D-Dimer in Patients With Clinically Suspected Pulmonary Embolism*

Transcription

1 D-Dimer in Patients With Clinically Suspected Pulmonary Embolism* JeffreyS. Ginsberg, M.D. F.C.C.P.; Patrick A. Brill-Edwards, M.D., F.C.C.P. ; Christine Demers, M.D.; Dianne Donovan, R.N.; and Akbar Panju, M.D. Study objective: To determine whether measurement of D dimer, using an enzyme-linked immunosorbent assay (ELISA) with a cutoff of3 ng/ml, and a latex agglutination assay with a cutoff of 5 ng/ml, is clinically useful in patients with suspected pulmonary embolism (). Design: Prospective cohort. Setting: Tertiary care referral center, university-affiliated hospital. Patients: Two hundred twenty-one consecutive patients with clinically suspected. Intervention: All patients had blood drawn to measure levels of and underwent ventilation/perfusion (V/ Q) lung scanning and bilateral impedance plethysmography (lpg); pulmonary angiography was performed in nine patients. Patients were classified as follows: (1) -positive; positive pulmonary angiography or high probability V/Q scan or non-high-probability V/Q scan and either abnormal lpg (either at presentation or on serial testing and confirmed by contrast venography) or symptomatic thromboembolic event within 3 months of presentation; or (2) negative; normal V/Q scan or normal pulmonary angiography or non-high-probability V/Q scan and either normal serial lpg or abnormal lpg with normal venography and absence of symptomatic venous thromboembolism within 3 months of followup. Forty-three patients were classified as positive and 17 patients were classified as negative. Measurements and results: The sensitivities, specificities, positive predictive values, and negative predictive values of the ELISA and latex agglutination assay were calculated for all patients and for the subgroup of patients with non- Q bjective testing is necessary to diagnose pulmonary embolism () because clinical diagnosis is inaccurate. 1 2 Pulmonary angiography is the gold standard test for the diagnosis of, but it is invasive and cannot be performed in all patients with clinically suspected. Radionuclide ventilation and perfusion (V/Q) lung scanning is useful in patients with clinically suspected 3 but has limitations because the results are frequently inconclusive. Two separate studies, one by the McMaster group and one by the PIOD investigators, have demonstrated that 4 to 6 percent of patients with clinically suspected have nondiagnostic lung scans; these are classified as "non-high" by the McMaster group and as indeterminate/intermediate or low probability by the PIOD group.u *From the Department of Medicine, McMaster Universit}; Hamilton, Canada. Dr. Ginsberg is a recipient of a Research Scholarship from the Heart and Stroke Foundation of Canada. Manuscript received January 21 ; revision accepted April2i. high-probability V/Q scans. The ELISA, using a cutoff of 3 ng/ml, showed a sensitivity and negative predictive value of 1 percent in all patients and patients with non-high-probability V/Q scans, but the corresponding specificities were only 26 percent and 13 percent, respectively. The latex agglutination assay for using a cutoff of 5 ng/ml showed the following: sensitivities of 4 percent in all patients and 9 percent in patients with nonhigh-probability scans, negative predictive values of 93 percent in all patients, and 9 percent in patients with nonhigh-probability scans and specificities of 56 percent in all patients and 55 percent in patients with non-high-probability scans. Conclusions: This study demonstrates that an ELISA D dimer result of less than 3 ng/ml excludes but occurs in a small proportion of patients with clinically suspected. The latex agglutination assay, using a cutoff of 5 ng/ ml, has potential clinical utility in excluding in the subgroup of patients with clinically suspected and nonhigh-probability V/Q scans. However, the 95 percent confidence interval on the observed sensitivity of the latex agglutination assay in patients with non-high-probability V/ Q scans is wide. Therefore, these promising results should be confirmed in a large clinical trial before the latex agglutination assay is used to make management decisions. (Chest 1993; 14:1679-4) DVf=deep venous thrombosis; ELISA=enzyme-linked immunosorbent assay; lpg = impedance plethysmography; = pulmonary embolism; V/Q =ventilation/perfusion The treatment of patients with such lung scans is problematic because the prevalence of in these patients is 14 to 4 percent. 4 5 Options in patients with nondiagnostic scans include the following: (1) performing pulmonary angiography for definitive diagnosis; (2) making a management decision on clinical grounds, an approach that is known to be inaccurate; or (3) investigating the legs for deep venous thrombosis (DVf) with contrast venography, real time B-mode ultrasonography, or impedance plethysmography (lpg). These tests are useful in patients with clinically suspected because DVf is commonly present in patients with pp and, the detection of ovr by one of these tests in patients with clinically suspected provides reasonable grounds for anticoagulant therapy. All of these options have limitations and disadvantages. A fourth approach is the use of assays that detect D dimer, a specific fibrin degradation product, as an adjunct for the diagnosis of. Recently high levels of, using latex agglutination assays and en- CHEST I 14 I 6 I DECEMBER

