Introduction

In December 2019, numerous patients had been diagnosed with unexplained pneumonia in Wuhan, China, which quickly spread to the other regions of the world. The disease, which was named as coronavirus disease 2019 (COVID-19) by the World Health Organization (WHO), has now been confirmed to have originated from a severe acute respiratory syndrome (SARS)-like coronavirus (SARS-CoV-2). By May 3rd, 2020, the pandemic had affected more than 200 countries, with 3,349,786 cases having been confirmed as COVID-19, including 238,628 deaths (data from COVID-19 Situation Report-104 by the WHO). The disease in most patients is characterized by mild to medium fever, dry cough, chest distress or dyspnea, with interstitial pneumonia features on chest computed tomography (CT) scan [1, 2]. The main routes of transmission are respiratory droplets and close contact transmission. The average of the reproductive number of COVID-19 is 3.28, based on a summary of 12 related studies, which is much higher than SARS and Middle East respiratory syndrome (MERS) [3].

As knowledge of the SARS-CoV-2 virus has accumulated, most researchers have agreed that COVID-19 is not just a respiratory disease and that it could affect other systems in human. There have recently been increasing studies on the neurological involvement of COVID-19, which could be associated with more severe symptoms and higher mortality. However, there has been no comprehensive review regarding on this topic until now. Here in this systematic review with meta-analysis, we collected and analyzed the clinical manifestations and evidence of neurological involvement in this disease.

Methods

Search strategy and study selection

This study was approved by the Ethics Committee of the Tongji Medical College, Huazhong University of Science and Technology, and the written consent was exempted because it was a secondary analysis of the published studies. In this systematic review and meta-analysis, we searched the English literature that demonstrated clinical manifestations and that provided evidence of neurological manifestations in COVID-19, according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA Checklist in Supplement materials). Given the novelty of the field, the search was conducted up to May 3rd, 2020, in two preprint databases (bioRxiv and medRxiv) and three popular medical databases (PubMed, Embase and Scopus). A combination of keywords related to COVID-19, neurological manifestations, neurological diseases and specific symptoms were used with Boolean operators for more efficient literature retrieval (search strategy in Supplement materials).

Eligibility criteria

The studies on clinical manifestations and evidence regarding neurological manifestations after COVID-19 were collected using Endnote X9 software (Clarivate Analytics), and the duplicates were removed accordingly. Case reports, case series, correspondence with relevant clinical data, and retrospective and cross-sectional studies were included for further analysis and review. The exclusion criteria included (1) clinical observation with no related neurological symptoms and a lack of related data; (2) the studies focused on a specific population (e.g., infants, children, patients with cancer, pregnant women); (3) repeated research on the same cohort of patients; and (4) other types of study (e.g., reviews, book chapters). All selected publications were comprehensively reviewed by two independent investigators to ensure their eligibility for inclusion in the review. If any disagreement occurred about eligibility of the literature, it was resolved by consensus after thoroughly discussing the publication with a third investigator.

Inclusion criteria, data extraction and quality evaluation for meta-analysis

Given that the articles reporting specific neurological symptoms (e.g., dysgeusia, agnosia) were limited, only articles that reported unspecific neurological symptoms (e.g., headache, weakness, respiratory failure) in patients with COVID-19 were selected for further meta-analysis. The quality assessment tool for case series studies designed by the National Institutes of Health (NIH) was applied to evaluate the included studies in the meta-analysis. For each item, we scored 1 point for “yes” and 0 points for “no,” “not available,” “not reported,” and “cannot determine.” The total score from all nine criteria (0–9) was generated accordingly to gauge the overall quality of the study.

Statistical analysis

Stata (version 15.1) was used to perform a meta-analysis of single arm studies. The combined prevalence and 95% confidence interval were calculated using the random-effects model. Egger’s test was performed to assess publication bias in all literature, and P < 0.05 was considered statistically significant. Subgroup analysis was done to discuss the heterogeneity of the meta-analysis.

Results

After a primary search, a total of 440 publications were retrieved, including 73 from Pubmed, 133 from Embase, 163 from Scopus, and 71 from medRxiv and bioRxiv. Among these, 142 were duplicates and were removed accordingly. We evaluated the eligibility of the remaining 298 papers according to title, abstract, article type, and full-texts (Fig. 1). The 11 papers from other resource (e.g., references list of existing papers) were added. Ultimately, 41 studies were included for systematic review and meta-analysis, including 20 studies on unspecific neurological manifestations [1, 2, 4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21], 20 studies on the specific neurological symptoms [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41], and 1 reporting both types of symptoms [42].

