The predictive characteristics of D-dimer testing in outpatients with suspected venous thromboembolism: a Bayesian approach

BACKGROUND: Despite the widespread use of quantitative methods to measure D-dimer, clinical decisions commonly are based only on binary test information (positive/negative). This study aimed to determine the significance of quantitative D-dimer results in the evaluation of venous thromboembolism (VTE) by means of a differentiated Bayesian approach. METHODS: Prospective study in 118 outpatients referred for workup of suspected pulmonary embolism (n = 75) or deep vein thrombosis (n = 43). The sensitivity and specificity of D-dimer results obtained by DD VIDAS (Biomerieux, France), STA Liatest (Diagnostica Stago, France), and D-dimer plus (Dade, US) were assessed for five different cut-offs. Further, predictive values and multilevel likelihood ratios were calculated in order to assess the operative test characteristics in excluding or confirming VTE. RESULTS: At a cut-off of 500 ng/ml and pretest probabilities < 47%, the VIDAS provides a negative predictive value (NPV) > 95%, whereas a positive predictive value (PPV) > 95% is obtained in patients with a D-dimer > 10,000 ng/ml and pretest probabilities > 50%. At a cut-off of 500 ng/ml and pretest probabilities < 33%, the Liatest exhibits a NPV > 95%, whereas a PPV > 95% is obtained in patients with a D-dimer >10,000 ng/ml and pretest probabilities > 37%. Finally, with the D-dimer plus, a NPV > 95% is seen at a cut-off of 150 ng/ml and pretest probabilities < 30%, whereas a PPV > 95% is obtained at a cut-off > 1000 ng/ml and pretest probabilities > 67%. CONCLUSIONS: D-dimer measurements in outpatients cannot only allow for exclusion but, in some situations, also for confirmation of venous thromboembolism. It is therefore advisable to conduct a quantitative interpretation of D-dimer results.