New-Onset Movement Disorders Associated with COVID-19

Introduction: Movement disorders are increasingly described in hospitalized and milder cases of SARS-CoV-2 infection, despite a very low prevalence compared to the total patients. Methods: We reviewed the scientific literature published in English, spanning from the initial descriptions of COVID-19 until January 25, 2021, in the PubMed/MEDLINE database. Results: We identified 93 new-onset movement disorders cases (44 articles) from 200 papers screened in the database or reference lists. Myoclonus was present in 63.4% (n = 59), ataxia in 38.7% (n = 36), action/postural tremor in 10.8% (n = 10), rigid-akinetic syndrome in 5.38% (n = 5), oculomotor abnormalities in 20.4% (n = 19), catatonia in 2.1% (n = 2), dystonia in 1.1% (n = 1), chorea in 1.1% (n = 1), functional (psychogenic) movement disorders in 3.2% (n = 3) of the reported COVID-19 cases with any movement disorder. Encephalopathy was a common association (n = 37, 39.78%). Discussion: Comprehensive neurophysiological, clinical, and neuroimaging descriptions of movement disorders in the setting of SARS-CoV-2 infection are still lacking, and their pathophysiology may be related to inflammatory, postinfectious, or even indirect mechanisms not specific to SARS-CoV-2, such as ischemic-hypoxic brain insults, drug effects, sepsis, kidney failure. Cortical/subcortical myoclonus, which the cited secondary mechanisms can largely cause, seems to be the most common hyperkinetic abnormal movement, and it might occur in association with encephalopathy and ataxia. Conclusion: This brief review contributes to the clinical description of SARS-CoV-2 potential neurological manifestations, assisting clinical neurologists in identifying features of these uncommon syndromes as a part of COVID-19 symptomatology. Highlights – Movement disorders are probably uncommon neurological manifestations in SARS-CoV-2 infection; – Myoclonus is the most reported movement disorder associated with COVID-19, its clinical complications or pharmacological management; – The pathophysiology is yet not well-understood but can include systemic inflammation, autoimmune mechanisms, or hypoxia.


BACKGROUND
Coronavirus disease 2019 (COVID-19) has been associated with various neurological symptoms [1,2]. Numerous putative dissemination mechanisms of the central nervous system (CNS) were hypothesized, including direct retrograde spread through the olfactory nerve via the transcribiform pathway or hematogenous spread via the blood-brain barrier. The neuronal damage could happen through different but not mutually exclusive mechanisms, including direct viral insult, cytokine release syndrome, hypoxia, immune-mediated neuroinflammation (post or para-infectious), coagulopathy, and endothelial dysfunction syndrome [1]. Headache, altered mental status (encephalopathy), chemosensory dysfunction (anosmia, ageusia), cerebrovascular events, and myalgia seem to be the most typical neurological manifestations [2]. However, the description of the isolated virus in the cerebrospinal fluid (CSF) is seemingly rare, suggesting that direct viral meningoencephalitis is not usual.
Movement disorders are increasingly being described not only among hospitalized patients but also in milder cases of SARS-CoV-2 infection, despite a possible very low prevalence in comparison to the total cases [3,4]. This study aims to summarize and describe, through a systematic procedure, relevant clinical and ancillary exam findings in patients with new-onset movement disorders associated with COVID-19.

LITERATURE SEARCH AND REVIEW
We first searched the PubMed/MEDLINE database for relevant literature published in the English language, between the initial descriptions of COVID-19 (December 31, 2019) and January 25, 2021, using the search terms "COVID-19" OR "SARS-CoV-2" OR "Coronavirus Disease 2019" OR "2019 n-CoV" OR "2019 Novel Coronavirus" AND ("movement disorders" OR "myoclonus" OR "ataxia" OR "parkinsonism" OR "Parkinson's disease" OR "chorea" OR "dystonia" OR "myoclonus" OR "catatonia" OR "tremor")" as search terms (n = 332). We added filters for Case Reports, Classical Articles, Clinical Studies, Clinical Trial, Comparative Studies, Original Journal Articles, Letters, Observational Studies, Human Species and identified 296 records for screening (titles/ abstracts). We conducted a prospection in the reference list of selected papers and other sources, identifying additional 19 items. After excluding 115 records with abstracts and titles irrelevant to this study's scope, we screened 200 records for eligibility. We evaluated 81 full-text papers for eligibility after withdrawing 119 articles for the following reasons: they were not written in English or were focused on social or healthcare impact of COVID-19 in patients already diagnosed with a movement disorder, dealt with aggravation of pre-existing movement disorders, telemedicine management strategies, or were listed as review/hypothesis/opinion articles ( Figure   1). A total of 44 studies were included for qualitative and quantitative synthesis as representative of neurological descriptions of new-onset movement disorders in the context of SARS-CoV-2 infection.

