Introduction
Defining Endotypes/Phenotypes
Background
Main Research Gap (Current State)
Panel Discussion of the Approach
-
Endotypes need to be characterized within the framework of the underlying, biological, mechanistic causes of the DoC, ranging from neural networks to biochemistry.
-
Endotypes will require widely available techniques; in addition, a role exists for advanced imaging, electrophysiological, and other specialized tools, some of which have yet to be developed. The approach will include comprehensively applying existing and novel tools (e.g., imaging, high-dimensional data sets, statistical models, analytic techniques) that accurately assess functional and structural neural pathways related to consciousness and generate appropriate biological hypotheses.
-
An example of the complexity of endotype characterization is the use of machine learning to analyze electroencephalography (EEG) or functional magnetic resonance imaging (fMRI) responses to diagnose CMD, which cannot be diagnosed by clinical examination alone. CMD is not an endotype per se, but it could be a feature characterizing a subgroup of patients who behaviorally are not overtly following commands. Taken together with subgroup-specific commonalities in network dysfunction, treatment response, and long-term outcomes, CMD could be a defining feature in one or several endotypes [7‐10, 21‐23]. Stated otherwise, CMD considered as a pure phenomenological state without connection to a specific biological mechanism would not alone constitute an endotype.
-
Predictive enrichment in clinical trials: developing endotypes to identify patients who have high or low likelihood to benefit from an intervention will allow for selective sampling for a specific RCT and will increase the probability that an effective intervention to treat DoC will be discovered.
Limitations/Challenges
-
Endotypes should be axiomatic (i.e., mechanistically driven, associated with outcomes and with a treatment response): endotypes not meeting these minimum criteria are unlikely to be useful in the clinical domain and should not be the focus of research unless novel therapeutics are likely to leverage these mechanisms.
-
Definitions and nomenclature: A review of the literature indicates that there is a high degree of inconsistency in the use of the terms (e.g., phenotype, subphenotype, endotype, endophenotype). We propose here an intuitive working definition of endotype that will need to be vetted, validated, and accepted. Conceptual validity: although the endotype paradigm has been explored in other medical domains and seems biologically plausible, research is needed to demonstrate their significance in patients with severe brain injury/DoC.
-
Practical considerations: the group raised the issue of feasibility, relevance, and implementation of endotypes in the clinical setting and in the coma science community.
-
There is a nonstatic nature of features within endotypes, in part, because they combine different biomarkers [24, 25] or biological processes into a single determinant factor (e.g., seizure activity, covert consciousness). To adjust for this dynamism when endotyping patients, researchers should implement clustering algorithms and be aware that a patient’s changing status may lead to endotype shifts [10, 21, 26‐30]. For example, the patient’s condition in the intensive care unit (ICU) is rapidly evolving because of dynamic changes in the underlying biological mechanisms, and that evolution could be reflected in changes in endotype. This dynamism also, however, affords potential advantages, such as monitoring of duration of coma as an intermediate biomarker or assessing of endotypes as time-varying covariates.
-
Validation: the group identified validation of endotypes, in light of the underlying biology, recovery probability, and response to intervention, as a major challenge.
Deliverables
Biomarkers
Background
Main Research Gap (Current State)
Panel Discussion of the Approach
-
Molecular and cellular biomarkers aim to detect neuronal and glial function, injury, death, recovery, and potential survival [45‐48]. Molecular and cellular biomarkers include genetic and epigenetic proteins and cells/cellular functional assays indicative of cellular function, viability, and death. There are too many molecular and cellular biomarkers that have been proposed or implicated to comprehensively review them here, but likely a combination of markers, rather than a single one, will end up being relevant.
