Pulmonary Embolism in COVID-19: The Actual Prevalence Remains Unclear

Editor:

Pulmonary embolism (PE) is a life-threatening complication in patients with coronavirus disease 2019 (COVID-19), but its actual prevalence is still unclear. Dr Suh and colleagues recently published a meta-analysis of observational studies in the February 2021 issue of Radiology, concluding that PE occurs in 16.5% of patients with COVID-19 and that d-dimer cutoffs from pre-existing guidelines appear to be appropriate to exclude PE in these patients (1). These conclusions, however, need to be treated with some caution.

First, the vast majority of included studies were retrospective observational studies in which CT pulmonary angiography was only performed in case of clinical suspicion for PE. Patients who did not undergo diagnostic imaging were assumed to have no PE. In our experience, ruling out PE on clinical grounds is difficult (if not impossible) given the fact that respiratory deterioration is not only a major symptom of PE but also of COVID-19. Therefore, such approach will underestimate the true prevalence of PE.

Second, the studies that Dr Suh and colleagues classified as “all patients underwent CT pulmonary angiography” are in fact retrospective analyses of imaging registries. As such, the number of CT pulmonary angiographic examinations positive for PE only reflect the prevalence of PE among selected patients who were actually referred for CT pulmonary angiography by their physician. This selection bias limits extrapolation of these numbers to patients not referred for CT pulmonary angiography, and this presumably overestimates the true prevalence of PE.

Third, the conclusion that specific d-dimer cutoffs can exclude PE is based on a self-fulfilling prophecy. CT pulmonary angiography, the standard test, was only performed in referred patients, whereas the nonreferred patients were considered to be negative for PE without testing. Because d-dimer levels presumably influenced the physicians’ decision to refer for CT pulmonary angiography, patients with low d-dimer levels are less likely to be referred for CT pulmonary angiography and may thus be incorrectly classified as negative for PE.

Unfortunately, a meta-analysis does not eliminate the internal bias present in individual (observational) studies. Therefore, the mentioned biases preclude drawing any conclusions on the prevalence of PE and the predictive value of d-dimer in patients with COVID-19. Hence, we call for well-designed studies that systematically evaluate PE in prespecified populations with COVID-19 to provide these urgently needed answers.