The Incidence of SARS-COV-2 Manifestations in the Central Nervous System: A Rapid Review and Meta-Analysis

Background: Coronavirus disease 2019 (COVID-19) is caused by SARS-CoV-2 and presents itself mainly as a respiratory tract infection. However, reports of associated central nervous system (CNS) manifestations are increasing. Methods: We conducted this rapid review to determine the frequency of CNS manifestations of COVID-19 (CNS symptoms, acute cerebrovascular disease, and infectious/inammatory CNS diseases) and to summarize the current evidence for direct invasion of the CNS by SARS-CoV-2. An information specialist searched Ovid MEDLINE, the CDC: COVID-19 Research Articles Downloadable and WHO COVID-19 Databases, CENTRAL, and Epistemonikos.org on May 13, 2020. Two reviewers screened abstracts and potentially relevant full-text publications independently. The data extraction, assessment of risk of bias,and certainty of evidence using GRADE was done by one reviewer and double-checked by another. If possible and reasonable, a meta-analysis was carried out. Results: We identied 13 relevant studies (four cohort studies, nine case studies) with a total of 866 COVID-19 patients.In a Chinese cohort, dizziness (16.8%; 36 of 214) and headache (13.1%; 28 of 214) were the most common CNS symptoms reported. A meta-analysis of four cohort studies including 851 COVID-19 patients showed an incidence of 3.3% (95% CI: 2.2–4.9) for ischemic stroke (follow-up: one to ve weeks). In 13 of 15 encephalitis case studies, PCR testing of the cerebrospinal uid did not detect any virus components. Conclusion: CNS manifestations occur frequently in patients with COVID-19. It is important to integrate neurologists into the multiprofessional COVID-19 treatment team to detect neurological complications early and to treat them correctly.


Background
Coronavirus disease 2019 (COVID- 19) is an infectious disease caused by the Severe Acute Respiratory Syndrome -Coronavirus 2 (SARS-CoV-2) [1]. The time between the rst case of illness (December 31, 2019, Hubei Province, China) and the declaration of a pandemic by the World Health Organization (WHO, March 11, 2020) was rather short [2]. Clinically, COVID-19 presents itself mainly as a respiratory tract infection with fever, fatigue, and dry cough and with in ltrates on chest X-ray [3]. Pneumonia and the subsequent development of acute respiratory distress syndrome are the most common severe manifestations of the disease [3]. As the number of infected people increased, more differentiated data on the symptoms emerged. More and more studies have been published on neurological manifestations such as the loss of the sense of smell and taste and also on serious complications such as encephalitis or stroke [4][5][6]. A yet unanswered question is whether the virus has a direct effect on the central nervous system (CNS), or whether these neurological manifestations should be considered parainfectious. The neuroinvasive potential of SARS-CoV-2 is still under debate. Proposed mechanisms for CNS invasion are hematogenous spread via an impaired blood-brain barrier, neuronal retrograde, or transcribial routes [7,8].
Neurological damage to the CNS was observed in earlier epidemics of diseases such as Severe Acute Respiratory Syndrome (SARS) or Middle East Respiratory Syndrome (MERS), which were also triggered by coronaviruses [9,10]. In SARS, the pathogen could be detected by a polymerase chain reaction (PCR) in the cerebrospinal uid (CSF) and brain tissue [11][12][13][14][15]. To determine the frequency of CNS complications and to investigate whether the virus infection affects the brain tissue directly, we conducted a rapid review. As manifestations of the CNS underlie different pathomechanisms, we phrased the following key questions (KQs) as:

KQ1
What is the incidence of central nervous symptoms in patients infected with SARS-CoV-2?

KQ2
What is the incidence of acute cerebrovascular disease in patients infected with SARS-CoV-2?

Methods
We adhere to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) throughout this manuscript [16].
Compared with the methods of a systematic review, the review team applied the following methodological shortcuts for this rapid review: no searches of grey literature; no dual independent data extraction; no dual independent risk of bias assessment and rating of the certainty of evidence.

