D-Dimer for venous thromboembolism diagnosis: 20 years later


Henri Bounameaux, Division of Angiology and Hemostasis, Department of Internal Medicine, Geneva University Hospital and Faculty of Medicine, 24 rue Micheli-du-Crest, CH-1211 Geneva 14, Switzerland.
Tel.: +41 22 372 92 92; fax: +41 22 372 92 99.
E-mail: henri.bounameaux@medecine.unige.ch


Summary.  Twenty years after its first use in the diagnostic workup of suspected venous thromboembolism (VTE), fibrin D-dimer (DD) testing has gained wide acceptance for ruling out this disease. The test is particularly useful in the outpatient population referred to the emergency department because of suspected deep vein thrombosis (DVT) or pulmonary embolism (PE), in which the ruling out capacity concerns every third patient clinically suspected of having the disease. This usefulness is based on the high sensitivity of the test to the presence of VTE, at least for some assays. Due to its poor specificity precluding its use for ruling in VTE, DD testing must be integrated in comprehensive, sequential diagnostic strategies that include clinical probability assessment and imaging techniques such as lower limb venous compression ultrasonography for suspected DVT or multi-slice helical computed tomography for suspected PE. The present narrative review updates the data available on the use of the various commercially available DD assays in the diagnostic approach of clinically suspected VTE in distinct patient populations or situations, including outpatients and inpatients, patients with cancer, older age, pregnancy, a suspected recurrent event, limited thrombus burden, and patients already on anticoagulant treatment.


Clinical suspicion of venous thromboembolism (VTE) mandates objective testing, such as compression ultrasonography or computed tomography to rule in or rule out the disease [1–3]. As VTE is frequent, it is often suspected. Moreover, the index of clinical suspicion has increased over the years and, as a consequence, the prevalence of the disease among suspected individuals has dramatically decreased, sometimes reaching an incidence of 10% or even less [4]. Therefore, the existence of a simple test allowing the exclusion of the disease without further testing in a substantial proportion of patients is of utmost interest. D-dimer (DD) fragments are specific cross-linked fibrin derivatives that can easily be measured in plasma and that fulfil this requirement, at least partially.

D-dimer testing was first proposed for that purpose almost 20 years ago in the settings of suspected deep vein thrombosis (DVT) [5,6] and pulmonary embolism (PE) [7,8]. In the present overview, we update a narrative review of 1997 [9], focusing mainly on clinical needs and basing our recommendations on the results of two recent meta-analyses [10,11].

D-dimer formation

D-dimer (DD) units are generated by action of factor XIIIa on fibrin monomers and polymers and when the endogenous fibrinolytic system degrades cross-linked fibrin present in the organism. These units consist of two identical subunits derived from two fibrin molecules. DD is the final fragment of the plasmin-mediated degradation of cross-linked fibrin and its molecular weight is around 180 000 daltons. Unlike fibrin/fibrinogen degradation products, which are derived from both fibrinogen and fibrin, DD fragments are end products of the action of plasmin on cross-linked fibrin; however, monoclonal antibodies used in the so-called DD assays recognize also many fragments from cross-linked fibrin without prior proteolysis by plasmin [12]. Because 2– 3% of plasma fibrinogen is physiologically converted to fibrin and then degraded, small amounts of D-dimer-containing species are detectable in the plasma of healthy individuals. However, concentration in blood is increased in all conditions associated with enhanced fibrin formation and subsequent degradation by plasmin. D-dimer represents the most frequently used laboratory marker of coagulation and fibrinolysis activation [13]. Plasma level is increased 8-fold on average in VTE compared with controls, with the level falling in parallel with the duration of symptoms and introduction of anticoagulant treatment [14]. Plasma half-life of DD fragments is approximately 8 h, and clearance occurs via the kidney and the reticulo-endothelial system [15].

Assays for measuring D-dimer level

Description of the different assays

Measurement of DD level has been enabled by the development of monoclonal antibodies, which bind to epitopes on DD fragments that are absent on fibrinogen and non-cross-linked fragments of fibrin [16]. Table 1 summarizes the main characteristics of the most commonly used DD tests. The detection of the resulting complexes occurs by enzyme-linked immunosorbent assay (ELISA), immunofiltration, sandwich-type or agglutination techniques. The classic microplate ELISA technique was considered the gold standard and this technique was used in early clinical studies to assess the value of DD for VTE diagnosis. The sensitivity and NPV were high enough to use the assay as an exclusion test in diagnostic strategies for DVT or PE. Unfortunately, this kind of test is cumbersome, labour intensive and appropriate only for batch analysis, and it is ill-suited for routine emergency use [17,18]. However, using these assays established the concept and the usefulness of DD measurement for VTE exclusion [6,19,20]. Subsequently, a single, fast (about 30 min) and fully automated ELISA was developed in which capture antibodies are coated on tips and reagents are located in a strip. Further modified ELISA assays that are more rapid and suited for single samples were developed later on. After the sample is transferred in the dedicated hole, the analysis proceeds automatically [21]. A batch calibration curve assay is performed by the manufacturer and is stored in a data base by bar-code reading. The usual cut-off for VTE exclusion, as determined in clinical studies by receiver operating characteristics (ROC) curve analysis, is the optimal point within the measuring range, which confers a good analytical sensitivity to that particular assay. VTE exclusion by a DD level below the cut-off has been validated in outcome studies, that is studies in which patients with clinically suspected VTE are left untreated on the basis of a normal DD test, and are systematically followed-up for a 3-month period to detect possible thromboembolic events [22,23]. These ELISA tests (e.g. Vidas® D-Dimer, bioMérieux, Marcy l’Etoile, France; Stratus® CS D-dimer Dade Behring, Newark, Delaware, USA; AxSYM® D-dimer, Axis-Shield Diagnostics, Dundee, UK) combine the ELISA technique with a final detection by fluorescence (D-dimer enzyme-linked immunofluorescence assay, ELFA), chemiluminescence (PATHFAST, Mitsubishi Kagabu Iatron, Chiba, Japan and Immulite, Siemens, Eschborn, Germany, in particular) or time-resolved fluorescence (Innotrac Aio!, Turku, Finland), are fully automated, provide a result within 15–35 min and can be used for single sample testing [24–26]. Their main limitation is the requirement of a dedicated immunoanalyzer.

