Introduction	Choose Top of pageABSTRACTIntroduction <<Subjects and MethodsResultsDiscussionReferencesCITING ARTICLES

Acute pulmonary embolism (PE) is a serious condition that affects approximately 69 persons per 100,000 individuals annually [1]. The approach to diagnosis of acute PE has changed dramatically [2]. With advances in imaging technology, the diagnosis of acute PE has become faster, more efficient, and more accurate. A 2005 systematic review [3–5] of a large number of studies showed that the clinical value of pulmonary CT angiography (CTA) is similar to that of the reference standard diagnostic test, conventional pulmonary angiography. In addition, normal findings at pulmonary CTA eliminate the need for anticoagulant therapy [3, 6]. In studies of the clinical use of pulmonary CTA in the evaluation of acute PE, some investigators [4, 7] have suggested that further research be directed at reducing overuse of pulmonary CTA owing to a high percentage of normal findings. In some studies, this percentage has been greater than 90% [8, 9]. The reasons for avoiding overuse of pulmonary CTA include high health care costs and unnecessary radiation exposure [3].

Use of a d-dimer assay is one possible means by which to reduce reliance on pulmonary CTA. d-dimers are cross-linked degradation products of fibrin, and their presence in human plasma is an indicator of fibrinolytic activity [10], which is a sign of acute thromboembolic disease [11]. d-dimer assays are highly sensitive and have a low rate of false-negative results. The assays have been found safe and effective when the clinical probability of PE is low [12]. When the d-dimer result is negative in patients with low clinical probability of having PE, the presence of acute PE can be safely ruled out without diagnostic imaging [13–15]. In one study [8], investigators observed patients with negative d-dimer results for 3 months and identified no false-negative results of the assay. A meta-analysis [16] of the cases of more than 6,800 patients showed that withholding anticoagulation is safe for patients who have low clinical probability of PE and a negative result of a qualitative or a quantitative d-dimer test.

Despite reports showing the high sensitivity and negative predictive value (NPV) of d-dimer testing, some clinicians have been reluctant to implement d-dimer testing into standard practice [5, 17, 18]. This reluctance may be based on results of studies [19–21] in which d-dimer assay was used as a standalone test and the belief that previous d-dimer assays, including the qualitative latex agglutination d-dimer test and bedside assays, lacked high enough sensitivity to exclude a condition as serious and potentially lethal as acute PE. Improved qualitative point-of-care d-dimer assays (e.g., Simplify d-dimer, Agen Biomedical) have become available. Many institutions, including ours, use a newer quantitative automated immunoturbidimetric assay (Advanced d-dimer, Dade Behring). The purpose of this study was to determine the sensitivity and NPV of use of the automated immunoturbidimetric assay in combination with clinical risk algorithms. The immunoturbidimetric assay has been found to yield results similar to those of the rapid enzyme-linked immunosorbent assay (ELISA), the most sensitive of the d-dimer assays [9, 17, 22–25]. Although substantial research has been devoted to ELISA, our study is the most extensive to date to our knowledge, of the sensitivity and NPV of the automated quantitative immunoturbidimetric d-dimer assay in conjunction with a clinical risk algorithm based on the revised Geneva score.

Subjects and Methods	Choose Top of pageABSTRACTIntroductionSubjects and Methods <<ResultsDiscussionReferencesCITING ARTICLES

This prospective study was performed in a 550-bed community teaching hospital, which also serves as a level 1 trauma center. The institution has 36,000 emergency department visits and 18,000 admissions annually. Our institutional review board approved this HIPAA-compliant study and waived the requirement for informed consent. From April 1, 2007, to March 31, 2008, patients who arrived in the emergency department with clinically suspected PE underwent an evaluation with the following three components: a clinical risk algorithm consisting of calculation of a revised Geneva score (a tested tool for assessment of clinical risk) and categorization of probability of PE as low, intermediate, or high; quantitative immunoturbidimetric d-dimer assay; and pulmonary CTA. Patients consecutively arriving in the emergency department were eligible for the study if acute PE was suspected because the patient had acute onset of new or worsening dyspnea or chest pain without another obvious cause. Patients were excluded from the study if they had renal insufficiency, were pregnant, or chose not to undergo pulmonary CTA.

