not receive an ICP-monitor. The most common reason reported was absence of radiological signs of a raised intracranial pressure. Although guideline adherence for ICP monitoring in patients with severe TBI was suboptimal (61.5%), we found that ICP monitoring had been conducted in 148/328 patients (45%) with moderate TBI and in 12 % of patients with mild TBI. Of all patients monitored, 14.2% were initially classified as mild TBI and 17.2% as moderate TBI. We found substantial between-center variation in ICP monitoring use (MOR: 2.5) duration of ICP monitoring (MOR: 3.2) and guideline adherence (MOR:2.5) (Huijben et al 2020). We conclude that clinical practice of ICP monitoring use in Europe is highly variable, and deviates from international Guidelines.
Physiological ICU data with high temporal resolution were obtained from 277 patients and provided important insights on the outcome impact of intracranial hypertension (Åkerlund et al 2020). While BTF Guidelines identify a single ICP threshold of 22 mmHg for treatment, we found that outcome was related to both the intensity and duration of ICP insults, and an ICP threshold threshold of 18 ± 4 mmHg, if maintained for long periods, was associated with poorer outcomes. This clear relationship between pressure/time dose (PTD) and outcome (mortality and unfavourable outcome were modulated by cerebrovascular autoregulatory status, with greater vulnerability to ICP insults when cerebral perfusion pressure was below the autoregulation lower limit of. There is a strong case for exploring individualised ICP thresholds for treatment, quantifying the detrimental pressure-time dose of ICP insults, and integrating autoregulatory status into patient management protocols.

Fluid management
A study of 2125 patients admitted to ICUs in Europe and Australia (Wiegers et al 2021) showed wide between-centre variations in mean daily fluid input (1·5 to 4·2L) and fluid balance (-0.9L to +1.5L). More positive daily mean fluid balance and fluid input were associated with higher mortality and worse functional outcome in both centre-level instrumental variable analysis and patient-level analyses. Each 0.1L increase in daily positive fluid balance was associated with an increased odds ratio (OR) for ICU mortality (range of ORs 1.03 - 1.1) and poor functional outcome (range of OR: 1.04 – 1.09). Higher daily fluid input was also associated with worse outcomes in patient level, but not centre level analyses. We conclude that maintaining neutral fluid balance is not general practice in European and Australian critical care units, and that Inappropriately high fluid input and positive fluid balance may be iatrogenic (and hence correctable) causes of poor outcome for critically ill TBI patients; addressing this could improve outcomes.

Deep Venous Thrombosis (DVT) prophylaxis
We analysed data on 2006 adult patients admitted to the ICU and found substantial variation in use of pharmacologic DVT prophylaxis between centres with an MOR of 2.7. A moderate association with better outcome was found at the centre-level (OR: 1.2 [0.7-2.1]) and on propensity matched analysis at the patient-level (OR: 1.4 [1.1-1.7]) which could not be explained by a reduction in number of thrombo-embolic events. These findings suggest that DVT prophylaxis may be associated with improved 6-month functional outcome and lower mortality rates, without CT progression.

Surgery for acute subdural haematoma (ASDH)
In an analysis of 1407 patients with ASDH, acute surgery was performed in 336 (24%) patients at a median of 3.8 hours (IQR 2.5-6.5) of whom 245 (73%) received a craniotomy and 91 (27%) additionally a (primary) decompressive craniectomy. The proportion acute surgery varied from 7 to 52% (IQR: 13-35%) between centers. Center preference for an acute surgical strategy compared to that of initial conservative treatment was not significantly associated with better outcome (odds ratio 0.92 [95% CI 0.77 to 1.08]). We conclude that an aggressive approach to acute surgical evacuation in patients for whom equipoise existed on surgical indication may not lead to a better outcome compared to a strategy favouring initial conservative treatment
Practice variation in the use of Decompressive craniectomy (DC) and its alignment with evidence
We analysed the harmonised data of the CENTER-TBI and OzENTER-TBI Core studies, which include patients admitted to participating ICUs in Europe the UK and Australia. Patients were compared between different regions and by injury characteristics. Of 2336 people admitted to ICUs following TBI, DC was performed in 320 (13·7%): in 4·5% (64/1422) of patients with diffuse TBI, and 30·5% (195/640) of patients with traumatic mass lesions. Substantial variation was found between regions. Most patients who underwent DC (258/320; 81%) had a primary decompressive craniectomy at the time of evacuation of a mass lesion, and would not have been eligible for either the DECRA or RescueICP trials – the 2 pivotal trials on DC. While current trials (such as the RESCUE-ASDH trial is addressing part of this question, there is currently no evidence to underpin this practice. Consistent with evidence, secondary DC was used infrequently in patients in whom it has been shown to be potentially harmful. However, our data show a large gap between current practice and available evidence.
Complications
We found substantial between-centre variation in the occurrence of complications, identified as potential quality indicators (see section 1.3.7): Median odds ratios ranged from 1.5 to 4.1 (Hyperglycaemia: 1.5; Hypoglycaemia: 2.4; decubitus ulcers: 2.5; ventilator-associated pneumonia (VAP): 4.1). We explored VAP, one of the commonest ICU complications, in more detail and found that 1/5 of ventilated patients (196/962) developed VAP at a median interval of 5 days after intubation (Robba et al 2020). Patients who developed VAP were younger, had a higher incidence of alcohol and drug abuse and more episodes of respiratory failure. Therapies as histamine-receptor antagonist intake could increase the risk of VAP, while antibiotic prophylaxis was associated with reduced risk. VAP was not associated with increased mortality or worse neurological outcome but did increase length of ICU stay. The large between-centre variation in some complication rates indicates substantial room for improving in the quality of care.

1.3.7 Develop performance indicators for quality assurance and improvement in TBI care (WP 13).
Quality measurement using appropriate indicators can guide quality improvement, for example, through identifying best practices and internal quality improvement initiatives. The potential of quality indicators to improve care has been demonstrated in other clinical areas, in sepsis, in stroke, and in children with TBI. However, quality indicators for general use in TBI are lacking. We explored two approaches to the development of performance indicators for TBI: First, we aimed to develop a set of quality indicators, and second, we explored benchmarking quality of care by comparing observed to predicted outcomes.
We developed a set of quality indicators that have the potential to improve quality of TBI care at European ICU’s. This set was developed in an extensive Delphi process, and consisted of 17 structure indicators, 16 process indicators and 9 outcome indicators (Huijben et al, 2019). The indicators were subsequently validated on data from 2006 adult patients enrolled to the ICU stratum of CENTER-TBI in 54 ICU’s (Huijben et al 2020). A total of 26 out of the initial 42 indicators could be validated in the CENTER data. The other 16 indicators related to organisational aspects, which could not be evaluated on the patient data of the CENTER core study. Significant between centre variation was found for 7 process and 5 outcome indicators with median odds ratios ranging from 1.51 to 4.14. Statistical uncertainty of outcome indicators was high, mainly due to low event rates. Validity of the indicators was rated according to pre-defined thresholds for feasibility and discriminability. Overall, nine structure and five process indicators showed potential for quality improvement purposes for TBI patients in the ICU. This does, however, not mean that the other indicators may not have value in other settings. Most indicators rated as low potential failed to meet the criteria for discriminability because of high consistence of centres in meeting the indicator standards. This may well be different in settings with fewer resources. The developed indicator set represents an important tool to support benchmarking and quality improvement programs for patients with TBI in the future.
Benchmarking quality of care can be based on comparisons between observed and expected outcomes, adjusting for differences in case-mix. Calculating expected risk requires the availability of robust prognostic models. Two robust and extensively validated models for predicting outcome are the CRASH and Impact models. The IMPACT models were validated on 1173 patients with a baseline GCS of 3-12 in the CENTER data and the CRASH model on a broader cohort of patients with a GCS<=14, consistent with the development populations (Dijkland et al 2020). For IMPACT, model discrimination was good, with AUCs ranging between 0.77-0.85 in 1173 patients and between 0.80-0.88 in the broader CRASH selection (n=1742). For CRASH, AUCs ranged between 0.82-0.88 in 1742 patients and between 0.66-0.80 in the stricter IMPACT selection (n=1173). Calibration of both models was moderate. These data support the use of these models for benchmarking quality of care. We further explored observed versus predicted outcome in the CENTER data. In ICU patients with moderate to severe TBI, the rate of unfavourable outcome (GOSE<5) was 55%, similar to that predicted by the IMPACT prognostic model (O/E ratio 1·06 [95% CI 0·97-1·14]), but mortality was lower than expected (O/E ratio 0·70 [95% CI 0·62-0·76]). These findings were recently replicated on analysis of 441 adult patients with moderate or severe TBI, enrolled to our sister study in the US, TRACK-TBI, and show that over time, mortality has decreased but that this comes at a cost of more survivors with severe disability. We note that the IMPACT models were developed on older data, some of which date back to the 1980’s. The lower mortality in current practice signals a need for continuous updating of prognostic models. Nevertheless, the potential of prognostic models for benchmarking quality of care has been demonstrated.

