Cognitive impairment after recovery from COVID-19: Frequency, profile, and relationships with clinical and laboratory indices

Cognitive impairment (CI) is regarded as a remarkable burden in COVID-19 survivors. Its prevalence and profile, and relationships with the disease clinical and laboratory indices, remain unclear. The present study investigated, in a large sample of patients recovered from COVID-19, the frequency of CI with both a face-to-face screening tool and comprehensive test battery (MCCB). The study also evaluated the profile of CI and its relationships with COVID-19 clinical and laboratory indices and with psychopathological features. Out of 1344 subjects assessed for eligibility, 736 completed the screening phase 11 months after the COVID-19 infection; 402 participated in the baseline phase and completed an in depth cognitive, clinical and laboratory assessment about one month later. More than one third of the screened subjects presented a CI (COG+); it was associated to age, education, male gender, COVID-19 severity, and presence of anosmia, dyspnea at rest and exertional dyspnea during the acute phase. COG+ subjects showed a higher severity of depression, anxiety and post-traumatic distress, and worse global functioning, than subjects without CI. The MCCB showed that 45% of the subjects had a CI involving attention, working memory, verbal learning, visual learning, and reasoning and problem solving. Finally, neurocognitive functioning was inversely correlated with LDH blood levels, a potential biomarker of disease severity. According to our findings, cognitive functioning should be routinely and periodically assessed in COVID-19 patients, especially in older subjects, who experienced more severe COVID-19 symptoms. In case of persisting dysfunctions cognitive training programs should be considered as treatment strategies.

Cognitive impairment (CI) is regarded as one of the most burdensome long-term consequences of COVID-19 survivors (Ritchie and Chan, 2021).However, despite the large number of studies on this topic, results about its prevalence and characterization are inconsistent, partly due the complexity of the disease clinical presentation, and partly due to study design limitations (Deng et al., 2021;Hassan et al., 2022;Hornsey, 2022;Perrottelli et al., 2022;Tzur Bitan et al., 2022).A recent meta-analysis, including over 13.000 subjects, showed that 3 months after the infection over 20% of the patients were cognitively impaired when assessed with either subjective/self-report assessment tools or objective tests; when considering only studies with an objective evaluation, a greater proportion of individuals (36%) presented a CI (Ceban et al., 2022).Most studies employing an objective, standardized evaluation used screening tools, such as the Montreal Cognitive Assessment (MoCA) or the Mini Mental State Examination, instead of comprehensive test batteries, and could only report a global score, instead of a detailed CI profile (Ceban et al., 2022;Crivelli et al., 2022aCrivelli et al., , 2022b;;Perrottelli et al., 2022).
Study sample sizes were often limited (Biagianti et al., 2022;Ceban et al., 2022;Crivelli et al., 2022aCrivelli et al., , 2022b;;Perrottelli et al., 2022).To our knowledge, only two studies included more than 800 subjects; both assessed subjects few months after the infection with a screening tool, the MoCA, that revealed a mild CI in 15% (Evans et al., 2021) and 25% of the sample (Hartung et al., 2022).Studies performing an extensive cognitive evaluation often involved less than 100 participants (Almeria et al., 2020;Cecchetti et al., 2022;García-Sánchez et al., 2022).Becker et al. (2021) assessed executive functioning, processing speed, memory and verbal learning in 740 subjects and found that 15-20% were impaired at least in one cognitive domain several months after the infection.However, subjects were tested online rather than during in-person evaluations, making the comparison with other studies difficult.
The great variability of the time interval (from one month to 24 months) between the infection and the cognitive assessment represents another potential reason for discrepancies (Almeria et al., 2020;Becker et al., 2021;Evans et al., 2021;Hartung et al., 2022;Mattioli et al., 2022;Miskowiak et al., 2021;Voruz et al., 2022).Overall, available evidence suggests that CI has a high prevalence in COVID-19 patients and can persist long beyond resolution of the acute infection.However, the precise pathways leading to the development of COVID-related CI have not been identified yet, and studies aiming at improving our knowledge of the prevalence, profile, and clinical as well as biological correlates, may contribute to the identification of potential biomarkers to predict its onset and persistence.
The present study was designed to: 1) assess, by means of a screening neurocognitive test administered face-to-face, the frequency of CI in a large sample of patients who either had been hospitalized for symptomatic COVID-19 infection, or experienced fever for at least five days or a body temperature higher than 38.0 • C for at least three days, and were followed as outpatients at University COVID centers at the time of study recruitment; 2) evaluate the relationships between COVID-19 severity in the acute and long-term phases and CI at a later stage; 3) characterize the patterns of CI by administering face-to-face a comprehensive cognitive battery; 4) assess the relationships between CI and psychopathology and the impairment in different cognitive domains; 5) assess the neurological, physical and laboratory correlates (including peripheral markers of chronic inflammation) of the various domains of CI.