The revised Geneva score is a clinical risk calculation composed of eight entirely clinically based variables with points assigned (Table 1). Low clinical probability was defined as a revised Geneva score of 0–3 points; intermediate clinical probability, 4–10 points; and high clinical probability, 11 points or more. According to clinical probability, the prevalence of PE in the derivation and validation set of the original study [26] was as follows: low, 7.9–9.0%; intermediate, 27.5–28.5%; and high, 71.7–73.7%. The overall prevalence of PE in the derivation and validation set was 23.1–25.6%. This prediction score has been internally and externally validated and awaits testing for clinical usefulness in an outcome study [27].

View Larger Version

TABLE 1: Revised Geneva Score

The d-dimer assay used at our center (Advanced d-dimer, Dade Behring) is an automated latex-enhanced quantitative immunoturbidimetric assay. The assay is performed with a Sysmex CA-1500 instrument (Sysmex America). A previous study [23] showed that when this assay is performed with a CA-1500 instrument, the sensitivity is 95.5% and the specificity 47.8% at a cutoff value of 1.1 mg/L. The department of pathology at our institution, in conjunction with scientific personnel from the vendor of the assay using independent testing, established a value of 1.2 mg/L as the NPV cutoff for venous thromboembolism (VTE) and acute PE. Testing was performed in accordance with guidelines set forth by the College of American Pathologists.

Testing for establishment of the NPV cutoff was conducted with two groups of patients from the emergency department, none of whom was included in this study. The first group had confirmed VTE or PE, and the second group had normal results of diagnostic studies for VTE and PE. d-dimer assays were performed on all patients. Sensitivity, specificity, NPV, and positive predictive value were calculated, and receiver operating characteristics curves were generated across a range of d-dimer values from 0.8 mg/L to 2.5 mg/L. The NPV cutoff value was then selected to maximize sensitivity and specificity and NPV. The value selected on the basis of results of this testing was 1.2 mg/L. The value of 1.1 mg/L suggested in a previous study [23] was rejected because it yielded lower specificity than the value of 1.2 mg/L while providing equal sensitivity. At our institution, 1.2 mg/L is the cutoff that determines whether a d-dimer result is categorized as positive or negative. A positive value (> 1.2 mg/L) necessitates additional testing for VTE and PE. A negative value (< 1.2 mg/L) would not prompt additional testing, according to the department of pathology. For the purposes of this study and daily clinical practice at our institution, 1.2 mg/L was the d-dimer value used to categorize values as positive or negative.

Pulmonary CTA was performed with a 16-MDCT scanner (Aquilion S16, Toshiba Medical Systems). All scans were acquired at 1-mm section thickness. The patients received 100 mL of iopamidol (Isovue 370, Bracco) at a rate of 4 mL/s IV. Fifty milliliters of normal saline solution were flushed IV after contrast administration. Imaging was performed approximately 15–20 seconds after contrast administration, determined with a precise contrast tracking system (SureStart, Toshiba Medical Systems). All final readings of the pulmonary CTA scans were rendered by board-certified radiologists with 2–20 years of experience. The final pulmonary CTA report was used to determine positive and negative CTA results for PE. Descriptive statistics were used, and sensitivity, specificity, negative predictive value, and positive predictive value were calculated for the d-dimer assay on the basis of CTA results.

