Impact of Aphasia on Brain Activation to Motor Commands in Patients with Acute Intracerebral Hemorrhage

Abstract

Background:

Brain activation to motor commands is seen in 15% of clinically unresponsive patients with acute brain injury. This state called cognitive motor dissociation (CMD) is detectable by electroencephalogram (EEG) or functional magnetic resonance imaging, predicts long-term recovery, and is recommended by recent guidelines to support prognostication. However, false negative CMD results are a particular concern, and occult aphasia in clinically unresponsive patients may be a major factor. This study aimed to quantify the impact of aphasia on CMD testing.

Methods:

We prospectively studied 61 intensive care unit patients admitted with acute primary intracerebral hemorrhage (ICH) who had behavioral evidence of command following or were able to mimic motor commands. All patients underwent an EEG-based motor command paradigm used to detect CMD and comprehensive aphasia assessments. Logistic regression was used to identify predictors of brain activation, including aphasia types and associations with recovery of independence (Glasgow Outcome Scale-Extended score ≥ 4).

Results:

Of 61 patients, 50 completed aphasia and the EEG-based motor command paradigm. A total of 72% (n = 36) were diagnosed with aphasia. Patients with impaired comprehension (i.e., receptive or global aphasia) were less likely to show brain activation than those with intact comprehension (odds ratio [OR] 0.23 [95% confidence interval 0.05–0.89], p = 0.04). Brain activation was independently associated with Glasgow Outcome Scale-Extended 4 by 12 months (OR 2.4 [95% confidence interval 1.2–5.0], p = 0.01) accounting for the Functional Outcome in Patients with Primary ICH score (OR1.3 [95% confidence interval 1.0–1.8], p = 0.01).

Conclusions:

Brain activation to motor commands is four times less likely for patients with primary ICH with impaired comprehension. False negative results due to occult receptive aphasia need to be considered when intepreting CMD testing. Early detection of brain activation may help predict long-term recovery in conscious patients with ICH.

Introduction

Level of consciousness assessed by bedside clinical examination is an established predictor of recovery from brain injury [1, 2] and is widely implemented in prediction scores affecting major decisions about aggressiveness of care and withdrawal of life-sustaining therapies [3–8]. More recently, brain activation to motor commands has been detected in clinically unresponsive patients with brain injury using functional magnetic resonance imaging (MRI) [9, 10] and machine learning analysis of electroencephalogram (EEG) [11, 12]. This state, which is called cognitive motor dissociation (CMD) and has been demonstrated in 15% of clinically unresponsive patients with acute brain injury, has been linked to long-term recovery of independence [11, 13]. Clinical assessments and CMD detection are typically performed by asking patients to act out or imagine a motor movement (e.g., “keep moving your right hand”). Understanding such a command may be challenging for patients with acute brain injury in a distracting intensive care unit (ICU) context, and particularly for patients with impaired speech comprehension (e.g., receptive aphasia) [14]. Recent guidelines for the assessment of patients with disorders of consciousness from the American [15] and European [16] Academies of Neurology have endorsed advanced assessments for responsiveness such as CMD, if available. However, the practical and moral risk of erroneously failing to detect CMD is enormous if CMD testing is introduced into clinical practice [17]. The concern for false negative CMD results has been raised in studies that included healthy volunteers [17–19], but the risk of occult aphasia in behaviorally unresponsive patients is poorly understood, as highlighted in a recent gap analysis [20]. Here, we aimed to provide estimates of false negative CMD testing among patients with aphasic in the ICU to allow care providers to manage this risk.

Brain activation using CMD paradigms has been reported in a wide range of brain injuries, including trauma [9], cardiac arrest [18], and stroke [13]. Patients with stroke are at particularly high risk of aphasia, ranging from 19 to 62% for acute ischemic and 9% to 32% for acute hemorrhagic strokes [19]. Our study, which examines the relationship between aphasia and brain activation in patients with spontaneous ICH, departs from previous CMD studies [11, 13] in focusing on patients who are awake and able to follow commands or mimic. This is a necessity because clinically evaluating for aphasia is impossible in behaviorally unresponsive patients, and a diagnosis cannot be made based on imaging alone given that anatomical localizations of aphasia types are increasingly understood to vary [20]. Our objective was to quantify the risk of failing to show brain activation in the context of aphasia. To achieve this, we performed a prospective study of patients with acute primary ICH who were able to follow verbal commands or mimic to visual cues [21] within 48 h of bleeding. All patients underwent standardized aphasia testing and assessments using an EEG-based motor command paradigm used to detect CMD while in the ICU. We tested the hypothesis that patients with impaired comprehension (e.g., receptive or global aphasia) would be less likely to show brain activation to the motor command paradigm.