1.3.8 To validate the common data elements (CDEs) for broader use in international settings, and to develop a user-friendly web based data entry instrument and case report form builder (WP 20, 22).
Standardization and harmonization of data collection in studies on traumatic brain injury (TBI) is of paramount importance to clinical research on TBI, that increasingly involves large scale studies, multicentre international collaborations and data sharing. This requires a “common language” for data collection, in terms of what variables to record and how to code them. The development of uniform data standards – termed “common data elements (CDEs)” – was initiated by the International Mission for Prognosis and Analysis of Clinical Trials in TBI (IMPACT) study group and taken forward by an international group of 149 institutes and agencies supported, among others, by the U.S. National Institute of Neurological Disorders and Stroke (NINDS), U.S. Department of Defense, and U.S. Department of Education and the US Department of Veteran’s Affairs. This consensus effort lead to Version 1 of the TBI CDEs (TBI-CDE v1), published in 2010. In 2012, a re-structuring was introduced with the overarching aim of creating a set of “Core” CDE elements suitable for use in all TBI studies. “Basic” elements (required for domain-specific studies) were defined according to the following domains: “Concussion/Mild TBI”, Acute Hospitalized (AH)”, “Moderate/Severe TBI: Rehabilitation (Rehab) and “Epidemiology.” A larger set of “Supplemental” elements was created to allow flexibility in adapting to unique study criteria and endpoints. This second version, TBI-CDE v2, is hosted and maintained by the National Institute of Neurological Disease and Stroke (NINDS, https://www.commondataelements.ninds.nih.gov/CDE.aspx). The TBI-CDEs have undergone several updates based on input from expert working groups, researchers and funding agencies. Despite international input, the TBI-CDE’s have remained US-centric. We consider the TBI-CDEs of such importance to the field, that they should become global standards. CENTER-TBI provided a prime opportunity to explore the validity of the CDEs in the context of achieving global applicability to support data sharing and international collaboration. In addition, we aimed to quantify the degree of harmonization between three large InTBIR studies: CENTER-TBI, TRACK-TBI and ADAPT. We found an agreement of 81% for CENTER elements with the TBI Core CDEs and 91% for Basic CDEs in the AH domain (Meeuws et al 2020). Non-harmonization was largely caused by absence of the elements in the studies. For elements present, the compatibility of coding with TBI CDEs was 90-99%. The degree of harmonization across the three IntBIR studies ranged from 75% to 87% for the AH domain and for the Rehab domain from 64% to 82%. For each domain the degree of harmonization was greatest between CENTER-TBI and TRACK-TBI. To our knowledge, this was the “first in its kind” study to systematically evaluate the implementation and harmonization of TBI-CDEs across large scale studies. The high degree of harmonization of study variables among these studies demonstrates the importance and utility of common data elements in TBI research. This was further re-enforced by the use of these CDEs in or linked studies in Australia (OzENTER), China (China CENTER registry) and India (CINTER-TBI, indicating global interest in their use. A critical appraisal of the TBI-CDEs, however, showed that their presentation on the NIH-NINDS website is not very user-friendly, and identified some major issues concerning global applicability: Some Core elements violate GDPR as they contain potential patient identifiers, e.g. date of birth is a required Core element. Two Core elements (Race and ethnicity) and 2 basic elements (educational level patient and caregiver) are US centric and not applicable to global use. Further, many of the outcome instruments, also those listed as Core elements, are not available outside the English language, or are copyrighted, hence limiting their broad use in international settings. Other issues include duplicates between Core and basic CDEs, overlap between the AH and rehab domains, listing of one variable as multiple elements and discrepancies in classification between domains. These issues have been communicated by e-mail to Dr Mendoza-Puccini, the NINDS lead on the CDE project, and were subsequently presented to the CDE steering committee in March 2020. We understand that work on an update of the TBI-CDEs will be initiated in the fall of 2021, and hope that this may be informed by our critical appraisal and by the empirical experience of the InTBIR studies, including CENTER-TBI. We conclude that, in their current form, the TBI-CDEs do not meet qualifications as global standards. The standardisation of data collection according to the CDEs for TBI is highly relevant to the field and we suggest that all efforts are made to upgrade these to global standards, thus facilitating meta-analyses across data collected in different parts of the world. Whilst we can provide input at an individual level, we strongly suggest formal input at an institutional level, either directly from the European Commission or through the
InTBIR collaboration.
Harmonization of data in preparation of meta-analyses, however, goes beyond coding issues. Initial collaborative efforts at meta-analyses between CENTER- and TRACK-TBI have made us recognize that harmonization also needs to address interpretation. For example, we learned that the approach to outcome assignment according to the GOSE was performed differently in the US compared to Europe: In the US, only TBI-related disabilities are considered, whilst in Europe the focus is more on “all-cause” disability (see also 1.3.4). Prior to our work, this substantial discordance had not been recognized, and may well explain why some studies in the US report better outcome compared to Europe. Issues have also been identified with regard to interpretation of coding for pre-injury morbidities and pre-injury alcohol use. “Deep harmonization” is required, and substantial efforts will be required in order to perform robust and meaningful meta-analyses. In Section 1.3.9 we describe the analytic platforms developed and implemented in CENTER-TBI. For meta-analyses across studies, we suggest a “Federated” approach. Data federation strikes a balance that protects patient privacy and supports clinical discoveries to improve patient outcomes. A federated approach will enable harmonization and analysis of virtually co-located studies, while respecting the need for patient data to remain safely behind institutional firewalls. In a cloud-based data federation pipeline, layers of data processing (a.k.a. data virtualization) are introduced between the data user and the primary source data, preventing its direct access. We have explored federated analyses between CENTER-TBI and CREACTIVE in collaboration with the Human Brain Project – MIP to validate prognostic models for TBI. From a technical perspective, these explorations were highly successful, but perhaps the most important lesson learned was the absolute need for “Deep Harmonisation”. Open-source tools and analysis pipelines were designed to incorporate legacy and future TBI datasets, and may be applied to other neurological conditions beyond TBI.

1.3.9 Development of an open source database compatible with FITBIR (WP20).
We aimed to develop, implement and maintain an open standards-based platform for the collection and storage of clinical data and neuroimaging and biomarker results based on Common Data Elements (CDEs). The goal was to develop a next generation open standards-based platform that would support advanced large-scale collaborative analytics and model building, also providing a model for future clinical studies on brain diseases and disorders. The informatics system aimed to (1) provide an open standards based platform where clinical (WP1) and repository (WPs2-6) data would be integrated in a data warehouse, and (2) facilitate collaborative analytics by providing standard interfaces to the platform through which various analytics tools can access the data. The definition of data in the warehouse was designed to be compatible with the FITBIR registry (https://fitbir.nih.gov/). An electronic data collection tool was developed and maintained in partnership with QuesGen Systems Inc. with additional funding from One Mind. All data were de-identified and coded by a Global Unique Patient Identifier (GUPI). Dates and times were rendered untraceable by re-coding all entries relative to UNIX epoch time. Potential identifiers in free text entries were deleted by manual screening and employing an AI algorithm. MR images were defaced upon upload. The CENTER-TBI clinical dataset is extremely complex, including a combination of over 2,400 distinct, discrete and longitudinal measurement concepts with the latter involving both regularly and irregularly sampled timepoints. A bespoke data access tool, Neurobot (details available on the SciCrunch Resource Identification Portal, using the Research Resource Identifier RRID:SCR_017004), was developed, maintained and updated by KI-INCF (Stockholm SE). This was no minor task given the large number of variables, the different types of variables and the inclusion of continuous and longitudinal data. For reference, we initially intended to implement TranSmart3 (originally developed by Johnson & Johnson with the Recombinant Data Corporation) as our data access platform, but that system failed to cope with the complexity of the CENTER-TBI data. Neurobot provides an easy to use web based front end, which allow investigators “shopping cart-like” access to free text searchable data elements. Within Neurobot, a convenient link is provided to the Data Dictionary, as well as to a description of the e-CRF design, aiming to assist the researchers in selecting data for any particular analysis. The data accessed through Neurobot were linked to a separate repository for high temporal resolution ICU data, stored in customized HDF5 format.
High-quality data are critical to the entire scientific enterprise, yet the complexity and effort involved in data curation are vastly under-appreciated. This is especially true for large observational, clinical studies because of the amount of multimodal data that is captured and the opportunity for addressing numerous research questions through analysis, either alone or in combination with other data sets. Substantial efforts were implemented to ensure high quality of the CENTER data: First, automated data checks were built into the e-CRF system to alert investigators to impossible or improbable values and to detect inconsistent data entries, providing immediate feedback to investigators. Second, source data verification (SDV) of major characteristics was performed by ICON (Paris, France). SDV was performed in 100% of cases for informed consent and in 28% of patients for major characteristics on a total of 13448 data points. Third, a team of three dedicated personnel was employed full-time to check completeness and accuracy of entered data. Fourth, a Data Curation Task force (DCTF) was formed to perform data curation at a higher level. The DCTF team examined data, not only for missingness and plausibility, but also for multivariate consistency by crosschecking variables with other related concepts in the database. Derived variables were introduced and where data quality problems were identified, these were investigated to identify if these were structural issues (e.g. variances in datatype that was unanticipated at design time), site specific issues (e.g. unanticipated variances in data element interpretation due to local or language related misinterpretation) or simply isolated random errors. Data quality problems were addressed in three broad ways: first, for a small number of systematic data entry inconsistencies, it was possible to transform data or unify concepts across time points and documented plans were created for this. Secondly, where systematic issues were identified, there was a robust process involving a dedicated team to go back to sites to understand and identify problems and implement solutions (including for example process validation at source or ongoing training/needs analysis). Where common, but unsystematic errors were identified, e-CRF rules were updated and the subjects reflagged as being incomplete so that sites could go back and make corrections.
All these efforts have resulted in a high-quality database with highly granular data – the CENTER-TBI database consists of 2829 data elements, 8 versions of datasets with overall file storage exceeding 2.65 TB. We developed a detailed Data Dictionary with frequency tables to help guide researchers to navigate and understand the complex CENTER-TBI data (https://www.center-tbi.eu/data/dictionary).
The amount of work involved in the data curation process was much larger than anticipated and prompted further reflection and action. We recognized that lack of details concerning data curation methods can result in unresolved questions about the robustness of the data, its utility for addressing specific research questions or hypotheses and how to interpret the results. In collaboration with InTBIR partners, we developed the Data Acquisition, Quality and Curation for Observational Research Designs (DAQCORD) Guidelines (Ercole et al 2020). These are the first comprehensive set of data quality indicators for large observational studies. They were developed around the needs of neuroscience projects, but we believe they are relevant and generalisable, in whole or in part, to other fields of health research, and also to smaller observational studies and preclinical research.
The Neurobot system was used in most of the CENTER-TBI analyses and proved its value and robustness. The system is adaptable to other disease states. Although CENTER-TBI formally ended on March 31, 2021, continued availability of the Neurobot tool and access to the data are guaranteed for at least another year. We have additionally uploaded the CENTER-TBI data onto an OPAL platform, hosted by LUMC in Leiden (NL). Opal offers a secure and multi-project web-based application where data sources are transformed into target models based on flexible defined views and provides mapping to ontologies, thus complying to the FAIR principles. Access to the data is facilitated via a research dashboard and authentication server. This platform potentially provides all features for research including data processing, harmonization and mapping of data to common data models as well as (federated) analysis across datasets. Maintenance and access to this platform is guaranteed for at least 3 years.
These platforms offer opportunities for external researchers to access and use the unique data of CENTER-TBI and its repositories in the years to come. CENTER-TBI is open to data-sharing and welcomes proposals from other researchers, thus optimizing the use of public funding that supported CENTER-TBI and advancing the care for patients with TBI. We wish to ensure “good use” of the data, and have implemented a study- and publication proposal platform (https://www.center-tbi.eu/data). Proposals are reviewed by the Management Committee for scientific rigor and feasibility (not all research questions can be answered from the data). Following approval of the proposal and signing of a data use agreement, access can be granted. To date, over 375 proposals have been submitted by internal and external researchers. We have found this platform to serve an additional important purpose in promoting collaborations and limiting the risk of redundancy of efforts.

