Use of different d‐dimer levels to exclude venous thromboembolism depending on clinical pretest probability

Abstract

Currently, the same D-dimer cut-off point is used to define a positive result for all patients with suspected venous thromboembolism, regardless of their pretest probability. However, use of a relatively high D-dimer cut-off point (lower sensitivity) for those with a low clinical pretest probability, and a low D-dimer cut-off point (higher sensitivity) for those with a high clinical pretest probability, may be preferable. To determine if using three different D-dimer cut-off points according to low, moderate or high clinical pretest probability has greater utility for exclusion of venous thromboembolism than using the same single D-dimer cut-off point in all patients. Data from a previously published study of 571 patients was used to identify the highest D-dimer cut-off point with a negative predictive value of at least 98% for the subgroup of patients with low and high pretest probability. The D-dimer cut-off point for those with moderate clinical pretest probability remained unchanged [0.5 fibrinogen equivalent units (FEU) microgram mL(-1)]. Accuracy of D-dimer testing for venous thromboembolism using three cut-off points vs. one cut-off point was than determined. D-dimer cut-off points of 0.2 and 2.1 FEU microgram mL(-1) were selected for the high and low pretest probability groups, respectively. When three pretest probability-specific cut-off points were used instead of the previously determined single D-dimer cut-off point (0.5 FEU microgram mL(-1)), sensitivity and negative predictive value were unchanged (95 and 98%, respectively), but specificity increased from 44.7 to 60.4% (P < 0.001). This resulted in exclusion of venous thromboembolism in 80 additional patients. Use of three pretest probability-specific D-dimer cut-off points rather than a single D-dimer cut-off point for all patients, has the potential to increase the utility of D-dimer testing for the diagnosis of venous thromboembolism.