Comparison of Diagnostic Performance of Three New Fast D-Dimer Assays in the Exclusion of Deep Vein Thrombosis

D-Dimer (DD), a soluble cross-linked fibrin degradation product, has been suggested as a reliable marker to rule out deep vein thrombosis (DVT) (1)(2)(3)(4). The disease can be ruled out in approximately one-third of outpatients suspected with DVT when the plasma concentration of DD is below an appropriate cutoff value (4). Measurement of plasma DD concentrations has been performed by two major methods: ELISA, which is highly sensitive but time-consuming; and latex agglutination, which is faster but less sensitive than ELISA (2)(4). SimpliRED, a latex assay, detects DD directly in whole blood, avoiding the plasma preparation step (5). However, this assay gives only qualitative results and adds an interindividual observation bias that may partly explain the discordant sensitivity values obtained by different investigators (3)(5)(6)(7). In the present study, we investigated the clinical utility of three new fast DD assays in the diagnosis of DVT. We compared 177 consecutive outpatients (117 women and 60 men) from the Montreal cohort in a bicentric prospective study (8). The mean age was 56 years (range, 19–87 years). The inclusion criteria were that subjects be ambulatory patients with a clinical suspicion of DVT and >16 years of age. The exclusion criteria were as follows: (a) recent surgery; (b) expected survival <3 months; (c) hospitalization >24 h before the onset of symptoms; (d) refusal or inability to give informed consent; (e) ongoing anticoagulant therapy at the onset of symptoms; and (f) follow-up not possible. Ultrasound and an uneventful 3-month follow-up were chosen for definitive diagnosis (reference standard). The institutional ethics committee of Hôpital Saint-Luc of the University of Montreal Hospital Center approved the protocol. Blood was drawn into sodium citrate tubes within 2–3 h of arrival to the emergency ward, and fresh plasma was tested immediately by the VIDAS ELISA method (bioMérieux). Frozen plasma aliquots (−70 °C) were used later to test DD concentrations by Tinaquant (Roche Diagnostics), which was performed on an Hitachi 911, and by Turbiquant (Behring), which is manual and uses a turbiditimer instrument (Behring) for DD quantification. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated with their confidence intervals for each assay. Sensitivities were compared for statistical significance by the McNemar test. The areas under the curve (AUCs) for ROC curves were calculated with SPSS software (Ver. 9) and compared by the method of Hanley and McNeil (9). The presence of DVT was confirmed by ultrasound in 36 patients (prevalence of 20%). No patient with a negative diagnosis had developed confirmed vein thromboembolic disease during the follow-up. VIDAS had the best sensitivity and NPV (97% and 98%, respectively), followed by Turbiquant, which had a sensitivity of 92% and NPV of 95%, whereas Tinaquant had a sensitivity of 83%, a NPV of 93%, and the best specificity (Table 1 ). Statistical comparison of sensitivities showed significant differences only between Tinaquant and VIDAS (P <0.025). Tinaquant ruled out DVT in 46% of the cohort vs 32% for VIDAS and Turbiquant. PPV was very low for the three assays (∼30%). The distribution of plasma DD results obtained with the three assays is shown in Fig. 1 . Values obtained with Turbiquant were 200-2970 μg/L, compared with 73–115 048 μg/L for VIDAS and 0–20 770 μg/L for Tinaquant. Values >9000 μg/L were represented as equal to 9000 μg/L. For VIDAS and Tinaquant, the correlation coefficient was 0.93. The correlation coefficients for Turbiquant and VIDAS (0.6) and Turbiquant and Tinaquant (0.65) weres considerably lower. The highest sensitivity-specificity pair for VIDAS (97%-44%) was obtained at the cutoff of 550 μg/L; similarly, the manufacturer reports sensitivity-specificity of 97%-40% at the recommended cutoff (500 μg/L). In contrast, the highest sensitivity-specificity for Tinaquant, 97%-31%, was at 250 μg/L, compared with 83%-58% at 500 μg/L. The NPV obtained by Tinaquant at 250 μg/L was 97% compared with 93% obtained at 500 μg/L. The highest sensitivity-specificity for Turbiquant was obtained at 235 μg/L (97%-34%), compared with 92%-40% obtained at the cutoff recommended by the manufacturer (250 μg/L). No significant differences were found between the AUCs of the ROCs of the three assays. The AUC ± SE was 0.876 ± 0.037 for VIDAS, 0.845 ± 0.042 for Tinaquant, and 0.809 ± 0.040 for Turbiquant. The performance we obtained with VIDAS was in accordance with previous studies reporting a sensitivity of 96–100% and NPV of 97–100% for DVT (3)(7)(10)(11) with similar performance in pulmonary embolism (12). In contrast, we obtained a low sensitivity with Tinaquant (83%) compared with the two recent investigations (3)(13) evaluating its performance in DVT (sensitivity of 99% and 98%, and NPV of 93% and 96%, respectively). However, the prevalence of DVT in these two studies was very high: 67% (3) and 46% (13) instead of 20–25%, as reported in the majority of studies of unselected patients. The rate of missed thromboses obtained by Tinaquant was very high (17%), but this test gave performance similar to that of VIDAS if the cutoff was fixed at 250 μg/L instead of 500 μg/L. The choice of the cutoff should be further studied in a large population. In pulmonary embolism, Turbiquant has been reported to have sensitivities of 72% and 89% at cutoff values of 99 and 66 μg/L, and a NPV of 97% at the same cutoff values (14). In DVT, we obtained a sensitivity (92%) and NPV (95%) that were lower than the diagnostic performance of VIDAS. Because of interference from intrinsic turbidity, ultracentrifugation was necessary for five plasma samples. This additional step could represent a limitation of the use of Turbiquant in the emergency ward. In summary, at the actual cutoff given by the manufacturers, VIDAS seems to have the best sensitivity and NPV, followed by Turbiquant. Tinaquant has a diagnostic performance similar to that of previously reported latex assays. However, at a cutoff of 250 μg/L, it shows performance similar to that of VIDAS. Tinaquant and Turbiquant should be further evaluated in large populations. Diagnostic performance of three D-dimer assays. Exclusion % is the proportion of patients in the entire cohort ruled out by a negative DD result. Values in parentheses are 95% confidence intervals. The NPV of the entire cohort (Geneva + Montreal) was published previously [see Ref. (8)]. Diagnostic performance of three D-dimer assays. Exclusion % is the proportion of patients in the entire cohort ruled out by a negative DD result. Values in parentheses are 95% confidence intervals. The NPV of the entire cohort (Geneva + Montreal) was published previously [see Ref. (8)]. Individual DD results obtained by VIDAS (A) and Tinaquant (C) at a cutoff of 500 μg/L and by Turbiquant (B) at a cutoff of 250 μg/L. N, negative; P, positive.