Evaluation of probability-adjusted D-dimer algorithms among patients imaged for pulmonary embolism in three Canadian...

D-dimer testing can reduce imaging utilization in the workup of pulmonary embolism, but the optimal cutoff remains unclear. The conventional D-dimer cutoff is < 500 µg/L Fibrinogen Equivalent Units, while the age-adjusted cutoff in patients over 50 is age × 10. Newer probability-adjusted strategies-the YEARS criteria and PEGeD algorithm-allow a higher threshold (D-dimer < 1000 µg/L) for select low-risk patients. We aimed to retrospectively compare the diagnostic accuracy of four evidence-based pathways to exclude pulmonary embolism without imaging among our emergency department patients who were imaged.

A historical patient cohort was generated including all adults who received computed tomography pulmonary angiography or ventilation-perfusion scans after D-dimer testing in three emergency departments in a large urban Canadian center. Electronic medical records were reviewed to retrospectively compare the test performance of four D-dimer pathways: (i) conventional, (ii) age-adjusted, (iii) YEARS, and (iv) PEGeD.

Of 1092 patients, 129 had pulmonary embolism. Conventional and age-adjusted cutoffs were both 100% sensitive (95% CI, 97.1-100.0), with specificities of 3.5% (2.5-4.9) and 6.4% (5.1-8.2). YEARS was 93.8% sensitive (88.2-96.8) and 30.4% specific (27.6-33.4), with PPV of 15.3% (13.0-18.0) and NPV of 97.3% (94.8-98.6). PEGeD was the least sensitive (92.2%, 86.3-95.7), but most specific (39.1%, 36.1-42.3), with PPV of 16.9% (14.3-19.8) and NPV of 97.4% (95.3-98.6). PEGeD would have resulted in the most patients managed without imaging (35.4%), followed by YEARS (27.6%), age-adjusted (5.7%), and the conventional cutoff (3.1%).

While implementation of a probability-adjusted D-dimer pathway, such as YEARS or PEGeD, would have substantially reduced imaging utilization, these strategies may miss some cases of pulmonary embolism detectable by both age-adjusted and conventional cutoffs. We add to the heterogeneity of safety data, suggesting that adjustment of D-dimer to clinical probability represents a trade-off between sensitivity and imaging utilization.