Table 1.   Characteristics of various commercial DD assays (non-exhaustive list)
  1. *No longer available.

Microplate ELISAAsserachrom® Ddi (Stago)
Enzygnost® (Dade-Behring)*
Dimertest Gold (Agen)*
HighLowConsidered as the gold standard; suitable for batch analysis and not useful for real-time single testing; observer-independent
ELISA and fluorescence (ELFA)Vidas® DD (bioMérieux)
AxSym® D-dimer (Abbott)
Stratus® D-dimer (Dade-Behring)
HighLowSimilar sensitivity as classic microplate ELISA; quantitative; suitable for real-time use; observer-independent
ELISA and chemiluminescenceImmulite (Siemens)
Pathfast (Mitsubishi)
HighLowSimilar sensitivity as classic microplate ELISA; quantitative; suitable for real-time use; observer-independent
ELISA and time-resolved fluorescenceInnotrac Aio!No clinical study available; observer-independent
Immunofiltration and sandwich-typeNycoCard® (Nycomed),High-intermediateLow-intermediateReduced sensitivity compared with classic microplate ELISAs; quantitative; suitable for real-time use; observer-independent
Cardiac D-dimer® (Roche)HighHighFew clinical studies available, promising results
Semi-quantitative latex agglutinationDimertest® latex (IL); Fibrinosticon® (bioMérieux)
Ddi latex® (Stago)
IntermediateIntermediateRapid, but insufficiently sensitive to be clinically useful; observer-dependent
Manual whole-blood agglutinationSimpliRED® (Agen)
Clearview Simplify D-dimer® (Agen)
High-intermediateIntermediateRapid, can be performed on whole blood. Exclusion of VTE only in association with low clinical probability; observer dependent
Second-generation latex agglutination (immuno-turbidimetric)TinaQuant® (Roche)
Liatest® (Stago)
Automated Dimertest® (Agen)
MDA® D-dimer (bioMérieux)
Turbiquant® (Dade-Behring)
Miniquant® D-dimer (Trinity)
HemosIL® Dimertest HS (IL)
Innovance D-dimer® (Dade-Behring)
HighIntermediateRapid and quantitative; comparable sensitivity to microplate ELISA; observer-independent

Another rapid format is immunofiltration, in which capture antibodies are coated onto a permeable membrane. The sample, washing solution and second antibody undergo passive filtration. A signal is generated by a colloid gold-labelled tag antibody and quantified with a reflectometer (NycoCard D-dimer, Axis-Shield, Oslo, Norway). A fully automated sandwich-type assay on a dedicated device, using colloid gold-labelled tag antibody, is available (Cardiac Reader D-dimer, Roche Diagnostics, Mannheim, Germany).

In some assay systems D-dimer is measured in anticoagulated whole blood (Cardiac Reader, Pathfast and Innotrac Aio!). In the Stratus assay, an anticoagulated whole blood sample is introduced into the device and centrifugation proceeds automatically.

Latex-based systems rely on the ability of the analyte to agglutinate latex beads coated with the antibody. This implies that fragments possessing either only one epitope or epitopes made unavailable by masking complexes, escape determination. Usually, latex assays are performed with native or slightly diluted plasma, whereas microplate or membrane ELISA techniques require high dilutions. Two types of latex tests are available: semi-quantitative slide assays or quantitative assays. The simplest and least expensive systems are semi-quantitative agglutination methods (Dimertest® latex, Instrumentation Laboratory, Lexington, MA, USA; Fibrinosticon®, bioMérieux, Marcy l’Etoile, France; Ddi latex®, Diagnostica Stago, Paris, France), because they are fast and do not require complicated instrumentation. However, as reading is most often visual, some inter-observer variability in estimating the presence of agglutination is unavoidable. Various clinical evaluations have shown a low sensitivity (between 51% and 96%, with an average around 80%) for acute DVT or PE [27,28], clearly insufficient for excluding these conditions.

More recently, quantitative, automated agglutination methods have been developed. These are photometric or turbidimetric methods, designed to be performed on routine coagulation or clinical chemistry analyzers. They do not need dedicated instruments (e.g. Tinaquant®, Roche Mannheim, Germany; STA Liatest®, Diagnostica Stago; IL® test, Instrumentation Laboratory; BC® D-dimer Plus, Dade-Behring, Marburg, Germany; MDA, Organon Teknika, Oss, The Netherlands) (Table 1). Results are usually available in 5–10 min, observer-independent, and full automation reduces other sources of variability. However, the analytical sensitivity and the low limit of detection may be a cause of concern. The calibration curve usually covers a wide range of concentrations, but the upper reference range of normal values and the detection limit for VTE exclusion often lies in the lower part of the calibration curve where the signal is weak [29]. In spite of these laboratory limitations, clinical performances of turbidimetric DD have been shown to be excellent and close to those of ELISA DD tests. In recent meta-analyses, the diagnostic performances for DD ELISA tests using latex turbidimetric methods appear to have operative characteristics similar to those of ELISA methods [30].

