Six patients with cerebral hemorrhages related to COVID-19 infection were selected. These cases were observed between 1st May 2020 and December 28th, 2020 at the Air Force Hospital of Belém. The patients had a confirmed diagnosis of COVID-19 through RT-PCR testing, and a diagnosis of ICH was established based on clinical-radiological aspects. Neuroimaging was done in all patients. After reviewing the neuroimaging reports for these patients, they were found to have documented radiographic evidence of hemorrhage. Neuroimaging for these patients was reviewed by a fellowship-trained neuroradiologist to verify the presence and type of hemorrhage. This diagnosis was further corroborated by neuroimaging tests, which included computed tomography (CT) or magnetic resonance imaging (MRI).

The six patients in this case series were not encompassed within the scope of the systematic review conducted in this study.

A comprehensive, systematic search of the literature published between December 19, 2021, and May 7, 2022, held on the MEDLINE, PubMed, and NCBI electronic databases was conducted using the following search terms: ( hemorrhagic encephalopathy ) OR ( intracranial bleeding ) OR ( subarachnoid hemorrhage ) OR ( subdural hemorrhage ) OR ( intracranial hemorrhage ) OR ( hemorrhagic stroke ) OR ( cerebral hemorrhagic complication ) OR ( cerebral hemorrhage ) AND ( SARS-CoV-2 virus ) OR ( SARS CoV 2 virus ) OR ( 2019-nCoV ) OR ( COVID-19 ) OR ( 2019 novel coronavirus ).

The search was limited to articles written in English. Articles identified by the initial search strategy were independently evaluated by two authors (WL and MP) according to the inclusion criteria: involving patients with COVID diagnosed by RT-PCR, a confirmed diagnosis of ICH, description of the cases with individual demographic characteristics, clinical-radiological aspects, interventions, and outcomes. Articles which were duplicates, those that had only the abstract available or were editorial letter articles, as well as those whose full-text was not in English, and those that involved pediatric patients (age < 18) and patients with predominantly non-spontaneous hemorrhages were excluded.

The Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia) was used to import all titles and abstracts of the articles identified and remove duplicate records. Potentially eligible articles were identified by screening the titles and abstracts. The full texts of the studies selected were then thoroughly reviewed for quality control by two authors (WL and MP) using the Newcastle-Ottawa scale, and the eligibility of each study was determined. Any disagreements between the investigators were resolved by consulting with the corresponding author (FP) ( Supplementary Material 1 https://www.arquivosdeneuropsiquiatria.org/wp-content/uploads/2023/10/ANP-2022.0230-Supplementary-Material-1.docx ).

The following information was collected from each study reviewed: surname of the first author and year of publication, study design, sample size, demographic characteristics, comorbidities, number of patients with hemorrhagic events in COVID-19 hospital admissions, time interval from admission/initial symptoms to radiological diagnosis, initial laboratory findings, antithrombotic therapy prior to onset of hemorrhagic event, type of ICH, clinicoradiological scales applied on admission and/or discharge, mortality rates, and discharge outcomes. Neuroimaging findings were divided into three major types: intraparenchymal hemorrhage (IPH), subdural hematoma (SDH), and subarachnoid hemorrhage (SAH). Additionally, regarding the hemorrhage distribution, we also classified three subtypes: focal intracerebral hemorrhage (FICH), multifocal intracerebral hemorrhage (MFIH), and multicompartmental hemorrhage (MCH). ^{9 10 17}

The synthesis of the data was performed with the aid of the Covidence and Excel (Microsoft Corp., Redmond, WA, USA) programs, where data extracted were compiled into tables with their respective categories. Primarily, the relevant findings on eligible cohort and case series studies reporting ICH in COVID-19 hospitalizations were presented in the form of a summary table ( Supplementary Material 2 , Table 1 ) accompanied by a narrative description. A concise overview of the attributes of the six patients featured in this case series has been incorporated into Supplementary Material 2 , Table 1 . The remaining case reports identified by the search were subsequently compared against our original case reports and stratified into additional tables by similar hemorrhagic events ( Supplementary Material 2 , Tables 2–4 ).

The cerebral hemorrhage causes identified in the six selected patients were as follows: IPV/IVH in two cases, both accompanied by indications of intracranial hypertension and uncus herniation; one case with SDH featuring mass effect on the right frontal, temporal, and parietal lobes, alongside indications of intracranial hypertension; two cases with CVT/IPH; and one case with IS/IPH. Detailed clinical and demographic attributes of the patients within this case series can be found in Supplementary Material 2 .