1.3.10 Networking activities and international collaborations in TBI (WP 16, 22).
TBI is a global problem that requires a global approach. We aimed to increase the scientific impact of CENTER-TBI by global collaborations, which offer opportunities for increasing patient numbers in joint analyses, strengthening research approaches, and provide a platform for involving the best scientists across the world. Specifically, we sought to foster collaboration with InTBIR partners, and to strengthen existing and initiate new collaborations on clinical TBI research worldwide. In addition, we welcomed proposals from external researchers to address research questions within the CENTER-TBI data and repositories, and established collaborations with industry. We present the multiple collaborations that CENTER-TBI developed outside of its core Consortium below:
Collaboration at the InTBIR level
The annual meetings of the International Initiative on TBI Research (InTBIR - http://intbir.nih.gov/) were attended by Lead investigators of CENTER-TBI, and our contributions based on ongoing work and emerging results from CENTER-TBI were extremely well received. Various investigators were active in the InTBIR working groups, and this led to a number of collaborative publications3, 4, 5. The platform provided by InTBIR is unique in bringing funders and researchers together on a regular basis and proved highly effective in stimulating networking and collaborations. From its inception in 2013, it was foreseen that the driving force provided by funding bodies would come to an end around 2020, when most major projects would be completed. InTBIR is currently transitioning towards a more investigator-driven initiative, and CENTER-TBI greatly welcomes the opportunity to take on a Lead role in the new organisation. Within the InTBIR group, CENTER-TBI developed close collaborations with TRACK-TBI (US), CREACTIVE (Europe) and developed the GAIN initiative on genetic analyses. Multiple interactions with TRACK-TBI occurred, led by the PI’s Prof Andrew Maas/Prof David Menon (CENTER-TBI) and Prof Geoff Manley (TRACK-TBI). These interactions have led to the initiation of meta-analyses across the two studies, and plans to formalize these towards the future. A prime example of the strength of such collaborations is the validation performed by CENTER-TBI on a clustering analysis of CT phenotypes and their association to adverse outcome in mild TBI developed by TRACK-TBI. Results obtained in the two studies were virtually identical (Yuh et al 2021). Collaboration with CREACTIVE focussed on utilizing the Human Brain Project Medical Informatics Platform (HBP-MIP: https://mip.humanbrainproject.eu/) for a validation study on the Core IMPACT prognostic model across CENTER-TBI and CREACTIVE. Preliminary results demonstrated the potential of a tool like MIP to federate large databases in the field of TBI. Discrimination of the model across datasets was excellent, but substantial differences were noted in calibration. These could be explained by differences in case-mix between the datasets. However, we also noted different approaches to scoring of some variables, highlighting the need for “deep harmonization” (see also section 1.3.8). The GAIN consortium (Genetic Associations in Neurotrauma) was formed to create a sufficiently large, high-quality, harmonized dataset of genetic, biomarker, imaging, and phenotypic data from victims of TBI with the aim to explore the influence of genetic variation on clinical outcome in TBI in a larger cohort than would be possible within the individual studies. Data from a provisional total of 5628 patients were included from 6 cohorts, providing the first GWAS/TWAS study in TBI (for details, see section 1.3.5)

Global collaborations
Collaborations with Australia, China and India resulted in linked data collections in these countries, using an identical data format as in CENTER-TBI (see also section 0). In Australia (Melbourne 2 sites), a total of 198 patients were recruited to the ICU stratum of OzENTER, and included in various comparative analyses which have to date resulted in three publications (Wiegers et al 2021 a,b; Gantner et al submitted). The collaboration with Australia was further intensified in the context of the MRFF Traumatic Brain Injury Mission Grant application, entitled “An informatics approach to predict outcomes and monitor intervention efficacy following moderate to severe TBI” which has been submitted by M. Fittzgerald with application number 2008223. Prof David Menon is a full Collaborative Investigator on this project and Prof Andrew Maas an Associate Investigator. Prof David Menon is also a full collaborator on another application on the same call, led by Prof Andrew Udy (PRECISION-TBI – Promoting evidence-based, data driven care for critically ill moderate-to-severe TBI patients). The CENTER China Registry collected data on patients with TBI admitted to hospitals across China in the same period and according to a similar format as the CENTER-TBI Registry. Data of 13138 patients from 52 hospitals in 22 provinces of China were analysed, and have resulted in 2 publications (Feng et al 2020; Gao et al 2020) with a third in preparation. Data collection in India was finalized with a total recruitment of 1017 patients to the Core study and 4903 to the Registry. The data have been transferred to the CENTER-TBI hub and are currently being curated, following which they will be entered into Neurobot. Comparative analysis between the European and Indian data will be performed in collaboration with our Indian partners. In collaboration with The Neurotraumatology Committee of the World Federation of Neurosurgical Societies and the NIHR Global Health Research Group (Professors Peter Hutchinson and David Menon, actions have been initiated towards the development of a global TBI Registry. The ERANET-NEURON Initiative has enabled additional collaboration with European and Canadian partners. Prof Anne Vik (Trondheim) coordinated the TAI-MRI project (A New Traumatic Axonal Injury Classification Scheme based on Clinical and Improved MR Imaging Biomarkers), which includes members of WP 3 (Icometrix/UZA) and WP 8 (Cambridge) as co-applicants; while Prof David Menon (Cambridge, WP3 and WP5) coordinated the ICON-TBI Project (International Collaboration On Neuroinflammation in Traumatic Brain Injury), which involves collaborations between partners in Canada (Calgary), Italy (Milan), and the UK (Glasgow).

Collaborations with academic and industrial partners
Multiple interactions have occurred with both academic and industrial partners. A total of 33 study proposals have been received from external Parties, of which approximately half were accepted. Of these, external collaborations were formalized with in several fields:
On Biomarker analyses:
• ABCDx SA (Geneva, Switzerland): We shared 2083 sample aliquots in this collaboration with the aims of validating point-of-care assay, including GFAP, H-FABP and Il-10 2), and undertaking assays of inflammatory markers in a substantial number of patients.
• University of Örebro (Örebro, Sweden): 2000 residual serum aliquots (50µl) were shared in this collaboration, with the aim of undertaking metabolomic and lipidomic analyses. This was the first study on TBI that included lipidomic analyses, and showed that lipid metabolites in particular were associated with TBI severity and were among the strongest predictors of patient outcomes (for details, see section 1.3.5).
• NanoDx Inc. (formerly BioDirection) (Southborough, MA, USA): We shared 120 leftover serum aliquots (~100µl) in this collaboration to validate point-of-care assays for GFAP, UCHL-1, and S100B.
• University of Edinburgh (Edinburgh, UK): We shared 300 leftover serum aliquots (20µl), with the aim of exploring prognostic/diagnostic value of infrared spectroscopic analysis of serum blood samples.
On Neuro-imaging
• Biogen (Cambridge, MA, USA): The aim of this collaborations was to explore the relationship between contusion characteristics following TBI, namely the volumetric growth of the hemorrhage and edema compartments, with functional outcome measures at 90 and 180 days post-injury.