Other available manual and semiquantitative DD tests that use whole-blood can be performed at the patient’s bedside. For example, the SimpliRED® test (AGEN Biomedical Ltd, Brisbane, Queensland, Australia) is a red blood cell agglutination assay designed for use with fresh capillary or venous whole blood. It provides a result in <5 min and is therefore suitable for point-of-care testing. This assay is based on the use of a hybrid monoclonal antibody, which reacts with DD and human erythrocytes, leading to hemagglutination in the presence of fibrin compounds containing the DD epitope. Study results with this assay have been variable, with an average sensitivity of 83% (95% CI, 67–93) for DVT and 87% (95% CI, 64–96) for PE according to the meta-analysis of Di Nisio et al. [11]. Published studies indicate that the sensitivity and NPV of this test are high enough to rule out VTE but only in the presence of a low clinical probability. As readings are visual, some inter-observer variability has been reported [31,32], and it may be that sensitivity is highest when the test is performed by experienced personnel.

Recently, a novel qualitative immunochromatographic method (Clearview Simplify D-dimer®; AGEN Biomedical Ltd) was described. This technique does not require instrumentation because reading is visual, which, however, raises concern regarding inter-observer variability [33]. The diagnostic performance of this test was compared with venography in a sample of 187 outpatients with suspected DVT in an emergency department [34]. Sensitivity and NPV were 94% and 95%, respectively, but confidence intervals remained large in relation to the small sample size [34]. Moreover, more recent studies have reported a definitely less good performance of this test [35,36].

Standardization of the DD assays

One of the main problems with DD measurement is the confusion raised by the multiplicity of commercial assays, with various techniques, cut-offs, systems of units (DD units or FEU, fibrinogen equivalent units; thus, the calibrators for DD units are purified DD fragments whereas the calibration material for FEU is obtained from controlled plasmin digestion of purified fibrinogen clotted in the presence of factor XIII), operational characteristics, and clinical validity. This demonstrated heterogeneity [37] led to sustained attempts to standardize or at least harmonize the expression of test results.

Standardization is based on the application of a comprehensive reference system, including a reference method, primary standard, and the use of SI units. Standardization is possible only if the entity measured in the assay can be clearly defined, which does not apply to DD, which is not a single entity but a complex mixture of degradation products of different sizes [38,39]. As a consequence, direct comparison of results obtained with different methods is impossible and each result should be considered method-specific. Another option would consist of harmonizing test results by applying a mathematical model to make test results comparable [40]. Some models of harmonizing DD tests have been proposed [39,41,42] but have not yielded a consensus yet.

D-dimer as a diagnostic test

D-Dimer for ruling out venous thromboembolism

The cut-off level used in diagnostic strategies for acute VTE does not represent the upper limit of the reference interval in a healthy population. Rather, each diagnostic system has its own cut-off level that helps to identify patients without the disease. This points out the need for clinicians to be aware of the particular test used, of its performances and, obviously, of its diagnostic cut-off.

Moreover, it underlines the fact that only DD assays that have been appropriately validated in prospective outcome studies, or compared with stored plasma samples from outcome studies should be used for clinical purpose. When choosing a DD assay to be used in diagnostic strategies for acute DVT or PE, utmost attention should be paid to the sensitivity and coefficient of variation at the cut-off level, as false-negative results can lead to potentially fatal consequences.

Nevertheless, specificity is also important because it determines the proportion of false-positive results and strongly influences the yield of the test. More specific assays theoretically allow ruling out VTE in a higher proportion of patients but this advantage is generally offset by a lower sensitivity restricting their use to patients with a low clinical probability of VTE. Also, specificity is affected by the patient population and clinical context, in addition to particular assay characteristics. Clinicians must be aware that DD level is increased in many conditions [43], such as infection, inflammation, cancer, surgery, trauma, extensive burns or bruises, ischemic heart disease, stroke, peripheral artery disease [44], ruptured aneurysm or aortic dissection [45], pregnancy, cerebral sinus thrombosis [46], in addition to acute VTE (DVT of the lower limbs and/or pulmonary embolism and DVT of the upper limbs) [43,47,48]. Furthermore, the diagnostic yield of DD is quite low in hospitalized patients [49,50], due to the variety of concomitant conditions potentially elevating DD in those patients, as well as in the elderly because of the rise of DD concentrations in a normal aging population [51]. A useful index allowing comparison of the diagnostic yield of various DD assays or of a specific assay in various clinical settings is the ‘number needed-to-test’ or ‘NNT’, that is the number of patients in whom DD must be measured to rule-out one DVT or one PE [52]. For instance, if the incidence of PE in a theoretical population of 100 patients is 20%, 80 patients will not have PE. Assuming a specificity of the DD assay of 40%, it will rule out PE in 32 of these 80 patients, that is 32% of the overall population. Thus, 3.1 patients will require testing in order to rule out one PE, and the NNT to rule out one PE will be 3.1.