Of a total of 1,624 articles retrieved in the literature search up to March 2022, 6 duplicate studies were removed, and 1,618 articles retained for screening of title and abstract. After exclusion of 1,421 non-relevant studies, 197 studies were retrieved, of which an additional 144 were subsequently excluded for the reasons presented in Figure 1 . The selection process resulted in a final total of 53 articles for inclusion in the review.

Considering 22 cohort selected articles, the prevalence of hemorrhagic cerebrovascular events among patients with COVID-19 was described in Supplementary Material 2 , along with epidemiological data. A total of 31 case reports and case series articles were included to report the sex, age, comorbidities, initial symptoms, diagnostic methods, radiological findings, treatment, and outcome. ( Supplementary Material 2 )

Interventions targeting ICH were reported in 10 cohort studies involving a total of 269 cases ^{17 18 19 20 22 24 25 26 27 33} ( Supplementary Material 1 , Table 1 ). Surgical management was performed in 29 patients (10.7%), 16 of whom had external ventricular shunt (EVD), 5 hematoma drainage with decompressive craniectomy, 2 aneurysm embolization by coiling or flow deviation, 2 had invasive monitoring of intracranial pressure (ICP), and the remaining reports were unclear on the type of surgical approach employed. Management solely by intensive care measures occurred in 45 patients (16.7%). However, the use of specific measures to control ICP or hemorrhage was not specified.

Among the cohort studies, 11 articles described the mortality rate, which ranged from 0 to 84.6% ^{17 18 19 20 22 23 24 25 26 27 28} ( Supplementary Material 2 ). The overall mortality rate in these studies was 44%, for a total of 313 patients out of 114,706 cases hospitalized with COVID-19. The mortality rate from cerebrovascular hemorrhagic events in hospitalized patients with a confirmed diagnosis of COVID-19 by the RT-PCR method was 0.12%. When case series studies and case reports were included, a total of 427 patients were obtained with an overall mortality rate of 46.3%.

Other outcomes were reported in 7 cohort studies, ^{17 18 20 22 24 25 27} with descriptions of hospital discharge (non-routine, to home or to rehabilitation) in 57.5% of a total of 245 patients. In addition, modified Rankin Scale (mRS) scores at discharge were variably reported and incomplete for the majority of the studies reviewed. Only 4 articles ^{20 22 24 26} mentioned a poor prognosis at discharge (mRS > 3), in 69.5% of a total of 23 patients who were evaluated at discharge with this scale ( Supplementary Material 2 ). Including case series and case report studies, hospital discharge was reported in 53.5% out of a total of 327 patients, while there was a poor prognosis (mRS > 3) in 73% of a total of 110 patients ( Supplementary Material 2 ).

In general, the most common neurological symptoms described in COVID-19 patients are headache, altered level of consciousness (ALC), dizziness, ageusia and anosmia, while other less common symptoms reported include visual impairment, CVD, seizures, occipital neuralgia, ataxia, tremor, and tics. ^{4 5 27} Severe infections were more likely in the presence of CVD and ALC ²⁷ and to be reported in hypertensive patients, who were older, had fewer typical symptoms, and were more likely to develop neurological manifestations, especially acute CVD. ⁴

Several of the articles reported the time between admission and neuroimaging, with mean values ranging from 11 to 29 days. ^{22 26 28 52} The interval between the onset of symptoms and diagnosis was similar, lying within the range 2 to 29 days. ^{19 24 27 29 52 53 54 55} Acute stroke signs were the initial manifestation of COVID-19 in only 11 to 44% of admissions. ^{26 29 32 44 50 52 54 56} Our 6 cases are in agreement with the literature as presenting a 10 days median time and 33% of neurologic symptoms onset. At admission, the main neurologic sign was the depressed level of conscience, and the NIHSS ranged from 4 to 12 points.