Embedded trials and associated studies
Embedded trials are defined as studies developed and initiated in collaboration with CENTER-TBI participants, addressing therapeutic interventions in CER trials. Associated studies refer to all other studies where principal investigators or sponsoring companies seek collaboration with CENTER-TBI – a potential win-win situation.
Embedded trials were
• Thromboelastometry in Acute Hemorrhage Induced by Traumatic Injury of the Brain (TAHITI-B). This pilot study provided preliminary evidence of efficacy and possible superiority of thromboelastometric over conventional coagulation tests, and has been published (Gratz et al 2019).
• Randomised Evaluation of Surgery with Craniectomy for patients Undergoing Evacuation of Acute SubDural Haematoma (RESCUE-ASDH). This trial enrolled 836 patients from 52 centres in 13 countries. The database has been locked. Analysis is underway, and results are awaited.
• Prophylactic Hypothermia Trial to Lessen Traumatic Brain Injury (POLAR-RCT). This trial showed that prophylactic hypothermia compared with normothermia did not improve neurologic outcomes at 6 months
• Protective Ventilatory Strategy in Severe Acute Brain Injury (PROLABI). A publication was anticipated in 2020, but has been delayed by the problems with the SARS-CoV-2 pandemic in Italy.
Five additional projects sought linkage CENTER-TBI as associated studies but have not yet reached publication. We remain in contact with the researchers who submitted these proposals, with a request for acknowledgement of CENTER-TBI should they successfully publish details of their analysis. These studies focus on 1) post-acute neurosurgical interventions, 2) Paroxysmal Sympathetic Hyperactivity (PSH) post TBI, 3) sedative management in TBI, 4) EEG predictors of outcome, 5) fluid management in severe TBI. Overall, the number of interactions and formalized collaborations are substantial and have added huge additional value to the CENTER-TBI project.

1.3.11 Dissemination of study results and management recommendations for TBI to health care professionals, policy makers and consumers, aiming to improve health care for TBI at individual and population levels (WP 18, 19).
Dissemination aimed for widespread knowledge and use of research results by the target population e.g. policy makers, health care professionals and patients. Approaches included publications in the scientific literature, presentations and interactions with policymakers, press releases and media communications, social media accounts, and interactions through the CENTER-TBI website. Below, we summarize our main outputs.

Publications in the scientific literature
In terms of scientific output, the CENTER-TBI Consortium has been highly productive with – to date – over 200 publications in peer-reviewed scientific journals, of which 26 were in journals with an impact factor > 10. Over the years of the Project duration, the number of publications and their citations has steadily increased, leading to date to a total IF value of 1111, calculated as the summated Impact factor value of each publication. A complete list of all publications generated by the Consortium and CENTER-TBI affiliates is available on the CENTER-TBI website (https://www.center-tbi.eu/publications/) and in section 2.1 of this Report . All publications based upon the data collected during the CENTER-TBI studies included a list of Group Contributors. We strongly felt that all Participants and Investigator sites should receive academic credits for the work performed in and for CENTER. Without the input of all Investigators who collected data, analyses and scientific output would not have been possible. Whilst large Group Contributor lists are common in many fields of science, the inclusion of our list in the medical domain proved highly challenging, as these lists appeared to be irregularly picked up by PubMed, the prime bibliography for medical domains. It first appeared that this varied by journal and could be related to the positioning of the Group Contributor list in the manuscript, e.g. at the end of the manuscript, in the acknowledgement section or in the supplementary material. It later turned out that PubMed changed their policy for extraction of authorship lists in 2016. Whilst prior that time, PubMed used to take care of this, working from information in the article or appendix listing, they stopped doing so without warning and pushed the work onto the Publishers. Some journals have the luxury of in-house production teams, but many do not. As a consequence, transfer of group Contributor lists to PubMed is often deficient. To alleviate these issues, the Lancet Journals, as example, now request authors to provide the names in a table clearly indicating forename and surname. The responsibility has therefore been shifted from PubMed to Publishers to authors – without any information on these changed procedures having been communicated to publishers or the broader academic community. To complicate matters even further, it turned out that academic bodies in some European countries, e.g. Norway, do not recognize Group Contributorship as meeting standards for obtaining academic credits. Some of the CENTER-TBI Investigators have suggested inclusion of principal investigators of high-enrolling sites in the main authorship listing. We did not consider this appropriate, as we strongly felt that all Investigators who contributed to the CENTER-TBI data should receive academic recognition for their efforts. We conclude that the current system of academic credits in the field of medicine should be critically appraised and a common EU approach implemented.

Interactions with Policymakers
Whilst publications in the scientific literature serve to improve the knowledge of and care provided by health care professionals, perhaps even greater advances can be obtained by implementing improvements at the policy level. CENTER-TBI developed various initiatives targeting policy makers. The publication of our Commissioned Issue on TBI in the Lancet Neurology (see also section 1.3.1) had a direct focus to inform policymakers on the huge burden posed by TBI to society, and summarized gaps in our knowledge. The Commission was presented at the European Parliament on Nov 7, 2017 – an occasion attended by a patient and his mother. Substantial advances in creating awareness for TBI at the policy level in the UK have been realized through the efforts of Prof David Menon in the All-Party Parliamentary Group on Acquired Brain Injury. This input drew heavily on the work undertaken in CENTER-TBI, and in particular the Lancet Neurology Commission on TBI, which was provided to all UK Members of Parliament in advance of discussions. A full report was published online on 18th Oct 2019, which concluded that The Government should bring together a taskforce to address the issues and recommendations as a matter of urgency (https://cdn.ymaws.com/ukabif.org.uk/resource/resmgr/campaigns/appg-abi_report_time-for-cha.pdf).
Media communications and press releases
CENTER-TBI has actively sought media attention by various initiatives. Press releases were broadly distributed around the presentation of the Lancet Neurology Commission at the European Parliament and the occasion summarized in a video (https://www.youtube.com/watch?v=VsUk_Q7qnWg). Forbes magazine featured the findings (https://www.forbes.com/sites/nicolefisher/2017/11/09/special-lancet-neurology-issue-targets-political-forum-to-combat-global-tbi/#59d4fd8675a8). CENTER-TBI attracted media interest across the globe, including Australia, China, Belgium, Germany, Hungary, Italy, the Netherlands, and the UK. EuroNews broadcast a special feature on CENTER-TBI in November 2019 (https://www.euronews.com/2019/02/25/i-was-not-who-i-was-researcher-into-new-care-for-traumatic-brain-injury-victims).

Social media
CENTER-TBI is present in Twitter (@CenterTBI) to inform the lay audience about the importance of TBI prevention and the results of the project. It also provides visibility to publications and scientific events where CENTER-TBI researchers are involved. The number of followers, mentions, and profile visits significantly increased over the course of the study, thus indicating a great interest in TBI both in clinicians who are not directly involved in the project and in lay audiences. During the report period, there have been 1312 tweets by 553 unique tweeters in 46 countries. This interest is widely diffused across the world showing that dissemination of our results exceeds the European boundaries (see Figure 14).
Public information platform
We developed and implemented an interactive public information platform explaining the impact and future developments of TBI research in lay language, within the public section of the CENTER-TBI website (https://www.center-tbi.eu/). This platform, active since month 13 after project start, aims to make the public active partners in research, clinical care, and policy development, provides links to patient organizations such as PatientsLikeMe (www.patientslikeme.com/) and includes a FAQ page. Development of a “knowledge commons” for TBI, integrating CENTER-TBI outputs into systematic reviews (WP18).

1.3.12 Development of a “knowledge commons” for TBI, integrating CENTER-TBI outputs into systematic reviews (WP18).
We established a “Knowledge commons” early on the project, consisting of the Partners involved in WP18. The Knowledge Commons was supported by a knowledge management project officer located at the National Trauma research institute (NTRI) and the Centre of Excellence in Traumatic Brain Injury Research in Melbourne. The aim was to develop a series of high-quality systematic reviews to summarize the evidence base underpinning our knowledge of TBI. The knowledge commons was responsible for overseeing the selection of systematic review topics, finding review authors, refining systematic review questions and instigating author training. We organized two three-day systematic review training courses (March 2014, Budapest and September 2015, Antwerp) for review authors, to which other interested researchers were also welcome.
We conducted a scoping review on trials in moderate and severe TBI, aiming to summarize the existing evidence from clinical trials, synthesize key RCT characteristics and findings, and determine their implications for clinical practice and future research (Bragge et al 2016). We identified 191 completed RCTs, of which only 26 (across 18 different interventions) were considered robust. Less than one-third of RCTs demonstrated low risk of bias. Less than a quarter of these RCTs used covariate adjustment, and only 7% employed an ordinal analysis approach. We concluded that considerable investment of resources in producing 191 completed RCTs for acute TBI management has resulted in very little translatable evidence. We further appraised the currency, completeness, and quality of evidence from systematic reviews (SRs) of acute management of moderate to severe TBI (Synnot et al 2018). A total of 85 SRs were identified, and of these less than half were judged as high quality (n = 38, 44.7%), and nearly 20% were low quality (n = 16, 18.8%). We concluded that a substantial number of SRs in acute management of moderate to severe TBI lack currency, completeness, and quality. Moreover, despite a substantial increase in number of systematic reviews, these are often outdated by the time they are published. The median time from primary study publication to its inclusion in a published systematic review ranges from 2.5 to 6.5 years. Considering the time required for conducting and publishing primary studies underpinning a systematic review, this means that by the time the SR is published, the information may already be over a decade old.
CENTER-TBI aimed to address this problem of lack of currency by introducing and pioneering the novel concept of Living Systematic reviews (LSRs). An LSR starts as a conventional SR, but then transitions into a “living” document, that is continually updated, incorporating relevant new evidence as it becomes available. By transforming the existing evidence into living reviews, the available evidence is optimized and remains current. A formal agreement was established with the chief editor and publishers of Journal of Neurotrauma. In brief, this agreement includes open access publication of the LSR’s (subject to peer review) as a separate manuscript category with a specific tab labelled ‘Living Systematic Review’.
During the project, 5 LSRs have been published:
• Epidemiology of Traumatic Brain Injury in Europe: A Living Systematic Review (Brazinova et al 2015).
• Adherence to Guidelines in Adult Patients with Traumatic Brain Injury: A Living Systematic Review (Cnossen et al 2021).
• Blood-based protein biomarkers for the management of traumatic brain injuries in adults presenting with mild head injury to emergency departments: a living systematic review and meta-analysis (Mondello et al 2021).
• Genetic Influences on Patient-Oriented Outcomes in Traumatic Brain Injury: A Living Systematic Review of Non-Apolipoprotein E Single-Nucleotide Polymorphisms (Zeiler et al 2019).
• The Apolipoprotein E4 polymorphism and outcomes from traumatic brain injury: a living systematic review and meta-analysis (McFadyen et al 2019).
Although literature searches can largely be automated, the maintenance and updating of LSRs required substantial efforts from the teams, and these efforts cannot all be continued after the end of the project. The development of LSR’s, has now been taken forward in broader context by the Cochrane collaboration. We are proud that the pioneering efforts of CENTER-TBI has contributed to the development of a “new evidence ecosystem”. In addition, CENTER-TBI has published a total of 19 systematic reviews across a broad range of topics, ranging from methodological approaches and design features to high level ICU care. In combination, the LSRs and conventional systematic reviews constitute a sound evidence base summarizing current knowledge on TBI and its treatment.

