Validity of D-dimer tests in the diagnosis of deep vein thrombosis: a prospective comparative study of three quantitative assays
Dr. Torben Bjerregaard Larsen Md, PhD Department of Clinical Biochemistry, Section for Thrombosis and Hemostasis Aalborg Hospital, Hobrovej 18 P.O. Box 365 DK91 Denmark. Phone: (+45) 9932 3158 Fax: (+45) 9813 1196 E-mail: firstname.lastname@example.org
Abstract. Larsen TB, Stoffersen E, Christensen CS, Laursen B (Aalborg Hospital, Aalborg, Denmark). Validity of D-dimer tests in the diagnosis of deep vein thrombosis: a prospective comparative study of three quantitative assays. J Intern Med 2002; 252: 36–40.
Objectives. To assess the diagnostic reliability of a new quantitative D-dimer assay (VIDAS New®) and an established quick test (Nycocard® D-dimer assay) in the diagnosis of deep vein thrombosis (DVT) compared with ultrasonography. A third assay (Auto Dimer®) became available during sample collection and has been included in the final assessment. The diagnostic performance of the Auto Dimer® assay was evaluated on three different coagulation analysers.
Design. A clinical prospective study of patients admitted to hospital for evaluation of DVT.
Setting. The admission ward at Aalborg Hospital.
Subjects. A total of 113 outpatients with suspected DVT.
Main outcome measures. Compression ultrasonography was used as the reference method for a diagnosis of DVT and compared with different D-dimer assays. The results were expressed as sensitivity, specificity, positive predictive value and negative predictive value (NPV).
Results. Deep vein thrombosis was established in 49 patients (43%). Two D-dimer assays (VIDAS New® and Auto Dimer®) showed sensitivities of 90 and 88%, specificities of 42 and 44%, and NPV's of 85 and 83%, respectively. The Nycocard® D-dimer assay showed a sensitivity of 63%, specificity of 67% and NPV of 71%.
Conclusions. The diagnostic performance of VIDAS New® and the Auto Dimer® D-dimer assays is almost identical, but this study suggests that neither of the D-dimer assays is suitable as the only screening method for DVT, in a situation with a high pretest probability of DVT. This call for a differential strategy that distinguishes between cases of low and high clinical probability using either a D-dimer test or ultrasonography.
Abbreviations DVT, deep venous thrombosis, NPV, negative predictive value, PPV, positive predictive value
Accurate diagnosis of deep vein thrombosis (DVT) and correct therapy are important to reduce the risk of long-term morbidity or mortality because of pulmonary embolism. As clinical signs and symptoms are nonspecific, DVT must be confirmed by objective methods.
Today, most hospitals use the noninvasive compression ultrasonography as the gold standard, with high sensitivity and specificity for proximal DVT, whilst the sensitivity for calf vein thrombosis is slightly lower [1-4].
Analysis of plasma D-dimer, a degradation product of cross-linked fibrin, has been used to support or exclude the tentative diagnosis of DVT, combined with clinical probability [5-8].
The main interest of a D-dimer assay is to rule out DVT in patients presenting with clinical symptoms of thrombosis and, hence, a test with a high negative predictive value (NPV) is needed. Standardization of D-dimer assays has not been possible, primarily because of the heterogeneity of the fibrin degradation products present, the lack of specificity of antibodies used in the assays, variability of calibrators and differences of format of the assay systems . Each assay must be validated in a clinical setting to obtain specificity, sensitivity and, in particular, the NPV. This setting should reflect a representative incidence of DVT in patients referred to the hospital.
The aim of this study was to assess the diagnostic reliability of a new quantitative D-dimer assay (VIDAS New®) and an established quick test (Nycocard® D-dimer assay) in the diagnosis of DVT, compared with ultrasonography. A third assay (Auto Dimer®) became available during sample collection and has been included in the final assessment. The diagnostic performance of the Auto Dimer® assay was evaluated on three different coagulation analysers.