-
Imaging biomarkers aim to map structural and functional elements of the brain, including large-scale networks. Fifteen to twenty percent of patients without signs of responsiveness, detected by either the Glasgow Coma Scale or Coma Recovery Scale, Revised, will demonstrate covert consciousness on fMRI [43]. Mapping of functional connectivity by using resting state MRI (e.g., characterization of default mode network connectivity) may help distinguish consciousness states [21, 49‐53] but alone is not sufficient to predict recovery of consciousness [50, 54, 55] and should be studied along with other networks that contribute to consciousness. Ongoing studies are using such connectivity biomarkers to select patients for targeted therapies to promote recovery of consciousness [56‐58].
-
Electrophysiological biomarkers can be used to detect CMD [59]. EEG can be used at the bedside to detect command-following [10, 40, 41, 60] and covert cortical processing [44, 61] in patients who do not show behavioral evidence of purposeful responses. EEG can also be used to develop biomarkers of cellular preservation that relate to outcomes, such as features within burst suppression patterns, that allow for the identification of patients with and without a chance to recover [62‐64].
-
TMS-EEG biomarkers aim to objectively measure brain responses to direct cortical stimulation. This technique bypasses afferent sensory and efferent motor systems, which may be damaged in patients with DoC. TMS-EEG is based on an information theory that posits that conscious experience is at once highly integrated and differentiated, with this combination resulting in a complex system [65]. To measure these complex systems, a perturbation approach is most helpful to track the system’s response [66]. In practice, the brain is stimulated noninvasively via transcranial magnetic stimulation (TMS), and the brain’s response is measured with EEG. Recent work has proposed a “perturbational complexity index,” in which the numerical value is correlated with changes in consciousness, suggesting value as a marker of consciousness [67‐69].
Limitations/Challenges
-
Ideal biomarkers for coma science should be safe (i.e., minimally invasive), feasible, practical (i.e., have a timely turnaround), and widely available. They should allow for accurate prognosis or goal-directed therapy and have therapeutic implications for either disease process or pathophysiology.
-
Biomarkers will need to have analytical validity, clinical validity, and clinical utility established [70].
-
The timing and source of sample collection, the molecules targeted, handling and processing procedures, and the end point of interest (e.g., etiology of coma, prognosis) need to be defined [71].
-
Confounders in biomarker measurement that need to be considered include the impact of physiologic states and interventions on biomarkers. Targeted temperature management may affect biomarker pharmacodynamics, and sedation may affect EEG and fMRI results.
-
Biases of investigators who are particularly invested in certain biomarkers need to be considered because they also may influence the goals of care decision-making processes. One of the challenges will be to temper the overinterpretation by families of results from proposed biomarker tests.
-
The development of new tools must be integrated within the context of existing, albeit insufficient, biomarkers. Experimental measures need to be carefully disaggregated from proven tools in the coma research repertoire so that the accepted measures can assist in validating the new tools. It will be vital to address whether, and when, information from new tools should be shared with families, because this creates a risk of biasing the results (e.g., detection of CMD may influence the family’s WLST decision).
-
Extrapolation from highly specialized centers to general practice needs to be taken into account. Coma science needs specialized care centers in which analytic tools are made widely available in a systematic, standardized way. Gaps in diagnostic, prognostic, treatment, and rehabilitation equipment and techniques extend beyond economically advantaged countries. Inequities from diagnosis to treatment and rehabilitation are a critical issue nationally and internationally.
Deliverables
Proof-of-Concept Clinical Trials
Background
Main Research Gap (Current State)
Panel Discussion of the Approach
-
Addressing the research gap: Researchers need to design proof-of-concept trials that focus on identifying biomarkers and neurophysiological techniques that will help diagnose, prognose, treat, and cure coma in patients. Interventional trials that aim at promoting recovery should be based on a mechanistic understanding of DoC.
-
Patient selection: Both endotypes and phenotypes should be taken into account in clinical trial design. For example, individuals in minimally conscious states are more responsive to treatment than those with unresponsive wakefulness syndrome. This observation highlights the opportunities of predictive enrichment strategies to maximize clinical trial success for specific populations.
