INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: pulmonary embolism; fibrin degradation products; diagnosis

The clinical diagnosis of pulmonary embolism (PE) is neither sensitive nor specific, and clinical signs or symptoms can only suggest the diagnosis. Therefore, objective diagnostic tests are necessary to establish the diagnosis of PE (1). Ideally, such a test would be able to rule in or rule out the diagnosis of PE with a high degree of certainty and would be rapid and widely available (5). This is clinically relevant because patients with clinically suspected PE are often being evaluated at the first aid department and a normal test result would imply that a patient is sent home without anticoagulant treatment.

A suitable noninvasive test is the measurement of the plasma concentration of D-dimer, a specific degradation product of cross-linked fibrin that is released when the endogenous fibrinolytic system attacks the fibrin matrix of thromboemboli. The plasma level of D-dimer is almost always increased in patients with acute PE or deep venous thrombosis (DVT). Because raised concentrations of D-dimer can be found in many conditions, a positive value should be interpreted in conjunction with more specific tests for PE. It has been suggested that a normal level of D-dimer measured with an enzyme-linked immunosorbent assay (ELISA) may accurately exclude PE (6, 7).

In a recent study by Perrier and coworkers, a noninvasive diagnostic strategy was evaluated in patients with suspected PE and DVT. In their study, 159 patients with suspected PE presenting at the emergency ward were left untreated on the basis of a normal D-dimer assay result, without any recurrent venous thromboembolism during 3-mo follow-up (8). However, despite this evidence of the safety of withholding anticoagulant treatment in patients with suspected PE and a normal D-dimer assay result, clinicians remain reluctant to use a D-dimer assay as the first and sole diagnostic test.

In view of these considerations, we postulate that clinicians require further understanding on which emboli are accurately detected by D-dimer testing only. A recent retrospective study by Sijens and coworkers in selected patients with proven PE suggested a relation between thrombus load and D-dimer assay accuracy and found the accuracy in the detection of PE in subsegmental arteries to be significantly lower (9). The aim of our study was to prospectively assess the relation between concentrations of D-dimer and the location of PE and to evaluate its influence on the diagnostic performance of the D-dimer assay in consecutive patients with suspected PE.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

This study was conducted from June 1997 through March 1998 as part of a large prospective multicenter study in six Dutch hospitals. Consecutive inpatients and outpatients with clinically suspected PE were eligible. Patients who had already undergone objective diagnostic examinations for their current symptoms were excluded, as were patients in whom the diagnostic workup could not be initiated within 24 h. Further exclusion criteria were pregnancy, age less than 18 yr, and the immediate need for thrombolytic therapy. The study was approved by the institutional review boards of all participating centers and informed consent was obtained from all participating patients.

Plasma Collection and D-dimer Assays

Before or within 24 h after the start of heparin therapy, venous blood was drawn using standard 4.5-ml citrate Vacutainer tubes (Becton Dickinson, Franklin Lakes, NJ). These were centrifugated at 4° C for 15 min at 2,500 g. Plasma was then aliquoted into 2-ml tubes, snap-frozen, and stored at 80° C. Quantitative D-dimer measurements were performed on samples, which had been thawed only once, using the Tinaquant assay (Roche Diagnostica, Mannheim, Germany) according to the manufacturer's instructions. Tinaquant D-dimer, an immunoturbidimetric assay for quantitative in vitro determination of fibrin degradation products (D-dimer and X-oligomers) was carried out on a Hitachi system (Hitachi Ltd., Tokyo, Japan). The manufacturer's advised cutoff value for the Tinaquant assay is 0.5 µg/ml. Technicians responsible for performing the assay were not aware of patient identity and diagnostic test results. Conversely, the D-dimer measurements were not made known to the interpreters of the diagnostic tests in this study.