Despite a wide variety of radiological findings in hospitalized cases (19), it is uncommon for patients to be diagnosed with COVID-19 using brain magnetic resonance imaging (MRI). Microhemorrhages, IS and ICHs are the most prevalent presentations. Other less common findings include hypoxic anoxic brain injury, encephalitis, acute disseminated encephalomyelitis (ADEM), leukoencephalopathy, transient perivascular inflammation of the carotid artery syndrome (TIPIC), and posterior reversible encephalopathy syndrome (PRES). ^{6 7 8 44 53 57 58}

The ICH group had the highest mortality rate, ^{7 44} followed by patients with leukoencephalopathy and IS, whereas patients with microhemorrhages or encephalitis as sole neuroimaging findings had the lowest mortality rates. ⁷ Patients in intensive care unit (ICU) had significantly higher incidences of cerebral microhemorrhages and encephalitis/encephalopathy. ⁵⁸

COVID-19 patients are also at a higher risk for hemorrhagic conversion of their stroke, accompanied by an increased mortality rate. ^{23 33} Nonetheless, multicompartmental hemorrhage is the ICH subtype with the highest mortality rate, followed by MFH presentations, while SDH had the lowest mortality rate. ¹⁷

Considering patients without COVID-19, lobar hemorrhages are often associated with structural changes such as cerebral amyloid angiopathy, arteriovenous malformations or brain tumors. ⁵⁹ Independent associated risk factors were anticoagulation, a prior history of IS and APOE e2 or e4 genotype, which had a specific association with lobar ICH. ⁶⁰

Hypertension is the leading attributable risk of non-lobar ICH, followed by prior history of IS and anticoagulation. Interestingly, hypercholesterolemia was less frequent in non-lobar ICH cases. ⁶⁰ The most common locations of hypertensive ICH are the basal ganglia (caudate nucleus and putamen), thalamus, cerebellum, midbrain, and pons. ^{59 60}

The results of the present review revealed a predominance of hypertension over DM2, dyslipidemia, and other comorbidities. Nevertheless, a distinct proportion of IPH was found, whereas patients had predominantly lobar locations (32%) as opposed to non-lobar (24%). In addition to the association with lobar ICH and ApoE e4e4 allele, recent findings suggest an increased risk of severe COVID-19 infection in this population, independent of preexisting dementia, hypertension, and DM2. ^{61 62} APOE ε4 carriers also present an increased susceptibility to SARS-CoV-2 infection with higher serum indicators of inflammation. ⁶³

Our case series had an expected predominance of hypertension as comorbidity, considering that all the intraparenchymal hemorrhages were in lobar locations and 40% in non-lobar. Therefore, these findings are consistent with the literature evidence found in this review.

Severe acute respiratory syndrome coronavirus 2-infected stroke patients exhibited particular clinical aspects compared with non-infected patients. The infection was associated with a higher prevalence of younger patients, hemorrhagic conversion of IS, ³³ severe NIHSS scores, elevated D-dimer levels, ^{26 33 64} thrombocytopenia, ^{33 43 64} elevated PTT, ⁶⁴ elevated INR, and in-hospital stroke. ⁴³

Considering only ICH, COVID-19 patients were younger with higher rates of malignancies, ²⁵ elevated INR, PTT, and fibrinogen levels, yet decreased frequency of hypertension. ²⁷ No significant changes were reported for other risk factors, such as DM2, dyslipidemia, smoking, ischemic heart disease, or atrial fibrillation. ^{20 25 26 27} These patients also had more severe NIHSS and ICH scores at admission. ^{26 27}

Although these cases present with higher median neutrophil-to-lymphocyte ratios, there was no significant difference when compared with control groups. ²⁷

The comparison between COVID-19 cases with and without ICH yielded more discrepant results. Those with hemorrhagic events were older and had higher rates of prior stroke, hypertension, DM2, dyslipidemia, congestive heart failure, ischemic heart disease, and smoking, ^{25 54} but a lower rate of atrial fibrillation ²⁸ and thrombocytopenia. ⁵⁴ When considering microbleeds alone, the only significant increase was in the rate of disturbance of consciousness prior to MRI, severity of lung computed tomography (CT), days of intubation, and duration of hospital or ICU stay. ¹⁸

The coronavirus, akin to other respiratory viruses, has neurotropism and the ability to invade the CNS in two ways: hematogenous and retrograde neuronal pathways. This ability to infect neurons from the olfactory bulb can also explain complaints of hyposmia and anosmia. The hematogenous route is identified as the main form of CNS infection, since the virus can infect endothelial capillary cells in the brain or infect leukocytes. Additionally, similarly to SARS-CoV, SARS-CoV-2 exploits the ACE2 receptor for cell entry. ^{14 66}