Methods

Study Design

From September 2019 to December 2023, we screened every patient admitted with a diagnosis of acute ICH to the neurological ICU at Columbia University Irving Medical Center in New York City. Inclusion criteria were the following: (1) diagnosis of primary ICH (e.g., hypertension, anticoagulation); (2) lobar (frontal, temporal, partial, or occipital) or subcortical (thalamic, capsular, or basal ganglia) location of ICH verified by computerize tomography and/or MRI; (3) age 18 years or older; (4) consciousness within 48 h of bleeding, defined as the ability to consistently follow a simple command (e.g., “stick out your tongue” or “show me two fingers”) and/or mimic the commanded action from visual cues (patients who made errors on some but not all of their responses were included); and (5) premorbid fluency in English or Spanish. Exclusion criteria were (1) secondary cause of ICH (e.g., trauma, neoplasm, vascular malformation, hemorrhagic transformation of ischemic stroke), (2) ICH location in the cerebellum or brainstem, (3) ongoing seizures, (4) severe cardiorespiratory compromise, (5) premorbid aphasia, (6) premorbid deafness, (7) pregnancy, (8) imprisonment, or (9) refusal to participate in the study on the part of the patient or health care agent. All patients meeting inclusion and exclusion criteria underwent an EEG-based motor command paradigm assessment developed to detect CMD [11] and formal aphasia testing on the same day and as soon as possible after admission to the ICU.

EEG-based motor command paradigm testing was performed using a standardized approach [11, 22]. Briefly, digital EEG was recorded using a standard 21-electrode montage, and commands were provided using single-use headphones (six blocks of eight commands each, duration approximately 25 min). Motor commands alternated between “keep opening and closing your right hand” and “stop opening and closing your right hand” for the right hand (as well as corresponding commands for the left hand). Digital EEG tracings was clipped (each into five 2 s long EEG segments following command presentations), and power spectral density in predefined frequency spectra (delta [1–3 Hz], theta [4–7 Hz], alpha [8–13 Hz], and beta [14–30 Hz]) were calculated for each recording electrode. EEG features were analyzed using a support vector machine learning algorithm with a linear kernel to identify systematic differences between clips following “keep …” and “stop …” commands. Performance of the support vector machine classifier was judged based on the area under the receiver operating characteristic curve. Significance testing involved a one-tailed permutation test (500 times randomly shuffled labels). Significance was set at area under the receiver operating characteristic curve > 0.5. Each patient only underwent one recording, so false discovery rate correction for multiple recordings was not required. All EEG analyses were performed with the use of open-source packages, including MNE-Python (www.martinos.org/mne/stable/index.html) and Scikitlearn (http://scikit-learn.org/stable/index.html).

Aphasia Testing

Aphasia testing was available in English and Spanish. Aphasia in English-speaking patients was diagnosed using three subtests of the Multilingual Aphasia Examination (MAE) [23, 24]—auditory comprehension, visual naming, and sentence repetition—together with the Western Aphasia Battery-Revised [25]. Aphasia in Spanish-speaking patients was diagnosed using the three MAE subtests alone (the Western Aphasia Battery-Revised is unavailable in Spanish). A team of researchers trained and supervised by a neuropsychologist (SA) administered the questionnaires at the bedside and recorded patients’ answers. The neuropsychologist reviewed the answers to determine which patients were aphasic and to classify patients by aphasia type (i.e., expressive, receptive, global, transcortical motor, transcortical sensory, transcortical mixed, conduction, or pure anomia) [26]. Second, each case of aphasia was rated as mild, moderate, or severe with respect to each affected domain—that is, comprehension, naming, and/or repetition. Ratings were based on a percentile rank [27] determined by comparison with a normative group of patients with aphasia described in the MAE manual [24]. Third, to capture subtle language changes beneath the threshold of clinical aphasia, the same scale was used to determine percentile ranks (although not mild, moderate, or severe rating) for patients without aphasia and for domains in patients with aphasia that were not frankly impaired.