DATA EXTRACTION, SYNTHESIS, AND ANALYSIS
Two authors (PB and TG) collected data independently, using a predefined protocol. One of the five authors (PB) checked the extracted data, and discrepancies were resolved by consensus between the two data collectors. Variables of interest included counts for single or combined movement disorders (myoclonus, dystonia, ataxia, parkinsonism, chorea, catatonia, action/postural tremors, oculomotor disturbances), associated clinical features (encephalopathy, stroke), summaries of ancillary exam results (FDG-PET, MRI/CT, ENMG, CSF, and EEG). The attempted treatment options (symptomatic treatment or immunotherapy, such as intravenous immunoglobulin -IVIg, steroids) were also counted. Clinical characteristics were quantified as case counts and ratios of case counts to total cases. Symptoms not initially reported in individual studies were considered absent (zero) rather than missing (NA).

Highlights
-Movement disorders are probably uncommon neurological manifestations in SARS-CoV-2 infection; -Myoclonus is the most reported movement disorder associated with COVID-19, its clinical complications or pharmacological management; -The pathophysiology is yet not well-understood but can include systemic inflammation, autoimmune mechanisms, or hypoxia.
R package was used to obtain the parameter values' Bayesian posterior distribution [5]. "Brms" uses the logitlink to convert the linear predictor zi (zero-inflated) to a probability. The logit-link accepts values between 0 and 1 and returns values on the real line. As a result, it enables the conversion of probabilities to linear predictors. Samples were drawn using No-U-Turn-Sampler (NUTS) sampling, the default algorithm used by the Stan program [6]. Rhat was used as a potential scale reduction factor on split chains (at convergence, Rhat = 1). We fit the model using four Hamiltonian Monte-Carlo (HMC) chains. We used Jeffrey's beta (0.5, 0.5) as the prior argument for zi rather than the default "flat" beta (1, 1) used by "brms". The intercept parameter prior was normal (0, 10). The Bayesian plot was created using the "bayesplot" R package (https://mc-stan. org/bayesplot/) [7]. Using counts as the measure of movement disorders frequency (main variables extracted from the articles), we performed an unsupervised grouping procedure (clustering) to separate and observe co-occurrence between the symptoms in the reported articles. The "NbClust" R package was used to determine the optimal number of clusters [8]. It provides 30 indices for that task and proposes the best clustering scheme by varying all combinations of several clusters, distance measures, and clustering methods (Supplementary Figure 1). We then employed a random forest (RF) clustering strategy to avoid variable transformation (e.g., categorical features). By forming two unique groups (as determined by the NbClust's method), RF clustering separated the original data from the virtual clusters. The similarity between two data instances was assessed using the proportion of trees that shared a single leaf node. The PAM (partition around medoids) technique [9] was used to assign each object to the closest medoid (object with the smallest dissimilarity) as a step to create the clusters.

AUTHORS
The major findings are depicted in Figure 3 as Bayesian binomial posterior proportions. Table 1 summarizes the clinical and ancillary exam data from each original study to serve as a simple reference guide for physicians and researchers.

CLINICAL CHARACTERISTICS
Of the 93 movement disorder cases identified in the literature search [3,4,, myoclonus was present    (Figure 2). Encephalopathy was a frequent association (n = 37, 39.8%), and stroke was detected in a minority of cases (n = 5, 5.4%).
Not surprisingly, confusional or encephalopathic states were highly prevalent in patients with action and stimulussensitive myoclonus. Myoclonus is a frequent involuntary movement in critically ill patients triggered or aggravated by several different drugs or by metabolic or hypoxia-related brain dysfunction [53]. Encephalopathy and delirium are very prevalent in critical care [54], and the reviewed sample has a significant proportion of hospitalized and critical patients. Three cases of new-onset rigid-akinetic parkinsonism were described in detail [10,11,26], one of which presented with encephalopathy and generalized myoclonus [11], two of which were preceded only by a mild respiratory syndrome [10,26].
Some de novo movement disorders were presumably drug-induced: two serotonin syndrome cases possibly related to lopinavir/ritonavir combined with drugs with serotonergic properties, and three tardive syndrome cases (described as rigid-akinetic syndromes combined to orobuccal stereotypies) [13]. Due to incomplete information, it was not feasible to establish a causal link between myoclonus and drug administration in critically ill mechanically ventilated patients, in whom sedative agents, especially opioids, or antipsychotic medications, may have contributed to myoclonus.
FDG-PET scan was performed in five patients, as described below. It showed normal glucose metabolism in a case with parkinsonism [10]. Two of these five scanned patients had myoclonus and ataxia [34,43], one had ataxia and other cerebellar signs [43], and one had orofacial myoclonus [47]. FDG-PET revealed hypermetabolic cerebellum in all four cases: two also had hypermetabolism in the basal ganglia, while the other three had cortical hypometabolism.
Antineuronal paraneoplastic antibodies (Abs) were not detected in any of the cases when a standardized Abs panel was tested (n = 9). Antibodies against Purkinje cells, striatal, and hippocampal neurons were observed in one patient with subacute cerebellar syndrome and myoclonus utilizing tissue immunostaining [34]. Franke et al. used indirect immunofluorescence to screen antibodies against novel CNS epitopes in unfixed mouse brain sections. They found staining associated with vessel endothelium, astrocytic proteins, and neuropil of basal ganglia, hippocampus, or olfactory bulb [47]. SARS-CoV-2 RT-qPCR positivity in CSF samples appears to be uncommon: it was detected in only one case of cerebellitis [23].

