ѻý

Brain Activity in 'Unresponsive' Patients May Predict Recovery

<ѻý class="mpt-content-deck">— EEG study points to possible prognostic tool early after acute brain injury
Last Updated June 28, 2019
MedpageToday
An unconscious woman in the hospital being comforted by her partner.

About one in six clinically unresponsive ICU patients showed electroencephalography (EEG) patterns of brain activity when spoken to soon after acute brain injury, a single-center study showed.

In 16 of 104 (15%) unresponsive patients, a machine-learning algorithm that analyzed EEG recordings detected brain activation following researchers' verbal commands a median of 4 days after injury, according to Jan Claassen, MD, of Columbia University in New York City, and colleagues.

Half of these 16 patients improved to the point that they were physically able to follow commands a median of 6 days later, the researchers reported in the . The injured patients who showed early brain activity were four times more likely to achieve partial independence at 12 months than similar patients with no activity, they added.

If confirmed, these findings "could inform prognostication of acute brain injury and potentially provide a means of communication with patients who seem unresponsive on the basis of a conventional clinical examination," wrote David Menon, MD, PhD, and Srivas Chennu, PhD, both of the University of Cambridge in England, in an .

Not being able to follow commands early after an acute, severe brain injury traditionally has been thought to mean either that damage is too severe for consciousness to return or that the ability to follow commands and other signs of consciousness may recover over time as injury and sedative effects fade, Menon and Chennu noted.

"However, a third possibility exists: there is activation of cerebral neural circuits for perception of commands that is not accompanied by corresponding motor responses," they wrote.

This phenomenon, known as cognitive–motor dissociation, is seen in about , with fMRI or EEG indicating cerebral activation during motor imagery tasks in patients who demonstrated no motor responses to commands. Other studies in chronic disorders of consciousness have shown a related phenomenon known as "covert cognition," when brain activation occurs in unresponsive patients during cognitive tasks that do not require the patient to attempt to move.

In this study, Claassen and colleagues looked for cognitive-motor dissociation in patients shortly after brain injury, since the absence of an ability to follow commands at that point may affect decisions about withdrawing life-sustaining therapies.

They identified 104 patients with acute brain injury from 2014 to 2107 who were unable to follow spoken commands, excluding patients with seizures, hyperglycemia, abnormal sodium, and renal or fulminant liver failure. Before each EEG assessment, they rated each patient on the 23-point Coma Recovery Scale–Revised (CRS-R), and classified sedation as minimal, low, or moderate. If deemed safe, sedation was reduced for an EEG assessment.

The researchers acquired a total of 240 EEG recordings while issuing verbal commands such as "open and close your hand." During half (52%) of the recordings patients were comatose; 54 recordings occurred (22%) while patients were in a vegetative state, and 60 (25%) while patients were in a minimally conscious state–minus category.

The machine-learning algorithm (support vector machine) detected cognitive–motor dissociation on at least one recording a median of 4 days after ICU admission in 16 of 104 (15%) patients. These 16 patients had injury from subarachnoid hemorrhage (five people), traumatic brain injury (three), intracerebral hemorrhage (four), cardiac arrest (two), neurosarcoidosis (one), and bupropion overdose (one).

Eight of these 16 patients (50%) and 23 of 88 patients (26%) without brain activity improved to the point where they were able to follow commands before discharge. At 12 months, 7 of 16 patients (44%) with brain activity and 12 of 84 patients (14%) without brain activity had a Glasgow Outcome Scale–Extended level of 4 or higher, indicating they were able to function independently for 8 hours (OR 4.6; 95% CI 1.2-17.1).

Six patients with cognitive–motor dissociation (38%) and 50 without (60%) were dead at 12 months. Four of six patients with cognitive–motor dissociation who died did so after life-sustaining therapy was withdrawn.

While some responses were inconsistent, the findings are intriguing, especially since EEG is more easily used and more widely available than fMRI in the ICU, Menon and Chennu wrote. The verbal commands used were similar to ones used in bedside clinical examination to determine whether a patient is unresponsive, they pointed out: "It is noteworthy that [the researchers'] methods were designed to reduce the risk of false discovery of cognitive–motor dissociation."

But questions remain and replicating this study is essential, they said. "Future studies should explore spatial patterns of EEG activation, quantify false positive detection rates of the EEG classifier, understand the basis of inconsistent responses within patients, and provide details and evaluation of the specifics of support vector algorithms," Menon and Chennu wrote. It's also unclear whether sedation or arousal played a role in the outcomes.

"A better understanding of the neural substrates for cognitive–motor dissociation could identify neurotransmitter systems as targets to restore behavioral responsiveness," they added.

Disclosures

This study was supported by the Dana Foundation and the James S. McDonnell Foundation.

Primary Source

New England Journal of Medicine

Claassen J, et al "Detection of Brain Activation in Unresponsive Patients with Acute Brain Injury" N Engl J Med 2019; DOI: 10.1056/NEJMoa1812757.

Secondary Source

New England Journal of Medicine

Menon D, Chennu S "Inverting the Turing Test -- Machine Learning to Detect Cognition in the ICU" N Engl J Med 2019; DOI: 10.1056/NEJMe1906061.