SARS-CoV-2 and Stroke in a New York Healthcare System

We obtained institutional review board approval from NYU Langone Health to perform the study, and informed consent was waived by the institutional review board. Data from the study are available for sharing upon reasonable request to the corresponding author. This is a retrospective observational study including patients admitted to one of 3 comprehensive stroke centers in the New York metropolitan area (NYU Langone Manhattan, NYU Langone Brooklyn in Sunset Park, Brooklyn, and NYU Langone Winthrop in Mineola, Long Island) with acute ischemic stroke hospitalized between March 15, 2020, and April 19, 2020. All consecutive patients with radiological confirmation of acute ischemic stroke during this time frame were included. In addition, we included consecutive patients from 2 of our facilities (NYU Langone Brooklyn and NYU Langone Manhattan) with a discharge diagnosis of ischemic stroke between March 15, 2019, and April 15, 2019, as a historical comparator group (historical control).

In general, patients underwent a standard diagnostic evaluation, including brain imaging, intracranial and extracranial vascular imaging, and cardiac evaluation, including ECG, in-house continuous cardiac telemetry for at least 24 hours, and transthoracic echocardiography per institutional protocol. Stroke subtype was classified based on the Trial of ORG 10172 in Acute Stroke Treatment classification.¹ Large artery atherosclerosis was defined as 50% or more narrowing in an artery supplying the ischemic infarct territory, small vessel disease was defined as a small (≤1.5 cm on head computed tomography or ≤2 cm on diffusion-weighted imaging sequence) subcortical infarct in patients with risk factors for small vessel disease, cardioembolic was defined as the presence of a major cardioembolic source, such as atrial fibrillation, endocarditis, or ejection fraction ≤30%, and other was defined as an alternative mechanism such as dissection or known hypercoagulability, excluding de novo hypercoagulability in the setting of a COVID-19. Cryptogenic was defined as cases not meeting criteria for any of the above stroke subtypes, including those with incomplete workup or with multiple competing high-risk mechanisms.¹ In addition, we defined embolic stroke of undetermined source according to the criteria proposed by the Cryptogenic Stroke/Embolic Stroke of Undetermined Source International Working Group.² Finally, we also classified strokes based on the ASCOD criteria (atheroclerosis, small-vessel disease, cardiac pathology, other cause, and dissection).³

The following variables were abstracted from the medical records of patients:

Screening for COVID-19 was performed at first provider contact and included evaluation for recent COVID-19 exposure, history of fever or respiratory symptoms, or chest radiographic findings. In general, patients with a negative screen do not undergo COVID-19 testing. A positive screen would trigger testing, and this is generally the case when screening could not be completed. Assays for COVID-19 were performed in accordance with standards established by the World Health Organization. A reverse-transcriptase polymerase chain reaction study was performed in each center’s laboratory using a sample obtained via a nasopharyngeal swab. The primary variable of interest was COVID-19 status (positive versus negative).

No sample-size calculations were performed. All patients (cases and control groups) were divided into 3 groups: group 1 included patients with ischemic stroke and COVID-19 (cases), group 2 included contemporary patients with ischemic stroke without COVID-19 (contemporary controls), and group 3 included historical patients with ischemic stroke without COVID-19 over the same time frame (historical controls). We compared baseline characteristics and stroke subtypes between the groups, using the Fischer exact test for categorical variables and nonparametric test for continuous variables. We then performed binary logistic regression analyses to determine baseline characteristics and laboratory values associated with stroke in the setting of COVID-19 compared with historical and contemporary controls. In these models, we included variables with a 2-sided P<0.05 on univariate models. In addition, we performed binary logistic regression analyses to determine the association between COVID-19 related stroke and in-hospital mortality adjusting for age and admission NIHSS score. Analysis was performed using SPSS version 25.0 (Chicago, IL), and a 2-sided P<0.05 was considered significant.

