To date, following the first report of coronavirus disease 2019 (COVID-19) in Wuhan in late December 2019, over 60 million people have acquired the disease worldwide. Just over one quarter of symptomatic patients needing hospitalization require intensive care support (¹).

While patients with severe pulmonary infections and the acute respiratory distress syndrome (ARDS) are at risk of thrombosis (^2–6), those with COVID-19 appear to be at particularly high risk, despite thromboprophylaxis (⁷). Thrombotic complications in critically ill patients are clinically difficult to detect and may go unrecognized.

Severe COVID-19 pneumonia is characterized by fulminant cytokine release leading to the activation of a coagulation cascade (⁸). A prothrombotic state is a recognized feature of severe COVID-19 infection, manifesting as venous and systemic or pulmonary arterial thrombus; however, the true prevalence of detectable (macrovascular) thrombus and associated complications is unknown (⁹). While typical pathologic features of ARDS are seen in patients with COVID-19 (¹), a recent study reported systemic thrombosis at microvascular level as an additional cause of respiratory failure (¹⁰).

The prevalence of image-diagnosed thrombosis appears higher in COVID-19 than in comparably ill patients with different etiologies. In a recent study, 22% of COVID-19 ICU patients had pulmonary embolism (without systematic imaging) compared with 7.5% in influenza patients, despite similar severity of respiratory disease (¹¹).

The aim of the present study was to evaluate the true prevalence of vascular thrombotic complications in patients with confirmed COVID-19 admitted to ICU for advanced ventilatory support, including those on extracorporeal membrane oxygenation (ECMO), as apparent on systematic CT imaging.

MATERIALS AND METHODS

We undertook a single-center, retrospective analysis of consecutive patients admitted to our ICU for critical care support caused by COVID-19 between March 19, 2020, and June 23, 2020. This study was undertaken following institutional board review and the requirement for informed consent requirement was waived.

Unselected patients with COVID-19 requiring ICU, with or without ECMO, were admitted to our tertiary center ICU via one of two possible pathways: the first group was admitted on the ECMO pathway as defined by the national ECMO guidelines, and the other patients were transferred from other hospitals due to local ICU capacity issues.

All patients had COVID-19 infection confirmed on reverse transcription-polymerase chain reaction testing prior to admission. Unless contraindicated (i.e., intracranial hemorrhage, n = 1), all patients received prophylactic low-molecular-weight heparin (LMWH) at admission and continued or escalated to a treatment dose LMWH as indicated (d-dimer level > 10 times the upper limit of normal [2,600 ng/mL] and a platelet count > 100 × 10⁹/L). Treatment was switched to unfractionated heparin (UFH) if the creatinine clearance fell below 30 mL/min, aiming for a heparin anti-Xa concentration of 0.3–0.7 international units (IU)/mL. Standard practice is to give a bolus dose of UFH at cannulation, followed by heparin infusion if there is no evidence of intracranial bleeding on the CT head within 24 hours of ECMO (¹²). The target heparin anti-Xa concentration was 0.2–0.3 IU/mL for patients on ECMO if there was no evidence of thrombosis or 0.3–0.5 U/mL if there was confirmed or high clinical suspicion of thrombosis. The target anti-Xa level in patients on LMWH was 05–1.0 IU/mL, and if there is evidence of thrombosis despite these levels, dose of LMWH increase by at least 20% and maintain anti-Xa levels of 1.0–1.2 IU/mL.

CT Scanning

Routine practice is to perform contrast-enhanced CT (CECT) in all patients with COVID-19 who are admitted to ICU in our hospital. Local standard operating protocols were followed to transfer patients to and from the CT scanner that was located in close proximity to ICU. The scans were performed on the day of admission or as soon as clinically feasible. Repeat scanning, to monitor progress and evaluate evolving clinical issues, was undertaken as clinically indicated. All CT examinations were performed on COVID-19 dedicated 128-slice, dual-source CT scanner (Definition FLASH; Siemens, Erlangen, Germany). The standard imaging protocol comprises an unenhanced CT of the head followed by CT angiogram of the thorax (following administration of 100 mL contrast agent [Visipaque 350; GE Healthcare AS, Nycoveien 1-2, NO-0401, Oslo, Norway]) to achieve adequate enhancement of the pulmonary arteries and the aorta) and, finally, portal venous phase acquisition of the abdomen and pelvis to assess the abdominal/pelvic viscera and vessels. All CT examinations were independently reviewed by two consultant cardiothoracic radiologists with disagreements resolved by consensus.