Results	Choose Top of pageABSTRACTIntroductionSubjects and MethodsResults <<DiscussionReferencesCITING ARTICLES

From April 1, 2007, through March 31, 2008, 745 patients with clinical evidence of acute PE were consecutively registered in the emergency department at our institution. In 118 cases, quantitative d-dimer assay was not performed as part of the emergency department diagnostic evaluation for acute PE, and these patients were excluded from the study. The 627 patients included in the study (414 women [66.0%], 213 men [34.0%]; mean age, 46.9 years; range, 15–94 years) underwent all three phases of evaluation: calculation of revised Geneva score, quantitative d-dimer assay, and pulmonary CTA. According to the revised Geneva score classification system, 281 of the 627 patients (44.8%) were categorized as having low clinical probability of PE, 330 patients (52.6%) as having intermediate probability of PE, and 16 patients (2.6%) as having high clinical probability of PE. Twenty-eight patients (4.5%) had CTA findings of PE (six in the low, 17 in the intermediate, and five in the high clinical probability group).

The sensitivity, NPV, and specificity of the d-dimer assay in the clinical probability groups were 100%, 100%, and 25% for low clinical probability; 100%, 100%, and 33% for intermediate clinical probability; and 80%, 80%, and 37% for high clinical probability (Table 2). The mean d-dimer value in patients with PE was 8.46 mg/L (range, 1.17–25.25 mg/L; median, 6.10 mg/L). One patient was given the diagnosis of acute PE with a d-dimer level less than 1.2 mg/L. This patient had systemic lupus erythematosus and was categorized as having high clinical probability for PE on the basis of the revised Geneva score. No patients in the low or intermediate clinical probability groups who had d-dimer values less than 1.2 mg/L had pulmonary CTA findings of PE.

View Larger Version

TABLE 2: d-Dimer Values and Pulmonary CT Angiographic Findings

Discussion	Choose Top of pageABSTRACTIntroductionSubjects and MethodsResultsDiscussion <<ReferencesCITING ARTICLES

The question of overuse of pulmonary CTA in the diagnosis of acute PE is one that has been studied at many institutions [9, 17]. Overuse increases health care costs and the radiation dose received by patients and delays disposition of emergency department patients [3, 13, 17, 28]. Although few studies have directly linked ionizing radiation from CT to induction of cancer, some have advocated prudent use of CT and minimization of exposure on the basis of knowledge of the effects of radiation [7, 28]. In one study [28], researchers specifically investigated the calculated effective minimum dose to the breasts of women undergoing pulmonary CTA and found the dose to be greater than 2.0 rad (20 mGy), which is nearly seven times the dose administered to the breasts during standard two-view mammography.

Many of the decisions regarding pulmonary CTA in the diagnosis of acute PE are made with clinical risk stratification models or, simply, with the history and physical examination [9, 29]. History and physical examination findings are unreliable [22], particularly because of the variety of presentations of VTE [29]. Many investigators have studied the sensitivity and NPV of d-dimer assays in the evaluation of acute PE. Typically, a d-dimer assay is used to exclude acute PE in patients with low clinical probability of having PE [12, 13, 16, 29]. Questions have been raised about the safety and efficacy of the d-dimer assay as a first-line or standalone test in the evaluation of patients with intermediate clinical probability of having PE. Some of this uncertainty, coupled with a desire to simplify risk scoring systems, has led to development and use of modified clinical risk algorithms (including the modified Wells criteria) in which a two-level clinical risk system (likely or unlikely to have VTE) is used instead of a traditional three-tiered system (low, intermediate, or high clinical probability of VTE).

Many options exist for stratification of clinical risk of PE. Examples include the Wells criteria, modified Wells criteria, pulmonary embolism rule-out criteria, Geneva score, and revised Geneva score. We gave careful consideration to these options. The Wells criteria are simple to evaluate and a widely used and studied clinical risk algorithm. They are not, however, specifically tailored to evaluation for PE, are associated with substantial interobserver variability because one of the categories is “alternative diagnosis as likely as deep venous thrombosis,” and can be affected by the level of training of the person performing the evaluation [30]. The pulmonary embolism rule-out criteria have been found [31] to have high sensitivity in populations at low and very low clinical risk but have not been extensively studied for evaluation of patients with intermediate and high clinical probability of having PE. The pulmonary embolism rule-out criteria also have very low specificity. The Geneva score has the advantage of being based on clinical variables but requires chest radiography and arterial blood gas measurement, two tests that are not always performed in the standard evaluation for PE [32].