1.3.13 Summary towards Global aims
The Global aims of CENTER-TBI were: [1] To improve characterization and classification of TBI; and [2] To identify the most effective clinical care, providing high quality evidence in support of treatment recommendations and guidelines. We addressed these aims using multiple strategies. These included:
• A comprehensive assessment of the existing evidence base. This resulted in authoritative reviews, including the development of conceptually new “Living Systematic Reviews”. The Lancet Neurology Commission on TBI, led by the CENTER-TBI coordinators, now a core reference for researchers, funders, and policy makers.
• Acquiring and analysing high quality, CDE-based, data obtained in detailed longitudinal observational studies (CENTER-TBI Core Study and Registry).The Core study was combined with a neuro-imaging repository and DNA and biosample repositories. For analyses, we used state-of-the-art statistical approaches, machine learning techniques and convolutional neural networks.
• Broadening the reach of CENTER-TBI through Embedded Trials and Associated Studies to interact with other research in the EU, and addressing the global perspective with linked studies in Australia, China and India.
• Undertaking Precision Medicine characterisation of TBI: In order to support work on this aim we established the largest neuro-imaging repository and biosample bank in the world in the field of TBI. These resources enabled us to perform the first ever large-scale GWAS/TWAS study in TBI, conduct the first lipidomic study in TBI, and perform extensive biomarker analyses in hitherto unprecedented numbers.
• Undertaking CER analyses: CER analyses were underpinned by performing extensive provider profiling of participating centres and applying instrumental variable analyses. Analyses were challenging as between centre differences in outcome were lower than anticipated. Nevertheless, we clearly identified best practices, including recommendations for fluid management and venous thrombosis prophylaxis in the ICU.
• Providing a legacy database and neuroimaging and biosamples repositories for future research
Overall, CENTER-TBI has been hugely productive (over 200 publications) and provided many novel insights towards accomplishment of our global aims. Some highlights of results from these analyses include:
• Demonstration of the added value of biomarkers for triaging patients with mild TBI for CT scanning;
• Showing the added value of (advanced) MRI in characterizing TBI (particularly at extremes of the severity spectrum);
• Refining the use of advanced ICU monitoring in individualizing treatment;
• Providing recommendations for best practices
• Development of quality indicators
We identified epidemiological changes and disparities in care of direct relevance to policymakers and health care professionals. We also identified gender disparities in TBI outcome. Our Precision medicine pipeline clearly showed added value of the use of emerging technologies and resulted in a novel multidimensional classification for initial injury severity and recommendations for the targeted application of selected outcome instruments in multidimensional approaches to outcome assessment. Our CER pipeline delivered recommendations for best practices in the ICU setting, and identified room for improvement in various aspects of care delivery and care paths. An overview of our findings and their impact on care and research is provided in section 1.4.6.
In summary, CENTER-TBI has accomplished what it set out to do and produced many results of direct relevance to citizens and patients, to policymakers, health care professionals and researchers alike. Formally, the project period ended on 31st March 2021. There will, however be “life after/with” CENTER-TBI: The clinical database and associated repositories constitute a unique and extremely rich resource which will enable addressing additional research questions both by CENTER researchers and external groups. Our plans include a major focus on meta-analysis between CENTER-TBI and other IntBIR studies. Initial analyses between CENTER- and TRACK-TBI are already being undertaken with own resources. We have accrued additional funding that will guarantee access to and maintenance of the CENTER data with options for support of data analyses. However, this funding will be insufficient to fully perform the required “deep harmonization” (see section 1.3.8) in preparation of federated meta-analyses and to subsequently perform detailed analyses. We are actively seeking support to realize the full potential of meta-analyses across studies and will be applying for definitive funding in partnership with TRACK-TBI in Q3 2021.
Potential Impact:
1.4.1 Background
TBI is a substantial global public health challenge. The annual global incidence of TBI is over 50 million, and it is estimated that about half the world’s population will sustain one or more TBIs at some point in their lifetime. Globally, TBI is a major cause of death and disability across all ages, and kills more young adults than any other disease. Survivors of TBI can be left with significant disability: at six months post-injury, half of survivors of severe TBI are severely disabled, and even with a mild TBI, 50% fails to make a complete recovery. TBI has a substantial impact in all countries, but imposes a disproportionate burden of disability and death in low- and middle-income countries (LMICs). It has been estimated that TBI costs the global economy approximately €325 billion annually, which means that around one in every €150 that the world makes is spent on the costs or consequences of TBI. Despite the magnitude of the problem posed by TBI, and its clear impact over many decades, efforts to tackle the problem have been fragmented and under-resourced, and the clinical care of TBI has been suboptimal. CENTER-TBI, along with other partners in the International Traumatic Brain Injury Research (InTBIR) initiative, was conceptualised as a means of generating knowledge that could contribute to reducing the individual, public health, and societal burden of TBI.

1.4.2 The CENTER-TBI study – matching study structure and outputs to desired impact
Details regarding the CENTER-TBI study are provided in section 1.2 of this report. The current discussion focuses on those aspects and approaches of the study that enabled us to deliver the impact we aspired to. CENTER-TBI addressed both broader issues of health care organisation for TBI, and more detailed aspects related to characterization and best clinical practices. We aimed to capture the “real world situation” in order to provide broadly applicable recommendations. This holds the greatest potential to improve current health care for TBI and its delivery at both population and individual levels. Key aspects of CENTER-TBI that are of relevance to our discussion of impact include:
Landscape of TBI care and research: Systematic Reviews and the Lancet Neurology Commission on TBI: Prior to start of CENTER-TBI, we identified major gaps in our knowledge and research priorities which informed our plans for the study. This analysis was subsequently consolidated in the form of systematic reviews, many of which used the novel methodology of Living Systematic reviews. Subsequently, the CENTER-TBI Coordinators (Maas & Menon) were invited by the Editors of the Lancet Neurology to lead a Commission on TBI, which was published in 2017, and is now a core reference resource for clinicians and researchers who work with TBI.
The CENTER-TBI Core study: The core study (in the EU and Israel) recruited ~4509 patients, expanded through recruitment in India and Australia to a total of 5726 patients. These patient numbers ensured that the study was appropriately powered for our planned analysis, and collected data with sufficient granularity to enable Precision Medicine and Comparative Effectiveness Research analyses of patient endotypes (clusters of patients with common presentation, disease course and outcome), and common interventions, respectively. We also ensured that the granular data collection included novel methods (advanced neuroimaging, high resolution ICU monitoring, blood biomarker measurement, and genomics), so that any definition of patient endotypes could move beyond existing schemes of classification based on the Glasgow Coma Score (GCS) as mild (GCS 13-15), moderate (GCS 9-12) and severe (GCS <8) TBI, or broad pathoanatomical categories based on computed tomography (CT) scanning.

The CENTER-TBI Registry: The CENTER-TBI registry (which comprised 22,849 patients from the EU and Israel, supplemented by patients from India and China to a total of 41,367 patients) collected far less granular data than the core study. However, it was important to explore broad epidemiological issues, and to address the effects of differences in systems of care. The effects of broad variations in care was also facilitated by the Provider Profiling exercise that we undertook, which characterised system and process variations in participating centres. In addition, comparison with the Core study data in each centre (and with collaborating national TBI audits) provided an excellent framework to assess generalisability of our study results, both nationally and internationally

The CENTER-TBI Database, Repositories and novel research tools: As part of our recruitment of patients to the Core study, we created a well curated and accessible database of clinical data, readily accessible through a novel tool created by INCF as part of the study (Neurobot). The CENTER- TBI study has also resulted in the largest biorepositories in the world for neuro-images (CT and MR images) and blood/serum samples for TBI. The establishment of these biorepositories and a well-characterized clinical database will facilitate legacy research extending well beyond the duration of this project. We developed several novel methodologies and tools, including AI based lesion and anatomical segmentation on CT and MR images, and FDA-approved diagnostic pipelines. These have been made widely available for academic use or developed as commercial pipelines.