-
Tested intervention: A mechanistic understanding of DoC should inform the tested intervention. One approach is use of the mesocircuit model, which allows researchers to identify areas affected by severe brain injury and the cascading effects on the system of consciousness. Depending on the affected area, different pharmacological and nonpharmacological treatments are available, including, but not limited to, amantadine [74], apomorphine, zolpidem [75, 76], TMS [77‐84], transcranial direct current stimulation [85‐91], deep brain stimulation [92, 93], low-intensity pulsed ultrasound, and vagal nerve or other sensory stimulation [94‐96]. Specifically, within trial designs, investigators should consider and evaluate the benefits and limitations of combination therapies [72, 90]. RCTs may focus on regeneration of lost or broken circuits after TBI [77‐79, 85, 86]. These techniques often seek to leverage the untapped synergistic potential of combination approaches. For example, TMS may be more efficacious when combined with a pharmacologic stimulant. Combination therapies to fix broken pathways could be defined on an individualized basis based on findings of a patient’s neurophysiologic, neuroimaging, genetic, or omics assessment (see “Defining Endotypes/Phenotypes” section for details) combined with an ever-increasing understanding of neurological pathways.
-
Timing of treatment: Although early treatment may be critical to the recovery of consciousness, researchers of chronic DoC may often allow for time after the injury to decrease the confounding possibility of spontaneous recovery [77‐80, 86‐89, 97]. Exact timing for treatment initiation is typically determined by clinicians and researchers on an individualized basis. It is influenced by a combination of factors (e.g., past experience, patient status, time from injury). During the hyperacute and acute phases, with active inflammatory changes and secondary injury, researchers often collect data but hesitate to begin treatment to avoid causing further harm. The optimal start time, duration, and number of repetitions of an intervention needs to be investigated [73]. When defining optimal timing, the individual patient’s neurologic injury and systemic illness need to be considered.
-
Biomarkers in clinical trials (see “Biomarkers” section for an overview): Clinical trials may be greatly improved by adding biomarkers, both as proxies to response and as identifiers for subpopulations (e.g., endotypes). Molecular and cellular biomarkers could serve researchers in conjunction with behavioral, electrophysiological, and imaging measures to both identify promising candidates to support recovery and track responses to investigated interventions. Although some common data elements (CDEs) may be shared with other similar disorders, many will be unique to coma. The larger the population-based group from which data are collected, the more versatile the data’s predictive capabilities. In addition, the sequence of tracking biomarkers from genome and molecules to populations is critical, and coma researchers can work from both ends: clinicians can start at the treatment population and work down toward genomic and molecular levels, whereas researchers can start at the genomic and molecular level and work up to the population of interest. This two-pronged translational and reverse-translational approach may expedite the process of identifying the best biomarkers for coma treatment and prognosis.
-
Outcome measures: A mechanistic approach to clinical trial design requires new tools to identify preserved brain network connections, the ability to map the human brain networks essential for consciousness, the ability to repeatedly assess brain networks starting from the acute care phase of treatment, personalized connectome mapping tools and biomarkers, and targeted personalized treatments. Researchers agree that a multimodal approach best tracks the outcomes of patients receiving treatments. Proof-of-concept trials with small sample sizes applying physiological proxy markers (e.g., EEG enrichment) as outcome measures may allow for identification of promising treatment avenues before larger trials requiring hundreds of patients and clinical outcomes become feasible. To then evaluate effects of treatment, the primary measure should include standardized evaluation of behavioral assessment, preferably the Coma Recovery Scale, Revised, and patient-centered outcomes, such as quality of life (QOL) measures. Neuroimaging and electrophysiological scans can then be used as secondary treatment assessments.
