A Latex D-Dimer Reliably Excludes Venous Thromboembolism

Background  D-Dimer, a cross-linked fibrin degradation product, has a high sensitivity in patients with suspected venous thrombosis. Traditional latex D-dimer assays, however, have not been sufficiently sensitive to exclude venous thromboembolism.

Methods  To determine the clinical utility of a latex D-dimer assay (MDA D-Dimer; Organon Teknika Corporation, Durham, NC) in patients with suspected venous thromboembolism, we conducted a retrospective cohort study involving 595 unselected patients at 4 tertiary care hospitals. Patients had blood drawn for performance of the D-dimer assay and underwent objective testing for venous thromboembolism. Pretest probability was determined using validated models in 571 patients. Patients were classified as venous thromboembolism positive or negative according to results of objective tests and 3-month follow-up. The sensitivities, specificities, predictive values, and negative likelihood ratios of the assay were calculated for all patients and for subgroups of patients with known cancer or a low, moderate, or high pretest probability of venous thromboembolism.

Results  The prevalence of venous thromboembolism was 19.0% (113/595). Of those who had a pretest probability assessment, 35.9% had a low pretest probability, 49.7% a moderate pretest probability, and 14.4% a high pretest probability. Using a discriminant value of 0.50 µg fibrinogen equivalent units per milliliter, the assay showed an overall sensitivity of 96%, a negative predictive value of 98%, a specificity of 45%, and a negative likelihood ratio of 0.09. In patients with a low or moderate pretest probability, the sensitivity, negative predictive value, and negative likelihood ratio were 97%, 99%, and 0.07, respectively.

Conclusions  The MDA D-Dimer assay is the first latex agglutination assay with sufficient sensitivity to be clinically useful in the exclusion of venous thromboembolism. A negative result has the potential to be used as the sole test to exclude venous thromboembolism in patients with a low or moderate pretest probability of disease.

DEEP VEIN thrombosis (DVT) and pulmonary embolism, collectively termed venous thromboembolism, are common causes of morbidity and mortality.¹ It is important that venous thromboembolism be accurately diagnosed because if left untreated it can be fatal,¹ and treatment with anticoagulants may cause serious complications.¹ Clinical diagnosis alone of venous thromboembolism is inaccurate, and most patients with suspected venous thromboembolism do not have the disease.^2,3

Although objective tests have been developed for the diagnosis of venous thromboembolism, several problems remain. Many of these tests are not available during the night and weekends and, therefore, patients who present with suspected venous thromboembolism during these times are often treated with empiric anticoagulants until diagnostic testing can be performed. This frequently results in unnecessary exposure to anticoagulant therapy because, in most cases, venous thromboembolism is ruled out.^2,3 The reference standard tests, venography and pulmonary angiography, are expensive, invasive, and have associated morbidity.^1,2,4-7 Noninvasive tests for DVT, such as impedance plethysmography (IPG) and compression ultrasonography (CUS), while sensitive and specific for occlusive proximal DVT, have a low sensitivity for calf vein thrombosis, which make up approximately 15% of thrombi.^1,2,8-10 Therefore, if the initial results are normal, these tests must be performed serially to exclude proximal extension of isolated calf DVT, which occurs in 20% to 30% of patients with isolated calf thrombus and predisposes them to pulmonary embolism.^1,2,8-11 This is an inconvenient and costly strategy.

Although radionuclide ventilation-perfusion ([UNK]) lung scanning is useful in patients with clinically suspected pulmonary embolism; the results are inconclusive in 40% to 60% of patients.^12,13 The management of cases with such nondiagnostic scans is problematic because the prevalence of pulmonary embolism in this population is approximately 25%.^12-14 This necessitates the performance of pulmonary angiography for definitive diagnosis or investigation of the legs for DVT with contrast venography¹⁴ or serial IPG or CUS.¹⁵ The diagnosis of recurrent DVT also remains problematic because the results of venography,¹⁶ CUS,¹⁷ and less frequently, IPG¹⁸ may be abnormal as a result of the previous DVT, which makes the diagnosis of new thrombosis difficult.