Promoting European, global, and academic-commercial collaboration: The CENTER-TBI dataset and repositories not only represent a valuable legacy resource, but also provide an important vehicle for ongoing collaboration as the study proceeded. Well characterised models of collaboration, including the use of Analysis and Publication summaries on the CENTER-TBI website, a mechanism for review of study proposals submitted by internal and external investigators, and practical Data Use Agreements, have all facilitated access to the rich data and sample repositories by investigators not originally part of CENTER-TBI. Many of these are new academic collaborators, while others are commercial entities particularly interested in biomarker development and study design for novel therapies. At a global level, CENTER-TBI has been one of the two main partners in InTBIR, driving internationalisation of CDEs, harmonisation of data, and joint analyses across studies.

1.4.3 Gaps in knowledge and Research priorities
Below, we first summarize gaps in current knowledge and research priorities identified at the start of the project. We then align these to key recommendations that we made in the Lancet Neurology Commission on TBI. This analysis provides a context to subsequent parts of this section where we describe the structured approach we used to evaluate the impact of our findings, our dissemination activities, and exploitation of results.
These gaps in knowledge, summarised in the CENTER-TBI funding application, were developed and refined as part of the Lancet Neurology Commission on TBI (http://www.thelancet.com/commissions/traumatic-brain-injury). The scope of the Commission was somewhat wider than our aims in CENTER-TBI, but there was substantial concordance in those sections of the recommendations that mapped onto our research plans. These sections, with the cognate recommendations that we made in in each case, are listed below. In some instances, the recommendations are amplified by commentary from the text of the CENTER-TBI application or the broader text of the TBI Commission.

• Systems of care for TBI: Health-care policies should aim to improve access to acute and postacute care to reduce the effects of TBI on patients, families, and society. Provision of post-acute care needs to address patients with mild TBI as well as those with more severe injuries.
• Clinical management of TBI: Robust evidence is needed to inform guidelines on medical, surgical, and rehabilitation interventions, and hence improve outcomes for patients with TBI. In addition to conventional randomized clinical trials (RCT), such evidence will be informed by epidemiological associations and comparative effectiveness research (CER)
• Characterisation of TBI: the path to Precision medicine: Research is needed to improve the precision of diagnosis, classification, and characterisation of TBI using multidomain approaches. Such classification needs to go beyond conventional approaches used to date, and make use of advanced neuroimaging, genomics, biomarkers, and novel methodology
• Assessment of TBI outcome – towards multidimensional approaches: Multidimensional outcome constructs that quantify the overall burden of disability from TBI need to be developed and validated to guide improved clinical management and support high-quality research. Such comprehensive outcome assessment must address practice variations, and be flexible enough to account for the entire spectrum of TBI outcomes.
• Prognosis in TBI – linking patient and injury characteristics to outcome: Efforts are needed to develop a set of quality indicators for TBI that includes structure, process, and outcome metrics. Currently, there are no TBI-specific indicators.
• New directions for acquiring and implementing evidence: Comparative effectiveness research should be supported to identify best practices and to improve the level of evidence for systems of care and diagnostic and therapeutic interventions. CER provides a valuable adjunct to conventional RCTs to seek evidence in areas where such trials are difficult to set up, and to set hypotheses for confirmation by subsequent trials
• Coordinated research efforts on a global basis: A commitment of governmental and non-governmental funding bodies, as well as industrial partners, is needed to foster global collaborations and to establish national and international biorepositories and databases that could facilitate future TBI research. This implies the need to develop a common vocabulary and syntax for describing phenotypic characterization of TBI: Common data Elements (CDEs)

1.4.4 A framework for describing the impact of CENTER-TBI
In our funding application, we structured our assessment of the potential impact of CENTER-TBI using a framework provided by the 2014 UK Research Excellence Framework (REF). The updated UK REF exercise (https://www.ref.ac.uk/publications/panel-criteria-and-working-methods-201902/) uses similar broad measures of impact assessment, but both have disadvantages, since they sought to parse impact across a broad range of academic endeavour – ranging from art and humanities, through physical and biological sciences, to clinical medicine and social sciences. These issues have been considered by other bodies, including the International School on Research Impact Assessment (ISRIA)6. ISRIA provides a useful framework, originally used by Alberta Innovates7, which more specifically addresses Impact Assessment in healthcare research. We therefore chose to structure our current analysis of impact according to ISRIA Framework.
• Capacity-building: Leveraged funding, research tools and methods, use of facilities and resources, career trajectory of researchers
• Advancing knowledge: Bibliometrics, engagements, esteem measures, collaborations and partnerships
• Informing decision-making: Influence on policies, practices, products, processes and behaviours (both in health and the determinants of health)
• Health: Medical and health interventions, health quality indicators, health status
• Economic and social benefits: Intellectual property and licensing, spin outs, economic returns, jobs, economic diversity and productivity
• Social engagement: Public involvement, dissemination, engagement with relevant patient or commissioning groups, culture and creativity
Regardless of the framework used, it is important to identify the stakeholders who will experience the benefits of such impact, since a broad consideration of these target groups will allow a more rational assessment and allocation of potential impact.

1.4.5 Stakeholders: Beneficiaries of impact from CENTER-TBI
A wide range of stakeholders could potentially benefit from the outputs of CENTER TBI. These include:
• Citizens/Patients: While the conventional description of this class of stakeholders/beneficiaries would be patients, we have chosen to use the term “citizens” since some of the epidemiological insights provided by CENTER-TBI may result in policies that prevent TBI, and hence prevent citizens ever becoming patients. In addition, our data suggest that individuals who sustain a TBI may exit the acute illness and be discharged from clinical care. Though such individuals are not seen as patients at that stage, our data suggest that a significant proportion have ongoing healthcare needs which are currently poorly addressed.
• Health care professionals and Researchers: Improvements in diagnosis, comparative effectiveness research, and characterisation of disease endotypes are all targeted outcomes from CENTER-TBI. If realised, these could provide more rapid and individualised treatments for patients through improved and harmonised guidelines – thus achieving the goal of precision medicine. The increases in knowledge delivered by CENTER-TBI provide a basis for setting new hypotheses which can be tested in subsequent clinical studies, while the clinical data, neuroimaging repositories, and biosample resources provide a rich substrate for secondary analysis and research. In addition, the novel insights obtained from analyses of clinical data could also inform basic research in a “bedside to bench” reverse translational pipeline.
• Policymakers: Insight into current epidemiological patterns of TBI across Member States will inform prevention campaigns, targeted to needs at national levels. Our focus on the impact of systems of care and organisational aspects of care delivery could yield substantial benefits. More efficient and targeted care and improved outcome will reduce costs. New performance indicators and improved prognostic models will facilitate benchmarking and assessments of quality of care.

1.4.6 Impacts achieved
Here we summarize the impact of CENTER-TBI using the domains of the ISRIA Framework.
Capacity building:
1) Leveraged funding: CENTER-TBI was a large scale collaborative project, supported by the FP7 Program of the European Union (Grant No 602150). We leveraged this funding by obtaining additional support from OneMind (US), the Hannelore Kohl foundation (DE), IntegraLifeSciences (US) and NeuroTrauma Sciences (US). OneMind was instrumental in supporting the development and implementation of the e-CRF, data collection procedures and the development of Neurobot, our data access tool. The Hannelore Kohl Foundation supported data collection and analyses by centres in Germany, and facilitated an extended Registry data collection in German sites. IntegraLifeSciences provided additional support towards the data curation. NeuroTraumaSciences provided support towards analyses, in particular during the no-cost extension period and will provide further support to maintain the CENTER-TBI infrastructure and to facilitate further data mining after the formal end of the FP7 funding period. In addition, CENTER-TBI participants have also been successful in securing follow on grants from the ERANET-NEURON program (see later)
2) Research tools: CENTER-TBI developed and implemented the following research tools:
o Translation, linguistic validation and psychometric evaluation of outcome instruments. In total, 237 translations and 211 linguistic validations were carried out in up to 20 languages; all are accessible in the public domain on the website of CENTER-TBI (https://www.center-tbi.eu/project/validated-translations-outcome-instruments/). We consider this a major output of CENTER-TBI as they provide a solid basis for future TBI research and clinical practice and allow for aggregation and data analysis across different countries and languages.
o A GOSE manual: Interactions in InTBIR highlighted variations in GOSE application by different groups. CENTER-TBI and TRACK-TBI collaborated to produce a definitive manual for its use (Wilson et al 2021).
o Support tool for Database access: Neurobot (details available on the SciCrunch Resource Identification Portal; Research Resource Identifier RRID/SCR_017004). This bespoke data access tool was developed and regularly updated by KI-INCF (Stockholm SE) - a substantial task given the large number and types of variables, and the inclusion of continuous and longitudinal data.
o Study- and publication proposal platform (https://www.center-tbi.eu/data). We implemented a study- and publication proposal platform (SPP) on the CENTER-TBI website. We wished to ensure “good use” of the data, promote collaborations and prevent redundancy of efforts. Proposals are reviewed by the Management Committee for scientific rigor and feasibility. To date, over 300 proposals have been submitted by internal and external researchers. All accepted proposals are listed on the website to be accessible in the CENTER-TBI community and the platform will remain operational in the upcoming years for further data sharing. We have found this platform to serve an additional important purpose in promoting collaborations and limiting the risk of redundancy of efforts.
o icobrain tbi (https://icometrix.com/services/icobrain-tbi): This tool offers automated reporting of acute CT scans in TBI, including automated volumetric analyses. It received FDA clearance 510(k) in Nov 2018.
o Icompanion is an app, originally developed by icometrix for self-reporting of outcome in patients with multiple sclerosis (https://icompanion.ms/) has now been expanded to include specific outcome domains relevant to TBI. This app provides patients with the opportunity to report their perception of outcome on a frequent basis.
o BLAST CT is a deep learning automated pipeline for lesion detection,segmentation, and quantitation in CT images following TBI, freely available on Github (https://github.com/biomedia-mira/blast-ct).
3) Career trajectory of researchers: CENTER-TBI provided a unique platform for stimulating the career of researchers and promoting interactions between research groups. We provided courses on evidence-based medicine and methodology for systematic reviews, and offered scholarships in conjunction with scientific societies to attract the best and brightest young researchers to neurotrauma and facilitate EU wide mobility by providing opportunities to work at leading TBI centres that participate in the project. A total of 7 PhD theses were successfully completed during the Project, and there are more in the pipeline.
4) Global standards: Standardisation of data collection is essential for research and the common data elements have shown their great value in IntBIR projects, including CENTER-TBI and its linked studies. However, they are currently mainly US-centric, and we identified major issues with regard to global application, including violation of privacy regulation. We conclude that efforts should be supported to upgrade the primarily US centric common data elements to global standards.

