Limitations of D-dimer testing in unselected inpatients with suspected venous thromboembolism

Purpose To determine the utility and limitations of D-dimer testing for the evaluation of venous thromboembolism in hospitalized patients. Methods We performed D-dimer testing by four different methods in unselected inpatients undergoing radiologic evaluation for possible venous thromboembolism. We included patients with a history of malignancy, recent surgery, thrombosis, and anticoagulation treatment. C-reactive protein levels were assayed as a measure of inflammation. Results Of 45 patients with radiographically proven proximal deep venous thrombosis or pulmonary embolism, 43 had elevated D-dimer levels by enzyme-linked immunosorbent assay (ELISA) (sensitivity, 96%); the specificity of the test was 23% (36/157). The qualitative non-ELISA tests had higher specificities, but their sensitivities were <70%. Nineteen patients (42%) with thrombosis had false-negative D-dimer tests by at least one assay. The specificity of the tests decreased with increasing duration of hospitalization, increasing age, and increasing C-reactive protein levels. D-dimer testing had little or no utility in distinguishing patients with thrombosis from those without in patients who had been hospitalized for more than 3 days, were older than 60 years, or had C-reactive protein levels in the highest quartile. Conclusion In unselected inpatients, D-dimer testing has limited clinical utility because of its poor specificity. This is particularly true for older patients, those who have undergone prolonged hospitalization, and those with markedly elevated C-reactive protein levels. In some patient subsets, a negative non-ELISA D-dimer test cannot discriminate between inpatients with and without thrombosis.