Fig. 1
figure 1

Study selection and characteristics

Full size image

Unspecific neurological symptoms associated with COVID-19 and meta-analysis

From a clinical aspect and the reported symptoms, headache, myalgia, fatigue, nausea, vomiting, confusion, anorexia, dizziness, malaise, dyspnea/shortness of breath were regarded as unspecific symptoms and were included in the subsequent meta-analysis.

Twenty-one eligible papers that reported unspecific neurological symptoms are summarized in Table 1, among which 19 (90.5%) were retrospective case series with one retrospective cohort (4.8%) and 1 prospective case series (4.8%). Nineteen studies (90.5%) enrolled patients from China, with the other two from Korea and India, respectively. There were 7 (33.3%) multicenter studies and 12 (57.1%) reports from a single institute. The number of enrolled patients ranged from 13 to 1099, and the sex ratio varied from 0.69 to 3.33 (male/female). The quality assessment for the literature is detailed in Supplement Table S1.

Table 1 Summarization of selected literature for meta-analysis
Full size table

We then meta-analyzed the prevalence of the nine unspecific neurologic COVID-19 manifestations in 3837 patients. For the common neurological manifestations (the number of the studies > 10 and total cases > 1500), fatigue (33.2% [23.1–43.3]) and dyspnea/shortness of breath (26.9% [19.2–34.6]) were the most prevalent symptoms, followed by myalgia (16.0% [12.3–19.8]), headache (9.2% [7.2–11.2]), and nausea/vomiting (5.1% [3.3–6.8]). Among the neurological manifestations that were reported sporadically (the number of the studies < 10 and total cases < 1500), the most common symptoms were anorexia (30.0% [23.2–36.9]), malaise (26.7% [13.3–40.1]), dizziness (10.0% [5.9–14.2]), and confusion (5.2% [− 1.7 to 12.2]), in descending order (Fig. 2). Significant publication bias was not observed in the common neurological manifestations including headache, myalgia, fatigue, nausea/vomiting, and dyspnea/shortness of breath (Fig. 2, all p > 0.05 by Egger’s test).

Fig. 2
figure 2

Meta-analysis of the prevalence of unspecific neurologic manifestations in COVID-19

Full size image

Substantial heterogeneity was observed in most symptoms within subgroups (Fig. 2 and Supplement Tables S2–5). In the subgroup analysis, dyspnea/shortness of breath was more prevalent in the studies performed in Wuhan, on elder patients (median or mean age > 47 years), with higher M/F ratio (M/F ratio > 1.38), and higher percentage of severe cases (> 15%) (p < 0.05). In addition, fatigue was more frequent in the severe cases and studies in Wuhan (p < 0.05), and younger patients tended to experience more headaches. No other potential sources of heterogeneity of these five unspecific neurological manifestations were identified.

Specific neurological symptoms associated with COVID-19

We retrieved 21 studies regarding the specific neurological symptoms and diseases after COVID-19 (summarized in Table 2), among which 6 (28.6%) studies were case series and the rest (71.4%) were cases reports. In the six cases series, two studies focused on a relatively wide spectrum of neurological manifestations (e.g., central nervous system [CNS], peripheral nervous system [PNS], neuromuscular disorders) [36, 42], with the other four focused on olfactory and gustatory dysfunction [23, 24, 29, 37]. In the 15 case reports, 5 were associated with Guillain–Barré syndrome [31,32,33,34,35], 5 focused on neurological inflammation (e.g., myelitis, encephalitis, meningitis) [22, 26, 28, 30, 39, 41], 2 on cerebrovascular diseases [27, 38], and the rest on specific neurological symptoms [25, 40].

Table 2 Summarization of studies on specific neurological manifestations
Full size table

Laboratory, electrophysiological, radiological, and pathological evidence of neurological manifestations after COVID-19

Eleven papers that demonstrated laboratory, electrophysiological, radiological, and pathological changes after COVID-19 were distilled from the summarized literature, including seven on the examination of cerebrospinal fluid [30, 32, 34,35,36, 39, 41], three on electroencephalogram [36, 39, 41], two on nerve conduction [32, 34], six on magnetic resonance imaging (MRI) scans [26, 27, 30, 32, 36, 40], two on CT images [27, 40], and one post-mortem examination [25] (summarized in Table 3).