Diagnostic Workup for PE

All patients underwent lung perfusion scintigraphy and extensive bilateral B-mode compression ultrasonography of the leg veins. A normal perfusion scintigram excluded PE, and no further examinations were performed. Ventilation scintigraphy and a spiral computed tomographic (CT) scan were performed after an abnormal perfusion result. Ventilation-perfusion (/) results were classified either as high probability for pulmonary embolism (defined as one or more segmental perfusion defect with locally normal ventilation) or as nondiagnostic (10). Pulmonary angiography was performed in patients with a nondiagnostic / scan and in patients with a high-probability / scan and a discordant normal spiral CT scan. The maximal time span between examinations was 24 h. A panel of independent experts interpreted all diagnostic tests. The final diagnosis of PE was thus established by a high-probability / scan with a concurrent abnormal spiral CT scan or by an abnormal pulmonary angiogram. PE was excluded on the basis of a normal perfusion scan or a normal pulmonary angiogram.

To ascertain the largest involved branch of the pulmonary artery, three experienced radiologists reviewed the pulmonary angiograms and the spiral CT scans of patients with the final diagnosis PE. Patients were categorized according to the largest pulmonary vessel in which PE was visible: central (main pulmonary trunk, left or right pulmonary artery, lobar artery), segmental, or subsegmental artery.

Lung perfusion scintigraphy was performed using technetium-99m macro-aggregated albumin particles, and images were obtained from six directions. Ventilation scintigraphy was performed using krypton-81m gas. Pulmonary angiography was performed using a digital subtraction technique, with a catheter positioned selectively in the left and right pulmonary artery. Spiral CT angiography was performed during a 32-s single breath hold. If patients were very dyspneic, scanning was performed during shallow breathing.

Statistical Analysis

The correlation between D-dimer concentration and embolus location was tested using the nonparametric Kruskal-Wallis test. Receiver operator characteristics (ROC) analysis and the area under the ROC curve (AUC) were constructed to assess the influence of embolus location on the accuracy of the assay for PE irrespective of the manufacturer-supplied cutoff value. An AUC of 1 would signify a perfect test, and an AUC of 0.5 would be found if the test results would be completely random. The sensitivity and specificity of the D-dimer assay were calculated at the cutoff value advised by the manufacturer.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

During the course of this study, 440 eligible patients gave informed consent and were included in the study. In 70 of the 440 patients, no final diagnosis regarding the presence or absence of PE could be made according to our strict study criteria (Table 1). Plasma was obtained within 24 h and processed according to protocol in 314 of the remaining 370 patients.

The mean age of the studied population was 50 yr (SD 18 yr), and 188 (60%) of the 314 patients were female. Of the 314 patients, 251 (80%) were initially seen on an outpatient basis; the remaining 63 (20%) developed symptoms of PE during hospital stay for other reasons. The clinical characteristics of the included and excluded patients were comparable (data not shown). The prevalence of PE was 32%.

Accuracy of D-dimer Assay

Using the manufacturer's advised cutoff value of 0.5 µg/ml, the overall sensitivity of the assay was 81% and the specificity was 63% (Table 2). Thus, the assay returned a false-negative result in 19 PE-positive patients. B-mode compression ultrasound of the leg veins was performed in all of these patients and revealed no DVT. AUC, a measure of the accuracy irrespective of the cutoff value, was 0.78 (95% confidence interval [CI] 0.73 to 0.84, Figure 1).

View larger version (13K): [in this window] [in a new window]   Figure 1.   ROC curve of the performance of D-dimer assay in all 314 patients with suspected pulmonary embolism.

Levels of D-dimer and Extent of PE

Figure 2 depicts the D-dimer concentrations measured in the subgroups categorized according to the largest involved pulmonary artery branch. The concentration of plasma D-dimer was shown to be dependent on the location of the largest embolus visible (p < 0.001), being highest in patients with emboli in the pulmonary trunk and lowest in the subsegmental group.

View larger version (20K): [in this window] [in a new window]   Figure 2.   Box and whisker plot of D-dimer concentration depicting quartiles, stratified according to largest involved pulmonary artery.

Influence of Embolus Location on Assay Performance

The influence of embolus location on the accuracy of the D-dimer assays is shown in Figures 3A and 3B. These figures show the ROC curve of the D-dimer assay in patients with PE in segmental or larger arteries (AUC = 0.86, 95% CI 0.80 to 0.91) and the curve in patients with PE in a subsegmental artery (AUC = 0.59, 95% CI 0.44 to 0.73). Using the manufacturer's advised cutoff value of 0.5 µg/ml, the corresponding sensitivity for segmental and larger PE was 93% (95% CI 90 to 96%). In contrast, the sensitivity for subsegmental PE was 50% (95% CI 44 to 56%).