2 zyme-linked immunosorbent assays (ELISA), have been reported in studies of patients with ovr and.6-15 These assays have been reported to have high negative predictive value for but most studies were small, there was potential for bias, and most studies did not evaluate patients with nondiagnostic lung scans. Therefore, the clinical utility of measuring in patients with clinically suspected and, in particular, in patients with nondiagnostic lung scans, is not known. To overcome these limitations, we have initiated a series of studies evaluating testing in patients with clinically suspected. 14 In the first study, an ELISA was used to measure. 14 The objective of the study was to determine whether a cutoff of the level of could be identified that had very high sensitivity and negative predictive value and reasonable specificity in patients with clinically suspected ; the results of that study generated the hypothesis that 3 nglml is such a cutoff. Subsequently, we have performed a second cohort study with three main objectives: (1) to validate our hypothesis that, using an ELISA technique to measure D dimer, 3 nglml is a clinically useful cutoff; (2) to determine whether a latex agglutination assay for D dimer, a test that is suitable for individual testing with a rapid turnaround time, is clinically useful; and (3) to determine the clinical utility of both assays in patients with nondiagnostic lung scans. In this study, consecutive patients with clinically suspected underwent objective testing for the diagnosis of and had blood taken at presentation for measurement of by ELISA and latex agglutination assay. lbtients METHODS Consecutive patients referred to the thromboembolism consultants at Chedoke-McMaster hospitals with clinically suspected between November 199 and June 1992 comprised the study population. Informed eonsent was obtained from all patients after the study had been explained. Clinical Intervention Patients with clinically suspected were seen by a consultant of the thromboembolism service and underwent history and physical examination. Ventilation and perfusion lung scanning were performed as previously described' within 24 h of presentation in all patients. The results of lung scanning were classified as normal, high probability (segmental or greater perfusion defects with normal ventilation), or non-high probability (segmental perfusion defects with matched ventilation defects, subsegm.ental perfusion defects with or without ventilation defects, or perfusion defects with mrresponding abnormalities on chest radiograph). Pulmonary angiography was performed in nine patients at the request of attending physicians. A pulmonary embolism was diagnosed on pulmonary angiography when a persistent intraluminal defect was seen in more than one view. Within 24 h of presentation, all patients underwent bilateral lpg that was performed and interpreted according to previously described methods. Patients with non-high-probability scans underwent serial lpg testing on days 2, 5, 7, 1, and 14 and 16 anticoagulant therapy was withheld provided the results remained normal. This approach has been validated by clinical trial." If the lpg became abnormal, venography was attempted and, if abnormal, the patient was treated with anti(.'(}agulants, whereas if venography was normal, anticoagulant therapy was withheld and the patient was followed up clinically. According to the results of the lung scanning, lpg, mntrast venography (when applicable), and pulmonary angiography (when applicable), patients were classified as follows: (1) -positive when one of the following occurred: (a) positive pulmonary anj.,'iography or (b) high-probability lung scan or (c) non-high-probability lung scan and either abnormal lpg (either at presentation or on serial testing and mnfirmed by venography) or symptomatic venous thromboembolic event, verified by objective testing, within 3 months of presentation; or (2) -negative when one of the following occurred: (a) normal lung scan or (b) normal pulmonary angiography or (c) non-high-probability lung scan and either normal serial lpg or abnormal lpg with normal venography and absence of symptomatic venous thromboembolism within 3 months offollowup. Patients were treated as follows: -positive patients were treated with full-dose intravenous heparin followed by long-term