Slow but Evident Recovery from Neocortical Dysfunction and Cognitive Impairment in a Series of Chronic COVID-19 Patients

Visual Abstract


INTRODUCTION
As the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic proceeds, neurocognitive long-term consequences are frequently observed (1): follow-up investigations of coronavirus disease-19 (COVID-19) patients two to four months after symptom onset report, among others, impaired memory (20-34%) (2,3), disturbed concentration (20-40%) (2,3) and cognitive problems (36%) (4). These cognitive deficits in the chronic stage are now frequently referred to as 'Long-COVIDsyndrome'. Recently, we described impairment of frontoparietal cognitive functions accompanied by frontoparietal dominant cortical hypometabolism on 18 F-FDG PET (as an established marker of neuronal function) in a relevant subset of subacute COVID-19 patients initially requiring inpatient treatment for non-neurological complications (5). By comparison of 18 F-FDG PET scans of those subacute COVID-19 inpatients to a control sample using voxel-wise principal components analysis, we established a COVID-19-related spatial covariance pattern, the expression of which was tightly correlated with performance in the Montreal Cognitive Assessment (MoCA) (6). Frontal and, to a lesser extent, temporoparietal cortical hypometabolism, which improved during follow-up at 1 and 6 months, were also confirmed as major findings in the acute phase of COVID-19-related encephalopathy by a recent study of Kas et al. (7).
Deviating from aforementioned results, Guedj et al. (8) reported a profile of hypometabolism in limbic/paralimbic regions extended to the brainstem and cerebellum in patients with 'Long-COVID' examined at about 3 months after symptom onset. Against this background, we investigated whether the frontoparietal hypometabolism might be a biological fingerprint of 'Long-COVID-syndrome' neurocognitive impairments. We re-assessed 18 F-FDG PET and MoCA performance in eight patients presenting for a follow-up in the chronic stage approximately six months after symptom onset.

Standard Protocol Approvals, Registrations, and Patient Consents
Patients were part of a prospective monocentric register (Neuro-COVID-19). The local ethics committee approved this study (EK 211/20) and all subjects gave written informed consent in accordance with the Declaration of Helsinki.

Study Design, Participants and Assessment of Cognitive Functions
The register enrolled patients with reverse transcription polymerase chain reaction-confirmed SARS-CoV-2 infection, at least one novel neurological symptom developed under COVID-19 and required inpatient treatment in the Department of Internal Medicine of the University Hospital Freiburg between April 20, 2020 and June 10, 2020 (for details see Hosp and colleagues (5)). During the acute stage a total of 31 subacute COVID-19 patients were assessed for impaired cognitive functions with the MoCA (German, version 7.1) (6). 17 of these patients had undergone an 18 F-FDG PET examination at the subacute stage. Of those, eight patients underwent a second examination with 18 F-FDG PET and MoCA (German, alternative version 7.2) at the chronic stage of the disease and were included in this study. Eight patients refused further investigations (no more self-perceived complaints: n = 6; long traveling distance: n = 1; bad physical condition: n = 1) and one patient died.
Of note, the present population does not represent a so-called Long-COVID (typically defined by long-lasting, not exclusively neurological complains at least 4 to 12 weeks after symptom onset (9)) since patients were enrolled into this prospective study based on at least one new neurological symptom at the subacute stage (on average 37 ± 19 days after COVID-19 symptom onset) and then followed-up to investigate the reversibility of symptoms. In fact, 4/8 (50%) patients did not have any more self-reported cognitive deficits at the time of the second examination ( Table 1).

F-FDG PET Imaging
PET emission data were acquired on a fully digital Vereos PET/CT scanner (Philips Healthcare, The Netherlands) 50 minutes after injection of 213 ± 9 MBq 18 F-FDG for 10 minutes. 18 F-FDG PET scans were spatially normalized to an in-house 18 F-FDG PET template in Montreal Neurologic Institute space, followed by a smoothing with an isotropic Gaussian kernel of 10 mm full width at half maximum. The topographic profile rating algorithm (10) was employed to derive each individual's pattern expression score (PES) of the previously established COVID-19-related spatial covariance pattern (5). For additional conventional analysis, a paired t test between the 18 F-FDG PET scans at the subacute and chronic stages was calculated after proportional scaling of individual voxel-wise 18 F-FDG uptake to white matter (given the obvious involvement of grey matter shown in previous study (5)). Voxel-wise two-sample t test was also applied to the 18 F-FDG PET scans of COVID-19 patients at the chronic stage compared to control cohort (n = 45 age-matched control patients in whom a somatic CNS disease was carefully excluded, see previous report (5)) in order to explore whether hypometabolism still remains at the chronic stage of disease. All processing steps were implemented with an in-house pipeline in MATLAB (The MathWorks, Inc., Natick, Massachusetts, United States) and Statistical Parametric Mapping (SPM12) software (www.fil.ion.ac.uk/spm).