D-Dimer, a specific cross-linked fibrin degradation product, has a high sensitivity and negative predictive value (NPV) in patients with suspected venous thromboembolism.^19-28 Three techniques are available to quantitate D-dimer.¹⁹ The clinical utility of the whole red-cell agglutination assay (SimpliRED; Agen Biomedical Ltd, Brisbane, Australia) has been the most extensively evaluated.^20-24 Although a number of studies suggest that this assay has a high sensitivity for venous thromboembolism,^20-23 not all centers have demonstrated similar results.²⁴ Moreover, a normal SimpliRED D-dimer result only reliably excludes DVT in patients with a low pretest probability or a normal noninvasive test,²² and is not useful in patients with cancer.²⁵ The time-consuming nature and low specificities¹⁹ of traditional enzyme-linked immunosorbant assays (ELISAs) have limited their use. Although more rapid ELISAs have been developed, the latter problem still remains.^26-28 Manual nonquantitative latex D-dimer assays are rapidly performed and their specificities have traditionally been higher than those of ELISA.¹⁹ In the past, however, these latex assays have not been sufficiently sensitive to exclude venous thromboembolism.¹⁹ The MDA D-Dimer (Organon Teknika Corporation, Durham NC) is a quantitative immunoassay that uses photo-optics to detect agglutination of latex microparticles as an index of binding of a specific monoclonal antibody to D-dimer and seems to be more sensitive than traditional visual latex assays. We therefore performed a retrospective cohort study to determine whether this D-dimer assay has a high sensitivity and NPV for venous thromboembolism, and if the specificity is sufficiently high to make the test clinically useful.

The study was performed between September 1997 and January 1999. Four university-affiliated tertiary care hospitals participated, 2 in Hamilton, Ontario (Henderson General Hospital and McMaster University Medical Centre, both of the Hamilton Health Sciences Corporation), and 2 in London, Ontario (University Hospital and Victoria Hospital, both of the London Health Sciences Centre). The study conformed to the guidelines set forth by each center's research ethics board, and all patients provided informed consent.

The study population consisted of unselected outpatients with clinically suspected DVT or pulmonary embolism referred to the thromboembolism service at the 2 Hamilton centers and the emergency department at the London centers. At the 2 Hamilton sites, more than 95% of outpatients with suspected venous thromboembolism were referred to the thromboembolism service for consultation.

Patients with suspected DVT or pulmonary embolism had their history taken and underwent physical examination. Prior to diagnostic testing, patients were assigned a pretest probability of venous thromboembolism using previously validated models that include an assessment of clinical symptoms and signs, risk factors for venous thromboembolism, and the presence of an alternative diagnosis.^29,30 Patients were evaluated with appropriate objective tests per the consulting physician. Ascending venography, IPG, or CUS were performed on all patients with suspected DVT, as previously described,^4,31,32 except for some of those with a low pretest probability and negative findings on the whole red-cell agglutination D-dimer assay. Patients with normal initial IPG or CUS results underwent serial testing approximately 1 week later, unless they had a negative red-cell agglutination D-dimer assay or a low pretest probability of venous thromboembolism. The safety of these latter approaches has been validated by previous clinical trials (personal communication, C. Kearon, November 1999).²²