NEUROPHYSIOLOGICAL TECHNIQUES (EEG AND EMG)
EEG findings were available in 29% (n = 27) of the cases. Of these, 59.3% (n = 16) had background slowing, while 33.3% (n = 9) had normal results. In one case, lateralized periodic discharges were observed in association with diaphragmatic myoclonus [31], while in another, frontal spikes were observed [21]. Two studies (three cases) used neurophysiological measures to characterize the possible CNS topographic origin of myoclonus [15,31], using a combination of EEG and EMG or jerk-locked registers. Two of the three cases had cortical/subcortical origin, and one, a subcortical source.

TREATMENT ATTEMPTS
In 39.8% of the cases (n = 37), the information about treatment attempts was incomplete or missing. There was immunotherapy with oral/intravenous steroids (n = 21, 37.5%). In 11 cases, the corticosteroid therapy targeted the movement disorder; in 5 cases, it targeted COVID-19related inflammatory reaction and, in 4 cases, the target was not clear. Immunoglobulin (IVIg) was used in 32.14% (n = 18) of the reported cases with available treatment information (n = 56). In all of these cases, IVIg therapy was targeting the movement disorder. Plasma exchange was tried in one case. Immunotherapy was used primarily in inpatient settings, but cases ranged from mild to critically ill in mechanical ventilation.

BAYESIAN REGRESSION USING ZERO-INFLATED POISSON DISTRIBUTION
In this analysis, we generated a regression using Bayesian zero-inflated Poisson (ZIP), a model for counts-based datasets with an excess of zero-valued data in the dependent variable (assuming a constant zero-inflation probability across observations). A value of "zero" indicates that nothing occurred, which may be because the events' rate is low or because the process that causes events failed to start.
Within the data extracted from the sample of COVID-19 associated with new-onset movement disorders, myoclonus was associated with normal neuroimaging results (head CT or brain MRI) and encephalopathy (left side of Figure 4)

