Acute Pulmonary Embolism in Critically Ill Patients with COVID-19

Methods

Starting in mid-March, we reviewed the number of patients who were critically ill from confirmed COVID-19 and required ICU or SDU level of care over a four-week period at Berkshire Medical Center in Pittsfield, MA. We compared the clinical features and laboratory test results of patients with and without PE. We were limited to descriptive analysis between the two groups given the small sample size. The study was approved by the institution’s IRB. Informed consent was waived based on criteria outlined in 45 CFR 46.104(d)(4).

Results

Of the 23 COVID-19 positive, critically ill, patients hospitalized during a four-week period at our institution, 7 (30%) were diagnosed with concomitant acute PE during their inpatient stay. The two groups (PE+, n=7; PE-, n=16), had similar features on admission (Table 1). All patients in both groups were either former or never smokers. Respiratory and gastrointestinal symptoms were the most common complaints, but the mean resting heart rate was noted to be modestly higher in the PE group (104 vs. 88). Both groups showed lymphopenia. Pro-BNP was higher in the PE group at the time of admission, but the comparison was limited by incomplete data for the two groups. D-dimer trended upward in the PE group compared to the control group during the hospitalization (Figure 1). Other risk factors for PE, including BMI, recent surgeries, and travel history were similar across both groups.

• Download figure
• Open in new tab

Figure 1

Of the 7 PE+ patients, one had no D-dimer measurements and one had a single D-dimer measurement (patient 6); of the 16 PE- patients, two had no D-dimer measurements and two had a single D-dimer measurement (patients 13 and 14). Solid linear regression lines represent PE+ patients (N=5) and dotted linear regression lines represent PE- patients (N=12). Decreasing linear regression lines only occurred in PE- patients (2 of 12 = 17%). Increasing linear regression lines occurred in 80% (4 of 5) of the PE+ patients and 33% (4 of 12) of the PE- patients. The remaining linear regression lines were essentially flat, occurring in 20% (1 of 5) of the PE+ patients and 50% (6 of 12) of the PE- patients.

Twenty one of the 23 patients were treated with DVT prophylaxis using either subcutaneous unfractionated heparin (5000 units BID, n=11) or low molecular weight heparin (40 mg daily, n=10) starting at the time of admission, while two patients were maintained on their home regimen of novel oral anticoagulants for atrial fibrillation (both in the PE-group). Of the seven patients who developed acute PE, the mean time to diagnosis was 9.4 days following hospitalization. The mean Wells score for the PE+ group was 5.7 and the PE-group 4.2, indicating a lack of discrimination with use of Wells criteria in these two groups. The diagnosis of acute PE was confirmed by CT pulmonary angiography in 6 patients and by transthoracic echocardiography in one patient, who suddenly became hypotensive on day five with severe right ventricular dysfunction and a positive McConnell sign. Despite treatment with thrombolytic therapy (100 mg intravenous tissue plasminogen activator) he expired three days later with multi-organ failure secondary to obstructive shock.

Of the 6 surviving PE patients, four were stratified as having minor (low risk) and two as having submassive (intermediate risk) pulmonary emboli. The mean pulmonary embolism severity index (PESI) score among our 7 PE patients was 119, indicative of high risk of mortality and morbidity (class IV, 4 to 11.4% 30-day mortality risk). Five received anticoagulation with either therapeutic doses of Heparin or Enoxaparin and were later converted to oral anticoagulation. One patient was found to have small, subsegmental PE on CTA, thought to be very low risk, and was maintained on prophylactic doses of subcutaneous Heparin (increased to 5000 units TID).

Discussion

Our study adds to the growing recognition that patients with COVID-19 appear to be at high risk of developing PE. Since a search for PE was not conducted in all our patients admitted to the ICU, the 30% incidence of PE identified in this study may be an underestimate. The mechanism of hypercoagulability in these patients is not known, but the emergence of antiphospholipid antibodies and/or triggering of disseminated intravascular coagulation in critically ill patients have been suggested as possible mechanisms^3,4. The latter is also supported by autopsy reports, which have shown the presence of frank hemorrhage along with small thrombi in sections of the lung parenchyma⁵.

In light of the growing incidence of acute PE in COVID-19 patients, despite standard DVT prophylaxis, intensified regimens of thromboprophylaxis may be indicated. D-dimer tended to rise during hospitalization in the PE group, which is consistent with studies correlating high D-dimer levels in patients at high-risk for PE⁶. However, the results were not consistent, so we could not conclude that initial or serial D-dimer assays could reliably predict who would subsequently develop PE and who would not.

The incidence of symptomatic PE in a general medical ICU is around 0.7-6%⁷, well below the 30% incidence reported in this series. Larger studies are needed to identify risk factors for PE and determine the level of anticoagulation required for specific subgroups. Meanwhile, clinicians should have a low threshold to suspect venous thromboembolism while treating patients with COVID-19 and consider routinely screening these patients and/or utilizing higher doses of thrombo-prophylactic regimens than the current standard for critically ill patients.