Advancing knowledge
CENTER-TBI was all about “advancing knowledge” on TBI and improving its treatment. Detailed information on “Game-changing” findings and recommendations are provided under the headings “Informing decision-making” and “Health” of this section. Here, we summarize the impact in terms of collaborations and bibliometrics.
1) Collaborations: Collaborations were fostered at the InTBIR level, with academic and industrial partners, and in a global context.
o IntBIR: CENTER-TBI was very active within IntBIR, participating in annual meetings, and being actively involved in various InTBIR working groups. IntBIR is currently transitioning to a more investigator-driven organization, and Prof David Menon (co-coordinator of CENTER-TBI) will likely become one of the co-leads. Within IntBIR, close collaborations exist with TRACK-TBI (“sister” study in the US to CENTER) and with CREACTIVE. We have utilized the Human Brain Project Medical Informatics Platform (HBP-MIP: https://mip.humanbrainproject.eu/) platform for a validation study on the Core IMPACT prognostic model across CENTER-TBI and CREACTIVE. We established the GAIN consortium (Genetic Associations in Neurotrauma) to explore the influence of genetic variation on clinical outcome in TBI in a larger cohort than would be possible within the individual studies. Data from a provisional total of 5628 patients were included from 6 cohorts: CENTER-TBI (n=3187), Cambridge (n=575), Turku (n=157), TRACK-TBI (n=1672), Mass general Brigham (n=409) to produce the first GWAS/TWAS study in TBI.
o Academic and industrial partners: Multiple interactions have occurred with both academic and industrial partners. A total of 33 study proposals have been received from external Parties, of which approximately half were accepted. Collaborations were formalized for biomarker analyses with the Universities of Edinburgh (Scotland) and Örebro (Sweden) as academic partners, and with ABCDx (Geneva, Switzerland) and NanoDx (Southborough, MA, USA) as industrial partners. We further established collaboration with Biogen (Cambridge, MA, USA) to explore the relationship between contusion characteristics and functional outcome.
o Global collaborations: Collaborations with Australia, China and India resulted in linked data collections (OzENTER: ICU stratum n=198); CENTER China Registry: n= 13138); India Core study: n=1017 and Registry: n=4903) using a similar data format as in CENTER-TBI, illustrating the global outreach of CENTER and highlighting the relevance of global data collection standards (CDEs).We established additional collaborations with European and Canadian partners through ERANET-NEURON(see also section 1.3.10).
2) Bibliometrics: CENTER-TBI has produced over 200 publications in peer-reviewed scientific journals to date.
Informing decision-making

Our large dataset, highly granular data collection, and advanced use of methodologies has allowed us to make important recommendations that are of relevance to a wide range of stakeholders. We list the key recommendations/conclusions below, and designate the stakeholders for whom the recommendation has impact relevance (CP: Citizens/Patients; Pol: Policymakers; HC: Health care professionals and researchers):
1) Health Care systems need to address the needs of older patients with TBI, and research should be stimulated to provide evidence in support of their treatments. Justification: TBI is a major cause of death and disability across all ages. In the EU, approximately 1.5 million patients are admitted to hospital each year for TBI. The epidemiology of TBI has changed: Median age is currently 50-55, and 28% are over 65 years of age – a substantial increase compared to one or two decades ago. Older patients have more comorbidities with associated medication. In the Core study, 11% had serious co-morbidities, that can adversely affect disease course. Both the CENTER Core study and the CENTER China registry showed that pre-injury use of anticoagulants is associated with higher mortality. The needs of older patients for post-acute care are different from younger patients. Most clinical trials to date have excluded patients > 65 years of age, and as a consequence little evidence exists to support their treatments. (Impact: CP/Pol/HC)
2) High energy transfer mechanisms should be de-emphasized for injury scene and ED triage of older people with TBI. Justification: We found that 40% of patients with TBI are injured by low energy falls. These mostly occur in older patients, and have similar rates of CT brain scan abnormalities and in-hospital mortality as those injured by other mechanisms, but are 50% less likely to receive critical care or emergency interventions. (Impact: Pol/HC).
3) Alcohol prevention campaigns should be expanded to increase awareness of the risk for serious injury from falls under the influence of alcohol. Justification: We found that alcohol use was reported in 28% of patients injured by incidental falls versus only 17% of those injured in road traffic incidents. These findings further illustrate the success of traffic-related alcohol prevention campaigns. (Impact: CP/Pol/HC).
4) Alcohol and substance abuse are major factors in violence-related TBI. Justification: Alcohol use was reported in 64% and cannabis use in 15% of violence-related TBI. (Impact: CP/Pol/HC).
5) Litigation procedures should not consider a normal CT scan at presentation as evidence of absence of structural brain damage, let alone absence of TBI. Justification: Our MR studies showed that 30% of patients with mild TBI and a normal CT scan demonstrated structural abnormalities on MR imaging. Moreover, at least in univariate analysis, serum levels of the brain-specific biomarker NFL were found to be significantly higher (p<0.05) in patients with mTBI and a normal CT scan on presentation who had residual complaints at 6 months, compared to those who had a full recovery, and this difference remained for samples obtained at 2-3 weeks. (Impact: CP/Pol).
6) Substantial health-economic benefits can be accrued by improving the care delivery (in particular structured follow-up) and developing new treatments for mTBI. Justification: Mild TBI is the most common form of TBI (82% in the CENTER-TBI registry) and poses the largest burden to patients and society. “Mild” TBI is not so mild: We found that 63% of patients report residual disability or complaints at 6 months after injury: 51% had a GOSE below 8, around 25% SF12v2 summary scores below threshold for impairment (scores <40) and 26% had RPQ scores ≥ 16, indicating significant postconcussion sypmtoms. Despite these high impairment rates, 90% of centres do not routinely schedule a follow-up appointment on discharge home from the Emergency Room after mild TBI, and only 46% do so on discharge of patients with mTBI from the ward. (Impact: CP/Pol/HC).
7) Access to and provision of care for individuals with moderate to severe disabilities after TBI needs to be improved. Health care systems should anticipate an increased need for rehabilitation after TBI Justification: Data on 1206 individuals enrolled into CENTER-TBI who had moderate to severe disability at 6 months after injury showed that 90% reported rehabilitation needs, but only 30% received in-patient rehabilitation and 15% out-patient rehabilitation. We found a much lower between-centre variation in mortality (MOR:1.2) compared to previous studies, and in patients with moderate to severe TBI mortality was lower than predicted from the IMPACT prognostic model (observed to expected ratio 0.70 [0.62– 0.76]) but unfavourable outcome (defined as a GOSE<5), was not (1.06 [95% CI 0.97– 1.14]). These data suggest that treatment has improved with fewer deaths, but at a cost of more survivors with disability. (Impact: CP/Pol/HC).
8) Efficiency of use of ICU resources may be improved by increasing resources to manage mild-TBI outside ICUs Justification: We found that 36% of patients admitted to the ICU with TBI are classified as “mild” TBI. Some of these are motivated by the presence of serious extracranial injuries, by secondary deterioration or by a substantial risk for deterioration due to possible progression of traumatic intracranial lesions, but it appears likely that ICU admission is motivated in some by lack of resources on other wards. (Impact: Pol/HC).
9) The co-occurrence of TBI with injuries to other parts of the body emphasizes the need for a multidisciplinary approach to treatment. Justification: Major extracranial injuries (abbreviated injury score ≥3) were reported in 422 (28%) patients in the admission stratum and in 1174 (55%) in the ICU stratum. The body region most commonly injured was thorax and chest (n=742 [35%]), and concomitant serious spinal injuries occurred in 374 (18%) patients. (Impact: HC).
10) Quality indicators should be used to benchmark quality of care between institutions. Justification: We have validated Quality Indicators to support benchmarking and quality improvement programs (Impact to Pol/HC).