Table 3 Laboratory, electrophysiological, radiological, and pathological evidence of neurological manifestations after COVID-19
Full size table

Discussion

To our knowledge, this is the first systematic review with meta-analysis of more than 41 studies involving approximately 4700 patients that provides a comprehensive view of neurological manifestations in COVID-19. In comparison with previous review and proposal on the topic, both clinical manifestations and related evidence were demonstrated to investigate multifaceted mechanisms underlying neurological involvement in COVID-19.

After the primary exploration, the neurological manifestations in COVID-19 were found to mainly fall into three categories: (1) neurological diseases comorbid with COVID-19, in which neurological symptoms occur prior to the infection that also make patients themselves susceptible to COVID-19 due to frequent exposure in medical facilities and suboptimal health status (e.g., cerebrovascular diseases, neural trauma); (2) unspecific neurological manifestations, which can be caused by systematic responses and partially by the neuroinvasive behavior of the infection (e.g., headache, myalgia, fatigue); (3) specific neurological symptoms and diseases that were due to neurological involvement in COVID-19 (e.g., encephalitis, myelitis, seizures). This review mainly focused on the last two categories of COVID-19 neurological manifestations.

Unspecific neurological manifestations in COVID-19

Unspecific neurological manifestations are various and insidious in COVID-19. It has been reported that unspecific neurological manifestations occurred in the early onset of the infection and could serve as the primary and only symptom when COVID-19 patients were admitted to the hospital [32]. Neurologic manifestations were also more prevalent in the patients with severe disease, which increased the risk of transmission and could lead to death if not handled properly and effectively [42]. However, there is a wide symptomatic spectrum of neurological manifestations in COVID-19, and it is easy to misdiagnose them during the epidemic due to the respiratory nature of the virus. After a systematic exploration of the literature, we found that the headache, myalgia, fatigue, nausea/vomiting, confusion, anorexia, dizziness, malaise, and dyspnea/shortness of breath were reported as unspecific neurological symptoms. Among them, the dominant clinical manifestations were fatigue, dyspnea/shortness of breath, anorexia, and malaise, which affected approximately one-third of patients with COVID-19.

Whether dyspnea/shortness of breath is a neurological symptom is still a matter of debate. Recently, several patients with dyspnea in an intense care unit (ICU) subjectively described that active and conscious breath was needed to maintain a normal breathing rhythm [43]. As we know, angiotensin-converting enzyme 2 (ACE2), an important SARS-CoV-2 receptor, is found to be expressed in the cardiorespiratory center of the medulla together with other brain regions [44, 45]. Given the high similarity among coronaviruses and the fact that other coronaviruses (e.g., SARS-CoV, MERS-CoV) could invade the nervous system [46], it is reasonable to speculate that SARS-CoV-2 virus might play a role in the dyspnea/shortness of breath via affecting the respiratory center of the medulla, especially in those patients with rapid disease progression but without severe respiratory symptoms. Thus, dyspnea/shortness of breath was included as one of the neurological manifestations in this review, and we observed that it was one of most prevalent neurological symptoms in COVID-19 infection. However, further clinical and experimental studies should be performed to differentiate the exact cause of dyspnea and to explore its potential association with neurological involvement.

Specific neurological manifestations and diseases associated with COVID-19

In this review, we found six case series focused on the specific neurological manifestations, including CNS-related symptoms (e.g., impaired consciousness, acute cerebrovascular disease, corticospinal tract signs, ataxia, and seizure), PNS-related symptoms (e.g., taste impairment, smell impairment, vision impairment, and nerve pain), and musculoskeletal injury [23, 24, 29, 36, 37, 42]. Interestingly, olfactory dysfunction and gustatory disorders were observed 30–80% in patients with mild COVID-19. Recently, Fodoulian et al. and Brann et al. had reported that ACE2 and SARS-CoV-2 entry genes were expressed by sustentacular cells in the human olfactory neuroepithelium [47, 48]. In addition, bilateral inflammatory response in the olfactory clefts was observed in a patient with COVID-19 with sudden olfactory loss but without nasal obstruction by MRI and CT scans, which suggests the virus might affect olfactory function via direct invasion of the olfactory neuroepithelium [40]. However, we lack definitive evidence regarding the potential association between the PNS symptoms and the risk of CNS infection.

