Clinical Utility of a Rapid Whole-Blood d-Dimer Assay in Patients with Cancer Who Present with Suspected Acute Deep Venous Thrombosis

Venous thromboembolism is a common complication that contributes to morbidity and mortality in patients with cancer (1, 2). An accurate and efficient diagnostic approach for confirming or excluding venous thrombosis in this subgroup is needed because it represents approximately 20% of all patients in whom deep venous thrombosis or pulmonary embolism is diagnosed (3-5). Noninvasive studies are the diagnostic tools of choice for initial screening for deep venous thrombosis because clinical diagnosis of this condition is inaccurate and because contrast venography, the reference standard, has associated morbidity and is costly. Although studies evaluating the accuracy of noninvasive diagnostic methods have included patients with cancer, it is not known whether these tests are as accurate in this subgroup (6, 7). Both the cancer and its treatments may reduce the accuracy of these tests. For example, extrinsic venous compression by tumor mass may cause false-positive results on impedance plethysmography and compression ultrasonography. In support of these concepts, the one study evaluating diagnostic tests for deep venous thrombosis in patients with cancer demonstrated that impedance plethysmography was less sensitive and specific in these patients than in patients without cancer (8). One of the most promising noninvasive methods for the diagnosis of deep venous thrombosis is d-dimer testing. d-dimer is a plasmin-derived degradation product of cross-linked fibrin whose levels are elevated in the plasma of patients with acute thrombosis and other prothrombotic or inflammatory conditions (9-11). Numerous commercial d-dimer assays are available, and their utility for excluding venous thromboembolism has been established (12-16). The SimpliRED assay (Agen Biomedical, Ltd., Brisbane, Australia) is a rapid d-dimer test that can be performed at the bedside by using 10 L of whole blood obtained from a capillary or venipuncture sample. The test reagent contains a bi-specific antibody formed by conjugating a high-affinity monoclonal antibody against d-dimer with an erythrocyte-binding antibody (17). In the presence of elevated d-dimer levels, erythrocyte agglutination is visible within 2 minutes, indicating a positive test result. In three previous studies, we evaluated the performance of the SimpliRED assay in outpatients presenting with clinically suspected deep venous thrombosis (15, 18, 19). The first study, which used contrast venography as the reference standard to diagnose deep venous thrombosis, demonstrated that the sensitivity, specificity, positive predictive value, and negative predictive value of this assay were 89%, 77%, 56%, and 95%, respectively (18). Other investigators have reported similarly high negative predictive values for this assay (20-22). In our subsequent two studies, d-dimer test results were used to make management decisions in patients with suspected deep venous thrombosis. These studies demonstrated the safety of withholding anticoagulants in patients who had a negative d-dimer test result and a low clinical probability of deep venous thrombosis or a negative d-dimer test result with a normal impedance plethysmogram (15, 19). However, these findings may not apply to patients with cancer for two reasons. First, because patients with cancer can have elevated plasma d-dimer levels in the absence of thrombosis, the specificity of the test may be lower (9-11). Second, because the prevalence of thrombosis is higher in patients with cancer than in those without cancer (1, 2), the predictive values of d-dimer testing may differ in the two patient populations (23, 24). To determine the accuracy and utility of the SimpliRED whole-blood d-dimer assay in patients with cancer, we performed a retrospective analysis of our three studies to compare the performance of this assay in patients with and those without cancer. Methods Patients Patients included in this retrospective analysis participated in three prospective studies conducted between 1992 and 1997 in the outpatient thrombosis units of two tertiary care hospitals in Hamilton, Ontario, Canada (15, 18, 19). In these studies, the accuracy and utility of the SimpliRED assay, in combination with impedance plethysmography or clinical assessment, were evaluated in consecutive outpatients with suspected deep venous thrombosis. These patients were referred directly from local general practitioners or oncologists from the Hamilton Regional Cancer Centre. All patients with a clinically suspected first episode of deep venous thrombosis were eligible. The number of patients, the exclusion criteria, and the reference diagnostic strategies (including venography, serial noninvasive studies, and clinical outcome) in these three studies are listed in Table 1. Table 1. Summary of Three Prospective Studies Evaluating the Performance of the SimpliREDd-dimer Assay in the Diagnosis of Deep Venous Thrombosis