Data Collection

Clinical characteristics, laboratory data, and outcomes were collected using the Electronic Patient Record. Imaging data were collated from the picture archiving and communication system (IMPACS-5.2; Agfa HealthCare, Mortsel, Belgium). Full blood count, biochemical profile, high-sensitivity C-reactive protein (hs-CRP), and coagulation tests were performed daily, including prothrombin time (PT), activated partial thromboplastin time (aPTT), fibrinogen, and d-dimer levels (with and without age-adjustment) (¹³). Only the data on ICU admission were used for analysis.

Endpoints Assessed

The primary endpoint was the detection of any venous or arterial thrombus or associated complication (deep vein thrombosis, pulmonary embolism, mesenteric ischemia, aortic or peripheral arterial thrombosis, or cerebral ischemic attack). We also recorded length of stay (LOS) and survival to hospital discharge.

Statistical Analysis

Among patients with COVID-19, we compared patients with, and without, thrombotic complications for the occurrence of thrombotic complications, survival, and LOS. Continuous variables are presented as mean and sd and were compared using Mann-Whitney U test.

Categorical variables are presented as numbers and proportions and were compared using Pearson chi-square tests or Fisher exact tests. Univariate and multivariate logistic regression analysis controlling for age, gender, diabetes, body mass index, and ethnicity were used to compare differences between patients with and without thrombotic complications, and survivors and nonsurvivors, as appropriate. A two-sided p value of less than 0.05 was considered statistically significant. Statistical analyses were done using SPSS v.10.0 (IBM Corp., Armonk, NY).

RESULTS

Patient Cohort

Seventy-two patients with COVID-19 were admitted to our ICU for advanced respiratory support during the study period (March 19, 2020, to June 23, 2020). The baseline clinical characteristics are shown in Table 1. Baseline biochemical and hematological data are shown in Tables 2 and 3.

Twenty-four patients (33%) had one CECT, 31 (43%) had two, 12 (17%) had three, and 5 (7%) had four scans. Sixty-one patients (85%) had at least one unenhanced CT of head. Systemic or pulmonary arterial thrombus or systemic venous thrombosis was diagnosed on the first CECT in 34 of 45 patients (76%) and on subsequent CECT in the remaining 11 patients. Thrombus was detected on CECT in 53% of patients within the first 3 days of arrival to ICU. Thrombus was detected on CECT in 38 patients (53%) within the first 3 days of arrival to ICU. The median time from the onset of disease and hospital admission to the diagnosis of thrombosis was 15 days (3–48 d) and 9 days (0–36 d). The time from hospital admission to first CT scan in patients with and without thrombus was 7.8 ± 6.8 and 7.1 ± 5.9 days, respectively (p = 0.701). The time from intubation and ventilation to first CT scan in patients with and without thrombus was 6 ± 6.3 and 6.1 ± 5.3 days, respectively (p = 0.896). Ninety-three percent of pulmonary thromboses (31/34) were identified on the first CECT with the remainder identified on follow-up CECT, requested due to poor clinical response. The mean time interval between the onset of symptoms and from admission to our ICU and the final status (discharged alive or death) was 41 days (± 20 d) and 32 days (± 18 d), respectively. Fifty-five patients (76%) survived and were discharged, 17 (24%) died as at 23 June 2020 (Table 4).

The biomarkers, outcomes, and the LOS in patients receiving ECMO are summarized in Supplementary Table 1 (https://links.lww.com/CCM/G184) and Supplementary Table 2 (https://links.lww.com/CCM/G185). The hematological and biochemical profiles of these patients at admission to ICU were similar, with the exception of fibrinogen that was higher and CRP that was lower in patients on ECMO. The prevalence of venous and arterial thrombotic complications was not significantly different among patients on ECMO compared with those patients who did not require ECMO (63% vs 54%; p = 0.695).

Thrombotic Complications

Examples of thrombotic complications are shown in Figure 1. Apart from one patient with evidence of intracranial hemorrhage, all received UFH if the patient was on ECMO. The prevalence of thrombotic complications was 54 in 42 of 72 patients (58%). Venous thrombosis was observed in 15 patients (21%) of whom 11 (73%) had thrombus in the iliac or femoral veins, and portal vein in one patient. There was evidence of head and neck vein thrombosis in four patients (27%). Thrombus was associated with a venous catheter in six patients.

Pulmonary artery thromboembolism was observed in 34 patients (47%), 12 (35%) of whom had thrombi in the main pulmonary artery and/or the proximal branches, while in the remaining 22 (65%) patients, thrombi were only visualized in the segmental and subsegmental pulmonary artery branches. Of the 34 patients with pulmonary artery thrombosis, 27 (77%) did not have radiological evidence of peripheral deep venous thrombosis. Of 15 patients with deep venous thrombosis, seven (47%) had no CT evidence of pulmonary artery thrombosis.