Participants
Participants were recruited at the infectious disease units of five Italian University Hospitals (University of Brescia; University of Campania Luigi Vanvitelli, Naples; University of Genoa; University of Rome "Tor Vergata"; and University of Salerno).The enrollment process started in March 2021 and ended in September 2022.
Inclusion criteria were: (a) age between 18 and 65 years; (b) a history of confirmed SARS-CoV-2 infection (i.e., positive RT-PCR test); (c) previous hospitalization due to COVID-19 disease or, if not hospitalized, having experienced either fever for at least five days or body temperature higher than 38.0 • C for at least three days; (d) being followed on an outpatient basis by one of the infectious disease units of the study centers; (e) absence of SARS-CoV-2 infection at the time of study recruitment; (f) willingness to sign the informed consent.Exclusion criteria were: (a) positive history of CI prior to COVID-19 infection, and (b) positive history of chronic psychosis, bipolar disorder, depressive major disorders, neurological disease, or head injury prior to COVID-19 infection.
The severity of COVID-19 disease at the time of symptomatic SARS-CoV-2 infection was assessed on a 1 to 4 scale (mild to critical) defined according to the US Centers for Disease Control and Prevention criteria (COVID-19 Treatment Guidelines Panel, 2023) (please refer to the Supplementary material for the description of the criteria).
The study was approved by the Ethics Committees of participating centers and was conducted in accordance with the principles of the Declaration of Helsinki (59th World Medical Association General Assembly; October 2008).

Assessment instruments
An ad hoc form was used to verify inclusion/exclusion criteria relevant to COVID-19 infection and to record information on the infection (e.g., date of the first appearance of symptoms, date and length of hospitalization, date of first COVID-19 negative test).Another ad hoc form was administered to verify the absence of cognitive impairment, psychiatric disorders, or any neurological disease before COVID-19 infection.
The COVID-19 screening form recorded the presence of COVID-19 related symptoms (fever, dysgeusia or ageusia, dysosmia or anosmia, arthromyalgia, dyspnea at rest and exertional dyspnea).The presence of CI (COG+) was assessed during the screening phase of the study using the MoCA, a 30-points test that, in approximately 15 minutes, evaluates executive functions, working memory, attention, language, abstraction, visuospatial ability, delayed recall, and orientation (Nasreddine et al., 2005).A total score below 26 indicates CI, with a one-point adjustment for subject with less of 12 years of education.
The Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognitive Battery (MCCB) was used at the baseline phase of the study to obtain a comprehensive assessment of cognition.It includes 10 tests that investigate seven cognitive domains (Speed of processing; Attention/vigilance; Working memory; Verbal learning; Visual learning; Reasoning and problem solving, and Social cognition) (Kern et al., 2008;Nuechterlein et al., 2008).Raw scores were converted to standardized T-scores, and corrected for age, gender and education using the Italian normative data.A Neurocognitive Composite T-score (using the six neurocognitive domains) was also obtained.
The Mini-International Neuropsychiatric Interview (MINI) 7.0.2(Italian version) was used to exclude psychiatric disorders before COVID-19 infection and detect the presence of one or more psychiatric diagnoses (PSY+), at study recruitment.If the MINI revealed the presence of a psychiatric disorder, the subject could be included in the study only if the onset of the identified disorder occurred after the COVID-19 infection (Sheehan et al., 1998).It is a structured interview covering the most common mental disorders according to the Diagnostic and Statistical Manual of Mental Disorders -Fifth Edition (DSM-5) and the International Classification of Diseases -Tenth Revision (ICD-10) (First et al., 2021).
The Hamilton Depression Rating Scale 17-item (HAM-D-17) was used to evaluate the severity of depressive symptoms for the week preceding the assessment (Hamilton, 1960).It is a 17 items scale (the higher the score, the more severe the depressive symptoms).The Hamilton Rating Scale for Anxiety (HAM-A) consists of 14 items and was used to assess the severity of both psychic and somatic anxiety symptoms (Hamilton, 1959).
The Impact of Event Scale -Revised (IES-R) (Weiss and Marmar, 1997) is a questionnaire measuring current subjective distress in response to a specific traumatic event.It includes 22 items, each rated on a 0-4 scale.A total score of 24 to 32 indicates mild psychological impact, 33 to 36 moderate, and from 37 on a severe impact.
The Insomnia Severity Index (ISI) (Bastien et al., 2001) is a questionnaire assessing the nature, severity, and impact of insomnia.It includes 7 items, each rated on a 0-4 scale .A total score of 8 to 14 indicates sub-threshold insomnia, 15 to 21 moderate, and 22 to 28 severe insomnia.
The Suicidal Ideation Attributes Scale (SIDAS) (van Spijker et al., 2014) is a questionnaire assessing suicidal ideation.It includes five items rated on a 0-10 scale.A global score of at least 21 indicates a high risk of suicidal behavior.
The Global Assessment of Functioning (GAF) scale assesses the individual's levels of social, occupational and psychological functioning (Endicott et al., 1976); scores are from 100 to 1, with scores above 60 indicating mild/no impairment.
The World Health Organization Quality of Life (WHOQOL-BREF) questionnaire evaluates the individual's perception of his/her health and well-being.It includes 26 items rated on a 1-5 scale (World Health Organization, 1996).The questionnaire covers four domains: physical health, psychological health, social relationships, and environment.