d-dimer testing was performed in all eligible, consenting patients at presentation. In the first study, all patients underwent impedance plethysmography and contrast venography (18); in the two subsequent studies, patients were managed on the basis of a priori diagnostic algorithms designed to assess the safety of withholding anticoagulants in patients with negative results on screening tests (18, 19). In these management studies, patients with concordant negative results (negative d-dimer test results and normal impedance plethysmograms or negative d-dimer test results and low clinical pretest probability of deep venous thrombosis) underwent no further testing. Anticoagulants were withheld, and these patients were followed prospectively for 3 months for development of venous thromboembolism. Patients with other combinations of initial test results (discordant results or concordant positive results) and those presenting with symptoms of thromboembolism during follow-up underwent compression ultrasonography, venography, or both to establish a definitive diagnosis. Patients were categorized as positive for thrombosis if deep venous thrombosis was confirmed by objective testing at presentation or during follow-up. All other patients were classified as negative for thrombosis. In all three studies, cancer status was recorded at enrollment and before diagnostic testing. Patients were considered to have active cancer if they had received a diagnosis of cancer, or had received treatment for cancer, in the 6 months before study enrollment; were receiving active or adjuvant treatment for cancer; or had recurrent or metastatic disease at the time of presentation. Patients in whom cancer was diagnosed after study enrollment were not considered to have active cancer at presentation. Patient Data Criteria To minimize bias, a priori criteria were used for data retrieval and analysis. Only data from patients whose initial d-dimer test result and final deep venous thrombosis status were recorded in the original databases were included. Patients whose cancer status was not confirmed on chart review were excluded from this analysis. To confirm cancer status and to obtain more information on cancer history, patient charts from the Hamilton Regional Cancer Centre and the thrombosis units were reviewed. A data collection sheet was developed a priori to facilitate and standardize information retrieval. Information collected included the type and extent of cancer at study enrollment, the time interval between cancer diagnosis and presentation with suspected deep venous thrombosis, and whether the patient was receiving cancer treatment (chemotherapy, hormonal therapy, radiation therapy, or combination therapy). To simplify information about the extent of cancer, three categories were used: localized or early disease, recurrent or metastatic disease, and other (for hematologic and central nervous system disease). Charts were reviewed independently by two reviewers who were blinded to d-dimer test results and final deep venous thrombosis status. Statistical Analysis Patients were analyzed according to their cancer status at the time of study enrollment. Patients in whom cancer was diagnosed after enrollment were considered patients without cancer for the purposes of analysis. To compare the performance of the d-dimer assay in patients with cancer and those without cancer, the sensitivity, specificity, positive predictive value, and negative predictive value were determined separately in the two patient groups (23, 24). Similarly, the likelihood ratios for a positive test result and a negative test result were calculated (23, 25). To combine the study-specific results overall and within the target groups, a simple summation across studies was used to compute the diagnostic indices. This provides the mean of the study-specific and group-specific estimates weighted by the number of patients in each of the studies or groups. For each index, 95% CIs were computed by using the exact binomial distribution for the proportional-type measures and the normal approximation based on the logarithm of the ratio of two proportions for the likelihood ratio measures. The formulas used to calculate the diagnostic indices are shown in Figure 1. Figure 1. Formulas for sensitivity, specificity, predictive values, and likelihood ratios. By using simple summation values, the relative risk for deep venous thrombosis in the presence of cancer was calculated as the ratio of the prevalence of deep venous thrombosis in patients with cancer to the prevalence in patients without cancer. The 95% CI was calculated by using the normal approximation based on the logarithm of the ratio of two proportions. The meta-analytic method used to compare the cancer and noncancer groups differed according to the type of index. For the five proportion measures (prevalence, sensitivity, specificity, positive predictive value, and negative predictive value), the differences between the proportions for each study were weighted by the reciprocal