Neurologic and neuroimaging findings in patients with COVID-19

Discussion

This is the first large, nationwide cohort of MRIs performed in patients with COVID-19 with neurologic manifestations.

The majority of patients had MRI abnormalities with serious and various findings beyond the severe respiratory disease (half of the patients had ARDS, and 11% died): cerebrovascular disease (especially ischemic stroke; large artery infarctions more frequently than watershed cerebral infarctions), encephalitis (including limbic encephalitis, radiologic ADEM, CLOCC, and radiologic acute hemorrhagic necrotizing encephalopathy), and focal or even diffuse LME. The heterogeneity of the imaging abnormalities was underpinned by heterogeneous clinical manifestations, ranging from anosmia, ageusia, or headache to more severe findings such as confusion with agitation, loss of consciousness, and corticospinal tract signs. Some correlations were found between clinical and imaging findings, allowing the identification of clinicoradiologic profiles of patients with COVID-19 displaying neurologic manifestations.

Ischemic stroke was 1 kind of clinicoradiologic phenotype; it was an acute event, arising in older patients who more frequently had corticospinal tract signs but was less frequently requiring oxygen or presenting an ARDS. This profile could be distinguished from the 2 others, LME and encephalitis. The encephalitis profile affected younger patients and seemed to be particularly severe because it was associated with the need for oxygen, ARDS, and death.

Stroke patients less frequently had ARDS, and this result is in agreement with the 6 patients with stroke previously reported in that only 2 patients were admitted to ICUs.¹⁶ One may hypothesize that strokes in patients with COVID-19 could be due to procoagulant events leading to thrombosis¹⁷ rather than to the systemic inflammatory process that accompanies ARDS, which is also directed against the CNS. However, this needs to be assessed by further studies dedicated to stroke and thrombosis in patients with COVID-19.

In our study, 17 (27%) patients had an acute ischemic stroke, as was previously reported with SARS-CoV,¹⁸ Middle East respiratory syndrome coronavirus (MERS-CoV),¹⁹ and SARS-CoV-2,^11,16,17 and among them, 11 had large artery infarctions (including 6 proximal artery occlusions, 2 cases of internal carotid artery dissection or occlusion) and 6 had watershed cerebral infarctions. For 15 of 17 patients with an ischemic stroke, the respiratory symptoms related to COVID-19 had begun before this acute event, while stroke preceded respiratory symptoms by 2 days for 2 patients.

Several mechanisms are likely to be associated. It is now known that SARS-CoV-2 could directly lead to myocardial injury, promoting cardiac arrhythmias associated with embolic events.²⁰ Six of our patients had arrhythmias, which were previously unknown for 3 of them. In the same way, a severe acute myocardial injury may be associated with a decrease in brain perfusion and therefore with watershed cerebral infarctions. As previously mentioned, it is acknowledged that viruses, particularly SARS-CoV-2,²¹ are associated with an increase of prothrombotic events such as ischemic stroke.^22,23 Thereby, viral infections may elevate procoagulant markers, leading to thrombosis, disseminated intravascular coagulation, and hemorrhagic events.²³ As Beyrouti et al.¹⁶ recently mentioned, all our patients tested showed elevated D-dimers, markedly elevated (>1000 µg/L) for 10 of the 11 patients tested.

Varicella zoster virus is also associated with direct vascular involvement and eventually arteritis promoting ischemic stroke events.²² Furthermore, the more severe critically ill patients hospitalized in ICUs have probably had hypoxemia and hypotension, leading to brain injuries.¹⁸

Of the 36 abnormal MRIs, 8 diagnoses of encephalitis were made, and LME was described, especially on postcontrast T1-weighted and FLAIR, in 11 patients, either focal (single focus vs multiple foci) or diffuse, and preferentially located in the posterior, supratentorial regions of the brain.

Two main pathophysiologic hypotheses can be advanced to explain the neuroimaging findings in the inflammatory profile group. First, SARS-CoV-2 may have a neuroinvasive potential similar to other hCoVs because this family of viruses is related to each other genetically. It is not clear if the virus can infiltrate the CNS with active replication and direct damage to brain cells (viral encephalitis), as observed with other viruses such as herpes simplex virus. Likewise, viral meningitis due to direct infiltration of the virus into the CSF may also be considered because it was previously reported with SARS-CoV-1. Indeed, SARS-CoV-1 RNA was detected in the CSF of 2 patients with seizures.^24,25 A recent case report¹⁰ has described for the first time a case of seizures with the detection of SARS-CoV-2 RNA in the CSF. Similarly, leptomeningeal inflammation may be present adjacent to subpial cortical lesions in the case of viral meningoencephalitis. In the case of viral encephalitis, CSF analysis usually demonstrates lymphocytic pleocytosis.²⁶ In addition, the CSF protein level is typically elevated in viral encephalitis. CSF viral RT-PCR is usually positive but can be negative if done too early. This was not the case in our series, and our results are in accordance with recently published data that demonstrated an absence of CSF pleocytosis and negative SARS-CoV-2 RT-PCR (5 of 5 patients) but elevated protein level (4 of 5 patients) in patients presenting cerebral MRI cortical abnormalities.¹³ MRI pictures of viral encephalitis vary with causal pathogen. Nevertheless, viral encephalitis usually involves cerebral GM with diffusion restriction and can be associated with intralesional and diffuse leptomeningeal contrast enhancement. This was not the case in our series and did not argue for the direct effect of the virus. It is all the more likely that the direct detection of SARS-CoV-2 RNA by RT-PCR was always negative in all our CSF samples.