Ventilation-perfusion lung scans were performed on all patients with suspected pulmonary embolism, as previously described,¹⁴ except for some who had a low pretest probability and negative findings on the red-cell agglutination D-dimer assay. A previous study²³ has demonstrated that the NPV of this combination of findings in patients with suspected pulmonary embolism is 99%. Patients with a nondiagnostic lung scan (segmental perfusion defects with matched ventilation defects, subsegmental perfusion defects with or without ventilation defects, or perfusion defects with corresponding abnormalities on chest radiograph) and normal initial CUS findings underwent serial CUS testing approximately 1 week later, unless they had negative results on the red-cell agglutination D-dimer assay. A previous study²³ found patients with the combination of a nondiagnostic [UNK] scan, negative results on a red-cell agglutination D-dimer assay, and normal findings on initial CUS to have an NPV of 99%. Patients with high or moderate pretest probability of pulmonary embolism and a high-probability [UNK] scan finding (segmental or greater perfusion defect with normal ventilation) were diagnosed with pulmonary embolism. Those with high-probability [UNK] scan results and a low pretest probability underwent either bilateral ascending venography or pulmonary angiography to confirm the diagnosis of pulmonary embolism; previous studies have shown that only about one half of such patients have pulmonary embolism.^12,13 Anticoagulant therapy was withheld in all patients with negative objective test results, and these patients were observed for 3 months for symptomatic venous thromboembolism. Patients in whom DVT or pulmonary embolism was diagnosed at presentation or during follow-up were treated with anticoagulants per local practice.

Patients were classified as venous thromboembolism positive or negative according to the results of objective testing and 3-month follow-up. Patients were considered DVT positive when 1 of the following occurred: (1) presence of an intraluminal filling defect evident in 2 or more views on ascending venography; (2) absence of compressibility of the common femoral vein and/or popliteal vein on CUS; or (3) symptomatic venous thromboembolic event verified by objective testing within 3 months of presentation with suspected DVT. Patients were considered DVT negative when 1 of the following results occurred along with an absence of symptomatic venous thromboembolism within 3 months of follow-up: (1) normal venography findings; (2) normal serial CUS results; (3) normal CUS or IPG results in the presence of a low pretest probability of DVT or negative red-cell agglutination D-dimer assay findings; (4) normal serial IPG results; or (5) low pretest probability and a negative D-dimer result.

Pulmonary embolism was diagnosed in the presence of 1 of the following: (1) a pulmonary angiogram with an intraluminal filling defect present on 2 or more views; (2) a high-probability [UNK] scan in patients with a moderate or high pretest probability of disease; (3) a nondiagnostic lung scan and either abnormal CUS findings or ascending venography results; (4) a high-probability [UNK] scan and abnormal ascending venography or CUS results in patients with a low pretest probability of pulmonary embolism; or (5) symptomatic venous thromboembolism verified by objective testing within 3 months of presentation with suspected pulmonary embolism. Pulmonary embolism was excluded if 1 of the following occurred along with an absence of symptomatic venous thromboembolism within 3 months of follow-up: (1) normal pulmonary angiography results; (2) normal perfusion lung scan findings; (3) a nondiagnostic lung scan without evidence of DVT; or (4) a low pretest probability and negative results on red-cell agglutination D-dimer assay. Patients were considered venous thromboembolism positive if DVT or pulmonary embolism was diagnosed according to the above criteria and venous thromboembolism negative if DVT and pulmonary embolism were excluded according to the above criteria.

At the time of referral, venous blood was collected in 5-mL Vacutainer tubes (BC Vacutainers; Becton Dickinson Co, Mountain View Calif) prefilled with 0.5 mL of 3.2% (0.105 mol/L) trisodium citrate-didihydrate. Specimens were centrifuged at 1700g for 15 minutes at room temperature. Platelet-poor plasma was then aliquotted into polystyrene tubes that were maintained at −70°C until assayed in batches. D-Dimer assays were performed according to manufacturer's instructions on the MDA 180 automated coagulometer (Organon Teknika Corporation) using a commercial kit (MDA D-Dimer).

The optimal discriminant value for the D-dimer assay was determined by receiver-operator curve analysis using the first 150 patients. A value of 0.50 µg fibrinogen equivalent units (FEU) per milliliter was chosen because it provided the highest sensitivity with a specificity approximating 50%. For the purposes of analysis, results were expressed as either negative (<0.50 µg FEU/mL) or positive (≥0.50 µg FEU/mL).

