The wide spectrum of COVID-19 neuropsychiatric complications within a multidisciplinary centre

Abstract A variety of neuropsychiatric complications has been described in association with COVID-19 infection. Large scale studies presenting a wider picture of these complications and their relative frequency are lacking. The objective of our study was to describe the spectrum of neurological and psychiatric complications in patients with COVID-19 seen in a multidisciplinary hospital centre over 6 months. We conducted a retrospective, observational study of all patients showing neurological or psychiatric symptoms in the context of COVID-19 seen in the medical and university neuroscience department of Assistance Publique Hopitaux de Paris—Sorbonne University. We collected demographic data, comorbidities, symptoms and severity of COVID-19 infection, neurological and psychiatric symptoms, neurological and psychiatric examination data and, when available, results from CSF analysis, MRI, EEG and EMG. A total of 249 COVID-19 patients with a de novo neurological or psychiatric manifestation were included in the database and 245 were included in the final analyses. One-hundred fourteen patients (47%) were admitted to the intensive care unit and 10 (4%) died. The most frequent neuropsychiatric complications diagnosed were encephalopathy (43%), critical illness polyneuropathy and myopathy (26%), isolated psychiatric disturbance (18%) and cerebrovascular disorders (16%). No patients showed CSF evidence of SARS-CoV-2. Encephalopathy was associated with older age and higher risk of death. Critical illness neuromyopathy was associated with an extended stay in the intensive care unit. The majority of these neuropsychiatric complications could be imputed to critical illness, intensive care and systemic inflammation, which contrasts with the paucity of more direct SARS-CoV-2-related complications or post-infection disorders.


Introduction
COVID-19, caused by the SARS-CoV-2 virus, has spread since December 2019 and was declared a pandemic by the World Health Organization. An initial Chinese cohort of 214 COVID-19 patients reported a high frequency of neurological symptoms (36%), including non-specific manifestations such as headache, confusion and myalgia, but also strokes and seizures. 1 Since then, a variety of neurological complications have been described, including cerebrovascular complications, 2 encephalopathy, 3 encephalitis, 4 seizures, 5 Guillain-Barré syndrome, 6 cranial nerve palsies, 7 anosmia and dysgeusia. 8 Several studies have raised concerns about a high prevalence of anxiety, mood disorders and post-traumatic stress disorders in COVID-19 patients. 9,10 Besides case reports and case series illustrating the pleiotropy of COVID-19 neurological manifestations, cohort studies and registries have highlighted the particular high prevalence of strokes, encephalopathy and neuromuscular complications. [11][12][13][14][15][16][17] More large-scale studies presenting a wide picture of these complications and of their relative frequency are needed.
We launched an observational study of the neuropsychiatric manifestations of COVID-19 at the onset of the first wave of the pandemic in France in March 2020. Here, we describe the detailed spectrum of neuropsychiatric disorders in 245 COVID-19 patients seen in the medical and university neuroscience department of APHP-Sorbonne University Hospitals over a 6-month period.

Study design and population
This is a retrospective, observational study conducted consecutively on all COVID-19 patients with neurological or psychiatric symptoms seen between March 1st and August 28th at the APHP -Sorbonne University medical and university neuroscience department, which groups the neurology and psychiatry departments of five Hospitals in Paris (Pitié-Salpêtrière, Saint-Antoine, Tenon, Rothschild, Charles-Foix hospitals). The study Investigators were physicians from the department, which includes all medical units in the field of adult neurology, neurovascular, neurological intensive care, neurorehabilitation, psychiatry, neurophysiology and neuropathology.
All consecutive in-or out-patients, aged 18 years or older, with COVID-19 and de novo neurological or psychiatric symptoms were included in the study. We included patients hospitalized or seen as outpatients in the neuroscience department, but also patients reported by physicians working in other departments involved in the care of COVID-19 patients after informing all physicians of our university hospital that patients with neurological or psychiatric symptoms should be reported. Patients showing uniquely anosmia and/or dysgeusia were not included. Patients presenting with acute exacerbations of known pre-existing neurological disorders were not included. We included patients with pre-existing psychiatric disorders only if they presented with acute decompensations with clear disruption in their clinical course.
The primary objective of the study was to describe the spectrum of neurological and psychiatric complications in COVID-19 patients.