oral anticoagulant therapy whereas -negative patients were not treated with anticoagulants. In 33 patients, overnight heparin therapy was instituted at the time of presentation until objective testing could be performed. However, once the tests were performed, patients were classified as positive or negative as described above and the appropriate treatment was instituted. The subgroup of -negative patients with non-high-probability V/Q scans and either normal serial lpg or abnormal lpg with normal venography were not treated with anticoagulants and were followed up for 3 months for the presence or absence of venous thromboembolism by clinical evaluation and repeated lpg. Laboratory Intervention Within 24 h of presentation, blood was obtained from all patients by venipuncture in a vacuum tube containing.15m sodium citrate. Specimens were centrifuged at 1,7xg for 15 min to obtain platelet-poor plasma. The plasma was divided into aliquots and frozen at -7oc for batch assay. This was performed at a later date by a technologist blinded to the clinical status of the patient. assays were performed using commercially available ELISA kits (Stago, Asnieres, France) and latex agglutination kits (Organon Teknica, Scarborough, Canada). Cutoffs of 3 nglml for the ELISA and 5 nglml for the latex agglutination assay were used; these results are expressed in fibrinogen equivalent units. The cutoff of3 nglml for the ELISA was chosen because it showed high sensitivity for in our "hypothesis-generating" study." A different cutoff (5 mglml) was chosen for the latex agglutination assay because it is the lowest titer associated with a positive test. The results of the latex agglutination assay are based on serial plasma dilutions and are semiquantitative; results of less than 5 nglml are simply reported as "negative." For the primary analysis, results were expressed as either negative (<5 nglml) or positive ( ~ 5nglml). For the comparison of the two assays, positive latex agglutination assays were grouped as 5 to 1, nf!/ ml, 1, to 2, nglml, 2, to 4, nglml, and more than 4, nglml. Analysis and Statistics The sensitivities, specificities, negative predictive values, and positive predictive values of the ELISA and latex agglutination assay were calculated (1) for all patients, (2) for patients with no comorbid conditions (see below), (3) for outpatients, (4) for outpatients with no comorbid conditions, and (5) for the subgroup of patients with non-high-probability V/Q scans. The (. ~ > m oconditions r b i d were decided upon a priori and included recent (within 1 days) surgery or trauma, recent (within 1 days) myocardial infarction or stroke, -imer in Clinically Suspected Pulmonary Embolism (Ginsberg at al)

3 Table I-ELISA D-Dimer Results* (a) All Patients (h) Patients Without Comorhid Conditions Sensitivity= 1%, 95% C % Specificity=26.4%, 95% CI % NPV = 1%, 95% CI % PPV=24.7%, 95% CI % Sensitivity= 1%, 95% CI.1-1% Specificity=42.l%, 95% CI % NPV= 1%,95% CI % PPV = 31.9%, 95% CI % (c) Outpatients (d) Outpatients Without Comorhid Conditions Sensitivity= 1%, 95% CI 6.9-1% Specificity= 35.5%, 95% CI % NPV= 1%, 95% CI % PPV = 24.5%, 95% CI % Sensitivity= 1%, 95% CI 4.-1% Specificity = 4.3%, 95% CI % NPV= 1%,95% CI % PPV = 31.3%, 95% CI % *CI =confidence interval; NPV =negative predictive value; PPV =positive predictive value; =pulmonary embolism. acute infection, disseminated intravascular coagulation, pregnancy or recent (within 1 days) delivery, active collagen vascular disease, and metastatic cancer. The <.X>rresponding exact 95 percent <. ~ > n f i dence intervals for sensitivities, specificities, and predictive values were calculated where indicated. Avoidnnce of Bias Results of assays were not disclosed to the clinicians caring for the patients so that management decisions were not made based on the results. Bias in the interpretation of lung scans, venography, and pulmonary angiography was avoided hy having these tests interpreted by an independent panel of experts that was unaware of the clinical status of the patients and results of the assays. Patients RESULTS Two hundred twenty-one patients (141 women and men) with a mean age of 54.2 years (range, 15 to 97 years) were entered into the study. Of the 221 patients, 43 (19 percent) were classified as positive (33 with high-probability lung scans, 1 with a nonhigh-probability lung scan and positive pulmonary angiography, with non-high-probability lung scans and an abnormal lpg with confirmatory venography, and 1 with a non-high-probability lung scan and venous thromboembolism in follow-up) and 17 patients (1 Table 2-Later Agglutination D-Dimer Results* (a) All Patients (h) Patients Without Comorbid Conditions Sensitivity= 3. 7%, 95% CI % Specificity= 55.6%, 95% CI % NPV=93.4%, 95% CI %.PPV=31.3%, 95% CI % Sensitivity=79.3%, 95% CI % Specificity=7l.O%, 95% CI % NPV = 92.7%, 95% CI % PPV = 42.6%, 95% Cl % (c) Outpatients (d) Outpatients Without Comorbid Conditions Sensitivity =.%, 95% CI % Specificity=67.7%, 95% CI % NPV=94.4%, 95% Cl % PPV=34.4%, 95% CI % Sensitivity=l.O%, 95% CI % Specificity=79.%, 95% CI % NPV=94.7%, 95% CI % PPV = 4.6%, 95% CI % *CI = mnfidence interval; NPV =negative predictive value; PPV =positive predictive value; =pulmonary embolism. CHEST I 14 I 6 I DECEMBER,