Other important characteristics include sample turnaround time, which should if possible be shorter than 30 min (to allow its use in the emergency room), and observer-dependence of results. Observer-independent methods are to be preferred because results obtained in clinical studies with well-trained readers can probably not be extrapolated to daily clinical practise with an emergency staff functioning 24 h a day [53].

A last important issue is the place of the test in the local diagnostic sequence. If used as a first-step with no additional tests in the presence of a negative result, the method of choice should be the most sensitive one, for example an ELISA test with sensitivity as close as possible to 100% in order to minimize the proportion of false-negative results. In contrast, if the DD test is used in association with other tests such as compression ultrasonography, a less sensitive test may be chosen.

D-Dimer for diagnosing (ruling in) venous thromboembolism

The positive predictive value for VTE rises as DD levels increase progressively above the chosen threshold. In a study evaluating 671 outpatients with suspected PE, the specificity of a DD test (Asserachrom® Ddi ELISA) was 93% when levels exceeded 4000 μg L−1, in the presence of a non-low clinical probability [54]. However, this resulted in a limited positive predictive value because of the relatively low incidence of PE (about 20%), which was even lower in patients with a low clinical probability. In the patients with intermediate to high clinical probability, the likelihood ratio for the presence of PE with a DD concentration above the cut-off was about 5, similar to that of single segmental ventilation/perfusion mismatch on a lung scan.

While ruling out DVT or PE on the basis of a normal D-dimer test is well supported by literature data and is now well accepted, clinicians would still be probably quite reluctant to accept diagnosing PE or DVT on the sole basis of a positive D-dimer test, even at high plasma concentration [1,2], and would probably require some imaging. Finally, although potentially elegant, the concept of using two different cut-offs for the same test for ruling out or ruling in the disease may be confusing in busy real-life settings and has not been shown to be safe in clinical practise. Therefore, at this stage, DD testing should only be used as an exclusion test for VTE, at least in our view.

Assessment of the various DD assays in the diagnostic work-up of DVT

The SimpliRED® test  As previously discussed, the SimpliRED® is a manual semiquantitative latex test using whole blood that can be performed at the patient’s bedside. In the first study assessing the performances of this test in patients with suspected DVT, the test had a sensitivity of 89%, a specificity of 77% and a NPV of 96% for the exclusion of DVT as compared with venography as the reference diagnostic criterion [55], but a sensitivity as low as 66% has also been reported [27]. In an outcome study, patients with a low clinical probability of DVT and a normal SimpliRED® DD test had a 3-month thromboembolic events rate of 1.8% (95% CI, 0.9–3.3%) [56]. In two Canadian studies with a 3-month follow-up, a negative SimpliRED test reduced the incidence of DVT from 2.5% and 5.6% in the low clinical probability class (according to Wells’ rule, see Table 2 for definition) to less than 1% during a 3-month follow-up [3,57]. Thus, although the test may not be sensitive enough as a stand-alone test for the exclusion of DVT, convincing data exist that demonstrate that the association of a low clinical probability and a negative test may safely rule out DVT.

Table 2.   Most commonly used clinical prediction rules for suspected PE
Wells score [105]Geneva score [106]Revised Geneva score [107]
Previous PE or DVT1.5Previous PE or DVT2Age > 65 years1
Heart rate > 1001.5Heart rate >1001Previous DVT or PE3
Recent surgery or immobilization1.5Recent surgery3Surgery or fracture within 1 month2
Clinical signs of DVT3AgeActive malignancy2
Alternative diagnosis less likely than PE3 60–791Unilateral lower limb pain3
Hemoptysis1 ≥802Hemoptysis2
Cancer1Arterial blood gases Heart rate
  CO2 (kPa) 75–943
   <4.82 ≥955
   4.8–5.191Pain on lower limb deep vein palpation and unilateral edema4
  O2 (kPa)  
  Chest X-ray   
  Elevated hemidiaphragm1  
Clinical probabilityClinical probabilityClinical probability
Dichotomized [71]    
PE unlikely≤ 4    
PE likely>4    

The Vidas® D-dimer test  A normal quantitative ELISA Vidas® D-dimer test (cut-off <500 μg L−1) was reported to have a 100% sensitivity when compared with phlebography in two studies [20,58]. In the large outcome studies of outpatients with suspected DVT, the sensitivity of this test varied between 98% and 99%, in 2239 patients, irrespective of clinical probability assessment [22,59–61].

The rapid automated turbidimetric tests  In two large outcome studies, the sensitivity of a normal turbidimetric assay (Tinaquant®, cut-off <500 μg L−1) for the exclusion of DVT varied from 91% to 98% and the specificity from 44% to 51% [60–62]. In one prospective outcome study, the combination of a normal Tinaquant® test result with a low clinical probability had a NPV of 99.4% [62].

Assessment of the different DD assays in the diagnostic work-up of PE

The SimpliRED® test  In a large prospective study of 1117 patients with suspected PE (incidence of PE, 17%), the SimpliRED® test had a sensitivity of 85%, a specificity of 68% and a NPV of 96% for exclusion of PE [63]. Using the Wells’ scoring system (Table 2), the incidence of PE in the patients with a low clinical probability (Wells score < 2 points) was very low (1.3–3.4%) in two prospective clinical outcome studies. Combining a negative SimpliRED® test result in those patients made it possible to rule out PE with a NPV of 99–99.5% [63,64]. The safety of this strategy has recently been confirmed in 373 patients with a low clinical probability and a negative SimpliRED® test in patients randomly assigned to no intervention or additional testing with lung scintigraphy [65]. None of 182 patients assigned to DD testing alone presented with a VTE event during follow-up after a negative result, compared with one patient in the additional perfusion lung scan group [65].