Literature search
An experienced information specialist conducted a systematic search of the literature published from January 1, 2019 to April 15, 2020 in Ovid MEDLINE, the CDC: COVID-19 Research Articles Downloadable Database, the WHO COVID-19 Database, CENTRAL (Cochrane Library/Wiley), and Epistemonikos.org. On May 13, 2020, the information specialist performed an updated search. The search strategies were modi ed based on the ndings of the initial searches, and the results were deduplicated with those of the rst search. We also conducted similar article searches in PubMed based on the rst 100 linked references for each seed article. Seed articles were potentially eligible studies identi ed during the preliminary searches or in the rst round of databases searches. The additional le provides the detailed search strategies [see Additional le 1]. Furthermore, the review authors screened the reference lists of the included studies to identify relevant citations that were not detected by searches of electronic databases.

Eligibility criteria
We included studies concerning central nervous symptoms (e.g., dizziness, headache, etc.), encephalopathy, acute cerebrovascular disease, and infectious (e.g., encephalitis, meningitis) or in ammatory CNS diseases in con rmed COVID-19 patients older than 18 years. Studies with suspected but not con rmed cases of COVID-19 and including children and patients with primary neurological diseases or taking additional immunosuppressive drugs (e.g., for multiple sclerosis) were excluded. Only articles in English and German were considered. After a quick presearch, we prespeci ed a best-evidence approach for the literature screening. For KQ1 and KQ2, case studies were excluded because observational studies were already available. Due to the lack of data on infectious/in ammatory CNS disorders in COVID-19 cases, there were no restrictions regarding the speci c study design for KQ3 and KQ3.1. However, case reports without SARS-CoV-2 PCR testing of the CSF were excluded.

Screening process
Two researchers independently screened all titles and abstracts of both searches based on the prede ned inclusion and exclusion criteria. Included abstracts were retrieved as full-text publications and independently screened by two reviewers. In cases of disagreements about eligibility, the two reviewers reached consensus by discussion. The team screened the literature using Covidence Systematic Review software (www.covidence.org).

Quality assessment
The review team assessed the risk of bias of the observational studies using an adapted version of the Risk Of Bias In Non-randomized Studies -of Interventions (ROBINS-I) tool [17]. A single reviewer rated the risk of bias for each relevant outcome of each study; a second reviewer checked the ratings. As there is no validated risk of bias checklist for case studies available, the review team checked whether SARS-CoV-2 PCR tests were carried out twice to reduce the risk of a false positive result.
To assess the certainty of the body of evidence for all the outcomes of interest, we applied the approach by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group [18]. A single review author applied GRADE, and a second review author veri ed all judgments. GRADE uses four categories to classify the certainty of evidence. A high certainty of evidence means that we were very con dent the estimated effect lies close to the true effect; a moderate certainty assumes it is likely to be close; with a low certainty rating, the true effect might substantially differ from the true effect; and a very low certainty means that we had no con dence in the effect estimate.

Data extraction
One reviewer extracted data from the included studies into standardized tables; a second reviewer checked the data extraction for completeness and correctness. We extracted the following data: report characteristics (year, authors), study design, setting, participant characteristics (age, preexisting medical conditions, main symptoms,) and results.