Data collection
Data were collected retrospectively by investigators from medical records and entered into a structured case report form. Items collected included demographic data, medical and treatment history, comorbidities, symptoms, date of onset, and severity of COVID-19 infection, neurological and psychiatric symptoms, neurological and psychiatric examination, and, when available, results from CSF analysis, brain imaging including MRI, EEG and electroneuromyography (ENMG).
COVID-19 was defined by at least one of the three following criteria: (i) positive SARS-CoV-2 polymerase chain reaction in swab or upper pulmonary samples, or positive serology; (ii) thoracic radiological findings typical of SARS-CoV2 infection; and (iii) suspected COVID-19 infection according to the World Health Organization guidance criteria. 18 The severity of COVID-19 was the status of the patient at the nadir of the disease according to the seven levels as defined by the World Health Organization: 1-not hospitalized, no limitation in daily life activity; 2-not hospitalized, with limitation in daily life activity; 3-hospitalized, no oxygen requirement; 4-hospitalized, necessitating oxygen; 5-hospitalized, necessitating non-invasive ventilation or Optiflow TM ; 6hospitalized, necessitating intubation or extracorporeal membrane oxygenation; and 7-death.
The investigators completed the database consisting of a predefined list of neurological and psychiatric characteristics (symptoms, clinical signs, date of onset). Each item was scored as present, absent or unknown (no clinical evaluation available), and a final neurological or psychiatric syndrome was determined. One investigator (C.D.) reviewed all cases, and classified the diagnoses according to criteria of the Liverpool Brain infections Group (Neuro Network). 19 We used the word 'encephalopathy' to refer to acute global disturbances in cognition (encompassing delirium, confusional states and altered mental status). 20 When a patient had more than one diagnosis, the diagnoses were classified as primary (pronounced) or secondary.

Standard protocol approvals
The study was conducted in accordance with good clinical practice, the French regulation for retrospective studies on clinical data, and was compliant with the European General Data Protection Regulation (GDPR) and the French Commission Nationale de l'Informatique et des Libertés (CNIL) rules. All patients (or their relatives in cases of impaired consciousness) received written information about the research, and consented to the use of their data. The study received the approval of the Sorbonne University Ethics Committee (N 2020 CER-202028). The study is registered on the clinicaltrials.gov website (NCT04362930).

Statistical analysis
For the analyses, we grouped the 7-level COVID-19 severity score into four categories: 1-not hospitalized (levels 1 and 2); 2-hospitalized without intensive care (levels 3 and 4); 3-hospitalized with intensive care (levels 5 and 6); and 4-death (level 7). Kruskal-Wallis tests and Fisher's Exact tests were used to compare groups. Pairwise comparisons were performed using pairwise Wilcoxon-Mann-Whitney tests and pairwise Fisher's Exact tests with Benjamini-Hochberg method to correct for multiple testing.
We performed adjacent category ordinal logistic regression on the four grouped categories of the 7-levels score with the risk factors age, gender, obesity (body mass index 30 kg/m 2 ) or other comorbidity.
Neuropsychiatric symptoms, signs and syndromes were assessed on non-missing data.
For the correlation matrix, we calculated the Pearson correlation for inter-relationships between the following variables: age, gender, presence of one comorbidity, intensive care unit (ICU) hospitalization and the various neurological syndromes. Hierarchical clustering was applied with the single linkage distance.
The risk factors for the most frequent neurological syndromes were analysed using the Classification And Regression Tree algorithm. The Classification And Regression Tree algorithm, also known as a 'decision tree', is a non-parametric supervised technique that combines variables in such a way as to best discriminate two groups. For each neurological syndrome, we trained one tree to depth 4 through entropy minimization.
Statistical analyses were performed using R 3.6.1. and package VGAM_(version 1.1-3) for the ordinal logistic regression (R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.) and using python 3.8 with the scikit learn 0.23.2 package for decision trees and correlation matrices. 21

Data availability
All data are available upon request to the corresponding author.