Statistical Analysis
Significance of differences between subacute and chronic stages for MoCA test scores and PES of COVID-19-related covariance pattern was assessed with paired t tests. Student's one-tailed t test was applied to test whether the PES in COVID-19 patients at the subacute stage was still significantly higher compared to the control cohort. Cohen's d was calculated for all pairwise comparisons. Strength of relationship between MoCA and PES of COVID-19-related covariance pattern was estimated within the stages by linear regression and across stages with a repeated measures correlation test (11). Statistical analyses were performed using R (https://www.R-project.org/).

Data Availability
The data generated and analyzed during the current study are not publicly available but could possibly be provided by the corresponding author on reasonable request and upon approval of the local ethics committee.

RESULTS
Demographics and patient characteristics are listed in Table 1. The eight patients presenting for a follow-up examination were not distinct from the rest of cohort at baseline examination (PES of COVID-19-related covariance pattern on 18 F-FDG PET, MoCA, and age were not significantly different between groups, all p > 0.1). MoCA-performance significantly improved over time from a mean (± standard deviation) global score of 19.1 ± 4.5 (maximum 30 points) at the subacute stage to 23.4 ± 3.6 at the chronic stage (d = 0.97, p = 0.03, table 1), which is however still below the frequently used cut-off value for detection of cognitive impairment (<26/30). Five of eight patients still were below this threshold (6). MoCA domain scores showed that orientation and attention were almost unimpaired at the chronic stage, but revealed persistent deficits in visuoconstructive and executive functions and, especially, memory ( Table 1). As previously shown (5), the PES of the COVID-19-related pattern (Fig. 1, panel A) in the subacute COVID-19 patients was significantly higher compared to the control cohort (44.3 vs. -11.3; d = 1.84, p = 6×10 -6 ). COVID-19 patients had significantly lower mean PES at the chronic than at the subacute stage (6.8 ± 32.6 vs. 44.3 ± 33.1; d = 1.06, p = 0.002), although still at trendlevel higher in comparison to the control cohort (6.8 vs. -11.3; d = 0.60, p = 0.06) (Fig. 1, panel B).
Exploratory correlation analysis revealed a significant relationship between cognitive assessment (MoCA global score adjusted for YoE) and PET (R 2 = 0.39, p = 0.01; i.e., lower PES was associated with better cognitive performance) over the subacute and chronic stages (Fig. 1, panel C). Moreover, changes in cognition (MoCA) seem to be associated with change in PES, which, however, failed to attain statistical

