By Alan Z. Segal, MD
Associate Professor of Clinical Neurology, Weill Cornell Medical College
SYNOPSIS: The COVID-19 pandemic presented us with an unprecedented number of critically ill patients with coma. These investigators determined that the degree of hypoxemia determined the depth and duration of coma, but recovery was much better than expected and could be delayed by several weeks.
SOURCE: Waldrop G, Safavynia SA, Barra ME, et al. Prolonged unconsciousness is common in COVID-19 and associated with hypoxemia. Ann Neurol 2022;91:740-755.
Neurological prognostication in comatose patients may be clouded by the “self-fulfilling prophecy” in which the withdrawal of life-sustaining treatments may negatively skew outcomes and portend a more dire prognosis than justified. The “self-fulfilling prophecy” most prominently is a factor in the prognostication of patients with anoxic-ischemic brain damage caused by cardiac arrest. Therapeutic nihilism has come under challenge, however, particularly in the so-called “hypothermia era,” as cooling and rewarming have produced an extension of assessment timelines. Of note, it recently was reported by Schiff and colleagues at Weill Cornell that a small subset of cardiac arrest patients (n = 3) could emerge from coma more than two weeks following their insult.1
In the current report, Nicholas Schiff at Weill Cornell, along with senior authors Jan Classen at Columbia and Brian Edlow at Massachusetts General Hospital, examined a cohort of patients with a different etiology for coma: hypoxic damage related to adult respiratory distress syndrome (ARDS) caused by COVID-19. The overall conclusion was that coma in COVID-19 is driven by hypoxia and that coma recovery can occur weeks after the cessation of mechanical ventilation.
Cases (n = 795) were reviewed from the three centers: Weill Cornell (n = 254), Columbia (n = 333), and Massachusetts General Hospital (n = 208). Gender and ethnic breakdown of the patients were representative of the population disproportionately affected by COVID-19: males (68%) and Hispanics (30%). This was influenced by the heavily Hispanic neighborhood of Washington Heights, where Columbia is located. Coma recovery was determined to have occurred if the patient regained the ability to follow commands (according to the motor score of 6 on the Glasgow Coma Scale [GCS]).
Two hypoxemia thresholds were used: PaO2 ≤ 55 mmHg, which is the lower limit of normal according to “ARDSNet” guidelines, and ≤ 70 mmHg, which is a more commonly used clinical target. Hypoxia was quantified according to the number of days when at least one PaO2 value below these thresholds was recorded. Overall, 571 patients recovered, with 336 (64%) regaining the ability to follow commands in a delayed fashion after extubation. These included 25% who emerged from coma ≥ 10 days after extubation and 10% who woke ≥ 23 days after cessation of mechanical ventilation. The median time to recovery for following commands was 30 days (interquartile range, 27-32 days).
Time to recovery was influenced directly by severity and duration of hypoxia. This interval was extended by nine days for patients with one day of PaO2 ≤ 55 mmHg and by 21 days for patients with two or more days with PaO2 ≤ 55 mmHg. The analysis considered multiple confounding variables that included: anemic hypoxia (hemoglobin ≤ 7 mg/dL) and ischemic hypoxia (mean arterial pressure ≤ 65 mmHg), as well as renal failure, hepatic failure, and CO2 narcosis. Adjustments were made for sedatives, which included a variety of benzodiazepines, opiates, dexmedetomidine, and ketamine.
Structural brain injury, such as an acute stroke or hemorrhage diagnosed on computed tomography (CT) or magnetic resonance imaging (MRI), did not have a measurable effect on these results. The effects of hypoxemia on recovery of consciousness persisted even when patients with radiological abnormalities were excluded. Also, the study included analysis of a subsequent smaller cohort of patients from the “second wave,” which closely paralleled the primary “first wave” study results.
COMMENTARY
These data would affect one of the most common neurology consults requested in critically ill patients — failure to awaken after sedation is withdrawn. The implication of this study is that persistent coma may be reversible because of neurons “hibernating” in a state of profound dysfunction — damaged, but not dead. As the authors noted, the pandemic created an “unprecedented population of comatose patients with a common underlying condition.” They further implied that these data could inform decision-making in other similar scenarios, most notably cardiac arrest. But the ARDS patients affected by COVID-19 can be distinguished from cardiac arrest survivors — the most common distinction being the trajectory of hypoxia, which in ARDS is partial and intermittent, whereas in cardiac arrest, it is sudden and complete.
This study may lack generalizability since it was performed at high-level tertiary care centers. Although the authors characterized the study as “multicenter,” it is drawn from only two hospitals — New York Presbyterian Hospital and the Massachusetts General Hospital. A broader perspective (at least in New York City) would be to include hospitals closer to the community setting; for example, Elmhurst Hospital in Queens, a site run by New York City Health and Hospitals, which shouldered an overwhelming initial burden of COVID-19 cases.
As the authors are aware, their chosen outcome (a motor score of 6 on the GCS) was a practical tool to assess emergence from coma, since these values were more uniformly available (documented regularly by nursing) and have well-validated inter-rater reliability. Other assessments, such as verbal responses (oriented vs. confused) would not be possible in patients with tracheostomies, and eye opening can occur in the persistent vegetative state. Ability to follow commands, however, hardly is a surrogate for a favorable neurological outcome. Even when a patient can follow commands, there can be a broad range of ultimate recoveries, from limited minimal consciousness to a near or complete neurological resolution.
This investigation suggests that there is a physiological connection between severity of hypoxemia and longer recovery periods and particularly focuses on patients with “negative brain imaging.” However, little can be said about the utility or predictive value of brain imaging in this study. The manuscript does not specify whether these patients underwent CT or MRI scanning, but given the logistics involved (especially given the massive burden of critically ill patients far beyond the usual intensive care unit census), it could be assumed there was a dearth of MRI scans. Certainly, in cardiac arrest, there often is a major discrepancy between CT imaging (which can be reassuringly normal) and actual damage (which still can be very severe). In contrast, MRI with diffusion-weighted imaging is much more sensitive, with lesions in locations such as the cortical ribbon being an important complement to laboratory and clinical predictors.
Further physiological questions remain as to why these patients looked so poor neurologically, eventually to show significant improvement. Delayed clearance of medications or gradual resolution of other problems affecting mental status (such as improved renal or hepatic function) clearly play a role, but the phenomenon observed here appears to be more complex. Postulated mechanisms suggested by the authors to explain this phenomenon included “diffuse leukoencephalopathy, brainstem injury, and global changes in brain network connectivity.” Of note, given that COVID-19 was reported to produce white matter injury (including our own reports of posterior reversible encephalopathy syndrome [PRES]), it is possible that impairment in consciousness was produced less by irreversible ischemic damage to neuronal cell bodies and more by communication breakdown within the brain as the result of demyelination of axonal tracts.
REFERENCE
- Forgacs PB, Devinsky O, Schiff ND. Independent functional outcomes after prolonged coma following cardiac arrest: A mechanistic hypothesis. Ann Neurol 2020;87:618-632.