Synthesis
We synthesized the results both narratively and in tabular form. If three or more studies were similar with respect to outcomes and populations and provided data for quantitative analyses, we conducted metaanalyses. Therefore, we conducted a test of heterogeneity (I 2 statistic, Cochran's q-test) and applied the DerSimonian and Laird method for random-effects models. A sensitivity analysis was performed by excluding high risk of bias studies. We planned to assess publication bias using funnel plots, Egger's regression intercept, and Kendall's S statistic. All statistical analyses were conducted using Comprehensive Meta-Analysis (CMA), version 2.2.050 (www.meta-analysis.com).
One of the four studies presented the data of 221 hospitalized COVID-19 patients admitted to hospital in Wuhan [6,22]. The data of these 221 patients were analyzed at two different time-points: the rst publication presented the data of 214 patients and focused on neurological symptoms [22]. In the preprint of the second analysis, seven patients were added, the observation period was extended by ten days, and additional data concerning cerebrovascular disease was released [6]. One French study was published in a "Letter to the Editor" and provided only limited data [19]. The baseline characteristics of the observational studies and the case studies are shown in Tables 1 and 2, respectively. Table 1 Characteristics, main results, and risk of bias of the observational studies Lodigiani 2020 21 Mao L 2020 22 Li Y 2020 6 Klok 2020 20     Four retrospective observational studies including 851 individuals reported on the incidence of acute cerebrovascular disease relating to COVID-19 [6,[19][20][21][22]. While three studies referred only to ischemic strokes [19][20][21], the Chinese study also included other acute cerebrovascular manifestations (cerebral hemorrhage, cerebral venous sinus thrombosis) [6,22]. The largest study from Italy contained only 15.7% severe cases of COVID-19 (61 of 388) [21]; the other three studies included 42.5% (94 of 221) of severe cases [6,22] or only considered patients with admission to an ICU for inclusion [19,20].
The Wuhan study was the only one of the included publications that provided the data of the total population and of the patients with an acute cerebrovascular event separately.   We performed a meta-analysis of four observational studies that included the data of 851 patients with COVID-19 to assess the incidence of ischemic strokes (Fig. 2). The overall incidence of ischemic strokes in COVID-19 patients was 3.3% (95% CI: 2.2-4.9; follow-up range one to ve weeks after hospital admission). A sensitivity analysis was conducted by excluding high risk of bias studies and showed no major changes in the incidence of ischemic strokes (3.4%; 95% CI: 1.16-7.2; follow-up range one to ve weeks after hospital admission).
However, it should be noted that the sensitivity analyses resulted in the loss of half the studies. Due to the small number of studies, an assessment of publication bias was not carried out.
Among the studies with a low risk of bias, in the study population with more severe cases of COVID-19, a higher patient proportion suffered an ischemic stroke than in the study population with less severe cases (5.0% vs. 2.3%) [6,21].
The Italian study reported that in six of nine cases, ischemic stroke was the primary reason for admission to the hospital. The Wuhan study measured a different period: the mean time from the onset of COVID-19 symptoms to acute ischemic stroke was 10.8 days. In the remaining studies, the ischemic strokes were diagnosed at the ICU. While the Dutch study did not mention whether the patients showed any focal neurological signs, in the French study, brain imaging was performed due to unexplained encephalopathy symptoms in patients without any focal neurological de cits [19].
Only the Wuhan study provided data concerning the etiology of the ischemic strokes: ve of the eleven ischemic strokes were classi ed as macroangiopathic, three as microangiopathic, and three as cardioembolic in terms of the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classi cation [6].

Other acute cerebrovascular diseases
The Chinese study also included other acute cerebrovascular manifestations [6]. Of the 221 reported patients, one had a sinus vein thrombosis (0.5%), and one had a cerebral hemorrhage (0.5%). Both patients were de ned as severe cases, and the cerebrovascular event occurred after more than nine days.
3.3 KQ3: Infectious/in ammatory CNS disease relating to SARS-CoV-2; KQ3.1: Neuroinvasion of SARS-COV-2 No case of an infectious/in ammatory CNS disease associated with SARS-CoV-2 was described in the cohort studies. That is why for KQ3 we also included case studies and case series. In nine articles, the data of 15 individuals was presented [5, 23-25, 27-31, 33]. The main symptoms and diagnostic results (neurological examination, brain imaging, CSF, electroencephalography, PCR tests) are included in Table 2. In two COVID-19 patients, the CSF was positive for SARS-CoV-2 in the PCR test [26][27][28]. One of these was a case of meningoencephalitis associated with COVID-19 and took place in Japan [28]. A 24-year-old man was admitted to the hospital because of generalized seizures and loss of consciousness. He already had shown symptoms of a general infection several days before. Neck stiffness was detected. The MRI supported the assumption of a meningoencephalitis. The nasopharyngeal swab tested negative for SARS-CoV-2 and the CSF sample positive.
Both tests repeatedly led to the same results.
General symptoms of an infection, such as headache and fever, and a new-onset seizure (not described whether generalized or focal) were also present in the case of a 41-year-old female in the USA [26,27]. Neck stiffness and photophobia were documented at the emergency department. In the course of the disease, she developed disorientation accompanied by hallucinations. In this case, both the nasopharyngeal swab and the CSF tested positive for SARS-CoV-2. A repetition of the tests was not stated.
In both cases, the CSF results were not su ciently presented to allow a more precise interpretation regarding an impairment of the blood-brain barrier. As for neurotropic viruses, only the herpes simplex virus and the varicella zoster virus were excluded. Additionally, in the second case, cerebral imaging was entirely missing.
In the other seven case studies and series with 13 individuals, the PCR tests for SARS-CoV-2 in the CSF samples were negative [23-25, 29-31, 33]. In total, only two studies repeated the PCR tests on the CSF samples [24,28].
CSF testing was also carried out in the previously mentioned study of Helms et al., in which seven patients with unexplained encephalopathy symptoms were lumbar punctured. In none of these patients could SARS-CoV-2 be detected by PCR tests [19]. However, whether cerebral imaging indicated an in ammatory process or the results of the liquor samples were released in detail to prove if there is any hint of SARS-CoV-2.
Notable is a case report of Xiang et al. that described consciousness impairment in a 56-year-old male COVID-19 patient [5]. The CSF of the patient was also positive for SARS-CoV-2 in the PCR test, but this was not included in this review due to the publication language (Chinese).