For purposes of primary analysis, patients with aphasia were grouped into those with impairment of language comprehension (i.e., receptive aphasia, global aphasia, transcortical sensory aphasia, or transcortical mixed aphasia) and those without (i.e., all other types of aphasia). Ambiguous cases were jointly adjudicated by the neuropsychologist and principal investigator (JC). The research team, the neuropsychologist, and the principal investigator were blinded to the results of the EEG-based motor command paradigm during data collection and data preparation. The neuropsychologist was masked to all demographic, clinical, imaging, and outcome data throughout the study.

Clinical and Outcome Measures

Basic demographic data were obtained directly from patients and families (e.g., primary language, level of education, and handedness). Admission head computerized tomography was used to obtain characteristics of each hemorrhage (e.g., laterality, volume, intraventricular extension). Outcome prediction scores (primary ICH score [8]; Functional Outcome in Patients with Primary ICH [FUNC] score [3]) were based on the ICU team’s initial calculation. Additional data were prospectively obtained from chart review including likely ICH etiology per SMASH-U [28] criteria, anxiolytic and sedative medications present during motor command paradigm testing, and attentiveness to the examiner during motor command paradigm testing. Outcomes were assessed at 3, 6, and 12 months after bleeding using the Glasgow Outcome Scale extended (GOS-E) [29] and modified Rankin Scale (mRS) [30]. Withdrawal of life-sustaining therapy and cause of death were recorded.

Statistical Analysis

Categorical variables are shown as counts (percentages), and continuous variables as means (standard deviations) or medians (interquartile ranges [IQRs]), as appropriate. Univariable associations were assessed with a Wilcoxon rank-sum test for quantitative variables or χ² test for qualitative variables. Variables found to be significantly associated with an outcome at the p < 0.05 level were then included in the multivariable analysis using logistic regression modeling. Statistical analyses were performed with R (version 4.0.3) statistical software.

Standard Protocol Approvals, Registrations, and Patient Consents

The protocol for the study was approved by the Institutional Review Board of Columbia University Irving Medical Center and registered as Recovery of Consciousness Following Intracerebral Hemorrhage (RECONFIG) under ClinicalTrials.gov identifier NCT03990558. Consent to participate in the study was obtained from all enrolled patients (or, in cases in which the patient lacked capacity to consent, from the patien’s health care agent). All elements of the study conformed to the World Medical Association Declaration of Helsinki. Enrolled patients, surrogates of the patients, and treating clinicians were blinded to the results of motor command paradigm testing throughout the hospital admission.

Results

Of 536 patients admitted for acute ICH who were screened (Supplemental Fig. 1), 61 met inclusion criteria and were enrolled in the present study (conscious arm of RECONFIG); another 54, who met these criteria but were unconscious within 48 h of bleeding, were enrolled in a sister study (unconscious arm of RECONFIG); the remaining 421 patients were excluded from both (Supplemental Table 1). Motor command paradigm and aphasia testing were successfully completed for 50 of the 61 enrolled patients. Of the remaining 11, 10 were too inattentive and/or uncooperative to complete aphasia testing but were still able to undergo motor command paradigm assessments. The family of one patient withdrew care after enrollment into the study and prior to completing motor command paradigm or aphasia testing. These 11 patients were excluded from the primary analyses because their aphasia diagnosis was uncertain.

In the remaining cohort (n = 50), the mean age was 64.2 ± 15.6 years, and 32% (n = 16) were women (Table 1). A total of 88% (n = 44) were right-handed, and 36% (n = 18) were primarily Spanish-speaking. The median primary ICH score was 1.0 (IQR 1.0–1.75), and the median FUNC score was 9 (IQR 8–10). ICH was < 30 cm³ in volume for 72% (n = 36) of patients and on the right side of the brain for 58% (n = 29). Deep hemorrhages (n = 28) were slightly more common than lobar (n = 20), and most deep hemorrhages were thalamic (n = 15; Supplemental Table 2).

Table 1.