DISTINCT SYNDROMES ASSOCIATED WITH MYOCLONUS
Cluster analysis was utilized to infer movement disorders syndromic grouping from the reviewed articles. According to a seven indices ensembling method (NbClust), we found two as the optimal number of clusters. The PAM method then evaluated dissimilarity among objects. By assembling two unique groups, the random forest (RF) clustering approach [53] identified 36 articles with a higher proportion of myoclonus reports associated with encephalopathy and eight articles with a higher proportion of myoclonus connected with ataxia or oculomotor disturbances.
We interpreted these findings as suggestive of at least two myoclonus-associated syndromes that might occur in COVID-19: (1) a syndrome that associates myoclonus and encephalopathic state (which, we speculate, may relate to with broader brain dysfunction and neuronal hyperexcitability due to clinical complications in critical care or pharmacological interactions); and (2) a post-viral syndrome with an autoimmune origin that links oculomotor disturbances, ataxia, and myoclonus, resembling an opsoclonus-myoclonus-ataxia syndrome.
It is critical to emphasize that in cases stated to be associated with COVID-19, several factors can contribute to the incidence of myoclonus: kidney failure, usage of provoking drugs in the context of intensive care, such as antibiotics, fentanyl, propofol, or phenytoin, as well as prolonged and sustained hypoxia [33,56].
Muccioli et al. conducted a comprehensive clinical and paraclinical assessment, suggesting that myoclonus could have a subcortical origin [30]. Interestingly, a proportion of myoclonus cases seemed to present in a delayedonset compared to the infectious syndrome [28][29][30]; concurrent ataxic manifestations were not unusual. Recent publications have emphasized the description of opsoclonus-myoclonus-ataxia (in its complete or incomplete clinical syndrome) and included clinical improvement after IVIg therapy [27,50,51]. Myoclonic jerks were usually symptomatically treated with antiepileptic drugs (levetiracetam and valproate) and benzodiazepines (clonazepam, lorazepam).
According to a published comment on this topic, myoclonus's underlying pathophysiology in COVID-19 may have a post-hypoxic or a postinfectious nature [56]. Although a putative physiopathology involving antigenic cross-reactivity with neuronal proteins from the brainstem or cerebellum remains hypothetical, an interesting case series from Germany provides initial support for this thesis [47]. The CNS's postulated role in the production of myoclonus could be attributed to brainstem hyperexcitability or even lack of cerebellar inhibitory output [56]. Normal structural neuroimaging findings are consistent with the hypothesis that some of these movement disorders may be mediated by post-viral immune processes rather than structural lesions.
In a small number of cases, the CSF profile was examined. A few patients presented with mild mononuclear pleocytosis. The most frequently observed phenotype was normal cell count, mild hyperproteinemia, and normal glucose levels. The rise in CSF protein levels may be caused by blood-CSF barrier disruption, intrathecal immunoglobulin synthesis, or even be secondary to pre-existing medical comorbidities. However, further research is needed to ascertain an intrathecal immunological or parainfectious reaction, using techniques such as Reiber diagram and comprehensive autoantibodies analysis [57].
FDG-PET results provide some insights: they indicate that in some cases, presenting as movement disorders, an active ongoing metabolic process involving primarily the cerebellum and basal ganglia is implicated [34]. Additionally, they support a subcortical origin for myoclonus in COVID-19. Further research is required to validate this idea and better understand these metabolic PET patterns [32].
Parkinsonism related to SARS-CoV2 seems to be a very unusual manifestation. Brundin et al. hypothesized that parkinsonism links with COVID-19 through the following mechanisms: (a) vascular insults to the nigrostriatal system (if with imaging compatible with stroke); (b) neuroinflammation triggered by systemic inflammation; (c) neuroinvasion and direct neuronal damage by the virus, that could gain access to the brain via olfactory nerve or gastrointestinal/respiratory tract via the vagus nerve [58].
The three initial cases linking COVID-19 and parkinsonism had acutely developed severe symptoms, which could suggest a direct link between nigrostriatal damage and SARS-CoV-2 but are not indeed able to prove a causal relationship between COVID-19 and Parkinson's disease [10,11,26]. Prodromal or presymptomatic PD could be unmasked when individuals are acutely ill. We could not reject that this might be the case in the reported patients with "acute onset" parkinsonism. Cohort studies of COVID-19 persistent anosmic cases, followed up for years, will help to disentangle such a link.
Serotonin syndrome was documented in two patients due to combined drugs for COVID-19 and other comorbidities [11]. Due to the high prevalence of psychiatric disorders, it is critical to consider this a possible differential diagnosis when myoclonus is between the clinical signs, as it is a treatable condition.
Neuropathological assessment of COVID-19 came initially from small case series that described hypoxic injury, blood-brain barrier breakdown, microscopic infarcts, hemorrhagic and demyelinating white matter lesions, and a scarcity of RT-qPCR positivity in brain samples [59][60][61]. Microglial activation, microglial nodules, and neuronophagia have also been described [62]. To our knowledge, no neuropathological study has been conducted to precisely examine the relationship between SARS-CoV-2 and movement disorders during acute or post-acute infection.
Our study has several limitations: heterogeneity of the sample (with a predominance of case reports and case series), a bias towards hospitalized patients with severe COVID-19 (requiring consultation with a neurology specialist), heterogeneity of available information, lack of a control group (composed of patients with similar infectious conditions, for example) for prevalence rates comparison. Even with a standardized data collection protocol, we could not address the issue of missing information. Thus, these findings produce hypotheses for future research rather than lead to causal inferences or overinterpretation.
Movement disorders were early (in the pandemic's course) predicted to manifest in COVID-19 patients. However, they represent uncommon or underreported phenomena concerning the infected population as a whole [63]. This brief study contributes to the clinical scenario of SARS-CoV-2. It guides general practicing neurologists to the potential identification of these uncommon movement disorders syndromes as part of COVID-19 symptomatology or a consequence of its pharmacological management or clinical complications. Myoclonus and ataxia deserve special attention in upcoming studies: these two manifestations were the most prevalent in the reviewed data, could appear in combination, and may have an immune origin. The direct link between movement disorders and SARS-CoV-2 (and its underlying mechanisms) will be established or disproved in still lacking comprehensive neurophysiological, molecular, pathological, and neuroimaging studies with large samples.

ADDITIONAL FILE
The additional file for this article can be found as follows:

COMPETING INTERESTS
The authors have no competing interests to declare.

AUTHOR CONTRIBUTIONS
PB and TG assisted in the research design, data analysis, and manuscript writing; DP, RM, and FC assisted in the data analysis and manuscript writing. The published version of the manuscript was read and approved by all authors.