Discussion
Initially, neurological manifestations were not the focus of attention in the scienti c discourse of SARS-CoV-2. In the meantime, an increasing number of articles on the neurological manifestations of SARS-CoV-2 have been published, even rst reviews that summarize these data [34][35][36]. While previously published reviews covered a wide range of neurological symptoms and included case reports for all outcomes, we focused on CNS disorders and prespeci ed a best-evidence approach. Although ours is a rapid review, it has methodological advantages such as a discriminated literature search, a risk of bias and GRADE assessment, and the inclusion of a quantitative analysis (regarding acute ischemic stroke). We consider these steps necessary to assess the available evidence.
Our meta-analysis of four cohort studies including a total of 851 patients with COVID-19 infections showed that 3.3% (95% CI: 2.2-4.9; follow-up range one to ve weeks) suffered from an ischemic stroke. The Wuhan study was the only one of the included publications that provided the data of the total population and of the stroke patients separately [6]. symptoms, such as dry cough and fever, than patients with mild cases of COVID-19. [22] In regard to the data provided, it is not possible to draw conclusions about the stroke incidence and, therefore, we decided against a post-hoc inclusion of this study. In contrast, in the two larger studies included in our meta-analysis, COVID-19 was laboratory proven in all patients, and the baseline characteristics of the population were described in detail [6,21].
In the past, respiratory infections in general were repeatedly associated with an increased incidence of ischemic stroke [40][41][42][43]. In 2018, Blackburn et al. presented their results of a time-series analysis of English hospital admissions for stroke and myocardial infarction [40]. They reported that respiratory viruses except parain uenza were signi cantly associated with ischemic stroke admission in the elderly (≥ 75 years). In particular, in uenza was widely discussed in the scienti c discourse because an in uenza vaccination is available and may possibly lead to a risk reduction [44]. However, the evidence is limited due to the lack of randomized controlled studies. A high concentration of C-reactive protein (CRP) was also discussed as a marker of elevated risk of ischemic stroke [45]. In the Wuhan study, the mean CRP concentration in patients with cerebrovascular events was signi cantly higher than in those without (51.1 mg/dl vs. 12.1 mg/dl) [6].
Nevertheless, the relevance of CRP to ischemic cerebrovascular disease remains unclear. Associations with ischemic stroke depend signi cantly on conventional risk factors and other laboratory signs of in ammation [45]. Of course, treatment teams must also consider laboratory signs and the general risks of hypercoagulability.
Further, confounding atrial brillation must be taken into account because previous studies showed that newonset atrial brillation occurs more frequently in the context of sepsis [46,47]. The presence of atrial brillation was not explicitly stated in any of the included studies of our meta-analysis.  [49]. In comparison, the proportion of severe cases in the two larger COVID-19 studies was 42.7% and 17.3%, respectively [6,21].
In SARS and MERS, there have also been isolated case reports of associated infectious/in ammatory brain diseases, of which only two showed positive PCR results in CSF analyses [12,15,50]. In addition, in SARS, the virus was detected in the brain tissue of an affected patient [14] and of autopsied patients [11,13]. The current pandemic involves several isolated cases of infectious/in ammatory brain diseases associated with SARS-CoV-2, however, only two cases in which the virus was detected in the patient's CSF [26][27][28]. It is generally accepted that a positive PCR result in the CSF is an indication of direct viral infection of the brain. However, one must bear in mind that during COVID-19, a marked systemic in ammatory response syndrome (SIRS) has been described [51]. In this context, the proin ammatory cytokine storm is likely to lead to an increased permeability of the blood-brain barrier [52]. The receptor by which SARS-Cov-2 enters its host cell is the angiotensinconverting enzyme 2 (ACE2) receptor. Immunohistochemistry for the ACE2 receptor in CNS tissue, though with limited description, failed to show neuronal or glial positivity but did con rm it in the brain vasculature [8].
Hence, viral replication in the brain's microvasculature in parallel with an open blood-brain barrier could explain the CSF detection of virus particles despite the absence of neuronal/glial cell invasion. An additional receptor binding the spike protein of SARS-Cov-2 has been possibly recognized in CD147 through an in vitro experiment, and CD147 is widely present in the CNS [53]. However, taken together with the paucity of reports on meningoencephalitis associated with COVID-19, direct CNS infection seems to be rare or con ned to very special patient/virus constellations during SARS-Cov-2.
In both reported cases referenced in the current review, the CSF results were not presented su ciently to allow a more precise interpretation regarding an impairment of the blood-brain barrier, and the results of important diagnostic tests were missing (e.g., brain imaging, exclusion of further neurotropic viruses, intrathecal synthesis of antibodies) [26][27][28]. At least, the following must be considered: CSF tests for SARS-CoV-2 were rarely reported, and the test accuracy of the various test kits was not speci ed. It is to be expected that a possible positive result can also be a false positive. That is why the data extraction of case reports included whether the CSF PCR tests were repeated. Overall, in only two case studies were the PCR tests on the CSF samples repeated [24,28], and in one the results were SARS-CoV-2-positive twice [28].
Overall, our results support the assumptions of other reviews. Considering the results in context with those of previously published studies, the following implications can be outlined: The treatment team for patients with an infection of SARS-CoV-2, particularly with severe disease progression, should be aware of the development of neurological signs and symptoms. The integration of neurologists into the multiprofessional COVID-19 treatment team can help detect neurological complications early. Introducing a suitable risk assessment concerning hypercoagulability in severe COVID-19 cases may be particularly important, following a recent metaanalysis that revealed that acute cerebrovascular disease in COVID-19 patients was associated with an increased poor composite outcome and mortality [54].
Furthermore, in the pandemic situation, even if the symptoms are exclusively neurological, a SARS-CoV-2 infection should be considered, and patients should be tested accordingly.
Obviously, more clinical studies are needed to improve the current evidence of the neurological CNS manifestations of SARS-CoV-2. In our search we also identi ed ongoing studies, which can provide more detailed information, for example, on the prevalence of acute encephalopathy in severely ill COVID-19 patients or on the long-term cognitive de cits in COVID-19 patients with acute neurological symptoms [55,56].

