Background
History and modern science have taught us that patients who appear unconscious may, in fact, not be unaware. One such example from recent history is that of Saul Bellow, a noble laureate who was assumed to be comatose during hospitalization, yet he was able to provide a stream of conscious narrative from his recollection during the period of presumed comatose state [
50,
51]. In a 2006 case report by Owen et al. [
7], the investigators used fMRI to demonstrate preserved conscious awareness in a patient who was behaviorally determined to be in a vegetative state. In this study, when the patient was asked to imagine playing tennis or moving around in her home, she was able to activate predicted cortical areas in a manner indistinguishable from that of healthy volunteers [
7]. In 2011, Bardin et al. [
52] also demonstrated dissociation between observed behavior and cognitive functioning noticed on fMRI. Subsequently, in 2015, Schiff et al. [
13] coined the phrase cognitive motor dissociation to describe the general discordance between phenotypic observation at the bedside and the actual state of awareness. In 2017, Edlow et al. [
11] reported that covert consciousness is not only present in patients with chronic DoC but also noted in patients with acute brain injury (ABI) in the ICU by using task-based fMRI and EEG techniques. We now know that about 15% of comatose patients in the ICU have covert consciousness and tend to have improved functional outcome at 1 year [
16]. Ironically, what science and neuroimaging has come to reveal recently, Saul Bellow had already conveyed to us through his words on a memoir of his illness [
53]. Although there is a culture of hegemony of science over humanities, it is important to acknowledge that all the things we face today from climate change and global pandemic to the discovery related to DoC, inevitably involves humanities along with science. Therefore, insights from history, philosophy and literature often provide promissory insights and ignoring such important perspectives could only be considered as wasted opportunities leading to delay in scientific discoveries.
Neuroprognostication refers to an understanding of the most likely prognostic trajectory in patients with neurologic injury by means of biomarkers, neuroimaging, clinical examination and other variables. Accurate neuroprognostication is important for effective communication with the surrogate decision-makers in terms of communicating the potential for recovery, likelihood of disability and the possibility of awakening from coma. During the 1st NIH symposium, five gaps in prognostication were identified: (1) end points for prognostic assessment have not been defined, (2) we do not have standard methods to reliably ascertain and statistically address withdrawal of life supporting therapy (WOLST), (3) we need to standardize and improve diagnostic evaluations and establish the right time window for neuroprognostication, (4) we need prediction tools that use comprehensive clinical information to create statistically appropriate models, and (5) prognostic communication and its impact on families’ decision-making in DOC is understudied.
The existing neuroprognostication research structure of DoC is fragmented, lacks the infrastructure to support collaboration, and lacks prognostic accuracy given categorization of disease based on overt presentation rather than mechanistic understandings. Existing limitations are related to the use of convenience samples with small sample size, lack of blinding and lack of external validation. The biggest limitation, however, is related to early WOLST. There is a need to develop consensus on how to study coma, either by DoC phenotype (coma/stupor, etc.), primary injury type (CA, SAH, ICH, traumatic brain injury [TBI], etc.) or neuroanatomic injury subtypes (brainstem vs. diffuse cortical injury vs. subcortical injury). Neuroprognostication parameters need to be approached as index tests based on relevant neurologic function that are directly related to the functional outcomes and affect the quality of life [
54]. There is also a need to understand the role of cortical plasticity, especially in the pediatric population, as return to baseline function may not indicate adequate recovery, especially when developmental milestones are considered over a long recovery period. Given these gaps, the neuroprognostication WG used the American Heart Association (AHA) scientific statement [
54] as their template for developing recommendations regarding prognostic indicators. To develop CDEs for high-quality prognostication, the WG proposed that we first develop understanding/consensus on several aspects surrounding prognostication, which include the following:
1.
Prognostic markers should not just be available but also applicable and feasible across various clinical settings and disease mechanisms. We need accurate, precise, and clinically applicable tests that can serve most patients after onset of coma.
2.
Timing of prognostication is extremely important. Current clinical practice recommendations suggest delaying prognosis for cardiac CA until 72 h and at least 2–3 days for other patients with ABI. However, data suggest that delayed awakening beyond 72 h [
55] can occur in up to 32% of patients with CA [
54]. There is an urgent need to move away from early prognostic dichotomization to guide interventions. We also need to determine the timing of completion of index test to adequately predict long-term prognosis. In this regard, the WG recommends allowing at least 30 days for neuroprognostication, whenever possible. Use of parcellated approach with stepwise evaluation of milestones throughout the course of recovery is encouraged. Further, the WG proposes delaying cognitive testing for at least 90 days, as acute encephalopathy post hospitalization may significantly alter findings.
3.
As coma recovery is a dynamic process, prognostic markers, specific to disease phenotypes should either have the flexibility to be applied across the DoC trajectory or we need to develop multiple tests specific to the stage of coma.