Laboratory assessments
The following laboratory assessments were performed: peripheral oxygen saturation (both at rest and after walking test), blood count parameters (white blood cells, lymphocytes and platelet), metabolic parameters (alanine aminotransferase, aspartate transferase, creatinine), cell damage and inflammatory markers such as lactate dehydrogenase, C-reactive protein, D-dimer, Interleukin-8 (IL-8), Tumor Necrosis Factor-α (TNF-α) and serum amyloid A (SAA) and levels of immunoglobulin M (IgM) and immunoglobulin G (IgG) antibodies against SARS-CoV-2.A detailed description of the procedures used to assess laboratory indices is reported in the Supplementary material section.

Screening phase
A comprehensive written explanation of the study procedures and objectives and a written informed consent form relevant to the participation in the study and personal data processing were provided to all subjects who met the inclusion/exclusion criteria reported above, as verified by two ad hoc forms.
The COVID-19 screening form was used to record the presence of COVID-19 related signs and symptoms and the severity of the disease during the acute phase.All patients who met inclusion and exclusion criteria participated in the face-to-face administration of the MoCA and the MINI.

Baseline phase
Subjects who completed the screening phase were invited for a baseline assessment visit, consisting of an in-depth investigation of cognitive functioning, psychiatric symptoms, infection sequelae, and laboratory tests.On the same day or at a variable time after the screening visit (mean ± s.d.= 39.2 ± 68.3 days after screening), subjects completed the baseline assessment.They were administered the comprehensive MCCB, the HAM-D-17, the HAM-A, the IES-R, the ISI, the SIDAS, the WHOQOL-BREF, and the GAF, and underwent a complete physical and neurological evaluation to assess the presence of COVID-19 related symptoms (fever, dysgeusia or ageusia, dysosmia or anosmia, arthromyalgia, dyspnea at rest, exertional dyspnea).They also underwent the laboratory assessments listed above.

Training of researchers
For each site, two researchers were trained for the psychopathological and psychosocial assessments, two for the cognitive assessment, and one for the collection of infection-related variables.To avoid halo effects, each researcher was trained for one area of assessment only.The inter-rater reliability was formally evaluated by Cohen's kappa for categorical variables, and intraclass correlation coefficient (ICC) for continuous variables.