-
Improving trial design and logistics: Adaptive clinical trials, in which results can change the design of subsequent doses or trial arms, hold promise for the evolving understanding of coma. Longitudinal test–retest designs allow for alteration of drug dose, duration of therapy, dose and duration of electricity, and range of potential stimulation targets in a Bayesian-adaptive manner. Such an adaptive study was conducted among patients with COVID-19 (Randomized Embedded Multifactorial Adaptive Platform for Community-acquired Pneumonia [REMAP-CAP]) [98] and allowed for rapid development of treatments on the basis of interim analysis. The established clinical trial networks allow for efficient and rapid testing of novel interventions. This approach would be useful in a cohort of patients with coma, particularly among proof-of-concept trials that are already seeking to point the field in the correct direction rather than to serve as a final study of potential prognostic tools or treatments. The concept of pragmatism in clinical trials is crucial because it promotes later translation of trial findings into clinical practice. A role for the NCS Curing Coma Campaign to facilitate use of a master protocol could facilitate the interoperability of adaptive study data.
-
Treatment standards: Establishing so-called baseline treatment standards for coma remains an important priority given current practice variations, particularly those surrounding WLST. As a criterion for enrollment in DoC and coma trials, requiring an intent to provide aggressive care and postpone WLST orders for a predetermined time frame after the injury leading to the DoC could alleviate concerns about early patient loss in acute-phase trials.
Limitations/Challenges
-
Novelty of treatments: Essentially, all treatments currently used in patients with coma are repurposed from other pathologies and disorders. Basic and preclinical efforts should focus on identifying promising novel therapeutic approaches to cure coma. Translational scientists should help introduce and evaluate these in the clinical context of the ICU and post-acute-care setting. Close bidirectional communication between scientists who focus on clinical, translational, preclinical, and basic science aspects of coma science will be fundamental in achieving the stated goals. This bidirectional dialogue may be conceptualized in a top-down and bottom-up analogy as a translational and reverse-translational approach that needs to be central to future funding initiatives.
-
Treatments may not work for all: Pharmacological treatments often work only in a small portion of the population treated [75], or they work to improve recovery only when they are being administered, and improvements disappear after treatment is stopped [74]. Nonpharmacological treatments often show limited clinical or neurophysiological effects or have highly variable effect sizes. Brain stimulation treatment trials have shown varied improvement rates depending on the treatment area, with the most promising target appearing to be the prefrontal cortex. In addition, findings from the stimulation studies suggest that individuals in minimally conscious states respond better than those in vegetative/unresponsive wakefulness states. Thus, personalized care approaches that take into account individual injuries, pathways, genetics, and omics are needed.
-
Variability in trials: Variability among patients and in the diagnostic and management approaches of different clinical centers is a major challenge in trials of coma treatments. Such variability was recently highlighted by the results of the Point PRevalence In Neurocritical CarE (PRINCE) study, an observational prospective investigation of the practice of neurocritical care, which included patients with coma [99, 100]. This variability may result from equipment or staffing availability or differences in comfort with procedures across clinical centers. Consensus about endotyping of patients will be required to minimize patient variability in clinical trials. On the other hand, it may be detrimental to force clinical centers to follow overly rigid and detailed protocols because this approach may lead to the failure of a treatment at one site that succeeded at another. Additionally, variability in research infrastructure and health care service delivery infrastructure that does not support investigations of treatments/prognosis throughout the continuum of care and across treatment settings is of concern. For example, not all health centers have on-site inpatient or outpatient brain injury rehabilitation units within their health system. Patients may be enrolled into a clinical trial in the critical care or acute care setting but then be transferred to rehabilitation settings outside health systems or to independent facilities that may not have the research infrastructure to support continued participation in a clinical trial. Local experts should be able to use the techniques with which they are most comfortable, but arrangements do need to be made so that outcomes will be comparable between centers.
Deliverables
Prognostication
Background
Main Research Gap (Current State)
Panel Discussion of the Approach
-
Standardization: There is a clear need to develop CDEs for established or promising diagnostic evaluations (both clinical and ancillary) and to define the optimal time window for these evaluations. There is a need to create or refine diagnostic modalities that add the greatest prognostic value at the lowest possible cost. No single test will have sufficient prognostic power to stand alone in predicting recovery from coma [116]. Therefore, many candidate tests and data will need to be evaluated simultaneously, with an aim toward multivariate model development.
