For more than 2 decades, the plasma D-dimer assay has been used in conjunction with clinical prediction scores to rule out pulmonary embolism (PE) among patients with a low pretest probability of having this condition without the need for more costly and invasive methods.1,2 The increased thrombotic risk among patients hospitalized with COVID-19 (ie, those with high pretest probability of PE) and increased D-dimer levels in the absence of thrombosis diverge considerably from the study population used to originally validate this assay.3 However, the availability of D-dimer samples routinely collected from patients hospitalized with COVID-19 and the heterogeneity of early, smaller studies generated uncertainty regarding the clinical utility of the assay in this setting.4 Therefore, we conducted a diagnostic accuracy study to characterize the performance of D-dimer using various threshold values to exclude PE among patients hospitalized with COVID-19.
This diagnostic study was approved by the University of South Florida institutional review board, which granted a waiver of informed consent because this study analyzed deidentified medical records. This study used the Standards for Reporting of Diagnostic Accuracy (STARD) reporting guideline for diagnostic studies.
A retrospective, cross-sectional sample of 1541 patients consecutively hospitalized with COVID-19 at a single hospital from January 1, 2020, to February 5, 2021, was collected. Plasma D-dimer concentrations from an automated, standardized assay (expressed as fibrinogen equivalent units) were compared with the criterion standard of computed tomographic pulmonary angiography (CTPA) among 287 patients with suspected PE. D-dimer distributions among patients with and without PE were compared using descriptive statistics and linear regression (after log-normal transformation at α = .05). The ability of plasma D-dimer concentrations collected the day of CTPA to correctly classify patients with PE was evaluated with a static threshold of 0.5 μg/mL or more (to convert to nanomoles per liter, multiply by 5.476) and an age-adjusted threshold (ie, D-dimer value, 0.01 × [age − 50 years]) for individuals aged older than 50 years.5 Receiver operator characteristic curves (ROCs) were evaluated for thresholds and an expanded set of D-dimer values ranging from 0.5 μg/mL to 20 μg/mL. Data were analyzed using Stata statistical software version 13 (StataCorp) from May through August 2021.
Among 287 patients with COVID-19 and suspected PE (177 [51.4%] men; mean [SD] age, 58.2 [16.1] years), 118 patients required intensive care unit levels of care (41.1%) and 27 patients died during hospitalization (9.4%). Of 287 patients with CTPA, 37 patients had radiographic evidence of PE (12.9%) and 250 patients did not (87.1%); 265 patients had plasma D-dimer levels of 0.05 μg/mL or more (92.3%), including all patients with PE and 225 of 250 patients without PE (91.2%). The median (IQR) D-dimer values were 1.0 (0.6-1.8) μg/mL for 250 patients without PE and 6.1 (2.0-19.4) μg/mL for 37 patients with PE. D-dimer values ranged from 0.2 μg/mL to 128 μg/mL among patients without PE and from 0.5 μg/mL to more than 10 000 μg/mL among patients with PE, and patients without PE had statistically significantly decreased mean (SD) D-dimer values (8.7 [11.6] μg/mL vs 1.2 [2.8] μg/mL; P < .001).
A D-dimer concentration of 0.05 μg/mL was associated with a sensitivity of 100%, specificity of 8.8%, negative predictive value (NPV) of 100%, positive predictive value (PPV) of 13.9%, and negative likelihood ratio (NLR) of less than 0.1. The age-adjusted threshold was associated with a sensitivity of 94.6%, specificity of 22.8%, NPV of 96.6%, PPV of 13.9%, and NLR of 0.24. The ROC analyses (Figure) yielded areas under the curve that were not statistically significantly different using the 0.5 μg/mL threshold compared with the age-specific threshold (0.81 vs 0.80; P = .67). Performance measures for thresholds 0.5 μg/mL (sensitivity, 100%; specificity, 9.3%; accuracy, 21.0%; positive likelihood ratio [PLR], 1.10; NLR, 0) to 20 μg/mL (sensitivity, 20.0%; specificity, 96.6%; accuracy, 86.7%; PLR, 5.90; NLR, 0.57) are presented in the Table.
This diagnostic study found that all hospitalized patients with COVID-19 and radiographic evidence of PE had plasma D-dimer levels of 0.05 μg/mL or greater. If using D-dimer to exclude patients with PE, the increased values we found among 92.3% of patients suggest that this assay would be less useful than in the populations in which it was originally validated, among which a minority of patients had increased D-dimer values. Setting higher D-dimer thresholds was associated with improved specificity at the cost of an increased false-negative rate that could be associated with an unacceptable patient safety risk. The inclusion of patients with D-dimer and CTPA results was necessary to estimate diagnostic performance; however, this may have introduced selection bias by excluding patients unable to undergo CTPA. Nonetheless, given the high pretest probability of PE and low specificity observed in this and other studies,6 these results suggest that the use of D-dimer levels to exclude PE among patients hospitalized with COVID-19 may be inappropriate and have limited clinical utility.
Accepted for Publication: August 7, 2021.
Published: October 8, 2021. doi:10.1001/jamanetworkopen.2021.28802
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Logothetis CN et al. JAMA Network Open.
Corresponding Author: Asa Oxner, MD, Department of Internal Medicine, Morsani College of Medicine, University of South Florida, 13330 USF Laurel Dr, MDC80, Tampa, FL 33612 (firstname.lastname@example.org).
Author Contributions: Drs Logothetis and Weppelmann had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Logothetis, Weppelmann, Jordan, Oxner.
Acquisition, analysis, or interpretation of data: Logothetis, Weppelmann, Jordan, Hanna, Zhang, Charkowick.
Drafting of the manuscript: Logothetis, Weppelmann, Jordan, Zhang, Oxner.
Critical revision of the manuscript for important intellectual content: Logothetis, Weppelmann, Jordan, Hanna, Charkowick, Oxner.
Statistical analysis: Logothetis, Weppelmann, Jordan.
Administrative, technical, or material support: Jordan, Hanna, Zhang, Oxner.
Supervision: Logothetis, Weppelmann, Oxner.
Conflict of Interest Disclosures: None declared.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the University of South Florida, Morsani College of Medicine, Tampa General Hospital, or any other organization.
et al. The original and simplified Wells rules and age-adjusted D-dimer testing to rule out pulmonary embolism: an individual patient data meta-analysis. J Thromb Haemost
. 2017;15(4):678-684. doi:10.1111/jth.13630PubMedGoogle ScholarCrossref
et al; ISARIC4C investigators. Features of 20 133 UK patients in hospital with COVID-19 using the ISARIC WHO Clinical Characterisation Protocol: prospective observational cohort study. BMJ
. 2020;369:m1985. doi:10.1136/bmj.m1985PubMedGoogle Scholar