Failure rate of D‐dimer testing in patients with high clinical probability of pulmonary embolism: Ancillary analysis of three European studies

Abstract

BackgroundIn patients with a high clinical probability of pulmonary embolism (PE), the high prevalence can lower the D‐dimer negative predictive value and increase the risk of diagnostic failure. It is therefore recommended that these high‐risk patients should undergo chest imaging without D‐dimer testing although no evidence supports this recommendation.ObjectiveThe objective was to evaluate the safety of ruling out PE based on D‐dimer testing among patients with a high clinical probability of PE.MethodsThis was a post hoc analysis of three European studies (PROPER, MODIGLIANI, and TRYSPEED). Patients were included if they presented a high clinical probability of PE (according to either the Wells or the revised Geneva score) and underwent D‐dimer testing. The D‐dimer–based strategy ruled out PE if the D‐dimer level was below the age‐adjusted threshold (i.e., <500 ng/mL in patients aged less than 50 and age × 10 ng/mL in patients older than 50).The primary endpoint was a thromboembolic event in patients with negative D‐dimer either at index visit or at 3‐month follow‐up. A Bayesian approach estimated the probability that the failure rate of the D‐dimer–based strategy was below 2% given observed data.ResultsAmong the 12,300 patients included in the PROPER, MODIGLIANI, and TRYSPEED studies, 651 patients (median age 68 years, 60% female) had D‐dimer testing and a high clinical probability of PE and were included in the study. PE prevalence was 31.3%. Seventy patients had D‐dimer levels under the age‐adjusted threshold, and none of them had a PE after follow‐up (failure rate 0.0% [95% CI 0.0%–6.5%]). Bayesian analysis reported a credible interval of 0.0%–4.1%, with a 76.2% posterior probability of a failure rate below 2%.ConclusionsIn this study, ruling out PE in high‐risk patients based on D‐dimer below the age‐adjusted threshold was safe, with no missed PE. However, the large CI of the primary endpoint precludes a definitive conclusion.