-
Accuracy: It is critical to reduce the uncertainty and imprecision of existing prediction models. In a study of outcome after intracerebral hemorrhage, nurses and attending physicians predicted 3-month patient outcome more reliably than the validated intracerebral hemorrhage score or predictors of functional outcome (e.g., Functional Outcome in Patients With Primary Intracerebral Hemorrhage (FUNC) Score) [117]. This difference in predictive accuracy may be due to a number of factors that models do not capture, including preexisting conditions, post acute care, and patient support systems. Models developed from clinical trial data are further biased by rigid exclusionary criteria, limiting generalizability across a heterogeneous population experiencing DoC. This is a particular issue for machine learning or other artificial intelligence approaches because algorithms created with biased data will produce biased results [118].
-
Validation: Prognostic tools need to be developed by using a sufficient volume of clinical information to ensure statistically robust models with adequate discrimination. Accomplishing this requires a comprehensive set of clinical data from both the acute- and post-acute-care environments across multiple centers. Coma science should practically focus on identifying predictors that incrementally enhance the prognostic accuracy of validated models.
-
Calibration: To overcome uncertainty within these models, prognostic tools require careful calibration [119], a measure of how closely observed outcome occurs in relation to model predictions. A systematic review of prediction models for outcome following TBI highlighted the substantial variability in reported measures of calibration, with nearly half of all models lacking calibration statistics [119]. Clinicians should be careful to avoid models for which calibration is not known, and coma science research will require transparency in both model validation and calibration.
-
Versioning: As the clinical field of coma science evolves, models may not account for updated treatment approaches (e.g., targeted temperature management) and subsequently may generate inaccurate or unreliable predictions. In this context, approaches should consider models that are adaptive to new data in real time and can continuously update their prognostic modeling [120].
-
Patient-centered outcomes: Coma research will benefit from moving away from using dichotomized outcomes as an end point [121]. Instead, future studies should examine more granular outcomes focused on measures that are important to patients and based ideally on each individual’s values and preferences. Patients’ perspectives are known to change before and after disability, resulting in a “disability paradox” [122]. Despite a high rate of functional disability, most patients treated with craniotomy for ischemic stroke reported being satisfied with life [123].
-
Demographic factors, such as age and sex, need to be taken into account for prognostic modeling.
-
Comorbidities: Medical comorbidities and their severity are critical determinants of patient outcome and should be taken into account in prognostic tools [124‐126]. Premorbid cognition, mental health, and personality are under-ascertained yet may similarly contribute to patient-centered outcomes [127, 128]. Coma science research will benefit from refining methods to comprehensively evaluate the patient’s comorbid status.
-
Early WLST: Methods to reliably ascertain and statistically address WLST in patients with DoC are important. Efforts to study outcomes in populations with high mortality [129] should include considerations for when death results from medical decision-making. Additional considerations include protocolized approaches to limit inappropriately early WLST within the context of clinical study design to avoid confounding results or designing studies for populations within cultures or societies that do not practice WLST to the same degree.
-
Communicating outcome: Clinicians face a schism between their implicit conceptualization of good outcome and the definitions brought by the patients and their families [108]. Clinicians may bring their personal biases to their perspective on what is considered a good outcome for patients [130, 131]. To avoid this bias, clinicians and researchers must increase their own awareness of the discordance between what clinicians think families should receive in terms of information and support and what families actual need [114]. Numerous factors may contribute to this discordance in communication and decision-making [114, 132], and the existing gaps in communication cannot be filled until clinicians have a better understanding of current prognostic communication practices. Clinicians should encourage families and surrogates to express a patient’s own values and preferences and then modify their prognostic communication on the basis of these [133].
Limitations/Challenges
-
Population-based data: Current prognostic tools are based on large aggregated patient cohorts and not on individualized measures. Hence, outcome assessments historically have lacked consideration of patient-specific outcomes of interest. Although population-based data are needed for statistical power, unmet challenges include curating comprehensive clinical data anchored by CDEs and a better understanding of biologically relevant and patient-centered end points.