4.
Communication strategies congruent with the values and preferences of patients and surrogates should be developed.
The World Health Organization international classification of functioning, disability, and health includes three components: (1) impairments, (2) functional ability, and (3) engagement in activity and participation. A variety of functional and other outcome measures should be included with granularity in prognostic research. Further, to address the numerous confounders of death reporting (e.g., WOLST, extracerebral organ failure), reporting should also include the mode of death, timing from injury, extent of support at the time of death, reason for WOLST and reason for primary injury at the time of presentation (if applicable).
Despite the widespread nihilism, particularly in DoC related to CA, available evidence suggests that patients with coma continue to improve post discharge and outcomes continue to evolve over 1–6 months after injury [
56,
57]. Patients ideally need to be out of the hospital to assess outcomes related to complex functional activities. Clinicians need to ensure that the acute symptoms have stabilized and that noncerebral confounders are minimized at the time of prognostication. Efforts should also be made to improve messaging in neuroprognostication guidelines to emphasize prognostic uncertainly given the current lack of definitive evidence. It is very important to avoid all sources of bias and continue full medical support until the right time of prognostic assessment as early DNR orders have been known to result in care limitations [
58]. We need a cultural shift to avoid poorly calibrated heuristics or nihilistic opinions by practitioners and work toward developing strict protocols for WOLST with focus on patient and family centric outcome measures. Additionally, as patients may be affected by cognitive, psychological, or physiological impairments, all factors need careful consideration in overall prognostication. The health-related quality of life reporting system should include reported outcomes from both patients and surrogates. Reporting scales should be pragmatic and easy to use and there is a need to work toward validating health-related quality of life for surrogate decision-makers of comatose patients. We should also ensure that all neuroprognostication studies follow the recommended standards of reporting.
In-depth understanding of limitations regarding available evidence for neuroprognostication will guide the design of future studies. Efforts should be focused on developing models with accurate calibration using large amount of clinical data. There are several existing candidate tests and data which need to be evaluated simultaneously with an aim toward development of multivariable models to allow incremental accuracy of neuroprognostication. To develop such models, we need a large multicenter prospective observational coma registry that includes key prognostic data including diverse etiologies of DoC with serial testing over time. Important variables to consider include the following: (1) clinical examination including phenotypes of DoC (2) somatosensory evoked potential (SSEP) to assess thalamocortical connectivity, (3) serological/cerebrospinal fluid/chemical/genetic biomarkers, (4) Functional neuroimaging pertaining to stages of coma, (5) EEG with quantified parameters, and (6) brain-computer interface and other novel technologies. Additionally, inclusion of baseline variables (age, race, ethnicity, zip code, comorbidities, lifestyle, frailty, social support network, socioeconomic status, etc.), intensity of management (acute and rehab phase) and inclusion of pediatric population is important. Also, there is a need for innovative methods of analysis to assess predictors from early to late phases of DoC. Accurate modeling of prognosis and disease trajectory for personalized prognostication could benefit from advanced statistical methods and novel tools such as ML technology for data assimilation. In such models, external validation of the outcome metrics in a large independent cohort is crucial.
Panel Discussion of the Approach
The panel discussed that there is a general sense of nihilism among clinical practitioners and one of the goals of CCC should be to help communicate the complexity of DoC prognosis and help develop an understanding of the confidence interval for neuroprognostication at various stages of DoC. A parcellated prognostication approach with stepwise milestones throughout the course of recovery would be more appropriate than a dichotomized prognostication of favorable versus unfavorable outcome. Panel reiterated the complexity of neuroprognostication in the pediatric population due to the continued neurodevelopment through the course of recovery and highlighted the need for robust longitudinal studies on recovery trajectory and prognostic approaches on all age groups.
The panel also discussed that one of the issues with clinical trials on neuroprognostication is the ethical aspect of continuation of life supporting therapy especially if it does not concur with the values of the patient and surrogates. Such patients and families should be excluded from clinical trials on prognostication. Widespread education endeavors are needed to change the culture of nihilism among providers and families. However, a trial that requires prolongation of DNR or WOLST wait times would require an in-depth discussion with the families in addition to detailed informed consent. Data from countries with cultures that support prolongation of life supporting measures would be helpful in determining the natural history of DoC. Within the United States, one of the limiting factors might also be related to the DNR orders in which some clinicians and families may perceive this to mean “do not provide aggressive medical therapy,” even though the intended use of this order is only for the clinical management of one specific clinical scenario (i.e., CA). The panel highlighted the need to develop protocols surrounding WOLST decisions and address confounders (e.g., sedative medications, systemic injury, etc.) that impact outcomes. Further, the panel discussed the need to identify key elements and strategies for communicating uncertainty in a tactful, emotionally supportive and culturally competent manner followed by verification of surrogates’ understanding of outcome communication.