Patient characteristics
During the study period, 1979 patients were admitted with a diagnosis of COVID-19 in our centre. A total of 249 COVID-19 patients with a de novo neurological or psychiatric manifestation were included in the database. Of these 249 patients, 24 were seen as outpatients and 225 were hospitalized (11.3% of all hospitalized COVID-19 patients). Three patients were excluded because they withdrew their consent, one patient was excluded because he did not fulfil the diagnosis criteria for COVID-19. The population analysed consisted of 245 cases.
The characteristics of the patients are presented in Table 1.
Brain MRI was performed in 119 patients and was abnormal in 81 (68%, 9 patients with missing data). MRI findings comprised ischaemic strokes, intracerebral haemorrhages, cerebral venous thrombosis, cytotoxic lesions, basal ganglia abnormalities and white matter enhancing lesions. They have been described elsewhere. 22 EEG was performed in 82 patients and was abnormal in 54 patients (77%, 12 with missing data). Detailed EEG data are presented in another paper. 23 Pathological EEG findings included focal abnormalities, metabolictoxic encephalopathy, periodic discharge and epileptic activity.
ENMG was carried out in 25 patients and was abnormal in 20 patients with evidence of peripheral nervous system impairment (87%, 2 patients with missing results).

Encephalopathy
Among patients with encephalopathy, 56% were hospitalized in the ICU. Twenty per cent of patients with encephalopathy had a pre-existing cognitive disorder. Seventy per cent of the patients with encephalopathy had at least one cardiological, respiratory or metabolic comorbidity. The most common clinical presentations of encephalopathy were confusion, and delayed awakening after stopping sedative drugs. Among the ten deceased patients, seven (70%) showed encephalopathy.
With respect to risk ratios, encephalopathy mainly affected patients over 60 years of age (Table 3).
In the correlation matrix, encephalopathy was associated with higher age, death, cardiac and diabetic comorbidities (Fig. 3).

Critical illness polyneuropathy and myopathy
Critical illness neuropathy or myopathy was diagnosed in the recovery phase after sedative drug reduction in the ICU. ENMG findings were available for 22% of these patients. Several patterns of neuromuscular injury were observed: (i) axonal sensorimotor polyneuropathy, (ii) myopathy, (iii) troncular nerve compressions (median, ulnar, peroneal, lateral femoral cutaneous nerves), and (iv) brachial plexopathy. Brachial plexopathy was only seen in patients after remaining prone for extended periods. Thirty-seven percent of the patients with critical illness polyneuropathy had pre-existing diabetes.
The decision tree showed that a stay exceeding 12.5 days was the strongest feature predictive of a neuropathy after entering the ICU (Supplementary Fig. 1).
Unusual symptoms at presentation (which were not explained by infarct location or metabolic disturbances)     were frequently found such as confusion (13/32, 40%) or apathy (5/38, 13%). A fraction of these patients has been already reported. 24

Seizures
Among the 14 patients with seizures, none had a previous history of epilepsy. One patient was under treatment for Parkinson's disease with dementia and one patient for glioblastoma. Three patients had a focal seizure without generalization, one patient a focal to bilateral generalized seizure, seven patients a generalized seizure, one patient a status epilepticus.

Cranial nerve palsies
Five patients presented with cranial nerve palsies. One patient had VI nerve palsy with normal MRI and CSF examination. As she was a heavy smoker, a thrombosis triggered by the COVID-19 infection was suspected. One patient presented with optic neuritis, in the setting of a possible inflammatory disorder of the CNS (inflammatory lesions on brain MRI, oligoclonal bands in the CSF). One patient presented with III nerve partial palsy after an ICU stay. One patient presented with unilateral hypoglossal nerve palsy, and one with combined homolateral X, XI and XII nerves palsies, while in the ICU, which were attributed to mechanical complications of positioning, intubation or jugular catheterizations.

Headaches
Isolated disabling primary headache was the primary diagnosis in 16 patients (7%). Headaches often had migraine characteristics. None of the patients had pre-existing migraines. Among patients with headache, eight underwent brain MRI, which was normal in seven patients and showed a non-specific lesion in one. Two had lumbar puncture (normal in both).
In the correlation matrix, headache is associated with younger patients who were not hospitalized and showed no associated comorbidities (Fig. 3).