The rapid quantitative DD assays  Both accuracy studies using pulmonary angiography or venography as reference diagnostic criteria and large studies with prospective data collection have shown sensitivity and NPV of 95–100% for the exclusion of DVT and PE using rapid ELISA Vidas® D dimer assay [20,22,58,60]. These data suggest that in the presence of a negative ELISA Vidas® D-dimer (cut-off <500 μg L−1), the post-test probability of VTE is close to zero, irrespective of the clinical score. Prospective outcome studies confirmed that a negative ELISA Vidas® D-dimer test predicts an uneventful 3-month follow-up. The test was first validated in 195 outpatients with suspected PE [66]. All 56 (29%) patients with a negative ELISA Vidas® D-dimer test did not have PE and had an event-free 3-month follow-up. In a subsequent prospective study of 444 outpatients with suspected PE, the sensitivity and the NPV of this test were 100% and the specificity 46% [22]. In a third study of 965 patients with suspected PE, a normal rapid ELISA Vidas® D-dimer level (<500 μg L−1) ruled out PE with a NPV of 99% and a specificity of 38% [67]. The test allowed ruling out PE in 29% of patients. Although the test was used in those studies irrespective of clinical probability, the NPV of a negative result in patients with a high clinical probability of the disease is not definitely established. Therefore, the latest British Thoracic Society guidelines [2] suggested the test should not be performed in patients with a high clinical probability.

Several recent prospective outcome studies confirmed that it is safe to rule out PE with a normal rapid ELISA Vidas® D-dimer test with a sensitivity and NPV of 100%, irrespective of clinical assessment [68,69]. Table 3 summarizes the diagnostic performances of Vidas® D-dimer test for suspected PE obtained in prospective outcome studies.

Table 3.   Exclusion of pulmonary embolism by a normal ELISA Vidas® D-dimer test in prospective studies
AuthorClinical probabilityCut-off (μg L−1)No. patientsIncidence of PE (%)Sensitivity (%)NPV (%)
  1. NPV, negative predictive value.

de Moerloose et al. [66]Low/intermediate/high<50019624100100
Perrier et al. [22]Low/intermediate/high<50044423100100
Kruip et al. [67]Low/intermediate/high<500234229899
Perrier et al. [72]Low/intermediate/high<50096521100100
Perrier et al. [69]Non-high (Geneva score < 9)<50075626100100
Kucher et al. [68]Non-high (Wells’ rule < 6)<78019145100100
Christopher study [71]Unlikely (revised Wells’ score ≤ 4)<50042820100100
Righini and Le Gal et al. [70]Non-high (revised Geneva score < 11)<500181921100100

In a recent multicenter study including 1819 patients, the DD test allowed exclusion of PE in 561 patients with a non-high clinical probability, and no thromboembolic events occurred in the 3-month follow-up of patients with a DD below 500 μg L−1 [70]. In the Christopher study [71], a prospective outcome study that enrolled 3306 patients with suspected PE, the diagnostic strategy was based on assessment of clinical probability assessment, DD test (two different assays) and spiral CT (mostly multi-slice), the combination of PE unlikely (Wells score ≤ 4 points) and a normal DD (either Vidas® ELISA or Tinaquant®) was present in 1057 patients (32% of the total group), of whom 1028 were not given anticoagulant treatment. Subsequent non-fatal VTE occurred in five patients (0.5%) and two were lost to follow-up. At the usual threshold (500 μg L−1), the Vidas® ELISA test had a sensitivity of 100% for the presence of PE. The five false-negative DD test results (0.8%) in the Christopher study all occurred in the group tested with Tinaquant® [71]. Lastly, in convenience samples of frozen plasma aliquots from an earlier outcome study [72], Reber et al. [24,73] showed that the Stratus CS D-dimer (Dade Behring) and the AxSYM D-dimer assay (Abbott), a microparticle enzyme immunoassay that runs on the Abbott analyzer, had diagnostic performances that matched those of the VIDAS test.

Summary of the diagnostic performances of the various DD assays in suspected VTE

In a recent extensive meta-analysis [11], Di Nisio et al. calculated the diagnostic performances of several DD tests, based on 113 individual studies. A summary of their results is presented in Table 4. Overall, compared with other DD assays, the ELFA, the micro-plate ELISA and the automated quantitative turbidimetric assays have a higher sensitivity but a lower specificity, resulting in a more confident exclusion of the disease at the expense of a potentially increased need for additional imaging [11]. The whole blood agglutination assays display a lower sensitivity (about 85% compared with 95% or more for the ELISAs, the ELFA, and the quantitative automated latex tests) but are more specific (about 70% compared with approximately 50% for the high sensitive assays), which allows safely ruling out VTE in populations with a low incidence of the disease. One year earlier, Stein et al. [10] had published similar results in a careful meta-analysis of 78 studies of DD assays used for either DVT or PE diagnosis.