We chose the three-tiered revised Geneva score over the other clinical risk algorithms because only clinical variables are used and the score is independent of a physician's implicit judgment [26]. Two studies [33, 34] have shown the revised Geneva score comparable with the Wells criteria in categorization of clinical risk of PE. After we completed this study, another study [35] was performed in which a two-tiered version of the revised Geneva score was not found to have decreased diagnostic accuracy or clinical utility relative to the traditional three-tiered system.

As for the d-dimer assay used in our study, it is important to note that although the early d-dimer methods were plagued with poor results [36], newer quantitative d-dimer tests have been found to have high sensitivity and high NPV in the exclusion of acute PE [9, 17]. A meta-analysis [18] showed the rapid ELISA had better sensitivity and NPV than slide agglutination methods. Although rapid ELISA, the current reference standard, is available at some institutions, the limitations include test turnaround time, technologist time, testing complexity, requirements for specific instruments, and cost [9, 17].

In several studies [9, 17, 22–25], the quantitative immunoturbidimetric assay has been found to have sensitivity and NPV similar to those of the rapid ELISA, the most sensitive of the d-dimer assays. Gupta et al. (Gupta RTet al., presented at the 2006 annual meeting of the Radiological Society of North America, 2006) performed a study with more than 700 patients and found that the quantitative immunoturbidimetric d-dimer assay used in our study has a sensitivity and NPV of 100% at a NPV cutoff of 1.2 mg/L. Those investigators also evaluated NPV cutoffs ranging from 1.2 mg/L to 1.8 mg/L to determine the value that maximized sensitivity and NPV. In that study, the optimal NPV cutoff was determined to be 1.2 mg/L.

The results of our study show that the quantitative immunoturbidimetric d-dimer assay had a sensitivity and NPV of 100% in the low clinical risk group. If the d-dimer assay had been used as a first-line test in this group, as suggested by results of previous studies, 11% of patients (69/627) might have avoided pulmonary CTA. It is important to note, however, that the d-dimer assay also had a sensitivity and NPV of 100% in the intermediate clinical risk group. If the d-dimer assay had been used as a first-line test in this group and in the low clinical probability group, more than 27% of patients (172/627) might have avoided pulmonary CTA without any false-negative d-dimer results (Table 3). This finding is important because most patients fall into one of these categories when the revised Geneva score is used. In this study, fewer than 3% of patients from a typical emergency department patient sample were outside of one of these two categories. This distribution of patients is similar to the distribution of patients in the group studied by the initial revised Geneva score investigators [26].

View Larger Version

TABLE 3: Sensitivity and Negative Predictive Value of d-Dimer Assay in Relation to Clinical Risk and Percentage of Pulmonary CT Angiographic Examinations That Might Have Been Avoided

As for the high clinical probability group, on the basis of currently available data and the small sample size in this study, we do not recommend use of the d-dimer assay as a firstline test. The one patient of the 627 in this study who was diagnosed with acute PE with a d-dimer value less than 1.2 mg/L was in the high clinical risk category, having a revised Geneva score of 11 and systemic lupus erythematosus. As for a possible cause of the false-negative d-dimer result in this patient, recent data [37] have suggested that the d-dimer values of patients with systemic lupus erythematosus may not be significantly different from those of age-matched controls, despite higher levels of plasma inflammatory markers in the blood. Although we acknowledge the results of a study [35] showing similar diagnostic accuracy of the two-tiered and three-tiered revised Geneva scores, this patient's case illustrates one situation in which we believe that a three-tiered system is valuable. The conventional three-tiered system allows greater stratification of patients, and the patients at highest risk can bypass d-dimer testing and proceed directly to pulmonary CTA. This method ensures that a potentially false-negative d-dimer result does not dissuade clinicians from pursuing further diagnostic testing for PE.