Regarding the neurological diseases, we found that aseptic neuroinflammation appeared to be related to COVID-19, including Guillain–Barré syndrome, Miller Fisher syndrome, myelitis, meningitis, and encephalitis [22, 30, 32, 34, 35, 39]. Notably the polymerase chain reaction (PCR) tests of cerebrospinal fluid (CSF) were negative in all cases. The mechanism for the neuroinflammation in the patients with COVID-19 is unclear. However, in a recent report on a rare condition (hemorrhagic necrotizing encephalopathy) in COVID-19, the author inferred that the breakdown of the blood–brain barrier (BBB) by cytokine storm could be the primary cause of disseminated brain necrosis and hemorrhage in this case [27]. Together with the increase in protein levels and immunoglobulin in CSF, we suspect that both the cytokine storm secondary to the systematic infection and the overactivation of the immune system might play a role in this process.

Evidence for neurological involvement of COVID-19

Currently, our knowledge regarding on neurological involvement of COVID-19 is still limited. The definitive evidence for the invasive behavior of COVID-19 in CNS is mainly from two neuropathological reports. Paniz-Mondolfi et al. had reported the presence of SARS-CoV-2 in brain tissue from post-mortem examination of a COVID-19 patient by employing a transmission electron microscope. Viral particles of approximately 80–110 nm, which were pleomorphic and spherical with distinct stalk-like projections and perfectly matched the structural features of SARS-CoV-2, were spotted in the frontal lobe section of the brain [25]. Notably, these viral particles were found to be located not only in the cytoplasmic vacuoles of neuronal cells but also in the small vesicles of endothelium cells, which suggested this direct neuroinvasive behavior via hematogenous pathways might be a leading cause for rapid progression of neurological symptoms [25]. In consistence with this finding, SARS-CoV-2 was also detected in CSF via PCR examinations in a male patient suffering from consciousness disturbance and transient generalized seizures, with typical meningitis and encephalitis characteristics shown on the MRI [30]. Interestingly, this case presented negative PCR result in nasopharyngeal swab at the time point, which indicated that CNS invasion might happen in the early phase of COVID-19 [30]. However, several similar studies had reported negative findings in the CSF examination, which suggests that direct neuroinvasion of SARS-CoV-2 is not a universal phenomenon for COVID-19 [32, 34]. But the interpretation of these negative results should be cautious due to the limitation of PCR examinations, which is currently the most prevalent test for viral detection. Its low sensitivity in CSF specimen may lead to repeated negative findings. Thus, more investigations are needed to further confirm this direct invasive behavior.

The adjuvant evidence for neurological involvement in COVID-19 includes CSF examination, electrophysiology, and radiological findings. The characteristic change in CSF after COVID-19 is the slight increase in cell counts and protein levels, especially immunoglobulins, which suggests its inflammatory or infective status [30, 32, 34,35,36, 39]. For electrophysiology, two cases presented unspecific alterations, another patient exhibited focal seizure in electroencephalogram [36, 39], and the rest showed delayed nerve conduction, attenuation of action potential amplitude, and absent F-Waves, which are associated with the damage to myelin and axons in Guillain–Barré Syndrome [32, 34]. In the radiological examination, MRI and CT scans had been commonly adopted to search for clues for neurological disorders after COVID-19, such as inflammation and cerebrovascular dysfunction, which suggest the direct invasion of SARS-CoV-2. In the patients diagnosed with encephalitis, myelitis, meningitis, and Guillain–Barré Syndrome, enhancement of lesion sites in meninges, brain, spinal cord, and nerve roots was detectable in MRI and correlated with individual clinical manifestations [30, 32, 36]. Poyiadji et al. reported a unique case of acute hemorrhagic necrotizing encephalopathy associated with COVID-19 and the patient presented bilaterally hemorrhagic lesions within multiple brain regions, which implied that the breakdown of the BBB might serve as an explanation of COVID-19-associated cerebrovascular diseases [27]. In addition, perfusion abnormalities appeared to be prevalent (100%) in the patients with COVID-19 with severe neurological symptoms and were not only found in the patients with cerebrovascular disorders [36].

Potential mechanism underlying the neurological manifestation in COVID-19

From the limited studies on the neurological involvement of COVID-19, SARS-CoV-2 could impair the nervous system via two potential pathways: (1) direct invasion of neural tissue and (2) maladaptive inflammatory responses (Fig. 3).