In the primary analysis, the accuracy indices (sensitivity, specificity, NPV, positive predictive value [PPV], and negative likelihood ratio) of the D-dimer assay were calculated for all patients with suspected venous thromboembolism. Because previous studies have shown that a number of D-dimer assays have similar sensitivity and specificity in patients with suspected DVT and in patients with suspected pulmonary embolism,³³ we reasoned that combining these patient populations was reasonable. In the secondary analysis, the same indices were calculated for the following subgroups: patients with suspected DVT; patients with suspected pulmonary embolism; patients with a high, moderate, or low pretest probability of disease; and patients with known cancer at the time of presentation. Where indicated, the corresponding 95% confidence intervals (CIs) for the accuracy indices were calculated according to the binomial distribution.

The technologists performing and interpreting the D-dimer assays were unaware of the results of the diagnostic tests for venous thromboembolism. Results of the D-dimer assays were not disclosed to clinicians caring for the patients, and the results were not used to make management decisions. Bias in the interpretation of IPG, CUS, ascending venography, lung scans, and pulmonary angiography was avoided by having these tests interpreted by physicians who were unaware of the results of the D-dimer assay.

Of 595 patients (352 women) enrolled in the study, 317 had suspected DVT and 278 had suspected pulmonary embolism. The prevalence of venous thromboembolism in the total study population was 19.0% (113/595). Sixty-five (20.5%) of the 317 patients with suspected DVT were classified as DVT-positive, while 48 (17.3%) of the 278 patients with suspected pulmonary embolism were classified as pulmonary embolism positive.

The distribution of D-dimer results is shown in Figure 1. The mean and median D-dimer levels were 6.81 µg FEU/mL and 2.28 µg FEU/mL, respectively, in those with confirmed venous thromboembolism and 1.14 µg FEU/mL and 0.54 µg FEU/mL, respectively, in those without venous thromboembolism. Duplicate precision for the patient samples for the D-dimer assay was calculated for all the samples and for those around the clinical discriminant value (0.40-0.60 µg FEU/mL). The standard deviation and coefficient of variation were 0.06 and 6.79%, respectively, for all samples and 0.02 and 3.98%, respectively, for those with results between 0.40 and 0.60 µg FEU/mL.

The sensitivity, specificity, NPV, and PPV of the D-dimer assay in the 150 patients used to determine the discriminant value were 97%, 51%, 99%, and 26%, respectively. Because these values were similar to those obtained in the remainder of the study population, these patients were included in the overall analysis.

The accuracy indices and corresponding 95% CIs of the D-dimer assay for the total study population are summarized in Table 1. The sensitivity and NPV are sufficiently high to reliably rule out venous thromboembolism. The corresponding values for the subgroups containing patients with suspected DVT and those with suspected pulmonary embolism are also given in Table 1. As expected, the results in these 2 patient populations are similar.

Of the 595 patients, 205 (34.5%) had a low pretest probability of venous thromboembolism, while 284 (47.7%) and 82 (13.8%) had a moderate or high pretest probability of venous thromboembolism, respectively. A priori determination of the pretest probability was not performed in 24 patients. The prevalence of venous thromboembolism was 5.9%, 16.9%, and 56.1% in the low, moderate, and high pretest probability populations, respectively. The accuracy indices and 95% CIs for the D-dimer in the various pretest probability subgroups are summarized in Table 2. While the sensitivity and NPV of this assay are high in patients with a low or moderate pretest probability of venous thromboembolism, they are lower in those with a high pretest probability of venous thromboembolism. When the low or moderate pretest probability categories are combined, the D-dimer has a sensitivity of 97% (95% CI, 89%-100%), a specificity of 46% (95% CI, 41%-51%), an NPV of 99% (95% CI, 96%-100%), and a PPV of 20% (95% CI, 16%-25%). The likelihood ratio of a negative test in this population is 0.07. Therefore, a D-dimer result of less than 0.50 µg FEU/mL essentially excludes venous thromboembolism in patients with a low or moderate pretest probability of venous thromboembolism; these patients made up 85.6% of the study population who had a pretest probability assessment.