Statistical analysis
Enrolled subjects were divided in two groups according to the presence (COG+) or absence (COG-) of CI.Demographic continuous variables were reported as mean ± standard deviation (SD), while categorical variables were reported as frequencies.
Comparisons between study groups on demographic variables, MoCA scores, severity of COVID-19 disease, length of hospitalization, time between COVID-19 symptoms appearance and date of the screening visit, as well as presence/absence of COVID-19 related symptoms during the acute phase, were performed through independent samples t-test and χ 2 tests.
Comparisons among the four groups on the same above-mentioned variables were performed through a one-way Analysis of Variance (ANOVA) and χ2 tests.Post-hoc analyses (Bonferroni's corrected) were performed when statistically significant group effects were detected.
To investigate the eventual influence of different psychiatric disorders on CI a one-way ANOVA was performed on MoCA total scores of the different diagnostic subgroups (major depressive disorders, anxiety disorders or comorbidity of the two), and a χ 2 test was conducted to evaluate differences in the frequency of CI among these subgroups.
Differences between COG+ and COG-in the presence/absence of COVID-19 related symptoms and laboratory tests at the baseline were calculated with χ2 tests and independent t-tests, while differences in neuropsychiatric symptoms, functioning and quality of life were calculated using independent samples t-tests.
A multivariate analysis of variance (MANOVA) was performed on MCCB neurocognitive composite score and individual domains T-scores, followed by post-hoc tests corrected for multiple comparison (p < 0.0062) to compare study subjects' performance with the Italian normative sample composed by 780 healthy subjects previously recruited for the Italian Network for Research on Psychosis study (Mucci et al., 2018).The severity of CI for each subject was calculated considering the magnitude of the deviation of T-scores from the normative mean in at least 2 domains: mild = 1 to 2 s.d.below the mean; moderate = 2 to 3 s.d.below the mean, and severe=3 or more s.d.below the mean.
The Pearson's correlation test was used to analyze the correlations of MCCB composite T-scores with laboratory tests indices and vital signs showing significant differences between COG+ and COG-subjects.Tests that showed significant associations were further investigated to identify which MCCB domain contributed to them.In addition, partial correlations were performed to exclude the influence of anxiety and depression on the above-mentioned correlations.
All analyses were performed using IBM SPSS, version 28.0.
Fig. 1.Study diagram displaying an overview of the included and excluded subjects.

Participants in the screening and baseline phases
One thousand three hundred forty-four subjects who had experienced COVID-19 disease and were followed-up at the outpatient units participating in the study were assessed for eligibility (Fig. 1).Fivehundred eighty-five subjects were not eligible (256 had an age > 65; 181 refused to participate in the study; 47 did not reach the severity threshold for COVID-19, and 103 had cognitive, psychiatric, or neurological conditions before the COVID-19 infection).Seven hundred and fifty-seven subjects met the inclusion criteria; 21 failed to complete the screening procedures, and 736 subjects completed them.The average time elapsed from COVID-19 infection until the screening visit was approximately 11 months (mean ± s.d.= 334.20 ± 255.54 days).
According to the MoCA results, 259 subjects (35.19%) presented a CI (COG+), and 61 (8.29%) also had a psychiatric disorder (COG+/PSY+).Among the COG+ subjects, 163 (22.15% of the whole sample) had a mild CI (MoCA 23 to 25) and 96 (13.04% of the whole sample) a moderate to severe one (MoCA score 22 or less).Four hundred and eleven subjects (55.84%) did not present either a psychiatric diagnosis or CI according to the MoCA and participated in further assessments as control subjects (COG-/PSY-).
Four-hundred and two subjects (181 COG+, 221 COG-) participated in the baseline phase of the study (Fig. 1).
A comparison between subjects who participated in the baseline phase of the study and those who dropped out after the screening phase revealed significant differences, after Bonferroni correction (p<0.001 for all comparisons): subjects who completed the baseline assessment were older, more frequently males, less educated, had lower MoCA total score and higher severity of the COVID-19 disease than those who dropped out.

Demographic and cognitive profiles
Differences in demographic characteristics between COG+ and COGgroups at the screening phase are reported in Table 1.COG+ were significantly older, less educated and had lower scores on the MoCA, as compared to COG-(p<0.001).They presented a higher male/female ratio (p<0.01)than COG-, while no difference was found in the length of hospitalization and time elapsed from the onset of COVID-19 symptoms to the screening visit (p>0.05).
The one-way ANOVA performed on the MoCA total scores revealed no difference among the main diagnostic subgroups (major depressive disorders, anxiety disorders or comorbidity of the two), indicating no significant influence of psychiatric diagnoses on the severity and frequency of CI (see supplementary tables S3 and S4).Therefore, in all remaining analyses the study sample was subdivided in COG+ or COGonly.