Limitations Of The Review
Due to the urgency of the COVID-19 pandemic, we conducted a rapid review and abbreviated certain methodological steps of the review process. Speci cally, we applied a single risk of bias assessment, data extraction, and certainty of evidence rating, with a second person checking for plausibility and correctness.
There was a limitation in regard to language; full texts only in German and English were included. This concerned only two case studies. It has been proved in previous studies that the exclusion of non-English publications from systematic reviews had a minimal effect on the overall conclusions [57]. We are con dent that none of these limitations changed the overall conclusions of this review.
However, the evidence of a review is only as solid as the underlying primary studies. Due to the low number of studies, we conducted a quantitative analysis for only one outcome (ischemic stroke). The study of Li et al. was a preprint and the manuscript was not peer-reviewed [6]. For infectious/in ammatory CNS disease associated with SARS-CoV-2, we referred to case studies, as this outcome was not described in any cohort study.
Nevertheless, our numbers, especially those of the cerebrovascular events, can be used as a rst orientation.

Conclusion
Central nervous complications occur frequently in patients with COVID-19 and are most likely of parainfectious origin. It is important to integrate neurologists into the multiprofessional COVID-19 treatment team in order to detect neurological complications early and to treat them correctly.

Competing interests
The authors declare that they have no competing interests.

Acknowledgements
We wish to thank Edith Kertesz from the Danube University Krems for administrative support.   Supplementary Files This is a list of supplementary les associated with this preprint. Click to download.