Patients and methods
In agreement with the local Medical Ethics Committee, the patients were not asked to give informed consent, because the study basically was a control of a laboratory method, and the patients were not exposed to any tests or therapy beyond standard procedures.
Patients and study design
Between May 2000 and March 2001, a total of 124 patients were referred to the medical admission ward at Aalborg Hospital, with a diagnosis of venous thromboembolism classified according to the Danish version of the International Classification of Diseases (ICD). Admission rates for venous thromboembolism were obtained, using the ICD-10 codes I800–I809 and I820–I829, respectively. Standard admission procedures were unchanged, i.e. the primary examination was completed by the physician on duty, and 118 symptomatic patients were included in the study. When ultrasonography of the symptomatic leg was decided, standard blood tests were taken.
The patients were not stratified according to pretest probability prior to admission, but with high clinical probability of proximal DVT; one dose (175 IU kg–1) of tinzaparin (Innohep®) was given on admission.
Diagnosis of deep venous thrombosis
Ultrasonography was performed by the radiologist on duty according to standard procedures (B-mode, grey scale ultrasonography with colour Doppler) at the Department of Radiology, Aalborg Hospital. The scan included the femoral veins, the popliteal vein down to the trifurcation of the calf veins to locate a proximal DVT and the calf veins to locate a distal DVT. The result of the D-dimer analysis was unknown to the radiologist who performed the procedure. Planned ultrasonography was cancelled in five patients, leaving 113 patients for evaluation. Subsequently, the results were compared with the D-dimer values by two independent investigators (Dr Larsen and Dr Laursen).
The blood samples for D-dimer analysis were obtained on admission, collected in 1/10 volume of 0.13 mol L–1 sodium citrate, centrifuged at 3500 g for 10 min to obtain platelet-poor plasma and stored at –80°C until simultaneous analysis with the different assays.
The samples were analysed by three different assays. The VIDAS New® D-dimer assay (bio-Mérieux, Lyon, France) is a new quantitative enzyme-linked immuno assay based on a well-established sandwich ELISA method  combined with a final fluorescent measurement on a dedicated analyser. The Auto Dimer® (Biopool® International, Umeå, Sweden) is a new quantitative latex test for cross-linked fibrin degradation products, involving a specific monoclonal antibody. Nycocard® D-dimer (Axis-Shield PoC AS, Oslo, Norway) is a quantitative rapid test based upon an immunometric flow-through principle utilizing an optical reading device, the Nycocard® READER. All tests were handled according to instructions from the manufacturer. The Auto Dimer® assay was originally manufactured for use on a Hitachi® 911/902 Analyser (Hoffmann-La Roche Ltd., Basel, Switzerland). In this study, the Auto Dimer® assay was adapted for the BCT™ and BCS™ Coagulation Analysers (both Dade Behring, Marburg, Germany) and the Thrombolyser™ Rack Rotor Analyser (Behnk Electronic, Norderstedt, Germany). The technologist analysing the samples was unaware of the result of the ultrasonography. The interserial precision for the Auto Dimer® and the VIDAS New® D-dimer assays was established prior to the study in 20 controls. In terms of coefficient of variations we found values between 4 and 5% and 5–6%, for the Auto Dimer® and the VIDAS New® D-dimer assay, respectively.
The results were expressed as sensitivity, specificity, positive predictive value (PPV) and NPV. This study followed the guidelines recommended for reporting studies of diagnostic accuracy of medical tests .