Table 4.   Summary estimates of sensitivity and specificity of DD methods (adapted from reference [11]) in suspected deep vein thrombosis or pulmonary embolism
 Deep vein thrombosisPulmonary embolism
Type of assaySensitivity (95% CI)Specificity (95% CI)Sensitivity (95% CI)Specificity (95% CI)
ELISA Microplate94 (86–97)53 (38–68)95 (84–99)50 (29–71)
ELFA96 (89–98)46 (31–61)97 (88–99)43 (23–65)
Membrane Immunofiltration89 (76–95)53 (37–68)91 (73–98)50 (29–72)
 Qualitative69 (27–93)99 (94–100)75 (25–96)99 (92–100)
 Semiquantitative85 (68–93)68 (53–81)88 (66–97)66 (43–83)
 Quantitative93 (89–95)53 (46–61)95 (88–98)50 (36–64)
Manual whole-blood assays83 (67–93)71 (57–82)87 (64–96)69 (48–84)

Due to its limited specificity, DD testing can only be part of more comprehensive diagnostic strategies that should be validated in appropriate studies. Figure 1 depicts two examples of such strategies in the settings of suspected DVT and PE. Of note, managing patients according to validated algorithms has been elegantly shown to be significantly safer than using not established management (thromboembolic event rate 3 months after exclusion of PE 1.2% vs. 7.7%, P < 0.001) [74].

Figure 1.

 Place of D-dimer measurement in a contemporary diagnostic algorithm for suspected deep vein thrombosis (DVT) or pulmonary embolism (PE).
1CUS (lower limb venous compression ultrasonography) in a case of suspected DVT.
2MSCT (multi-slice helical computed tomography) in a case of suspected PE.
3In a case of negative CUS or MSCT and high prior clinical probability, consider additional imaging, for example venography (suspected DVT) or lung ventilation/perfusion scintigraphy or pulmonary angiography (suspected PE).
Rx stays for treatment; D-dimer refers to highly sensitive D-dimer assays. If less sensitive assays are used, a negative test result rules out DVT or PE only in patients with low (or unlikely) clinical probability.

Use of DD testing in special populations

A reduced diagnostic yield of DD testing has been observed in several clinical settings, such as in inpatients [43,50], postoperative patients [50], during pregnancy or postpartum [75,76], in patients with a high clinical probability [77] or with previous venous thromboembolic disease [78], and in elderly patients [79]. Indeed, in these situations, although the sensitivity and NPV of the test remain high enough to safely rule out VTE, its clinical usefulness – that is the proportion of patients in which VTE can be ruled out on the basis of a negative DD – is greatly diminished. In other words, the NNT becomes quite high (10 or more). Some of these situations are described hereafter and Table 5 displays the NNT to rule out one PE in various medical conditions.

Table 5.   Number needed to test (NNT) to rule out one PE in various medical situations
AuthorClinical conditionsNNT to rule out one PE
  1. The number NNT reflects the number of patients in whom DD measurement has to be performed to rule out one pulmonary embolism. These figures concern a high-sensitivity, low-specificity test. *No patient with negative DD result in this cohort.

Perrier et al. [22]Outpatients with suspected PE3.3
Righini et al. [77]High clinical probability9.1
Non-high clinical probability2.2
Righini et al. [87]Cancer9.1
No cancer3.1
Le Gal et al. [78]Previous VTE6.3
No previous VTE3.1
Chabloz et al. [75]Pregnancy before the 30th week2.6
Pregnancy between weeks 30 and 424
Righini et al. [79]Elderly outpatients of more than 80 years20
Miron et al. [50]Inpatients (non-surgical patients)30
Inpatients (surgical patients)Infinite*

Elderly patients

Several investigators have reported a significant decrease of specificity and hence of clinical usefulness of the DD test with increasing age [51,80]. In a study from our group [79], a cohort of 1029 patients was divided into six classes of age (by decade) and the diagnostic performances of the test (ELISA) were calculated for each decade. DD allowed ruling out PE in almost two-thirds of patients aged less than 40 years, but in only 5% of patients above 80 years of age [81]. In a cost-effectiveness analysis in relation to age [82], we recently could show that using DD is cost-saving until the age of 79 years. After the age of 80, there was no clear economic advantage of measuring DD. However, DD might still be useful in patients above 80 if the availability of other diagnostic tests is limited or if the risk of imaging using helical CT is considered too high because of impaired renal function.

Some authors attempted to increase the specificity, and hence the clinical usefulness, of DD in elderly people by increasing the diagnostic threshold [83], but this turned out to be unsafe in a diagnostic algorithm in which DD measurement was used as the initial step, PE being ruled out in patients with a concentration below the cut-off [83].

Cancer patients

In cancer patients with clinically suspected deep vein thrombosis (DVT), two recently published studies have drawn opposite conclusions about the clinical usefulness of the same whole blood agglutination DD assay [84,85]. ten Wolde et al. [84] published the results of an outcome study, in which a moderately sensitive DD assay (SimpliRED®) was used to rule out DVT in cancer and non-cancer patients. Of 217 cancer patients, 63 (29%) had normal DD test and a normal ultrasound at the day of referral and were, therefore, not treated by anticoagulants. Only one thromboembolic event occurred during follow-up (1.6%; 95% CI, 0– 8.5%). In contrast, in the study by Lee et al. [85], the NPV of DD was only 79% in cancer patients compared with 97% in patients without cancer. However, in that retrospective analysis of three prospective series [85], the sensitivity of DD was similar in patients with (86%) and without cancer (83%) and the lower NPV was entirely due to a 3-fold higher incidence of DVT in cancer patients (49% vs. 15%).