Fig. 3
figure 3

The potential mechanism underlying the neurological manifestation in COVID-19. Based on the current evidence, there are two routes for the neurological involvement of COVID-19. (1) SARS-CoV-2 could direct infect nervous system via hematogenous and neural retrograde pathways; (2) SARS-CoV-2 overactivates the immune system, and secondary cytokines storm and immunoglobins impair the nervous system

Full size image

In terms of direct invasion, SARS-CoV-2 could affect the CNS mainly through the hematogenous and the neuronal retrograde routes. First, ACE2 was found to be enriched in the endothelium cells of the CNS. Thus, viruses might invade nervous systems by infecting the endothelial cells of the BBB and the blood–CSF barrier in the choroid plexus. This hypothesis was strongly supported by the study conducted by Paniz-Mondolfi et al. [25]. The authors captured the viral particles using a cytoplasmic vacuole at the endothelial neural cell interface in a transmission electron microscope, which suggests that SARS-CoV-2 could bind to vascular endothelium, penetrate the BBB, and invade nervous tissues through hematogenous pathways. However, some researchers believe that the hematogenous route is not the only pathway, and that the neuronal retrograde transmission serves as another potential mechanism. In this route, SARS-CoV-2 might first invade peripheral nerve terminals and infect the CNS retrogradely. Previous animal studies on SARS-CoV and OC43-CoV had suggested the virus could enter the brain by disruption of the nasal epithelium and lead to subsequent neuronal dissemination when administered intranasally [46] [49]. Due to the high similarity among SARS-CoV, SARS-CoV-2, and OC43-CoV, it is likely that this same route also works in COVID-19. Notably, anosmia and dysgeusia were reported recently to be prevalent among patients with COVID-19 despite a lack of upper respiratory symptoms [23, 37]. In addition, ACE2 and other entry genes of SARS-CoV-2 are found to be highly expressed by sustentacular and non-neuronal cells in the human olfactory neuroepithelium [47, 48]. However, definitive evidence for this transmission route is lacking, and more solid pathological and experimental investigations should be conducted to validate this pathway.

Maladaptive inflammatory responses can also contribute to the neurological manifestations of COVID-19, especially in those patients in whom no virus RNA is detected in CSF. In the early phase of COVID-19, the increased secretion of inflammatory cytokines (e.g., IL-1β, IFN-γ, TNF-α, IL-4, IL-10) was a signature feature caused by rapid viral replication and secondary cellular injury. In addition, higher plasma levels of these inflammatory cytokines were observed in the ICU patients, which suggests that a cytokine storm presented in the severe patients [12]. This cytokine storm could induce various neurological symptoms and was previously reported to be able to disrupt the BBB and induce neuroinflammation in the sepsis [27, 50, 51]. In addition, the neutralizing antibodies (anti-S protein lgG), which are assumed to clear the virus and promote the recovery, can also cause severe secondary injury by altering inflammatory responses. In a previous study on SARS-CoV, anti-S protein lgG was found to facilitate severe tissue injury in other organs, such as the lung [52, 53]. In a recent case series of neurologic features in severe SARS-CoV-2 infection, oligoclonal bands and elevated IgG were detected in the CSF, which suggests that the activation of the immune system might be a double-edged sword [36].

Limitations

This systematic review with meta-analysis has several limitations. First, in the meta-analysis of the unspecific manifestations we included all the related symptoms, some of which were not classical neurological presentations. Neurological involvement could contribute partially to these symptoms, and an overall analysis may exaggerate its role in COVID-19. Second, in the systematic review, most of the included papers on clinical characteristics were of retrospective design and were performed in China, which could introduce potential bias. Third, the studies on specific neurological symptoms and diseases were mostly case reports and were limited. Fourth, during the outbreak of COVID-19, the influx of numerous COVID-19 patients made advanced neuroimaging, CSF tests, electrophysiological tests, and pathological examinations impractical due to high risk of cross-infection, which limited the evidence of neurological involvement. Fifth, this study was conducted during an ongoing outbreak. Many related studies have not yet been published, which could influence the results. The potential mechanism underlying the neurological manifestation in COVID-19 will be updated along with emerging evidence.

Conclusions

The neurological manifestations are various and prevalent in COVID-19, but are usually underestimated. Emerging clinical evidence suggests that neurological involvement is an important aspect of the disease. The multifaceted mechanisms are involved in its neurological impact, which includes both direct invasion and maladaptive inflammatory response. Nevertheless, more clinical and experimental research should be conducted to further explore the role of neurological manifestations on the disease progression and its underlying mechanism.