Rates of brain activation to an EEG-based CMD paradigm in conscious patients with ICH (N = 50)

	Total (N = 50)	Activation present (N = 17)	Activation absent (N = 43)	OR (95%-CI)	P
Demographics
Age	64.2 ± 15.6	63.3 ± 11.8	64.6 ± 17.1	0.82 (0.24–2.79)	N.S
Sex, female	16 (32)	3 (20)	13 (37.1)	2.36 (0.61–11.79)	N.S
Primary language					N.S
English	31 (62)	9 (60)	22 (62.9)	Ref
Spanish	18 (36)	6 (40)	12 (34.3)	1.22 (0.33–4.2)
Highest level of education					N.S
Did not graduate high school	13 (30)	3 (20)	10 (28.6)	0.36 (0.02–4.64)
Graduated high school graduate	15 (26)	3 (20)	12 (34.3)	0.3 (0.01–6.3)
Graduated college	19 (38)	9 (60)	10 (28.6)	Ref
Unknown	3 (6)	0	3 (8.57)	0.15 (0.02–1.22)
Right handedness	44 (88)	14 (93.3)	30 (85.7)	2.3 (0.33–46.93)	N.S
ICH characteristics
Volume of ICH					N.S
< 30	36 (72)	12 (80)	24 (68.6)	Ref
30–60	11 (22)	3 (20)	8 (22.9)	0.8 (0.45–1.42)
> 60	3 (6.0)	0	3 (8.57)	0.28 (0.06–1.16)
Laterality of ICH					N.S
R brain	29 (58)	9 (60)	20 (57.1)	Ref
L brain	21 (42)	6 (40)	15 (42.9)	0.88 (0.25–3.02)
ICH etiology*					N.S
Hypertension	34 (68)	12 (80)	22 (62.9)	Ref
Amyloid	8 (16)	2 (13.3)	6 (17.1)	0.69 (0.29–1.61)
Medication	3 (6.0)	0	3 (8.57)	0.25 (0.05–1.18)
Unclassifiable/Others	5 (10.0)	1 (6.67)	4 (11.4)	0.6 (0.23–1.52)
Admission
Primary ICH score	1 [1–1.75]	1 [0–1]	1 [1, 2]	0.62 (0.12–2.5)	N.S
FUNC score	9 [8–10]	10 [9, 10]	9 [8–10]	3.3 (0.98–12.99)	N.S
Confounders
Sedative	6 (12)	1 (6.67)	5 (14.3)	0.42 (0.02–3)	N.S
Inattention/Noncooperation	18 (36)	3 (20)	15 (42.9)	0.33 (0.06–1.27)	N.S
CMD assessment
Days from ICH onset	2 ± 2.58	2.0 ± 1.63	2.0 ± 2.88	0.92 (0.27–3.1)	N.S
Clinical responsiveness					N.S
Command following	42 (84)	14 (93.3)	28 (80)	Ref
No command following but mimicking	8 (16)	1 (6.67)	7 (20)	0.28 (0–1.83)
Aphasia diagnosis
Any aphasia					N.S
Not present	14 (28)	6 (40)	8 (22.9)	Ref
Present	36 (72)	9 (60)	27 (77.1)	0.44 (0.11–1.66)
Global or receptive aphasia					0.04
Not present**	29 (48)	12 (80)	17 (48.6)	Ref
Present	21 (42)	3 (20)	18 (51.4)	0.23 (0.05–0.89)

^*More than one hemorrhage location was seen in a number of patients; etiology was assigned according to SMASH-U criteria

^**Includes patients without aphasia and any aphasia that does not have a receptive deficit; Statistical comparison based on univariable analysis. Data reported as N (%), mean +/− SD, or median (IQR)