Group comparisons on COVID-19 disease severity and frequency of symptoms at the time of hospitalization or first visit
According to NIH criteria, 54% of the screened subjects had a mild illness at hospitalization or during the acute phase of the disorder, while 18%, 21.9% and 6.1% had moderate, severe and critical illness, respectively.
Three COVID-19 symptoms, i.e., dysosmia or anosmia, dyspnea at rest and exertional dyspnea, had a significantly higher frequency in the COG+ group as compared to the COG-group (p<0.008;p<0.001 and p < 0.001, respectively; Table 2).

Comparison between COG+ and COG-on the frequency of COVID-19 symptoms and laboratory tests results at the baseline assessment
The results of the χ2 tests showed that the frequency of some COVID-19 symptoms differed between the two study groups even when the acute phase of the disorder had ended: asthenia (p < 0.001) and exertional dyspnea (p < 0.001) at the baseline visit were more frequent in COG+ subjects than in the COG-ones (Table 3).
The results of the independent sample t-tests showed that COG+ subjects presented higher levels of blood oxygen saturation at rest and of LDH (p < 0.001), as compared to COG-subjects (Table S4).The values in brackets display the percentages of subjects in the corresponding group (COG+ / COG-) presenting that symptom at hospitalization/first outpatient visit.* p-value survived correction for multiple tests (p<0.0071):significant higher frequency in COG+ as compared to COG-.

Comparison between COG+ and COG-on neuropsychiatric symptoms, functioning and quality of life at the baseline assessment
COG+ subjects had a higher severity of depression and anxiety symptoms, higher levels of post-traumatic distress and worse global functioning than COG-subjects (p < 0.005).No difference was found between the two groups for insomnia, suicidal ideation, and quality of life domains (Table S5).

Comprehensive cognitive assessment at the baseline
According to the MCCB, 45% percent of the study sample (n = 182) was cognitively impaired as compared with the Italian normative sample.Thirty-three percent (n = 134) had a mild impairment (1-2 s.d.below norms in at least two domains), 11% (n = 45) had a moderate (2-3 s.d.below norms in at least two cognitive domains) and 1% (n = 3) a severe CI (i.e., 3-4 s.d.below norms in at least two domains) (Fig. 2).
The MANOVA showed a significant difference on the MCCB scores between the whole COVID-19 group assessed at the baseline (n = 402) and the MCCB Italian normative sample (n = 780) [F (9,1117) = 13.383,p < 0.001]: post-hoc comparisons showed that the COVID-19 subjects had significantly lower composite neurocognitive score (effect size = 0.45) and lower scores in the following five cognitive domains: attention/vigilance; working memory, verbal learning, visual learning and reasoning and problem solving (p < 0.01) (Table 4).No difference was detected for the domains Speed of Processing and Social Cognition (Table 4).The highest degree of impairment was observed for working memory (effect size = 0.42), verbal learning (effect size = 0.44) and visual learning (effect size = 0.47).
When comparing COG+ with COG-subjects and each COG group with the Italian normative sample a significant group effect emerged [F (16,2234) = 15.357,p < 0.001].The post-hoc analysis showed that the neurocognitive composite score and the scores of all cognitive domains, except social cognition, were significantly more impaired in the COG+ group as compared to the other two groups (p < 0.001); working memory, verbal learning and visual learning were the most impaired domains (Fig. 3).Differences between the COG+ and COG-groups remained significant after controlling for the effects of anxiety and depression.

Correlations between cognitive domains and infection-related variables
The MCCB neurocognitive composite scores were significantly correlated with the baseline oxygen saturation at rest (r = 0.174; p = 0.003) and the lactate dehydrogenase (LDH) blood levels (r = − 0.365; p < 0.001).However, only the inverse correlation between the LDH blood levels and the neurocognitive composite scores (the higher the LDH levels the poorer the neurocognitive performance) survived the correction for multiple tests and remained significant also after covarying for The values in brackets display the percentages of subjects in the corresponding group (COG+ / COG-) presenting that symptom at baseline assessment.
* p-value survived correction for multiple tests (p<0.0071):significant higher frequency in COG+ as compared to COG-.anxiety and depression scores.
The LDH blood levels showed an inverse significant correlation with each cognitive domain except attention/vigilance and social cognition domains (Table S6).Most of these correlations survived the correction for the anxiety and depression scores (Table S6).