Thirty patients with suspected DVT (14 with a low or moderate pretest probability of DVT) and 36 patients with suspected pulmonary embolism (28 with a low or moderate pretest probability of pulmonary embolism) were known to have active cancer (diagnosed within 6 months, receiving treatment at the time of the study or within the previous 6 months, or receiving palliative treatment only at the time of the study) at presentation. The prevalence of DVT in the first group was 53% (16/30), while the prevalence of pulmonary embolism in the second group was 36% (13/36). Overall, in those with cancer, the D-dimer assay had a sensitivity of 97% (95% CI, 82%-100%), a specificity of 46% (95% CI, 30%-63%), an NPV of 94% (95% CI, 73%-100%), and a PPV of 58% (95% CI, 43%-72%). The accuracy indices and 95% CIs for the assay in the various subgroups of patients with cancer are given in Table 3. The NPV and sensitivity of the assay remained high in patients with cancer and a low or moderate pretest probability of venous thromboembolism.

In total, 5 patients with suspected venous thromboembolism (2 with a moderate pretest probability and 3 with a high pretest probability) had false-negative D-dimer results when a discriminant value of 0.50 µg FEU/mL was used. D-Dimer levels in these patients ranged from 0.23 to 0.43 µg FEU/mL. The duration of patient symptoms was between 1 and 4 days, and prior treatment with anticoagulants (heparin for 12 hours) was confirmed in only 1 patient.

This study is one of the first to demonstrate that a D-dimer assay has the potential to be used as the sole diagnostic test to exclude venous thromboembolism. Although other D-dimer assays have proven useful, most reliably exclude venous thromboembolism only in certain patient subgroups; for example, a negative SimpliRED D-dimer result has been shown to exclude DVT only in those patients with a normal IPG finding or a low pretest probability²² and pulmonary embolism in the subgroup of patients with nondiagnostic lung scan and a normal CUS result, and in patients with a low probability of pulmonary embolism.²³ The Instant 1A D-Dimer (Diagnostica Stago, Snieres, France) has been demonstrated to exclude DVT only in the presence of a normal CUS finding.²⁸ When used with a discriminant value of 0.50 µg FEU/mL, the MDA D-Dimer assay, an automated test with a turnaround time of less than 30 minutes (including 15 minutes for plasma preparation), has a sensitivity of 96% and an NPV of 98% in patients with suspected venous thromboembolism. This NPV compares favorably with that of ascending venography,⁵ serial IPG,^9,10 and serial CUS,^33,34 3 widely accepted diagnostic strategies for DVT, as well as pulmonary angiography, the reference standard for pulmonary embolism,¹³ all of which approximate 98% to 99%. In our study, 37% of the population had a negative D-dimer result.

In patients with a low or moderate pretest probability of DVT or pulmonary embolism, the NPV of this assay is 99%. More than 80% of patients in our study had a low or moderate pretest probability of venous thromboembolism, and approximately 40% of these patients had a D-dimer result of less than 0.50 µg FEU/mL. The relatively low NPV (77%) of this assay in patients with a high pretest probability of venous thromboembolism is noteworthy. This observation is almost certainly the result of the high prevalence of venous thromboembolism in this subgroup (56.1%), as the sensitivity (93%) is not inconsistent with that in the other subgroups, and NPV is critically dependent on both sensitivity and prevalence. For example, in a population with a prevalence of venous thromboembolism of 50%, a D-dimer assay with a sensitivity of 98% and a specificity of 50% would have an NPV of only 96%. Conversely, an assay like the SimpliRED D-Dimer, which has a sensitivity of approximately 90% and a specificity of approximately 75%, would be expected to have an NPV of only 76% in a patient subgroup with a prevalence of venous thromboembolism of 50%. Therefore, in patients with a high pretest probability of venous thromboembolism, we do not recommend obviating further testing in those with a negative D-dimer result.