Encephalitis
Only seven patients fulfilled criteria for encephalitis. Among those seven patients, one patient had positive polymerase chain reaction for varicella-zoster-virus in the CSF, suggesting concomitant varicella-zoster-virus encephalitis. One patient had an encephalopathy with myoclonus and inflammatory brain lesions, one patient had cognitive disturbances with myoclonus and CSF pleiocytosis, one patient had brainstem impairment, movement disorders and dysautonomia with white matter lesions on MRI, two patients showed alterations of consciousness with white matter lesions on MRI, one patient had confusion with MRI features of limbic encephalitis.
Several of these patients have been already reported. 25

Guillain-Barré syndrome
Five patients were diagnosed with Guillain-Barré syndrome, three of them requiring hospitalization in the ICU. One patient had a pre-existing demyelinating Charcot-Marie-Tooth disease. In four patients, Guillain-Barré syndrome presented with severe motor weakness of all four limbs. One of the patients had associated bilateral facial nerve palsy. CSF showed elevated protein in two patients (1.1 g/l and 1.65 g/l). ENMG showed typical characteristics of acute inflammatory demyelinating polyneuropathy in all of them.

Discussion
We describe a wide range of neuropsychiatric symptoms and syndromes occurring in COVID-19 patients seen in a single multidisciplinary centre over a 6-month period. Our patient cohort primarily presented with encephalopathy, critical illness polyneuropathy and myopathy, psychiatric disturbances, and cerebrovascular complications. The prevalence of neuropsychiatric symptoms among patients hospitalized in our COVID-19 centre was 11%.
The prevalence of neurological signs reported in previous studies has been very variable, depending on inclusion criteria and evaluation methods, varying between 4.2% 26 and 57.4%. 27 The relatively lower prevalence in our study may be due to the strict inclusion criteria and the exclusion of patients with isolated symptoms of anosmia or dysgeusia. Compared to those with non-severe disease, the patients in ICU were more likely to have comorbidities, including hypertension, diabetes, cancer, cardiac or kidney disease. Our cohort highlights the high frequency of neuropsychiatric complications related to critical illness and intensive care. Almost half of the patients in our cohort were hospitalized in the ICU (47%). Our findings emphasize that encephalopathy is a major issue in patients with COVID-19. This has been previously reported by previous teams which variably referred to encephalopathy, 28 delirium 29 and altered mental status, 30 which we chose to regroup under the term encephalopathy. 20 The presence of encephalopathy was strongly associated with COVID-19 infection severity and the presence of comorbidities, which is in keeping with complications of hypoxia and critical illness. The high prevalence of encephalopathy in patients with COVID-19 patients has already been reported, and seems to be more common in patients with more severe COVID-19-related respiratory disease, associated comorbidities, evidence of multi-organ system dysfunction, including hypoxaemia, renal and hepatic impairment, and elevated markers of systemic inflammation. 1,31,32 The association of encephalopathy with greater age and comorbidities has already been reported. 33,34 A recent large-scale cohort also emphasized the frequency of non-specific complications, including toxic/metabolic encephalopathy and hypoxic/ischaemic brain injury. 13 The prevalence of encephalopathy is very high in patients with COVID-19 hospitalized in the ICU. 35 As in previous works, we found that encephalopathy was associated with the risk of death. 34,36,37 Critical care polyneuropathy and myopathy were particularly common in our patients. Besides classical ICU polyneuropathy, many patients presented with mechanic plexopathy and nerve compressions, probably secondary to prolonged sedation and lying prone. The occurrence of critical illness polyneuropathy and myopathy was statistically associated with a longer stay in the ICU. The long duration of critical care, the requirement for high doses of anaesthetics and curare, and the frequent association with diabetes could partly explain the particularly high frequency of ICU neuropathy. Two patients presented with mechanical cranial nerve injuries (XII in one patient; X, XI and XII in one patient) in the setting of critical care, a complication known as Tapia syndrome. 38 Acute cerebrovascular disorders consisted mostly in ischaemic strokes. The risk of cerebrovascular disorders was highly associated with the presence of comorbidities. The presence of unusual stroke symptoms at presentation, such as encephalopathy without focal deficit, is noteworthy and was also reported by other authors. 