ACE2 (is a critical enzyme in the renin-angiotensin-aldosterone (RAAS) system that regulates blood pressure, fluid and electrolyte balance, and vascular resistance. This enzyme is extensively expressed in alveolar epithelial cells (type 2 pneumocytes), oral and esophageal mucosa, as well as in vascular endothelial cells, smooth muscle, glial cells, and in some neurons, including those in the cardiorespiratory center of the brainstem. ^{11 14 67}

Severe acute respiratory syndrome coronavirus 2 infection in humans is mediated by S (spike) glycoprotein binding, by the receptor-binding domain (RBD) to ACE2 receptors in host cells, which leads to downregulation of ACE2 expression. This negative regulation during SARS-CoV-2 infection can increase serum levels of angiotensin II, causing endothelial function impairment and blood pressure dysregulation. Therefore, blood pressure fluctuations with an increased risk of hemorrhagic cerebrovascular events can occur. ^{11 14}

The affinity of the SARS-CoV-2 spike protein to ACE2 receptors in brain capillary endothelium can also cause direct vascular injury. The explanation for this involves the process of binding of viral particles by the endothelial cells and, subsequently, damage to the endothelial lining that can cause ruptures and bleeding. This same process can occur within neurons from the viral invasion of the CNS. ^{11 67}

There is a release of cytokines and proteases that accompanies the immune response to SARS-CoV-2 infection, involving massively increased levels of interleukin 6 (IL-6), IL-7, IL-10, IL-1β, interferon-gamma (IFN) -γ), and tumor necrosis factor α (TNF-α), while there is a reduction in CD4 + and CD8 + T cells, indicating that the cytokine storm attenuates adaptive immunity against SARS-CoV infection. ^{13 68} In critically-ill patients with COVID-19, higher serum levels of inflammatory markers (e.g., C-reactive protein and D-dimers) and an increase in neutrophil-lymphocyte ratio can be seen, also present in the inflammatory process of ICH. ^{27 69 70}

The cytokine storm usually starts in the second week of infection, with the activation of macrophages, dendritic cells, other immune cells, and subsequent massive release of proinflammatory cytokines. ⁶⁸ Consequently, via a mechanism that is still unclear, changes in the permeability of the BBB can be impaired, facilitating the influx of inflammatory molecules to activate C macrophages and microglia. Ultimately, these cells become hyper-activated and start producing their own set of inflammatory molecules, which can lead to cerebral edema and even hemorrhagic events. ^{56 68}

Thus, BBB breakdown is a possible additional mechanism for several cerebrovascular events associated with this infection, such as hemorrhagic transformation of IS, ICHs, and cases of PRES reported in some patients with COVID-19. ^{71 72 73}

The binding of spike protein may also promote a downregulation of ACE 2 expression in the brain, thereby triggering an increase in local angiotensin II levels and reduction in the vasodilator heptapeptide (angiotensin 1–7). Ang 1–7 acts as a neuroprotective factor by stimulating the release of prostaglandin and nitric oxide, as well as inhibiting the growth of smooth muscle cells and action of catecholamines. ^{67 74}

Patients with hypertension normally have low ACE2 expression, which is further reinforced with SARS-CoV-2 infection, increasing the risk of stroke. ⁷⁵ The intrinsic relationship between systolic BP variability and poor prognosis of cerebral hemorrhage should be pointed out, as a high variation in BP during the first 24 hours of admission was associated with an unfavorable hospital prognosis in patients with ICH. The lack of BP control might be explained by autonomic dysfunction, with sympathetic predominance, associated with the production of proinflammatory cytokines, hyperglycemia, and increased permeability of the BBB, which are present in SARS Cov2 infection. ^{75 76}

Diabetic patients with COVID-19 are at increased risk of serious complications. The possible mechanisms that lead to an increased risk of stroke in these patients include excessive proinflammatory responses and reduced ACE2 expression by advanced glycosylation, leading to increases in angiotensin I and II. ⁷⁷

Coagulation disorders may be a plausible hypothesis to explain how SARS-CoV-2 infection can induce brain hemorrhage, as patients with COVID-19 may suffer from consumption coagulopathy with prolonged prothrombin time and reduced fibrinogen, both of which also contribute to secondary cerebral hemorrhage. ⁷⁵

The older population has several aggravating factors for the development of intravascular hemorrhages, such as cerebral microembolism, white matter lesions, vascular basement membrane thickening, and increased BBB permeability, which promote endothelial damage, changes in elasticity, and subsequent fluctuations in blood flow and pressure causing loss of self-regulation and increase in ICH risk. ⁷⁸