Diagnostic tests are considered useful if treatment decisions are affected by the results. When evaluating patients for pulmonary embolism (PE), the diagnostic threshold is low because the results of a missed diagnosis are so serious. The initial evaluation is based on algorithms that classify the various signs and symptoms of PE into low, moderate, and high pretest probabilities. A ventilation-perfusion (V/Q) lung scan is then obtained to identify V/Q mismatches. Alternative diagnostic options include helical computed tomographic (CT) scans, pulmonary angiography, and flow studies of the extremities. A new diagnostic tool, d-dimer, measures a fibrin degradation product that is often increased when thromboembolism occurs. Although this measurement is helpful, the level also can be elevated in common inflammatory and infectious diseases and after trauma or surgery. Brown and associates used a meta-analysis of prospective, quantitative studies using enzyme-linked immunosorbent assay (ELISA) to determine the accuracy of d-dimer measurements used to diagnose PE in the emergency department.
Acceptable criteria for diagnosing PE included a high-probability V/Q scan, CT scan results that were positive for PE, and positive lower-extremity imaging study results. Acceptable criteria for a negative diagnosis included a normal or very-low-probability V/Q scan, negative angiography results, or negative follow-up for at least three months. Eleven studies were included, with a study population totaling 2,126 patients.
The pooled data resulted in a sensitivity of 0.95 (95 percent confidence interval [CI], 0.90 to 0.98) and a specificity of 0.45 (95 percent CI, 0.38 to 0.52). Subgroup analysis was limited, but among older patients (70 or more years of age), d-dimer testing had a lower specificity. Among patients with symptoms for more than three days, both sensitivity and specificity showed a lower trend.
The authors conclude that a highly sensitive test, like the ELISA d-dimer, can help rule out a diagnosis and is particularly useful in patients with a low pretest probability. Further testing may not be needed in many of these patients. These results cannot be generalized to other types of d-dimer analysis. A newer, latex-based testing method has accuracy similar to the estimate derived in this meta-analysis.