Along the same lines, implementing the d-dimer assay into diagnostic algorithms for the diagnosis of acute PE requires appropriate use of the test on the basis of available data. The d-dimer assay should be performed not in isolation but in conjunction with clinical risk stratification of patients with suspected acute PE [38]. The d-dimer assay should not be performed unless a sufficiently high suspicion of acute PE is present to warrant pulmonary CTA or a ventilation–perfusion scan [17]. The d-dimer value can be elevated in many conditions other than VTE and acute PE. Because of the low specificity of the test, positive values are not helpful in ruling in acute PE. If a d-dimer assay is performed for patients in whom PE is not suspected on clinical grounds, elevated results can lead to unnecessary follow-up diagnostic testing.

Although the sensitivity and NPV of the assay used in this study are high, we acknowledge possible limitations. First, 118 patients had to be excluded from the study because a d-dimer assay was not performed. Even though the departments of radiology and emergency medicine collaborated on this project, physicians in the emergency department were not required to order d-dimer tests for patients with suspected acute PE. Rather, the d-dimer assay could be bypassed on the basis of clinical judgment, and patients could proceed directly to pulmonary CTA. One can assume that on the basis of previous findings, the patient population for whom d-dimer assays are not ordered is at high risk in the opinion of clinicians. As such, the rate of PE in this group would be expected to be higher than that in the group who did not undergo d-dimer assays. Of note, the prevalence of PE in the excluded group who went directly to pulmonary CTA was not significantly greater than that of the group included in the study (5.1% vs 4.5%).

A second possible limitation was the way in which the NPV cutoff was determined. This value was established through independent testing of samples from patients with known VTE and those without. These values are established and usually are institution-specific. As described earlier, some of the authors had conducted a retrospective study in which a variety of NPV cutoff values from 1.2 mg/L to 1.8 mg/L were analyzed in more than 700 patients (Gupta RT et al., 2006 RSNA meeting). Values less than 1.2 mg/L were not analyzed because sensitivity and NPV were 100% for this value, and values less than 1.2 mg/L would simply yield decreased specificity. Ideally, the NPV cutoff should yield a sensitivity and NPV of 100% while maximizing specificity [11]. Although a value of 1.2 mg/L is used at our institution, each institution using this assay undertakes similar testing and generation of curves to determine the value that best maximizes the sensitivity of this assay in the specific patient population. Having previously analyzed a range of NPV cutoff values, including those greater than are used at our institution, we believe that these results can be relevant and applicable to patients outside our institution.

Finally, the rate of positive findings of PE in our study sample was 4.5%. Because this value is lower than the rate of positive PE in some other similar studies, this finding suggests that our study might have been performed in a setting in which there is higher than normal use of pulmonary CTA and that there might have been concomitant overestimation of the number of pulmonary CTA examinations that could have been avoided in this sample. Although this supposition is theoretically true, three similar studies [39–41], also performed in urban emergency departments, had similar rates of positive findings of PE, ranging from 3.8% to 5.8%. Therefore, we believe that these results can be generalized to other similar populations.

We conclude that our data support the use of a quantitative immunoturbidimetric d-dimer as a first-line test in the evaluation of patients with intermediate or low clinical probability of having PE. Appropriate use of the quantitative immunoturbidimetric d-dimer assay can have a substantial influence in reducing health care costs and, more important, reducing unnecessary radiation dose to patients. In the same way that previous studies have led to guidelines on the appropriate use of ELISA d-dimer assays, the findings in this study in conjunction with those of ongoing similar studies should give guidance to those who use immunoturbidimetric d-dimer assays.