A majority of patients were diagnosed with aphasia (36 of 50, 72%), and 58% of patients with aphasia had aphasia types affecting language comprehension (receptive or global aphasia, 21 of 36; Fig. 1). Motor command paradigm testing revealed brain activation in 19% (3 of 16) of patients with global aphasia, 27% (3 of 11) of patients with expressive aphasia, all with conduction aphasia (n = 3), none with receptive aphasia (n = 5), none with pure anomia (n = 1), and 43% (6 of 14) of those with no aphasia. Patients with impaired comprehension (i.e., receptive or global aphasia) were less likely to show brain activation than those with intact comprehension (odds ratio [OR] 0.23 [95% confidence interval 0.05–0.89], p = 0.04; Table 1). In contrast, brain activation was not associated with demographic characteristics (age, sex, primary language, education, or handedness), ICH characteristics (volume, laterality, or etiology), outcome prediction scores (FUNC or ICH), or potential clinical confounders (sedative/anxiolytic medication or inattention/noncooperation at the time of motor command paradigm testing). Patients with a higher percentile rank for impaired comprehension showed a trend toward lower rates of brain activation (p = 0.07, OR 3.1 [95% confidence interval 0.7–15.8]), whereas higher percentile ranks for impaired naming (p = 0.18, OR 1.75 [95% confidence interval 0.42–7.43]) and impaired repetition (p = 0.61, OR 0.9 [95% confidence interval 0.18–4.34]) did not (Supplemental Figs. 2 and 3). Aphasia and motor command paradigm testing were performed on the same day in all patients.

Fig. 1.

Rates of brain activation to an EEG-based motor command brain activation paradigm by aphasia diagnosis. Distribution of aphasia types (outer circle) and brain activation rates for each aphasia type (inner circle, striped sections)

A total of 72% (n = 36) of patients had recovery to a GOS-E score of 4 or more (i.e., ability to remain at home alone for at least 8 h), and 72% (n = 36) of patients had recovery from disability and handicap (mRS ≤ 3) within 12 months of bleeding. Brain activation was associated with recovery of independence (GOS-E ≥ 4) by 12 months (OR 2.7 [95% confidence interval 1.3–5.5], p = 0.004). The association between brain activation and GOS-E score ≥ 4 remained significant (OR 2.4 [95% confidence interval 1.2–5], p = 0.01) when accounting for FUNC score (OR 1.3 [95% confidence interval 1–1.8], p = 0.01). Brain activation was associated with recovery from disability and handicap (mRS ≤ 3) by 12 months (OR 2.3 [95% confidence interval 1.1–4.8], p = 0.01). There was a trend for brain activation to predict recovery defined as mRS ≤ 3 (OR 1.9 [0.9–4], p = 0.07) when modeled together with FUNC score (OR 1.4 [95% confidence interval 1–1.8], p = 0.01). Among the subgroup of 21 patients with global/receptive aphasia, all 3 patients with brain activation had a GOS-E score ≥ 4 (GOS-E 6, 7, and 8; median GOS-E with brain activation was 7 [IQR 7–7] vs. 4 [IQR 3–4] for those without, p = 0.002), and 3 patients had an mRS ≤ 3 (mRS 0, 1, and 2; median mRS with brain activation was 1 [IQR 1–1] vs. 3 [IQR 3–4] for those without, p = 0.003).

Discussion

Our study demonstrates that brain activation assessed with an EEG-based motor command paradigm is less common in patients with impaired language comprehension (i.e., receptive or global aphasia). The paradigm used for this study is identical to the one used to detect CMD [9, 11, 13, 22] in behaviorally unresponsive patients. This failure of brain activation in the context of receptive or global aphasia needs to be taken into account when considering CMD in making prognostic statements. CMD paradigms can detect correlates of consciousness in clinically unresponsive patients, and these correlates have been linked to long-term outcomes [11, 13]. As CMD becomes more widely used in prognosis [20] and given the challenges of diagnosing aphasia in unconscious patients, occult receptive aphasia must be considered as a possible source of false negative results [14, 17]. To counter this risk of false negative results using language stimuli, additional paradigms including nonlanguage-based stimuli should be explored and taken into account when implementing CMD testing in clinical practice.

The patient cohort was limited to those with primary ICH to increase homogeneity of the patient population, as underlying pathologies in patients with secondary ICH could impact outcomes, applicability of prediction scales (i.e., FUNC score), and imaging and electrophysiological predictors. We decided to include patients if they correctly followed a command at least once in response to verbal prompting or mimicking. However, patients with ICH may have ideomotor apraxia with or without aphasia [31], making it challenging to consistently and reliably follow commands. The rate of aphasia (76%) is quite high in our study and may in part be a result of focusing on patients with lobar or subcortical ICH location as well as detecting even mild cases of aphasia due to detailed aphasia testing by a neuropsychologist. None of the included patients had developmental disabilities.