Arterial thrombosis and/or systemic embolism were observed in five patients (7%). Aortic macrothrombosis was present in two patients. Embolic ischemic changes were observed in five patients (splenic infarction, n = 2; bowel ischemia, n = 1; ischemic stroke, n = 2; and renal ischemia, n = 1). All patients with arterial thrombosis also had pulmonary thrombosis, but none had peripheral venous thrombosis. Intracerebral hemorrhage was present in two patients. There was no difference in the mean length of ICU stay in patients with and without thrombotic complications (32 ± 18 vs 31 ± 15 d, respectively [p > 0.53]; Table 4). Patients with thrombotic complications were more likely to die, and patients without thrombotic complications were more likely to be discharged alive (p = 0.022; Table 4). There was no significant difference in LOS in patients with or without ECMO (p = 0.546) and in those with or without thrombosis (p = 0.53) (Table 4; and Supplementary Table 2, https://links.lww.com/CCM/G185). From 20 cases on ECMO and imaging evidence of thrombotic complications, thrombosis was detected in 14 patients (70%) on the admission CT scan, immediately before or after the ECMO initiation. In six patients, the mean time from ECMO to the detection of thrombosis was 9 days (+4 d; 3–14 d). This indicates that in a significant number of cases, the thrombosis was likely present prior to ECMO initiation. In the other 30%, thrombosis occurred later and despite full anticoagulation.

Relation of Clinical Characteristics and Biomarkers to Thrombotic Complications

Specific demographics (i.e., male gender, non-Caucasian ethnicity, history of hypertension, and diabetes) were not associated with a higher risk of thrombotic complications (Table 1). The following biomarkers on ICU admission were interrogated for their relationship to subsequent thrombotic complications: d-dimer (with or without age-adjustment), aPTT, international normalized ratio, platelet count, WBC count, lymphocyte count, hs-CRP, and fibrinogen (Tables 2 and 3). None of these variables alone, or in univariate analysis, or in combination as part of multivariate analysis, was predictive of thrombotic complications.

DISCUSSION

This study confirmed a relatively high prevalence of thrombotic (arterial and venous) complications in patients admitted with severe COVID-19. Our study is unique for two reasons. To the best of our knowledge, this represents the first report of thrombotic complications in high-risk ICU patients, in whom regular systematic whole-body CT scanning was undertaken. Second, we report on the prevalence of thrombotic complications in relatively large number of patients with severe COVID-19 receiving ECMO.

In this cohort, pulmonary artery thromboses were present in 47%. To put this into context, a review of the CT imaging data from a group of patients with infectious pneumonia-related ARDS (many due to viral pneumonitis) on routine thromboprophylaxis, admitted to our ICU for respiratory support in 2018, demonstrated a prevalence of pulmonary thromboembolism that was less than half of that seen in patients with COVID-19 (14/64, 22%; male:female = 34:30; mean age 47 ± 15 yr; 75% on ECMO) (unpublished data).

Prevalence of Thrombosis and Importance of Systematic Imaging

Our study reports a much higher rate of thrombotic complications than previously reported. The recent report from three articles on a combined total of 441 ICU-treated patients with COVID-19 receiving standard anticoagulant prophylaxis revealed a pooled 16% rate of pulmonary thrombotic complications (^7,11,14) and a 3.7% arterial thrombotic event rate (⁷). The main difference in these studies, compared with ours, is that their assessment with CT, or ultrasonography, was only performed for clinical indications, without routine systematic evaluation, and may therefore not have captured all thrombotic complications. There has only been one prior report of routine ultrasound imaging in a small case series of 26 patients at admission to ICU who received anticoagulant thromboprophylaxis. Venous thrombosis was observed in 69%, 100% of the patients developed thrombosis while on prophylactic dose anticoagulation but only in 56% of those on therapeutic anticoagulation. This study highlights the importance of systematic screening (¹⁵).

A further report on 198 hospitalized patients with COVID-19 (75 on ICU) receiving thromboprophylaxis showed the prevalence of thrombotic complications increased over time and was linked to increased mortality (¹⁶). Autopsy findings in 12 consecutive COVID-19 deaths revealed deep vein thrombosis in seven patients in whom thromboembolism was not suspected antemortem, with pulmonary embolism being the direct cause of death in four patients (¹⁷).