An alternative hypothesis appears more relevant: the neurovirulence of hCoVs, especially MERS-CoV, seems to be a consequence of immune-mediated processes.^19,27,–,29 Indeed, ADEM and Bickerstaff brainstem encephalitis, which were described with MERS-CoV,^18,–,27 are 2 examples of autoimmune demyelinating diseases resulting from a postinfectious or parainfectious process. A recent case report⁹ has described a case of acute hemorrhagic necrotizing encephalopathy (which is the most severe form of ADEM) associated with COVID-19, which strengthens the hypothesis of a predominant immunologic mechanism. In our cohort, we also described 2 cases of limbic encephalitis and 1 case of CLOCC. Concerning the latter, no treatment had been started or stopped in previous weeks. Thus, 75% of encephalitis cases in our cohort (limbic encephalitis, CLOCC, radiologic ADEM, radiologic acute hemorrhagic necrotizing encephalopathy) are possibly considered immune-mediated diseases. Therefore, SARS-CoV-2–mediated disease is driven largely by immunologic processes, and the same mechanisms could explain LME. Indeed, the arachnoid and pial membranes form the leptomeninges, and a significant amount of antigen-presenting cells are present in the subarachnoid space.³⁰ Thus, initial immune activation occurs in the subarachnoid space, which is a site for antigen presentation, lymphocyte accumulation and proliferation, and antibody production.³⁰ All lead to an important local inflammatory infiltrate with a blood-CSF barrier disruption, which can also explain LME. Thus, LME can be linked to inflammatory or immunologic responses, as previously described in animal models of autoimmune encephalomyelitis,³¹ with some neurotropic viruses such as human T-cell leukemia virus and HIV³¹ and in several immune-mediated neurologic diseases.³¹ Moreover, a very similar intracranial pattern has been described in an animal study in piglets affected by gastroenteritis coronavirus.³² Indeed, histology demonstrated diffuse pia matter gliosis with mild congestion of the meningeal and parenchymal vessels with neuronal degeneration, located mostly in posterior parietooccipital lobes.²⁴ Postcontrast FLAIR is the best sequence to highlight LME, but a potential pitfall is inhalation of increased levels of oxygen by patients who are intubated, which may increase subarachnoid space signal intensity noted on FLAIR images within the basal cisterns and sulcus along the cerebral convexities.³³ That is why we have chosen in this situation to acquire systematically precontrast and postcontrast FLAIR sequences to avoid misinterpretation of FLAIR hyperintensity within the subarachnoid spaces and not to misdescribe a meningeal enhancement. For our 11 patients with LME, no precontrast FLAIR signal abnormalities were visible in the subarachnoid spaces.

Our study has several limitations, mainly due to its multicenter and retrospective design. First, brain MRI protocols were dictated by clinical need, and the patterns of working could be different among the 11 centers, even if neuroradiologists are used to work together. Thus, 36% of the MRIs were performed without administration of gadolinium-based contrast agents, notably preventing the detection of LME, therefore probably underestimating it. Moreover, LMEs are very subtle neuroimaging findings, which need the realization of delayed postcontrast FLAIR sequences, which were not done in the vast majority of the centers. Second, we did not realize a follow-up MRI, and some abnormalities could have appeared during follow-up. Third, outcomes were not available for all patients at the time of writing. Thus, the mortality rate is probably underestimated in our cohort. Fourth, although patients with COVID-19 may develop a wide range of neurologic symptoms, which can be associated with ischemic stroke or encephalitis, it is difficult to state without a controlled general neurologic population that such events are more frequent in patients with COVID-19. However, it now seems well established that thrombotic events are particularly frequent in patients with COVID-19.²¹ Moreover, encephalitis, which is a rare disorder in the general population,³⁴ was surprisingly frequent in our population.

In this large multicentric national cohort, most patients with COVID-19 (56%) with neurologic manifestations had brain MRI abnormalities such as ischemic stroke, LME, and encephalitis. Because the detection of LME suggests the abnormality of brain MRI, it would be interesting to realize systematically postcontrast FLAIR acquisition in patients with COVID-19.

Concerning the cases of encephalitis and LME, even if a direct viral origin cannot be eliminated, the pathophysiology of brain damage related to SARS-COV-2 seems rather to involve an inflammatory or autoimmune response. The knowledge of the various clinicoradiologic manifestations of COVID-19 could be helpful for the best care of the patients and our understanding of the disease. In addition to SARS, cytokine storm syndrome, and heart failure, brain injuries may contribute to increased mortality. This highlights the importance of neurologic evaluation, especially for patients hospitalized in ICU with clinical deterioration or worsening of their symptoms, which can be associated with an acute neurologic event.