Patient characteristics of 113 outpatients (58 females and 55 males) with suspected DVT are shown in Table 1. Ultrasonography was performed within 24 h in 70 patients, within 48 h in 39 and within 60 h in four patients. A diagnosis of lower limb DVT was established in 49 patients (43%), of which seven (14%) were distal to the popliteal vein. Using the discriminant values suggested by the manufacturer, two D-dimer assays (VIDAS New® and Auto Dimer®) showed sensitivities of 90 and 88%, specificities of 42 and 44%, and NPVs of 85 and 83%, respectively. The Nycocard® D-dimer assay showed a sensitivity of 63%, a specificity of 67% and a NPV of 71%. Alternative cut-off points did not affect the diagnostic performance significantly (Table 2. The best diagnostic performance for the Auto Dimer® assay was obtained using the BCT™ Coagulation Analyser and a cut-off point of 0.20 mg L–1. The Thrombolyser™ and the BCS™ analysers had similar performance, but the latter performed better utilizing a cut-off point of 0.30 mg L–1 (details in Table 3.
Diagnostic performance of three D-dimer assays at different cut-off levels
Diagnostic performance of the Auto Dimer®
D-dimer assay using three different instruments
The use of one of these three D-dimer assays as the only parameter in excluding DVT in outpatients with a pretest probability of DVT of 43% will result in a false negative result (1 – NPV) in 15–29% of the patients. In an emergency ward, the pretest probability of DVT is usually high in patients referred for ultrasonography, and in this study we found a higher pretest probability compared with similar studies [10,12-14]. These studies report NPVs from 90 to 95%, which is considerably higher than the performance found in the present study. Using the VIDAS New® D-dimer assay as an example, and a pretest probability for disease only half (21.5%) of that in the present situation, the NPV would have increased from 85 to 95%, but the PPV would decrease from 54 to 30% (causing a high false positive rate). Thus, lowering the pretest probability is not the appropriate way to improve a diagnostic test. Good clinical practice must be the principal objective.
Several strategies for the exclusion of deep venous thrombosis have been suggested . One study used the standard VIDAS® D-dimer assay as the initial screening in 474 patients with suspected deep venous thrombosis, combined with ultrasonography . This combined strategy could demonstrate a NPV of 99.3% in a population with a low prevalence of venous thromboembolism of 23%. Using the reversed strategy, i.e. ultrasonography as the initial screening method combined with a D-dimer assay, another study of 476 patients with suspected deep venous thrombosis revealed a prevalence of 27.5% and a NPV of 99.6% . This calls for a differential strategy; in cases of low clinical probability, the initial test could be a D-dimer test with a high NPV . In cases with a high clinical probability of disease, a test with a high PPV should be chosen, for example, ultrasonography.
In the present study, the diagnostic performance of VIDAS New® and the Auto Dimer® D-dimer assays was almost identical, with no significant difference in performance (Table 2). The latter requires speciality equipment provided by the manufacturer, whereas the Auto Dimer® assay does not. Although, the VIDAS New® assay apparently has the best performance, the Auto Dimer® assay is faster and can work in many routine settings on existing equipment. Thus, the choice of assay probably depends on different circumstances in the single setting.
The present study has strengths and limitations. The clinical scenario is probably comparable with daily practice in many emergency wards in Western Europe, and a pretest probability of about 40–50% is common in Scandinavian countries [17-19]. The study was designed as a blinded, prospective study, and assessment of pretest probability of DVT prior to admission was based exclusively on a clinical assessment of the patient. In the present study, 94% had proximal thrombosis, but a higher proportion of distal thrombosis would have a profound effect on the diagnostic accuracy.
In this study, we did not include phlebography as the gold standard, but based on results from a previous study  and our own in-house examination of the performance of phlebography versus compression ultrasonography, the latter has become the routine procedure in our hospital (personal communication with Dr A Nielsen, Department of Radiology, Aalborg Hospital). The main purpose of this study was to evaluate how the use of D-dimer assays compares with ultrasonography in confirmation or exclusion of DVT, reflecting a routine scenario to the clinicians.
In conclusion, this prospective study suggests that neither of the D-dimer assays is suitable as the only screening method for DVT in an emergency ward with a high pretest probability of DVT, but calls for a differential strategy that distinguishes between cases of low and high clinical probability. For practical reasons this should be a very simple strategy [16, 20].