In support of this finding, a previous study has reported that the NPV of a whole-blood agglutination D-dimer test is significantly lower in patients with cancer than in those without cancer, again almost certainly because the prevalence of venous thromboembolism in patients with cancer is high (approximately 50%).²⁵ However, in a relatively small subgroup analysis of patients with known cancer at presentation, the NPV and sensitivity of the MDA D-Dimer assay remained high in patients with a low or moderate pretest probability of venous thromboembolism, albeit with wide 95% CIs. This observation should be verified in larger studies.

The results of our study are valid for this assay and cannot be extrapolated to other assay systems. The results obtained should be generalizable to other patient populations because unselected patients were evaluated and the prevalence of venous thromboembolism is consistent with that reported in other studies.^{9,20-23,27-30} Although the assays were performed on frozen rather than fresh samples, stable results have been obtained with this D-dimer assay in samples frozen for up to 2 months at −20°C or lower (data not shown). The potential for bias was eliminated by having objective tests interpreted by clinicians unaware of the D-dimer results and assays performed by technologists unaware of the clinical status of the patients. Although the reference standard tests of venography and pulmonary angiography were not performed in all patients with suspected DVT and pulmonary embolism, respectively, the classification of patients in this study as venous thromboembolism positive and venous thromboembolism negative was corroborated by long-term clinical outcome. This approach has successfully been used to validate the use of serial IPG^9,10 and CUS^23,24 in patients with suspected DVT, as well as in those with suspected pulmonary embolism and nondiagnostic [UNK] lung scans.¹⁵ However, it is still likely that a small proportion of patients who truly had calf DVT or small pulmonary emboli were misclassified as DVT negative and pulmonary embolism negative, respectively.

While the sensitivity and NPV of the MDA D-Dimer assay are similar to those of ELISA, the specificity of this latex assay seems somewhat higher. Based on these promising results, further clinical trials should be performed to determine the safety of withholding anticoagulant therapy in patients with a low or moderate pretest probability of venous thromboembolism and an MDA D-Dimer result of less than 0.50 µg FEU/mL. If this approach is safe, it would reduce health care costs by allowing many patients who present with suspected DVT or pulmonary embolism to be discharged without further expensive and invasive testing.

Accepted for publication August 31, 2000.

The MDA D-Dimer kits used in this study were supplied by Organon Teknika Corporation, Durham, NC. This study was funded by an unrestricted grant from Organon Teknika Corporation.

Shannon M. Bates, MD, CM, and Anne Grand'Maison, MD, are both recipients of a Research Fellowships from the Heart and Stroke Foundation of Ontario, Ottowa. Michael J. Kovacs, MD, is a University of Western Ontario (London) Department of Medicine Scholar. Jeffrey S. Ginsberg, MD, is a recipient of a Career Investigator Award from the Heart and Stroke Foundation of Ontario, Toronto.

The authors would like to thank Nicola Booker, RN; Jo-ann Bennett; Sue Smale; Pamela Stevens, RN; Jody Joval, RN; Joanne McGiniss, RN; Karen MacKinnon, MLT; and Andrea Willoughby, RN, for their help collecting plasma samples and clinical information. We thank Terri Finch, BA, and Donna McCarty, BA, as well for additional assistance in collecting clinical information.

Corresponding author: Shannon M. Bates, MD, CM, McMaster University Medical Centre, Thromboembolism Unit, HSC 3W15, 1200 Main St W, Hamilton, Ontario, Canada L8V 1C3 (e-mail: batesm@mcmaster.ca).