-
Bias: Unmeasured bias in clinician to decision-maker relationships and communication may confound predictions. As new prognostic tools are developed, researchers need to avoid introducing personal or data biases during the design process. A critical source of bias that touches both limitations may occur during communication between clinicians and patient surrogates. New tools are needed to harmonize prognostic communication to ensure that patients’ cultural and social desires are considered during assessment and treatment decision-making [133, 134]. Prognostic tools are meant to serve not only clinicians but also families, who must make decisions about the continued care of their loved ones [133]. One qualitative study of patients with TBI found that patient surrogates preferred numeric prognostic probabilities, whereas physicians tended to provide qualitative prognosis, in part, because of their underestimation of families’ ability to understand them [114]. Adding to this complexity are unspoken frustrations surrogates feel about uncertainty and unintentional perceptions of certainty created by physicians [114]. For patients with DoC, this is particularly important, and the best methods to reduce the bias introduced by how prognostic communication is approached are currently unknown.
-
Communication: On the basis of a policy statement from leading critical care societies, communication with surrogates and families should not be limited to a single meeting [133]. Instead, clinicians should build trust and form relationships through frequent contact with families, helping them understand the evolving condition of the patient and discussing both treatment options and potential outcomes [133, 135]. Meeting regularly with families may be difficult for clinicians who are already overburdened in their tasks and overstretched in their time. However, this open communication is recommended by critical care societies and the Institute of Medicine [133, 136]. The clinician–family relationship should include the option for families to obtain second opinions from unbiased clinicians. Another key element of open communication, commonly requested by families, is clinician humility in admitting what information is uncertain about a patient’s prognosis or treatment [114, 137]. Finally, clinicians should strive to be open with families—not only about treatment but also about any issues (such as cultural differences, past experiences, racism, and more) that may complicate communication itself [138].
-
Nonmedical factors in clinical care: Clinicians are not ideally equipped to determine nonmedical aspects of prognosis, which include social support, spiritual beliefs, and socioeconomic factors. An interprofessional approach (including social workers and pastoral care) has been recognized as a key component in providing high-quality critical care to complex and diverse patients [139‐141]. Institutional support for these aspects of care are often limited, with care burden falling to patients and their families. The process of organ transplantation includes a framework for considering the nonmedical aspects of care. [142]. The evaluation of psychosocial support structures available to transplant patients has been well established. By contrast, there is little understanding of the structures of support for patients with acute brain injury [143].
-
Nonmedical factors in coma science: coma science similarly should seek to better quantify social support, spiritual and religious influences, personal priorities, social and cultural values, and socioeconomic status when considering prognosis.
Deliverables
Long-Term Recovery
Background
Main Research Gap (Current State)
Overall Goal
Panel Discussion of the Approach
-
Combating pessimism: A challenge of long-term recovery is promoting optimism in the clinical team responsible for caring for the patient. The neurocritical care team should partner with other disciplines, including neurorehabilitation specialists and social workers. These partnerships extend the breadth of knowledge necessary to ensure accurate diagnosis, improve outcome prediction and prognostic counseling, identify short- and long-term care needs, and establish comprehensive treatment regimens. This approach fosters a more thorough understanding of the patient’s condition and the probability of further recovery, which can boost the optimism of the neurocritical care team and sustain high engagement in care.
-
Longitudinal outcome data: Investigations of long-term outcome following severe brain injury have historically employed 6- or 12-month study end points. In 1994, the authors of a major review [149, 155] were able to find outcome data later than 12 months in less than 50 patients with TBI. The challenges associated with obtaining long-term outcome data from patients with DoC make it difficult to build large data sets necessary for tracking the natural history of recovery. Point-of-care or other carefully designed longitudinal studies are needed to acquire long-term data, and study end points should exceed 12 months because there is growing evidence that meaningful recovery can continue for at least 10 years [153].
-
Conducting pragmatic clinical trials that investigate the context of care delivery (i.e., inpatient rehabilitation facility vs. home vs. nursing home facility) on long-term recovery of patients with DoC.