Discussion
In a large sample of COVID-19 subjects, who had experienced COVID-19 illness about 11 months before the assessment, we found that 35.19% presented a CI according to the MoCA.This finding is close to the one reported by a recent meta-analysis (36%) (Ceban et al., 2022), and is in line with two other studies that included the MoCA assessment in more than 700 participants: the multicenter study conducted by Evans et al. (2021) reported that 16.9% of the subjects presented a MoCA score lower than 23, which is comparable to our 13% with a score below 23, and the study by Hartung et al. (2022) reported a CI in 27% of the sample, not far from our 31.2%,although only 6% of their study sample were hospitalized subjects, as compared to 72% of our sample.
Heterogeneity in findings relevant to the frequency of cognitive deficits in COVID-19 patients has often been attributed to the time between the infection and the cognitive assessment.In our study, the average time elapsed from the start of COVID-19 symptoms to the cognitive assessment was almost one year and did not differ between COG+ and COG-subjects.Our results do not support the assumption of a higher chance to identify a CI short after the start of the COVID-19 infection, and are in line with previous findings showing that these deficits can be recorded both few weeks and several months after the resolution of the infection (Biagianti et al., 2022;Ceban et al., 2022;Crivelli et al., 2022aCrivelli et al., , 2022b;;Perrottelli et al., 2022).As also noted by other investigators, this evidence suggests the opportunity to assess cognition periodically in subjects recovered from COVID-19 (Biagianti et al., 2022;Geddes, 2021;Miskowiak et al., 2021;Schomerus et al., 2021).Indeed, a recent meta-analysis showed that the prevalence of neurocognitive impairments could even increase from mid-to long-term follow-up evaluations (Premraj et al., 2022).
Among subjects who accepted to participate in the baseline assessment, and completed the MCCB, 45% performed at least 1 standard deviation below the Italian normative population in at least two cognitive domains.This might suggest a higher sensitivity of the comprehensive cognitive battery in the detection of cognitive dysfunctions, as compared to standardized screening tools such as the MoCA.However, MCCB completers had lower MoCA scores and higher severity of the COVID-19 disease as compared to those who left the study after the screening assessment.
According to our MCCB findings, almost all the assessed domains, except Speed of processing and Social cognition, were impaired in subjects affected by COVID-19 as compared to the Italian normative sample.The effect size of the deficit in the five domains significantly impaired ranged from small to medium, with working memory, verbal learning and visual learning showing the highest degree of impairment.
MCCB findings strongly suggest that, although subjects with CI assessed by MoCA (COG+) were older, less educated, and more frequently males (as also found by Amalakanti et al., 2021;Crivelli et al., 2022aCrivelli et al., , 2022b;;Evans et al., 2021;Hartung et al., 2022;Valdes et al., 2022), these features cannot fully explain the development of COVID-related CI as MCCB t-scores are corrected for age, gender and education.
Other studies, at odds with our findings, reported that processing speed was one of the cognitive domains most frequently impaired following the COVID-19 infection (Becker et al., 2021;Cecchetti et al., 2022;Ferrucci et al., 2022;Henneghan et al., 2022;Mazza et al., 2021;Santoyo-Mora et al., 2022;Vakani et al., 2023;Vannorsdall et al., 2022;Velichkovsky et al., 2023).The discrepancy with our findings might be explained by the different tests employed to evaluate the same cognitive domain.In fact, the MCCB includes the combination of three different tests to estimate the speed of processing: the Symbol-Coding test, the Category Fluency animal naming test, and the Trail Making Test-A (TMT-A).Previous studies reporting a deficit in this domain used different tests, such as the simple reaction time task (Santoyo-Mora et al., 2022;Vakani et al., 2023) or the Paced Auditory Serial Addition Test (PASAT) (Ferrucci et al., 2022), or one of the tests included in the MCCB for the evaluation of this domain, such as the TMT-A (Becker et al., 2021;Vannorsdall et al., 2022) or the symbol-coding test (Mazza et al., 2021).
According to our findings, COG+ subjects showed a higher severity of the COVID-19 disease during the acute phase, as compared to COG-.Previous research did not report univocal results on the relationship between severity of the acute clinical presentation and the onset of CI, and a recent meta-analysis found that illness severity and/or duration of hospitalization were not associated with the development of CI (Jacot de Alcântara et al., 2023).We might hypothesize that, more than the overall severity of the clinical picture, specific symptoms may drive the development of CI.In fact, we found that dysosmia or anosmia, dyspnea at rest, and exertional dyspnea during the acute infection phase were significantly more frequent in subjects who developed cognitive dysfunctions.An association between anosmia during the acute phase of the illness and cognitive deficit evaluated both at hospitalization (Pirker--Kees et al., 2021) and at different follow-up visits was previously reported (Cecchetti et al., 2022;Cristillo et al., 2021;Ferrucci et al., 2022;Llana et al., 2023).It has been linked to a direct damage of the olfactory tract and the entorhinal cortex, that are anatomically and functionally associated with the hippocampus and limbic cortical areas, considered as key regions for memory processes (Cecchetti et al., 2022;Ferrucci et al., 2022).However, when considering the COVID-19 symptoms assessed at the baseline evaluation, we found that the COG+ group, as compared to the COG-one, had a higher frequency of exertional dyspnea and asthenia, but no longer of dysosmia or anosmia, or dyspnea at rest.This result might be explained by the fact that exertional dyspnea, asthenia and CI tend to persist for a long time after the resolution of the infection (Chen et al., 2022;Fumagalli et al., 2022;Han et al., 2022), and are regarded as features of the long-term syndrome, often present simultaneously, possibly due to a shared pathophysiology.However, the study by Hartung et al. (Hartung et al., 2022), although confirming that exertional dyspnea, asthenia, and CI are the most persistent symptoms of COVID-19, failed to find an association among them and suggested that these symptoms are separate consequences of COVID-19.