39 COVID-19 might facilitate ischaemic strokes via at least three non-exclusive pathogenic mechanisms: atherosclerotic plaque vulnerability, a hypercoagulable state, and cerebral microvasculature injury (endothelitis). [40][41][42] Indeed,   COVID-19 is commonly complicated by sepsis-induced coagulopathy, induced by a systemic inflammatory response involving endothelial dysfunction and microthrombosis often associated with multi organ failure. 43 These observations contrast with the rarity of syndromes potentially linked to viral neuroinvasion. The question of the neuroinvasiveness of the SARS-CoV-2 is an ongoing debate. Its detection in the CSF has only been reported in a few case studies. 4,44 The majority of neuropathological observations favour the role of critical illness complications or described non-specific inflammatory brain lesions, [45][46][47] although some recent findings have shed new light on the neuroinvasive potential of SARS-CoV-2. 41,48 A recent cohort with 606 patients with neurological signs of COVID-19 reported no encephalitis. 13 Only seven patients in our cohort fulfilled the diagnostic criteria for encephalitis, and none had evidence of SARS-CoV-2 detection in the CSF. 19 The CSF pleiocytosis without SARS-CoV-2 detection and brain inflammatory lesions seen in our patients and other reported cases of COVID-19 might result from immune-mediated inflammatory mechanisms rather than direct viral invasion. The role of cytokine activation has been widely reported for COVID-19. 49,50 Headaches are also proposed to be secondary to inflammatory cytokinic mechanisms, 51 which may also play and important role in psychiatric manifestations. 52 Other immune-mediated complications, such as Guillain-Barré syndrome and myelitis occurred anecdotally in our cohort. Cranial nerve impairments could often be explained by alternative mechanisms (pre-existing inflammatory disorder, thrombotic mechanism, mechanical compressions).
Although our sample of patients who developed psychiatric symptoms during COVID-19 is small, our findings are in accordance with previously reported data. 53 The occurrence of psychiatric symptoms in patients with COVID-19 was remarkably high, which is in keeping with previous studies showing that psychiatric symptoms are more frequent in patients with COVID-19 than with other infections or health events. 54 Indeed, our patients mainly suffered from anxiety disorder or depression, and none developed obsessive compulsive disorder or bipolar disorder. Five patients presented with acute psychosis. The difficulty to distinguish acute psychosis and delirium has been highlighted by some authors. 55 In our multidisciplinary centre, each patient suspected of experiencing a psychotic episode was evaluated by a psychiatrist at several time points; thus, we paid a special attention to know if the patient met the DSM-5 criteria of delirium (disturbance in attention and awareness; disturbance develops over a short period of time and tends to fluctuate in severity during the course of a day; an additional disturbance in cognition) or the criteria for an acute psychotic episode. Long-term follow-up data will be valuable to have an estimation of the risk of developing posttraumatic stress disorder, especially for patients with adjustment disorders and acute stress syndrome.
Our study has several limitations. Although the study is retrospective, it was implemented early at the onset of the COVID-19 pandemic and data could be collected prospectively by the investigators as they examined new patients. Furthermore, the absence of a control group of patients with COVID-19 without neurological or psychiatric manifestations limits the overall interpretation of the results. We may have an over representation of complications due to critical illness and critical care as there was a large number of COVID-19-dedicated ICU beds in our institution. We cannot exclude reporting bias and lack of exhaustivity, but the involvement of all physicians of the Department implicated in every step of patient care (ICU, acute hospitalization departments, rehabilitation departments, imaging departments, neurophysiology) limits this bias, especially for the most severe forms. In stroke patients, one limitation of the study is that we could not determine for all patients the aetiology of ischaemic strokes and in case of a negative outcome, whether COVID-19 was the trigger.
In conclusion, we report the broad landscape of neuropsychiatric complications in a large cohort of COVID-19 patients. The majority of these complications could be attributed to critical illness, intensive care and systemic inflammation, which contrasts with the paucity of more direct SARS-CoV2-related complications or post-infectious disorders. Further studies are needed to better disentangle the different mechanisms underlying these various symptoms, and to explore potential long-term complications.

Supplementary material
Supplementary material is available at Brain Communications online.