-
Large multicenter clinical trials: The field needs large RCTs to assess any of the proposed mechanisms of treatment for both short- and long-term outcomes. However, large trials alone will not solve a number of structural data insufficiencies, such as survivor bias, diversity of injury etiologies, and the many aspects of social determinants of health. These can be addressed by using strategies that may include oversampling for underrepresented groups, provision of rehabilitation for all enrolled, and standardized acute- and post-acute-care treatment models. Recruiting large samples of patients with DoC will require a network of sites for recruitment, especially to ensure a large enough population to apply endotypes.
-
Ethical questions of equity in clinical trial participation of patients with DoC, autonomy of decision-making, and the implications of CMD need to be discussed, particularly if long-term care is required.
-
Impact of cultural perspectives on neurorecovery and disability: The DoC patient population is culturally diverse, which means that patients and surrogates have distinct sets of values, customs, and cultures and a variety of perspectives on health, wellness, life, and disability. These beliefs impact multiple aspects of care, including access to services, treatment opportunities, community engagement, and QOL. This work may require the assistance of a cultural representative who can support the conversation by translating a family’s needs to a clinician.
Limitations/Challenges
-
Early WLST in patients who would otherwise go on to attain significant recovery of function.
-
Discriminatory payer policies limit access to specialized post-acute-care facilities that have been shown to reduce complications in patients with DoC.
-
Variability in clinician knowledge and approach to communicating prognostic information may negatively impact caregiver understanding, further increasing the probability of early withdrawal of aggressive treatment.
-
Difficulty integrating data from different systems of care that are not comparable and do not permit data sharing.
-
Ethical implications of clinical trial participation for long-term recovery of DoC.
-
Limited and regionally variable psychosocial support for caregivers following discharge from the acute care setting. Some consumer-created, well-organized caregiver support and advocacy groups do exist; however, most resources are available for patients with TBI rather than for those with non-TBI. The impact of advocacy groups in promoting access to care has not been formally studied and remains unknown.
Deliverables
Large Data Sets
Background
Trial/database | University/sponsor | Description | Population | Sample size | Patients with coma | MRI | EEG | Behavioral assessment | Outcome scale | Outcome time points |
---|---|---|---|---|---|---|---|---|---|---|
University of California, San Francisco; NINDS/private | Multicenter observational | TBI | 3211 | 150–200 | Yes (2 weeks, 6 months) | No | CRS-R | GOS-E, DRS | 2 weeks; 3, 6, and 12 months; 7 years | |
European Union | Multicenter observational | TBI | 4509 (47% ICU) | 968 | Yes (subset) | No | GCS, CRS-R | GOS-E | 6 months (< 5% 2 years) | |
ProReTro Database [172] | French Ministry of Health, France | Multicenter observational | Acute brain injury | 310 | 310 | Yes (subset) | Yes (n = 270) | GCS | GOS, mini mental state | 1 months |
Assistance Publique–Hôpitaux de Paris, France | Multicenter observational | CA, ICH, SAH, TBI | 218 | 218 | Yes | No | Not answering simple orders | GOS-E | 1 years | |
RECONFIG [175] | Columbia University, NINDS | Multicenter observational | ICH | 120 | 120 | Yes (acute) | Yes | CRS-R | GOS-E, mRS, neuro-QoL, TICS | 6 months, 3 and 5 years |
CONSCIOUSNESS [10] | Columbia University | Single-center observational | Acute brain injury | 150 | 150 | Yes (subset acute) | Yes | CRS-R | GOS-E, mRS | 3, 6, and 12 months; 3 and 5 years |
RESPONSE [176] | Massachusetts General Hospital | Single-center observational | TBI | 75 | 75 | Yes (acute, 6 months, 1 and 3 years) | Yes (acute, 6 months) | GCS, CRS-R | GOS-E Revised, DRS, BTACT | 6 months; 1, 3, and 5 years |
SAHIT [177] | Canadian Institutes for Health Research | Mixed (9 RCTs, 5 cohort studies) | SAH | 11,443 | 1250 | No | No | No | GOS-E, mRS | 1 years |
University of California, San Francisco, US Department of Defense | 8 studies | TBI | 6814 | Subset | Yes (subset) | No | CRS-R (TRACK-TBI only) | GOS-E, DRS | 1 years |
Main Research Gap
Panel Discussion of the Approach
-
To accomplish this goal, researchers must develop a database that incorporates the following: a high-throughput neuroinformatics platform [163] with real-time clinical and research input, automated imaging segmentation and analytical tools for quantification [164], remote computation and Web-based user interfaces, a biobank, long-term follow-up data, data sharing agreements, and an international open access governance structure.