View larger version (14K): [in this window] [in a new window]   View larger version (14K): [in this window] [in a new window]   Figure 3.   ROC curves of the performance of the assay for (A) segmental and larger emboli and (B) subsegmental emboli.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main finding of our study is that there is indeed a strong correlation between plasma D-dimer concentration and embolus location. Furthermore, our data indicate that the accuracy of D-dimer measurement to exclude PE depends strongly on embolus location. At the manufacturer's advised cutoff value, sensitivity was raised from 81% overall, to 93% for segmental and larger PE. The accuracy for smaller subsegmental emboli, however, was shown to be only 50%. In our series, 20% of all patients had emboli confined to subsegmental arteries only, which is well in the range found in previous studies (6%, 20%, and 30%) (11). None of the 19 PE-positive patients with a negative D-dimer assay had concomitant DVT as assessed with bilateral B-mode compression ultrasonography.

This study provides important knowledge about the consequence of a negative D-dimer result when used as a first and sole test in patients with suspected PE. A negative D-dimer result rules out virtually all large PE and half of subsegmental PE. Although no studies are available in which patients with proven (subsegmental) PE are left untreated, indirect evidence suggests that this might indeed be safe (9, 14, 15). It has been suggested that patients with smaller emboli and good cardiorespiratory reserve could suffice with follow-up, using screening tests for DVT as a possible source of new emboli. Hull and coworkers obtained 6-mo follow-up in 627 patients with a nondiagnostic lung scan (14). These patients were serially tested for the presence of DVT and, if none was detected, were not treated with anticoagulants. Venous thromboembolism during follow-up occurred in only 12 patients (1.9%). Similarly, in a study by Wells and coworkers 665 patients with a nondiagnostic lung scan and no DVT upon serial testing with ultrasonography were left untreated (15). Only three patients (0.5%) had recurrent venous thromboembolism during 3 mo follow-up.

Of note, in the latter two studies a nondiagnostic lung scan could easily be obtained in the presence of larger than subsegmental PE (11, 12), and in these studies only testing for the presence of DVT in the leg veins as a possible source of recurrent PE was performed. In contrast, a negative D-dimer assay result, using a carefully chosen cutoff value, could potentially exclude DVT not only from the leg veins, but also from elsewhere in the body and would thus be a more extensive test than ultrasound of the leg veins.

Although D-dimer testing does not detect all PE, the missed emboli are small and their clinical relevance is unclear. The accuracy of other diagnostic modalities with respect to these small emboli is similarly unclear. Ventilation-perfusion scintigraphy requires at least a segmental mismatch for it to be "high probability" for PE. Smaller defects need further investigation with other techniques and only a normal perfusion scintigram rules out PE (10). Spiral computed tomography of the pulmonary arteries, a highly propagated new test in the diagnosis of PE, is less accurate at identifying subsegmental emboli as well (16). Finally, the interobserver variability of pulmonary angiographythe gold standard diagnostic modality for PEis much higher for these small subsegmental emboli than it is for larger PE (19, 20).

A few methodological points should be considered. The rigid protocol and the later independent blinded adjudication of the reference tests reduced the number of patients in which a final diagnosis could be established. However, it is unlikely that this has biased our study because this did not lead to differences in clinical characteristics between the included and the excluded patients (data not shown) and the prevalence of PE is similar to that reported in previous studies. Furthermore, embolus location as measured by pulmonary artery size is not similar to clot burden. However, an estimate of clot burden is difficult and prone to a high observer variability. On the other hand, the vascular anatomy of the lungs is clearly depicted by both pulmonary angiography and spiral CT and will give more reproducible results.

In conclusion, D-dimer plasma concentration is clearly correlated with the location of PE. Thus, the accuracy of D-dimer measurement in patients with suspected PE depends on the embolus present in the largest pulmonary artery. Although highly accurate for segmental and larger PE, D-dimer measurement can miss subsegmental PE. It is however uncertain whether these small emboli really need anticoagulant treatment. Our data could support physicians to take management decisions in patients who present at the emergency clinic with clinically suspected pulmonary embolism and have a normal D-dimer blood test.