Our data highlights a number of important issues; first, the prevalence of thrombotic complications is very high in COVID-19 patients on ICU, despite chemical thromboprophylaxis; second, clinicians cannot rely on clinical features to determine thromboembolic disease; and third, biomarkers do not appear to be predictive of thrombotic complications in this cohort. Although ultrasound can be used at the bedside to exclude peripheral venous thrombosis, it has limited application in these patients as it cannot map systemic and pulmonary artery thrombosis. We would recommend that systematic imaging should be considered in all COVID-19 ICU-treated patients to adequately guide treatment decisions. We feel that the additional radiation and contrast burden is justified in this cohort to enable diagnosis and treatment of thrombotic complications that adversely impact on outcome. This would understandably limit the use of a dedicated scanner to reduce infection risks to staff and other patients. We acknowledge that the routine imaging of these patients may not be possible in all settings due to the availability of CT scanners and safety concerns pertaining to transfer of infected patients. It may be therefore necessary to work closely with infection control teams and carefully risk assess each patient and consider institutional logistics.

Clinical Relevance of Identifying Thrombotic Complications

As with previous publications, we demonstrate that the presence of thrombotic complications in patients with COVID-19 is related to adverse outcome. The benefits of anticoagulant thromboprophylaxis in hospitalized COVID-19 patients are now well recognized. The International Society on Thrombosis and Haemostasis and the American Society of Hematology recommends that all hospitalized COVID-19 patients should receive prophylactic dose LMWH unless contraindicated (²⁰). More recently, a report on 3,000 patients with COVID-19 in New York reported that anticoagulation improved survival (not the thrombosis risk), particularly in patients on mechanical ventilation, in whom inhospital mortality fell from 62.7% to 29.1%, and the median survival jumped from 9 to 21 days (²¹). Bleeding was similar in both groups.

Usefulness of Biomarkers in Predicting or Diagnosing Thrombotic Complications on the ICU

One of the emerging hallmarks of severe COVID-19 is a coagulopathy that is detectable through markers of coagulation and inflammation in peripheral blood. In the original Wuhan cohort of 919 patients, lymphopenia, leukocytosis, and elevated alanine aminotransferase, lactate dehydrogenase, d-dimer, and PT were reported and related to increased mortality (¹). Since then, severe coagulation abnormalities have been reported in some 20% of COVID-19 patients and in almost all patients with very severe disease (^22,23).

It has been suggested that d-dimer should be used as a guide to indicate pulmonary embolism (e.g., ≥ 500 mg/L, or ≥ 1,000 mg/L when no clinical for pulmonary embolism are present (^18,19). However, estimation of d-dimer levels for predicting thrombosis risk is generally not helpful, given the significant baseline elevations in ICU-treated COVID-19 patients (²⁴). In our study, d-dimer levels did not discriminate patients with or without thrombosis as the d-dimer levels (even after age-adjustment) were highly elevated in most patients. This does not exclude the screening role of d-dimer (with or without age-adjustment) in earlier stages of the disease.

The finding that coagulation and inflammatory markers on ICU admission did not correlate with thrombotic complications should be interpreted with caution as we only analyzed the admission results in a limited number of patients. The main message is that in this setting, the biomarkers did not discriminate patients with thrombosis.

Pathologic Mechanism of Thrombotic Complications

While significant disturbances in coagulation markers were seen in our cohort, these did not correlate with the occurrence of thrombosis. This could suggest that the pathologic mechanism in these patients may be more complex than a simple procoagulant state, with inflammation and endothelial dysfunction playing important roles, particularly in some vascular beds (^25,26,27). The finding that 77% of patients in our cohort had CT evidence of pulmonary thrombosis without evidence of venous thrombosis is similar to that reported by Poissy et al (¹¹) and may suggest that in some cases, the pulmonary arterial filling defect represents in situ thrombosis.

Limitations

As a large observational study of COVID-19 patients, some patients remain in hospital, so the prevalence of thrombotic complications and its relationship to mortality may be under-estimated. In addition, it was difficult to assess when exactly patients developed the thrombosis and these were likely present prior to admission to ICU. This might have affected the predictive value of blood biomarkers as shown in this study.

CONCLUSIONS

Among COVID-19 patients needing ventilatory support on ICU, arterial and venous thrombosis was observed in nearly three in five patients. Thromboses were related to adverse outcome, and importantly the presence of these thromboses was not predicted based on usual biomarkers of coagulation at admission to ICU. Since many thrombotic complications are clinically silent, we propose that systematic CT imaging should be considered in all ICU-treated COVID-19 patients and may improve patient outcome if implemented early and routinely.

Supplemental Digital Content

• CCM_49_5_2020_12_23_MIRSADRAEE_CCMED-D-01827_SDC1.docx; [Word] (21 KB)
• CCM_49_5_2020_12_23_MIRSADRAEE_CCMED-D-01827_SDC2.docx; [Word] (16 KB)