Out of 3556 patients hospitalized with COVID-19 infection during the study period, we identified a total of 32 patients (0.9%) who had radiologically proven ischemic stroke. Among contemporary controls, 70% (32/46) underwent the COVID-19 test and the rest screened negative and were not tested.

Of the 32 patients (Table 1), stroke was the reason for admission in 43.8%, and COVID-19 symptoms were the reason for admission in 56.2%, with index stroke occurring during the hospital stay. The number of strokes in hospitalized patients with COVID-19 in our health care system, all COVID-19 new hospitalizations in our health care system, and all COVID-19 new hospitalizations in New York City are shown in Figure 1. Among the 32 patients with COVID-19 and stroke, the median (interquartile range [IQR]) age was 62.5 (52.0–69.25) years and 71.9% (23/32) were men; 70.0% (21/30) were white, 20.0% (6/30) were black, 10.0% (3/30) were Asian, 13.3% (4/30) were Hispanic; 65.6% (21/32) were of cryptogenic subtype and 34.4% (11/32) met the criteria for embolic stroke of undetermined source. The median (IQR) time from first COVID-19 symptoms to identification of stroke was 10 (5–16.5) days. The most prominent clinical features were cough (84.4%), fever (71.9%), and hypoxia (78.1%). At last follow-up, 81.3% (26/32) met the criteria for severe disease; 75.0% (24/32) were dead or critically ill. The median (IQR) D-dimer level closest to the time of the stroke was 3913 ng/mL (2549–10 000), and the median (IQR) C-reactive protein level was 101.1 mg/L (38.8–214.3). Figure 2 shows brain and chest imaging of 2 patients with COVID-19 and cryptogenic stroke subtype. Treatments before stroke symptoms/diagnosis included hydroxychloroquine (40.6%, n=13), lopinavir-ritonavir (3.1%, n=1), and tocilizumab (6.3%, n=2).

When compared with contemporary stroke controls, patients with COVID-19 and stroke were younger (median [IQR] in years: 63 [17] versus 70 [18], P=0.001), had higher admission NIHSS score (median [IQR] NIHSS score: 19 [23] versus 8 [12], P=0.007), had higher peak D-dimer level (median [IQR] in ng/mL: >10 000 [7427] versus 526 [2752], P=0.023, were more likely to be treated with anticoagulation (78.1% versus 23.9%, P<0.001), more likely to have a cryptogenic stroke subtype (65.6% versus 30.4%, P=0.003), and higher inpatient mortality (63.6% versus 9.3%, P<0.001; Table 2).

When compared with historical stroke controls, patients with COVID-19 and stroke were more likely to be men (71.9% versus 45.0%, P=0.012), had higher admission NIHSS score (median [IQR] NIHSS score: 19 [23] versus 3 [12], P=0.001), had higher erythrocyte sedimentation rate level (median [IQR]: 79 [53] versus 41 [52], P=0.001), were more likely to have positive troponin levels (45.2% versus 8.1%, P<0.001), were more likely to have proximal large vessel occlusion (45.5% versus 20.3%, P=0.026), were more likely to be treated with anticoagulation (78.1% versus 25.0%, P<0.001), were more likely to have a cryptogenic stroke subtype (65.6% versus 25.0%, P<0.001), and higher inpatient mortality (63.6% versus 6.3%, P<0.001; Table 2). Conversely, patients with COVID-19 and stroke were less likely to have hypertension (56.3% versus 80.0%, P=0.017) and history of prior stroke or transient ischemic attack (3.1% versus 25.0%, P=0.007).

Among patients with ischemic stroke and COVID-19, cryptogenic stroke subtype was associated with a trend towards higher d-dimer median (IQR) level in ng/mL at the time of the stroke (4841 [7064] versus 2087 [6306], P=0.109) but not with a significantly higher median (IQR) C-reactive protein level mg/L (141.8 [169.1] versus 74 [122.8], P=0.303).