In cancer patients with suspected PE, limited data are available about the clinical usefulness of DD testing [86]. To the best of our knowledge, only two recently published studies analyzed this particular issue. The first study reported a high sensitivity (100%; 95% CI, 82–100%) and NPV (100%; 95% CI, 72–100%) in a sample of 72 cancer patients [86]. These data were confirmed in a second larger study [87], which studied a large cohort of 1721 outpatients with (n = 164) and without cancer previously included in two outcome studies on the diagnosis of PE [69,72]. Accordingly, the NNT to rule out one PE rose from about 3 in patients without cancer to 9 in those with an active malignancy. In patients in whom PE was considered ruled out on the basis of a negative DD test and who did not receive oral anticoagulant treatment, no thromboembolic event was observed during the 3-month follow-up, either in patients with (0/18; 0%; 95% CI, 0–18%) or in patients without (0/469; 0%; 95% CI, 0–0.8%) cancer. Therefore, sensitivity and NPV of DD were 100% in both cancer and non-cancer patients. However, the 95% CIs in cancer patients were much wider, owing to the small number of patients. Specificity of DD test was definitely lower in patients with cancer: 16% (95% CI, 11–24%) vs. 41% (95% CI, 39–44%) in patients without cancer.

Table 6 displays the diagnostic performances of various DD tests in cancer and non-cancer patients. In summary, these data suggest that ELISA DD assays are probably safe to rule out PE in patients with cancer in the emergency department. However, the test’s clinical usefulness is limited because only one of ten patients with cancer and suspected PE will have a negative DD result [87].

Table 6.   Diagnostic performances of various DD assays in cancer and non-cancer patients
AuthorPatient typeSample sizeDD test Prevalence of DVT/PE (%) Sensitivity (%) Specificity (%)NPV %, (95% CI, when available)
Suspected DVT
 Wells et al. [55]No cancer184SimpliRED®25788292
 Ginsberg et al. [108]No cancer350SimpliRed®13878598
 Kearon et al. [109]No cancer413SimpliRED®11828097
 ten Wolde et al. [84]No cancer217SimpliRED®20936497 (96–98)
Cancer152237984897 (89–100)
 Lee et al. [85]No cancer947SimpliRED®15838296 (95–98)
Cancer12149864879 (63–90)
Suspected PE
 Di Nisio et al. [86]No cancer447Tinaquant® DD19935397 (95–99)
Cancer722610021100 (72–100)
 Righini et al. [87]No cancer1554Vidas® D-dimer3310041100 (99–100)
Cancer1642310016100 (82–100)

Patients with previous DVT/PE

Diagnosing recurrent venous thromboembolism (VTE) is a difficult challenge [88,89]. DD measurements in plasma are widely used as a first-line approach in patients with suspected DVT or PE. In the setting of DVT, Rathbun et al. [90] recently reported the results of an outcome study, in which a highly sensitive DD test (the STA Liatest D-Di) was used to exclude DVT in 300 patients with a previous episode of DVT and suspected of a recurrent event. The final incidence of DVT was 54/300 (18%), and 134 (45%) patients had a negative DD test. During the 3-month follow-up, only one of them experienced proven recurrent VTE (0.8%; 95% CI, 0.0–4%). Admittedly, VTE on follow-up could not be definitively excluded in six patients because of the lack of a diagnostic reference standard for recurrent DVT.

A recent study assessed the safety and usefulness of DD in patients suspected of PE who had experienced VTE in the past [78]. In this study, PE was ruled out by a negative DD test in only 15.9% (49/308) of patients with previous VTE, compared with 32.7% (462/1411) in patients without previous VTE (P < 0.0001). The 3-month thromboembolic risk was 0% (95% confidence interval 0.0–7.9%) in patients with previous VTE and a negative DD test who did not receive anticoagulant treatment. The 2-fold lower risk of obtaining a negative DD result in patients with previous VTE was independent from older age, active malignancy, fever and recent surgery. These data suggest that in patients suspected of PE with a history of previous VTE, a negative ELISA DD test makes it possible to safely rule out the diagnosis. However, the usefulness of the test is lower because the proportion of patients with a history of VTE and a negative DD test tends to be lower than in patients without previous VTE.

Pregnant women

It is well known that DD concentration increases as pregnancy progresses, which should obviously diminish the specificity and clinical usefulness of the test. Nevertheless, the proportion of pregnant women with a normal D-dimer value was still 39% before the 30th week of gestation and 25% before the 42nd week of gestation [75,76]. Therefore, quite a substantial proportion of DD results may be negative during pregnancy and could spare further radiological imaging, even though the NNT is increased during pregnancy.

A recent prospective study suggested a high sensitivity (100%; 95% CI, 77–100%) and NPV (100%; 95% CI, 95–100%) in 149 pregnant women with suspected DVT using another D-dimer test (SimpliRED® D-dimer, AGEN Biomedical). Even though this study was not an outcome study and the confidence interval for sensitivity was large owing to the small number of patients, it confirms the high clinical utility that can be obtained with a whole blood agglutination test in low-risk populations (DVT incidence was 8.7%) [91]. However, there is no evidence from any of the situations in which DD has a lower specificity (patients with cancer, or with previous VTE or in the elderly [78,79,87]) that sensitivity might also be affected. Therefore, although there is presently no prospective outcome study establishing the safety of not treating a pregnant woman with suspected VTE on the basis of a normal DD test result, it is probably safe to do so.