DISCUSSION
In the present follow-up study, we demonstrate essential reversibility of decreased neocortical glucose metabolism assessed by 18 F-FDG PET accompanied by an improvement of cognitive functions in COVID-19 patients from the subacute to the chronic stage after a SARS-CoV-2 infection. The expression of the previously established COVID-19-related spatial covariance pattern at the chronic stage was significantly reduced compared to the subacute stage. However, in comparison to a control cohort, chronic COVID-19 patients still exhibited a slightly higher pattern expression (at trend-level) and residual hypometabolism indicating a shift towards normal levels, but no definite return. Although we observed a significant improvement in the cognitive screening test (MoCA), the average performance was still within the range of mild cognitive impairment (6). This slow but evident recovery provides fundamental and novel insights into the pathophysiology of cognitive deficits associated with COVID-19.
Recent neuropathological examinations of patients who died from COVID-19 due to nonneurological causes shed light on potential pathophysiological mechanisms underlying the sustained cortical hypometabolism and impairment (12): SARS-CoV-2 RNA or proteins could be detected in 53% of patients with a predominance for caudal brainstem and cranial nerves highlighting the known neuroinvasive propensity of human Beta-Coronavirus clades (13,14). However, major histopathological findings were astrogliosis, microglia activation and mild infiltration by cytotoxic T lymphocytes with an emphasis on brainstem and cerebellum (12) that were unrelated to the presence of SARS-CoV-2.
Therefore, these changes are more likely caused by a systemic inflammatory response or cytokine release (15). As the cortical grey matter is largely spared from damage and inflammatory changes (5,12), the reduced glucose metabolism is likely secondary, e.g. as a consequence of a functional decoupling from aminergic brainstem nuclei (16,17). This inflammation-trigged process could have outlasted the acute infection and only partly recovered over the contemplated period of six month. Of note, changes in cognition (MoCA) seem to be associated with change in PES although this observation failed to attain statistical significance (r = -0.54, p = 0.16). We did not observe an association between changes of PES and MoCA on one hand and time to follow-up examinations on the other. This may also be due to the limited number of subjects and relatively narrow and late time range of follow-up examinations. Still, the regions with residual hypometabolism are those with the most prominent decreases during the acute stage and may thus take a longer period to fully recover. Consequently, the slow reversibility of post-COVID-19 cerebral hypometabolism and cognitive impairment described in the present study would be in accordance with a lasting perturbation of cortical function caused by a subcortical peri-inflammatory process as correlate of the 'Long-COVID-syndrome'.
The comparison of the present study to other studies employing 18 F-FDG PET for assessment of COVID-19-associated metabolic deficits is hampered by various factors. For instance, Guedj et al. (8) reported a cohort of patients examined at highly variable time points (about 1 to 5 months after COVID-19, on average 96 ± 31 days). Given the apparent time dependency of cognitive and metabolic changes, such pooling of patients at presumably different stages precludes a comparison to studies of selected time points like ours. Moreover, given the obvious alterations in cortical metabolism observed in our cohort and the study by Kas et al. (7), the use of cortical regions for count rate normalization of PET data appears problematic. In fact, this is why we selected an approach (i.e., principal components analysis) that does not require an a priori definition of a reference region. Such factors (among others) might led to discordant results and contra-intuitive findings like decreasing glucose metabolism with longer time after first COVID-19 symptoms (8), which is in contrary to both, the study from Kas et al. (7)  Finally, Kas et al. (7) also conducted an assessment by a comprehensive cognitive test battery but no direct association to PET data was reported.
A limitation of the present study is the small sample size with only eight out of 17 initial patients receiving a follow-up 18 F-FDG PET and MoCA examination. Obviously, patients with actual cognitive complaints are more likely to adhere to a follow-up program than subjectively healthy patients. In fact, six patients declined follow-up with reference to the lack of self-perceived complaints, whereas only four out of eight patients denied cognitive impairments in our actual sample. However, this "selection bias" is inherently linked to the major finding of the present study of slow but evident recovery of cognitive impairment. In addition, the selection of the initial cohort (only inpatient, but not dominant ICU treatment) limits generalizability of our findings (especially to outpatients, representing the majority of COVID-19 patients). Furthermore, there may be premorbid conditions or risk factors rendering subgroups of patients particularly susceptible for COVID-19-associated cognitive impairments. No obvious factors were identified in the present prospective single-center study (including the initial sample) (5). However, future larger, population-based studies are needed to address this question.

CONCLUSION
Given the current pandemic situation and still tremendous uncertainty concerning the long-term sequelae of COVID-19, the present study provides novel insights of highest medical and socioeconomic relevance. We provide evidence of longer lasting metabolic and accompanying cognitive deficits after COVID-19. Although a significant recovery of regional neuronal function and cognition can be clearly stated, residuals are still measurable in some patients six months after manifestation of COVID-19. In consequence, post-COVID-19 patients with persistent cognitive complaints should be presented to a neurologist and possibly allocated to cognitive rehabilitation programs.

DISCLOSURE
The authors declare they did not receive any funding for this study. PTM received honoraria from GE (presentation, consultancy) and Philips (presentation). All declared interests are outside of the submitted work. of the COVID-19-related spatial covariance pattern is lower in COVID-19 patients at the chronic stage compared to the subacute stage, but still at trend-level higher compared to the control cohort. Boxplots (grey) as well as individual values for COVID-19 patients (colored) and the control cohort (grey) are displayed. Repeated measures for each patient are connected by the line. *** p < 0.001 (two-sample t test; see previous study (5)); ** p < 0.005 (two-tailed paired t test); § p = 0.06 (one-tailed two-sample t test). C: Association between the PES and the Montreal Cognitive Assessment score adjusted for years of education. Each dot represents an individual patient's data; the lines (shaded areas) correspond to the fit of a linear regression (95% confidence interval) for each disease stage separately (p = 0.07 and 0.12 for the subacute and chronic stage, respectively). Repeated-measures R 2 and P value represent the correlation between variables with both stages pooled.