In binary logistic regression models (Table 3), when compared with contemporary controls, patients with COVID-19 had non-significantly higher admission NIHSS score (odds ratio [OR] per point increase 1.13 [95% CI, 0.98–1.30], P=0.085) and higher peak d-dimer level (OR per 100 ng/mL increase 1.02 [95% CI, 1.00–1.05], P=0.093). In addition, when adjusting for age and NIHSS score, patients with COVID-19 and stroke had a significantly higher in-hospital mortality than contemporary controls (adjusted OR 64.87 [95% CI, 4.44–987.28], P=0.002).

In binary logistic regression models (Table 3), when compared with historical controls, COVID-19 positive patients were non-significantly more likely to be younger (OR per year increase 0.89 [95% CI, 0.78–1.01], P=0.062), be men (OR 13.48 [95% CI, 0.91–200.27], P=0.059), have elevated troponin levels (OR 17.25 [95% CI, 0.58–510.35], P=0.099), a significantly higher admission NIHSS score (OR per point increase 1.23 [95% CI, 1.05–1.44], P=0.010), and a significantly higher erythrocyte sedimentation rate level (OR per 1 unit increase 1.03 [95% CI, 1.00–1.06], P=0.041). In addition, when adjusting for age and NIHSS score, patients with COVID-19 and stroke had a significantly higher in-hospital mortality than historical controls (adjusted OR 40.27 [95% CI, 5.44–298.01], P<0.001).

In this multi-ethnic study, we report key demographic and clinical characteristics of patients who develop ischemic stroke associated with acute severe acute respiratory syndrome CoV-2 coronavirus infection. The observed rate of imaging-confirmed acute ischemic stroke in hospitalized patients with COVID-19 of 0.9% was lower compared with prior reports from Chinese COVID-19 studies. Reasons for difference are unknown but could possibly be related to differences in the patient population studied in our patient population as compared to the other studies and other studies including hemorrhagic stroke and venous sinus thrombosis patients. In addition, the rate of ischemic stroke in our study may be an underestimate as the detection of ischemic stroke symptoms is challenging in those critically ill with COVID-19 infection who are intubated and sedated.

When classified according to the Trial of ORG 10172 in Acute Stroke Treatment criteria,¹ a majority (65.6%) of these patients were classified as cryptogenic stroke and 34.4% met embolic stroke of undetermined source criteria. In contrast, 30.4% of the contemporary COVID negative control group and 25.0% of the historical control group were classified as cryptogenic stroke, in keeping with other modern stroke cohorts.⁴ Patients with COVID-19 and stroke were very ill as a group; 68.8% of patients required mechanical ventilation and 81.3% had severe illness graded according to the American Thoracic Society/Infectious Diseases Society of America Criteria for pneumonia severity.⁵ Our findings are congruent with other studies which reported an increased prevalence of neurological disorders in those with more severe infection.⁶

Our study also shows that the number of COVID-19 positive ischemic strokes has increased initially but seems to have peaked and then decreased. This finding may be related to the overall reduction in COVID-19 admissions, likely due to social distancing and stay at home order (Figure 1). In addition, a therapeutic anticoagulation protocol was instated in our institution the week of April sixth, 2020, which suggests the use of therapeutic anticoagulation in patients with high D-dimer levels. This may have led to a lower rate to thrombotic complications including ischemic stroke in hospitalized COVID-19 positive patients.

Furthermore, the number of patients with stroke hospitalized in the study period were less than historical controls. This witnessed low volume of acute emergencies during the COVID-19 pandemic has been observed in other institutions as well.⁷ The reasons for this are unclear but possibly that patients with stroke and mild symptoms are staying at home and not presenting to the emergency department for stroke treatment.

There are multiple, not mutually exclusive, possible mechanisms associating COVID-19 with ischemic stroke. In patients with COVID-19 requiring invasive respiratory support, the median duration of mechanical ventilation has been reported 11 days,⁸ which renders them vulnerable to complications associated with critical illness and a prolonged intensive care unit stay including the risk for (1) hypotension and inadequate cerebral perfusion; (2) relative hypertension leading to posterior reversible encephalopathy syndrome; (3) septic embolization in the case of superimposed bacterial infection; (4) stress cardiomyopathy and an attendant reduction in left ventricular ejection fraction; and (5) atrial fibrillation with or without a rapid ventricular response.