Health
The conclusions that we detail below also translate into important insights for clinicians and researchers, which will inform ongoing patient management and research in TBI
1) The biomarker GFAP should be included in decision rules for triaging patients with mild TBI for CT scanning. In patients with mild TBI, GFAP showed incremental diagnostic value: discrimination increased from 0.84 [95%CI: 0.83-0.86] to 0.89 [95%CI: 0.87-0.90] when GFAP was included. Combinations with other biomarkers showed no added value. These results further challenge the utility and cost effectiveness of combined biomarker assays.
2) The estimated heritability in our GWAS studies was 0.28 suggesting that common genetic variation significantly contributes to inter-individual variability in host response and outcome.
3) Qualitative MR imaging adds to the accuracy of outcome prediction beyond conventional clinical and CT characteristics, and quantitative volumetric MRI and DTI metrics provides further added value.
4) DTI metrics can help predict emergence from coma in patients with very severe disturbances of consciousness.
5) Criteria for in-hospital intubation should be broadened to include patients with a GCS of 9 or 10.
6) Early tracheostomy (within one week) for patients requiring ventilator support is associated with better outcome and reduced length of stay in hospital and ICU.
7) The recommended threshold in the Guidelines of 22 mmHg for treating raised intracranial pressure is not absolute. We found a threshold of 18 +/- 4 mm Hg.
8) Treatment for raised ICP should be individualized, taking autoregulatory status into account.
9) Maintaining a neutral fluid balance in ICU patients is associated with better outcome, but is not common practice. We found an increased for poorer outcome per 0·1L increase of fluid balance with an OR of 1·10 [95%CI:1·07–1·13] for ICU mortality and 1·03 [95%CI:1·02–1·05] for functional outcome.
10) Outcome predictors differ between mild and moderate/severe TBI. In mod/severe TBI, outcome is mainly dependent on injury severity, whilst in mild TBI it is more “what the patient brings to the injury” (e.g. pre-injury health and psychiatric history). Existing models for predicting outcome in mod/severe TBI were validated and updated and a new model for mild TBI developed.

Economic and social benefits
Many of the outputs of CENTER-TBI, described under the headings “Informing decision-making” and “Health” carry the potential for substantial economic and social benefits. For example, we identified various disparities in care provision for patients with TBI (e.g. needs of older patients, lower care for patients injured by low energy mechanisms, lack of structured follow-up and post-acute care for patients with mild TBI, rehabilitation needs for patients with moderate to severe disability), and addressing these will result in large economic and social benefits. We also identified gender disparities in outcome after mild TBI. Whilst males are more prone to TBI, compared to men, women with mild TBI had worse outcomes (OR 1.4 95% CI: 1.2-1.6) lower generic and disease-specific HRQoL, and more severe PCS, depression, and anxiety.

Social engagement
1) Public information platform: We developed and implemented an interactive public information platform explaining the impact and future developments of TBI research in lay language on the CENTER-TBI website (https://www.center-tbi.eu/). This platform aims to make the public active partners in research, clinical care, and policy development, and provides links to patient organizations, such as PatientsLikeMe (www.patientslikeme.com/). Patient requests have frequently been received and answered about disease characteristics, rehabilitation possibilities and referrals to TBI specialists worldwide. Interestingly, various of these requests originated from outside Europe. This illustrates the great need of patients across the world for guidance and help in seeking appropriate treatment for their TBI.
2) Media attention: CENTER-TBI has actively sought media attention by various initiatives. Press releases were broadly distributed around the presentation of the Lancet neurology Commission on TBI at the European Parliament and the occasion summarized in a video (https://www.youtube.com/watch?v=VsUk_Q7qnWg). Forbes magazine featured the findings (https://www.forbes.com/sites/nicolefisher/2017/11/09/special-lancet-neurology-issue-targets-political-forum-to-combat-global-tbi/#59d4fd8675a8). CENTER-TBI attracted media interest across the globe, including Australia, China, Belgium, Germany, Hungary, Italy, the Netherlands, and the UK. EuroNews broadcast a special feature on CENTER-TBI in November 2019 (https://www.euronews.com/2019/02/25/i-was-not-who-i-was-researcher-into-new-care-for-traumatic-brain-injury-victims). Public engagement was specifically sought in the UK by the All-Party Parliamentary Group on Acquired Brain Injury. A full report (Executive editor: Prof David Menon, joint coordinator of CENTER-TBI) was published online on 18th October 2019 (https://cdn.ymaws.com/ukabif.org.uk/resource/resmgr/campaigns/appg-abi_report_time-for-cha.pdf).
3) Social media: CENTER-TBI is present on Twitter (@CenterTBI), providing visibility to publications and scientific events where investigators and other clinicians interested in CENTER-TBI are involved. The number of followers, mentions and profile visits has significantly increased during the years, and interest is widely diffused across the world showing that dissemination of our results exceeds the European boundaries.

1.4.7 Dissemination and exploitation
1) Knowledge Commons: Within CENTER-TBI we established a “Knowledge commons” with the aim to develop high-quality systematic reviews to summarize the evidence base underpinning our knowledge of TBI. We conducted a scoping review on trials in moderate and severe TBI, and published 19 systematic reviews and 5 Living systematic reviews (see section 1.3.12). CENTER-TBI pioneered the implementation of Living systematic reviews, in which the evidence is continually updated, incorporating relevant new evidence as it becomes available. These pioneering efforts have contributed to the development of a “new evidence ecosystem”, that is currently being pursued by the Cochrane Collaboration.
2) Dissemination of study results: Dissemination aimed for widespread knowledge and use of research results by the target population. We conducted a range of approaches targeting policy makers, health care professionals and patients. Approaches included publications in the scientific literature, presentations and interactions with policymakers, press releases and media communications, social media accounts, and interactions through the CENTER-TBI website (see also section 1.4.6). Here, we focus on dissemination in the scientific literature. The CENTER-TBI Consortium has been highly productive with – to date – over 200 publications in peer-reviewed scientific journals, of which 26 were in journals with an impact factor > 10. A complete list of all publications generated by the Consortium and CENTER-TBI affiliates is available on the CENTER-TBI website (https://www.center-tbi.eu/publications/). CENTER-TBI was designed as team collaborative effort, and most publications spanned various research groups.
3) Exploitation: We developed and implemented an open standards-based platform (Neurobot) for the collection and storage of clinical data and neuroimaging and biomarker results based on Common Data Elements (CDEs). Collaborative analytics are facilitated by providing standard interfaces to the platform through which various analytics tools can access the data. A second platform (Opal) was implemented, offering additional analytical tools and options to facilitate meta-analyses across different studies. These platforms offer opportunities for external researchers to access and use the unique data of CENTER-TBI and its repositories in the years to come. CENTER-TBI is open to data-sharing and welcomes proposals from other researchers, thus optimizing the use of public funding that supported CENTER-TBI and advancing the care for patients with TBI. The linguistically validated translations of outcome instruments in up to 20 languages are accessible in the public domain on the website of CENTER-TBI (https://www.center-tbi.eu/project/validated-translations-outcome-instruments/) and provide a solid basis for future TBI research and clinical practice in international settings. Icobrain tbi, developed for automated segmentation and volumetric analyses of CT images, received FDA clearance 510(k) in November 2018 and now offers radiologists, neurosurgeons, and neurologists easy access to clinically important metrics to better characterize and inform management of TBI in the acute clinical setting.

References:
1) Steyerberg EW, Mushkudiani N, Perel P,et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008 Aug 5;5(8):e165; discussion e165. doi: 10.1371/journal.pmed.0050165. PMID: 18684008; PMCID: PMC2494563.
2) MRC CRASH Trial Collaborators, Perel P, Arango M, Clayton T, Edwards P, Komolafe E, Poccock S, Roberts I, Shakur H, Steyerberg E, Yutthakasemsunt S. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008 Feb 23;336(7641):425-9. doi: 10.1136/bmj.39461.643438.25. Epub 2008 Feb 12. PMID: 18270239; PMCID: PMC2249681.
3) Scheufele E, Aronzon D, Coopersmith R, McDuffie MT, Kapoor M, Uhrich CA, Avitabile JE, Liu J, Housman D, Palchuk MB. tranSMART: An Open Source Knowledge Management and High Content Data Analytics Platform. AMIA Jt Summits Transl Sci Proc. 2014 Apr 7;2014:96-101. PMID: 25717408; PMCID: PMC4333702.
4) Retel Helmrich IRA, Lingsma HF, Turgeon AF, Yamal JM, Steyerberg EW. Prognostic Research in Traumatic Brain Injury: Markers, Modeling, and Methodological Principles. J Neurotrauma. 2020 May 20. doi: 10.1089/neu.2019.6708. Epub ahead of print. PMID: 32316847
5) Huie JR, Mondello S, Lindsell CJ, et al: Biomarkers for Traumatic Brain Injury: Data Standards and Statistical Considerations. J Neurotrauma. 2020 Apr 1. doi: 10.1089/neu.2019.6762. Epub ahead of print. PMID: 32046588.
6) Adam, P, Ovseiko, PV, Grant, J. et al. ISRIA statement: ten-point guidelines for an effective process of research impact assessment. Health Res Policy Sys 16, 8 (2018). https://doi.org/10.1186/s12961-018-0281-5
7) Graham KER, Chorzempa HL, Valentine PA, Magnan J. Evaluating health research impact: Development and implementation of the Alberta Innovates – Health Solutions impact framework. Res Eval. 2012;21(5):354–67
List of Websites:
Prof. Tomas Menovsky – Antwerp University Hospital
Tel: +32 38 21 45 39
E-mail: center-tbi@uza.be
Project website address: www.center-tbi.eu