Patients with a small thrombus burden

It has been shown that DD levels are correlated with thrombus extension. Obviously, a large fibrin thrombus will result in higher DD levels than a small thrombus. This may explain why sensitivity of DD has been reported to be lower in the presence of distal (infra-popliteal) DVT [55,92] or subsegmental PE [93]. In a prospective study of 314 consecutive inpatients and outpatients with suspected PE, De Monyéet al. [93] studied the relation between the diagnostic accuracy of DD plasma concentration and pulmonary embolus location. A strict protocol of ventilation-perfusion scintigraphy, spiral computed tomography and pulmonary angiography was used to reach a final diagnosis and to assess the largest pulmonary artery in which embolus was visible. There was a strong correlation between plasma DD concentration and embolus location (Kruskal–Wallis, P < 0.001), and the assay showed greater accuracy in excluding segmental or larger emboli (sensitivity = 93%) than subsegmental emboli (sensitivity = 50%) [93]. In another study, Galle et al. [94] suggested that DD levels above 4000 μg L−1 were associated with perfusion defects of more than 50% on ventilation-perfusion lung scan. The situation is identical in the presence of suspected DVT. In a study where the diagnostic performances of the LIA test Stago and of a quantitative microlatex test measurement were compared with contrast venography, the two tests had a diminished sensitivity for distal DVT. In that study, the sensitivity and NPV were 98.5% and 95.6%, respectively, for diagnosing proximal DVT, but only 83.8% and 84.6%, respectively, for diagnosing distal DVT [95]. However, despite this limited sensitivity for calf DVT, many outcome studies using ELISA DD tests confirmed the safety of not treating patients with suspected DVT and normal DD levels [22,66].

Patients with a prolonged duration of symptoms

There is an inverse relation between DD plasma level and duration of symptoms. DD concentration tends to decrease when the patient has been presenting symptoms for several days before testing. In a study published more than 10 years ago, dealing with suspected DVT, the time elapsed from the onset of symptoms was negatively associated with DD levels both in patients with and in those without DVT. In these patients, the DD values already reached 25% of the initial value after 1–2 weeks [14].

Patients already on anticoagulant treatment

Anticoagulant therapy, both with heparin and with vitamin K antagonists, reduces the formation and deposition of fibrin, and thus decreases DD levels. Some authors calculated on the basis of a literature review that the mean DD concentration decreases by 25% 24 h after starting heparin therapy in a patient with acute VTE [96]. This 25% decrease resulted in a diminution in sensitivity from 95.5% (95% CI, 90–99%) to 89.4 (95% CI, 84–95%). Therefore, DD measurement below the diagnostic cut-off performed after starting anticoagulation should be interpreted with caution, as a higher frequency of false-negative results must be anticipated under these circumstances [96,97].

D-dimer as a prognostic marker in patients with VTE

As previously mentioned, DD levels parallel the extent and burden of thromboembolic disease. It has been suggested, though not fully established, that risk assessment may be important for selecting the appropriate management in patients with PE. In particular, selected patients with a low risk of recurrence or of unfavourable outcome could be treated as outpatients, while patients with poorer prognosis might benefit from more aggressive treatments (surgical or endovascular embolectomy, or thrombolysis). Risk stratification may be based on clinical prediction rules [98,99], biomarkers or right ventricular imaging by echocardiography or CT [100–103]. DD has also been associated with mortality after symptomatic PE. In one study, patients who died had higher median DD levels than patients who survived (4578 vs. 2946 μg L−1; P = 0.005) [104]. When DD concentrations were classified in quartiles, mortality increased with increasing DD classes, rising from 1.1% in the first quartile (<1500 μg L−1) to 9.1% in the fourth quartile (>5500 μg L−1). Sensitivity and NPV for mortality of a DD level less than 1500 μg L−1 were 95% (95% CI, 74–100%) and 99% (95% CI, 94–100%), respectively. Thus, DD, possibly in combination with other prognostic markers, might play a role in identifying low-risk patients who could be treated on an outpatient basis, but this remains to be substantiated in further studies [104].


Twenty years after its first use in the diagnostic workup of suspected VTE, DD testing has gained wide acceptance for ruling out the disease, at least in the outpatient population referred to the emergency department. Several accuracy and outcome studies have demonstrated that ELISA DD assays and automated latex turbidimetric tests are associated with the highest sensitivity and with virtually no interobserver variability. Manual latex and semiquantitative whole blood tests are less sensitive but have the advantages of a quick bedside realization and being less labour intensive. Some ELISA DD assays have been shown to reliably rule out PE and/or DVT, irrespective of clinical assessment. However, the theoretical concern of a lower NPV in patients with a high clinical probability and hence a higher incidence of the disease, explains why recent guidelines suggest that these tests should be used to rule out VTE only in non-high clinical probability patients. Less sensitive tests can also be used, but they should be restricted to ruling out VTE in patients with a low clinical probability, and even so only provided the true incidence of VTE in that clinical probability category does not exceed 5%.

The main limitation of DD testing for diagnostic purposes is a reduced usefulness in specific patient categories or clinical settings. Indeed, specificity and hence clinical usefulness diminishes in elderly patients, in cancer patients, in patient with previous VTE, and in pregnant and postpartum women, although the test retains its high sensitivity in those situations.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.