Additionally, severe COVID-19 has been associated with a hyperinflammatory state (cytokine storm⁹) and hyperviscosity. Progression to disseminated intravascular coagulation is more common in COVID-19 than in other forms of critical illness; one case series reported an incidence of 8.7% with associated 94% mortality.¹⁰ Mortality is associated with higher fibrin-degradation product levels and prolonged prothrombin and activated partial thromboplastin times.¹¹ D-dimer was elevated in 36% of patients with COVID-19 in Wuhan,¹² which was associated with a higher risk of mortality,¹³ an association suggested to be driven at least partially by increased thrombotic complications. Preliminary reports from China describe patients with COVID-19 who developed multiple, bilateral ischemic cerebral infarcts, antiphospholipid antibodies, and hematologic indices suggestive of an acquired thrombophilia.¹⁴ During the first SARS outbreak in the early 2000s, postmortem studies demonstrated a florid vasculitis in multiple arterial beds,¹⁵ and it is not known whether this disease pattern occurs with severe acute respiratory syndrome CoV-2 coronavirus infection.

There are 2 main implications for clinical care that arise from our data. First, many institutions are currently attempting to balance the benefits of rapid, structured neurological evaluations for patients with COVID-19 exhibiting new neurological symptoms with the risks of exposing multiple team members to infection.¹⁶ As centers develop protocols for the prompt triage and assessment of patients with COVID-19, the co-occurrence of stroke and COVID-19 should be considered when weighing these risks. Second, stroke in the setting of COVID-19 could be a manifestation of systemic hypercoagulability as shown in our patient population with higher D-dimer levels when compared with contemporary controls. Further study is required whether therapeutic anticoagulation in this setting mitigates ischemic stroke risk. In fact, PROTECT COVID (A Randomized Trial of Anticoagulation Strategies in COVID-19) is an ongoing randomized clinical trial testing the safety and efficacy of therapeutic versus prophylactic anticoagulation in patients with COVID-19 infection and mild to moderate elevation in D-dimer level (URL: https://www.clinicaltrials.gov. Unique identifier: NCT04359277).

Our study reports the clinical characteristics of a diverse patient population with ischemic stroke in the setting of COVID-19. We report subject level data, including key clinical variables, markers of disease severity, and diagnostic workup for ischemic stroke.

Our findings should be interpreted with caution in the context of a number of important limitations. First, our study was a relatively small, retrospective, observational study with potential for selection bias. Second, we did not have outcome data on all patients as some are still admitted receiving active clinical care. Third, we do not have complete laboratory investigations or diagnostic imaging for all study subjects, and therefore, some cryptogenic strokes may be related to another undiagnosed mechanism. This likely contributed to an increased prevalence of cryptogenic stroke subtype in patients with COVID-19 infection. Fourth, our study may not be fully representative of all patients with stroke in our healthcare system; patients who are critically ill may not be diagnosed with stroke due to impaired consciousness, confounding systemic illness, or withdrawal of life-sustaining therapies. Fifth, since this report focused on ischemic stroke, we did not provide information on hemorrhagic stroke and venous sinus thrombosis occurring in patients with COVID-19. These complications in the setting of COVID-19 need further study. Finally, since not all patients were tested for COVID-19, it is possible that some asymptomatic patients with stroke may have been COVID positive but were included in the control group.

We observed a relatively low rate of imaging proven ischemic stroke in hospitalized patients with COVID-19 infection. In patients with COVID-19 and ischemic stroke, a majority of strokes were classified as cryptogenic, possibly related to an acquired hypercoagulability, and were associated with increased mortality. Ongoing studies are testing the utility of therapeutic anticoagulation for stroke and other thrombotic event prevention, in select patients with COVID-19 and laboratory evidence suggestive of hypercoagulability.