COG+ and COG-subjects showed few significant differences on laboratory test results and vital signs: lower O2 saturation at rest and higher LDH levels in COG+ than in COG-.The findings confirm, once again, a more severe COVID-19 disease in subjects who develop a CI, but, in the absence of significant differences involving inflammatory indices, fail to provide a strong support to the hypothesis that an abnormal and prolonged inflammatory state following the infection might play a role in the onset of post-infection cognitive dysfunctions (Boldrini et al., 2021;Mazza et al., 2021;Zhou et al., 2020).
Our data did not show significant associations of COVID-related CI with physical and laboratory indices, except for a significant association between elevated blood levels of LDH and worse performance on speed of processing, visual learning, working memory, and reasoning and problem solving.An increase in LDH activity was reported in many cases of severe COVID-19.It is probably linked to cell damage, as well as impaired blood flow and oxygen delivery (Russo et al., 2023).Elevated concentrations of LDH were observed in patients with encephalitis, ischemic stroke, and head injuries (Valvona et al., 2016).The LDH increase in COVID-19 has been regarded as a potential biomarker/predictor of severe respiratory pathology, admission to ICU and death (Fialek et al., 2022;Terpos et al., 2020).Interestingly, one study reported that loss of gray matter volume in the right cingulate was associated to memory impairment and higher LDH values (Lu et al., 2020).This might suggest that the reduction in the volume of this cerebral area in patients with high LDH is the result of an atrophy due to a severe inflammatory response, which could also lead to the onset of cognitive dysfunctions (Penninx, 2021).In addition, a study focusing on CI after an acute ischemic stroke suggested that elevated levels of this biomarker are associated with the development of cognitive impairment because of neuronal damage following prolonged states of inflammation (Xu et al., 2022).Of course, studies including neuroimaging and inflammatory indices might contribute to validate these hypotheses in COVID-19 subjects with cognitive impairment.
We also investigated the associations of CI with psychopathology, psychosocial functioning, and quality of life.As reported in the literature, COVID-19 can cause the onset of several psychiatric symptoms, which can persist even after two years since the start of the infection (Greenberg and Rafferty, 2021;Holt-Lunstad, 2021;Lenze et al., 2022;Mei et al., 2021).In line with previous studies, we found an association of CI with the severity of depression (Mendez et al., 2022;Miskowiak et al., 2021), anxiety symptoms (Mendez et al., 2022;Miskowiak et al., 2021) and post-traumatic distress (Mendez et al., 2022), and with lower functioning at the 1-year follow-up evaluation (Mendez et al., 2022;Miskowiak et al., 2021).We found no association of CI with insomnia and suicidal ideation.These symptoms are frequently reported in COVID-19 subjects; however, their association with CI has never been investigated before.We did not find an association of CI with poor quality of life, possibly due to both the recent onset and the relatively mild degree of cognitive dysfunction in most subjects.Future studies should stratify the sample according to the degree of impairment to exclude that COVID-related CI has no significant impact on the quality of life.
The following study limitations should be acknowledged: a) the exclusion of cognitive impairment before COVID-19 infection was retrospectively carried out by means of a questionnaire, which may affect accuracy of this assessment; b) we cannot exclude that patients experiencing cognitive impairment after the infection might have been more inclined to accept the participation in the study due to health concerns, thus producing a potential bias on the findings on frequency of cognitive deficits in subjects with long COVID.However, the prevalence in the present study is close to the one reported in previous papers.
In conclusion, our study confirms that persistent CI affects more than one third of patients recovered from the COVID-19 infection, thus placing a consistent burden on affected persons and healthcare systems.It involves most of the cognitive domains assessed by comprehensive cognitive batteries, but not social cognition, and is related to the severity of specific symptoms of the disease in the acute stage.It is also associated with severity of depression, anxiety and post-traumatic distress, with poor global functioning, with indices of cellular damage, in particular the LDH blood levels, regarded as a potential biomarker/ predictor of severe respiratory pathology, admission to intensive care units and death.
In COVID-19 patients cognitive functioning should be routinely and periodically assessed in clinical practice, especially in patients who experienced more severe COVID-19 symptoms.In the light of the persistence of cognitive impairment in some cases long after the resolution of COVID, cognitive training programs, and eventually other interventions, such as the transcranial direct stimulation, should be considered as potential treatment strategies for patients showing a cognitive impairment.
Prof. Antonio Vita, in the last three years, received advisory board/ consultant fees and support for clinical studies or trials, conferences, and congress presentations from the following drug companies: Alkermes, Angelini, Boehringer Ingelheim, Janssen-Cilag, Lundbeck, Otsuka, Roche, Rovi Pharma.
Prof. Armida Mucci, in the last three years, received advisory board/ consultant fees from the following drug companies: Angelini, Boehringer Ingelheim, Pierre Fabre and Rovi Pharma.
Prof. Stefano Barlati, in the last three years, received advisory board/ consultant fees and support for clinical studies or trials, conferences, and congress presentations from the following drug companies: Angelini, Lundbeck, and Otsuka.
All other authors have no conflicts of interest to declare.