In our cohort of 50 patients who underwent assessments with the EEG-based motor command paradigm and aphasia testing, 3 patients with global and none with receptive aphasia had brain activation. It will remain uncertain why those three patients demonstrated brain activation despite impaired comprehension but the severity of their comprehension impairment on formal testing did not differ from other patients with global aphasia. Of note, acute aphasia syndromes are not static [32], and many patients pass transiently through a state of global aphasia (the most common aphasia type in our study) to later settle into receptive or expressive aphasia syndromes [33]. Even though in our study aphasia and motor command paradigm testing were performed at the same time, we do not have follow-up aphasia testing available, and it is possible that, over time, those three patients with brain activation would have transitioned into aphasia syndromes with intact language comprehension. Even though we don’t have aphasia follow-up testing available for any of the patients, these three patients raise the question of whether full language comprehension is required for demonstrating brain activation using the motor command paradigm that is used for CMD detection.

Secondarily, our study also implicates brain activation to a motor command as a possible predictor of recovery in conscious patients with ICH. Unconscious patients with brain activation to a CMD paradigm are known to have a higher rate of recovery of independence by 1 year after injury (44% vs. 14%) [11], and brain activation in unconscious patients remains an independent predictor of time to recovery of independence after accounting for age, injury mechanism, and baseline function [13]. Our study shows for the first time that even conscious patients with brain activation to a CMD paradigm have a higher rate of recovery. The significance of this observation is uncertain, however, because the association remained significant after accounting for established prediction scores of functional outcomes (i.e., FUNC score). Prediction scores for the recovery after ICH have limited precision [6, 34, 35], and future studies of conscious patients with ICH may consider more comprehensively exploring the prognostic relevance of assessing brain responses to standardized stimuli (such as the motor activation paradigm).

It is important to recognize that the motor command paradigm assesses willful modulation of brain activity. However, unaltered consciousness in a healthy individual and, in a broader sense, intact cognition entails much more than ability to pass the motor command paradigm test. Fascinating questions are raised, including the nature of consciousness and cognition in patients with severe language impairments or cognitive decline, all of which are highly relevant in the context of CMD [36] but are beyond the scope of this article. Assessing the nature of cognition in a patient with CMD will only be possible if highly functioning brain computer interfaces can be developed [37]. For the purposes of this study, it is most important to recognize that, in behaviorally unresponsive patients, language comprehension must be considered as a potential confounder in patients for whom a CMD paradigm generates a negative result.

Additionally, several limitations deserve mention. First, a reliable clinical diagnosis of aphasia in behaviorally unconscious patients is not possible because patient participation with the testing battery is required. Therefore, we have to extrapolate the impact of aphasia from conscious to unconscious patients. Second, we only enrolled patients who spoke English or Spanish, so conclusions beyond these populations should be made with caution. Third, to create a homogenous patient cohort, we restricted enrollment to patients with supratentorial primary ICH, limiting applicability of our results to patients with infratentorial or secondary ICH. Fourth, our overall sample size is small and limits our ability to determine brain activation rates and predictors in less common aphasia subtypes (e.g., transcortical aphasia). Only 6 of 14 conscious patients with acute brain injury without aphasia demonstrated brain activation to the motor command paradigm. Although prior studies did not investigate brain activation in this patient population, rates using an EEG-based motor imagery paradigm in healthy volunteers range between 55 and 80% [12, 17, 38]. Only one small study investigated rates of brain activation using the motor activation paradigm and found that all ten healthy volunteers had evidence of brain activation [11]. Of note, prior to the present investigation, no study determined rates of brain activation of conscious patients with acute brain injury in the ICU context. Fifth, patients with akinetic mutism or abulia could have a similar clinical presentation to patients with aphasia but in this study are unlikely to play a major role because patients were required to demonstrate some command following to verbal or visual prompts. Lastly, we did not repeat aphasia testing on follow-up. Larger multicenter studies should more comprehensively assess the relationship between brain activation and aphasia. Functional MRI studies may provide insights into the spatial distribution of brain activation to CMD testing paradigms [10].

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1007/s12028-024-02086-z.

Data Availability

The data on which this study is based on can be made available to other investigators attempting to replicate the results, subject to legal and ethical constraints on the disclosure of personal health information.