-
Such a database would seek to capture data that cover the trajectory of the disorder. This scope may require enrolling healthy populations to track patients who experience DoC from initial onset to death in a longitudinal model similar to the Framingham Heart Study [165].
-
The database would collect clinically meaningful data at standardized intervals in real time and provide space for raw data (e.g., neuroimaging, neurophysiology, serological, pathological, multimodal). The associated biobank would collect clinically meaningful samples at standardized intervals using standardized practices (i.e., sampling, handling, processing) for delivery to a centralized repository with disaster-proof storage. The biobank may include a brain bank for whole-brain specimens.
-
The database platform should be automated for real-time data input into a structured data architecture. Automation requires no manual data entry; instead, a tool would retrieve information from standardized electronic health records. The database should require only periodic human maintenance, with manual validation of archived data to ensure reliability. Statistical control processes existing in other industries would be used to check data quality for completeness and consistency.
-
Its governance model, which would aim to promote equal stakeholder access, would include a central oversight board and incorporate FAIR (Findable, Accessible, Interoperable, Reusable) and open science principles. Any legal liability for using the database would lie with the end user, but oversight would discourage unethical use.
-
The database should be accessible through an open access, secure Internet-based interface that would be intuitive and user-friendly. The output should be flexible to allow for project-specific programming by the end user. In cases in which project-specific programs may prove useful to the community at large, the platform may also incorporate new technologies developed by individual users. A federated approach for data archival, analysis, and result sharing would be ideal.
-
Initial steps: The Curing Coma Campaign is not currently prepared to create such a complex system without a robust foundational framework. (1) To start, a deep and broad survey of the current database landscape is needed to provide a better understanding of what resources are available and which require creation. (2) In parallel, CDEs must be defined. (3) Once language, resources, wants, and needs are established, a simple database to obtain data from patients with coma can be built that begins standardizing data input and structure, types of patients and disorders covered, and current diagnostic and management strategies. (4) Once the initial database is established, existing frameworks could guide its expansion (e.g., the Alzheimer’s Disease Neuroimaging Initiative [166] and Transforming Research and Clinical Knowledge in Traumatic Brain Injury [167]) to include increasingly complex data and automated data input. (5) The final result will be a comprehensive international repository for research on coma and DoC. This ultimate version of the database must balance comprehensiveness and flexibility to incorporate future progress as the field advances its understanding.
Limitations/Challenges
-
Data elements need to be harmonized.
-
Long-term data capture will require infrastructures that may otherwise not be available globally.
-
The Curing Coma Campaign will need to create guidance for updating and upgrading both the technology and the clinical practices of global sites treating patients with coma.
-
Existing databases do not specifically address DoC or coma and will require careful data mining but may provide a launching point for the larger database. Thus, another approach is to start to collate existing neurological research databases.
-
Balancing simplicity to allow for global access and data entry, with complexity and depth of data required to address scientific questions: researchers will need to leverage innovative technology to create a self-maintained system that can retrieve data from electronic medical records and laboratory reports and make it accessible and available globally.
-
Lack of data sharing and master trial agreements may require Curing Coma Campaign researchers to assist sites with their creation to match the new technological requirements.
-
Obtaining funding for this enterprise will require government support and global lobbying.