Fig. 2 .
Fig. 2. Cognitive profile at baseline of the whole COVID-19 patients assessed with MATRICS Consensus Cognitive Battery MCCB) SD = standard deviation of the distribution of each cognitive domain scores.Mild = 1-2 SD in at least 2 cognitive domains; Moderate = 2-3 SD in at least 2 domains; Severe = 3 SD in at least 2 domains.

Fig. 3 .
Fig. 3. Cognitive profile at baseline of subjects with and without cognitive impairment at screening Mean z-scores on the MATRICS Consensus Cognitive Battery (MCCB) using the Italian normative sample to calculate the z-scores COGþ: subjects with cognitive impairment at screening; COG-: subjects without cognitive impairment at screening AV: attention/vigilance; RPS: reasoning and problem solving; SC: social cognition; SoP: speed of processing; Vrbl Lrng: verbal learning; Vis Lrng: visual learning; WM working memory.

Table 1
Characterization of the screened study sample based on the presence or absence of cognitive impairment (n = 736).

(n ¼ 259) COG-(n ¼ 477) t-χ 2 / p- value
Mean ± Standard Deviation for age, education, MoCA score, number of days between COVID-19 infection and screening time and length of hospitalization and frequency for gender.COGþ/: subjects who presented a MoCA score<26; COG-: subjects who scored ≥ 26 on MoCA evaluation.*statistically significant difference between groups after Bonferroni correction for multiple tests (p < 0.0071).

Table 2
Frequency of COVID-19 symptoms recorded at time of the acute infection (n = 644).

Table 4
MATRICS Consensus Cognitive Battery MCCB) in the whole COVID-19 study group and comparison with the Italian Normative Sample -Baseline T-scores.