4 2 16 -l E "' 12» c: < en e ::i w ~ 4 1 ~ <5 9 g 5-1 ~ ~ a.jioo Latex ng/ml percent) were classified as negative (1 with normal lung scans, with normal pulmonary angiography, and 9 with non-high-probability lung scans and either normal serial lpg or abnormal lpg with normal venography and absence of symptomatic venous thromboembolism within 3 months of follow-up). ELISA The ELISA showed sensitivities and negative predictive values of 1 percent, confirming the results of our previous study (Table 1). The specificity was only 26 percent in all patients, but increased to 4 percent in the subgroup of outpatients without comorbid conditions. All but two -positive patients had -dimer results of greater than 1, nglml, but, in order to exclude in all patients, a cutoff of 3 ngl ml is required (one -positive patient had a result of 362 nglml). Using the ELISA, a cutoff of 5 nglml showed a sensitivity of 9 percent and a specificity of 39 percent, whereas a cutoff of 1, nglml shows a sensitivity of 95 percent and a specificity of 6 percent. On the other hand, the finding of a D dimer result of greater than 1, nglml does not diagnose because only 41 of 113 patients with such a result were classified as positive (positive predictive value of 36 percent). >4 FIGURE I. A <:omparison of by ELISA and latex agglutination assay in study patients. Closed circles = positive; open circles= negative; ELISA results of 2, ngfml or more are recorded as 2, ngfml. Latex Agglutination Assay The latex agglutination assay showed a lower sensitivity than the ELISA (4 percent) in all patients, but a higher specificity (56 percent) (Table 2). In a similar way to the ELISA, the specificity increased to percent in outpatients with no comorbid conditions with little change in sensitivity (1 percent). A cutoff of greater than 5 nglml is not useful clinically since the sensitivity is too low (62 percent using 1, ngl ml). On the other hand, the finding of a very high level of using this assay does not reliably diagnose since only 23 of 47 patients with levels of greater than 2, nglml had (positive predictive value of 49 percent). A summary of individual results for the ELISA and latex agglutination assay is provided in Figure l. This shows that there are many patients with negative by latex agglutination assay who had elevated by ELISA and vice versa. Moreover, of the -positive patients with latex results ofless than 5 nglml, 5 had ELISA results of greater than 1, nglml, 1 had an ELISA of 76 nglml, and 1 had an ELISA of 362 nglml. This suggests that the differences in sensitivities and specificities for of the two assays are not simply due to the different cutoffs used but rather due to different properties of Table 3-D-Dimer Results in Patienta With Non-High-Probability VIQ Scan ELISA Sensitivity= 1%, 95% Cl69.2-1% Specificity= 13.4%, 95% Cl % NPV = 1%, 95% Cl % PPV= 1.6%, 95% CI % Latex Sensitivity= 9.%, 95% Cl % Specificity=54.6%, 95% Cl % NPV=9.1%, 95% Cl % PPV = 17.%, 95% Cl % *Cl = <.'Onfidence interval; NPV =negative predictive value; PPV =positive predictive value; =pulmonary embolism imer In CWnically Suspected Pulmonary Embolism (Ginsberg eta/)

5 the assays. l.htients With Non-High-Probability Scans The results of the -dimer assays in the subgroup of patients with non-high-probability scans are summarized in Table 3. The sensitivity and negative predictive value of the ELISA are 1 percent but the specificity is so low (13 percent) that it severely limits the clinical utility of the assay in this group of patients. On the other hand, the latex agglutination assay appears promising since its sensitivity and negative predictive value are reasonable (9 percent and 9 percent, respectively) and the specificity is sufficiently high to make the test potentially useful (55 percent). DISCUSSION The results of our study validate the results of our previous study demonstrating that, using a cutoff of 3 nglml, the ELISA has a high sensitivity and negative predictive value for. Unfortunately, the low specificity and positive predictive value limit the clinical utility of this test, although the test has higher specificity in patients without comorbid conditions and/or in outpatients. The clinical implications of these results (Fig 2) are that can be excluded in patients with levels of less than 3 nglml (using the ELISA) but that such levels are measured in the minority of patients without. On the other hand, further investigation is required in patients with clinically suspected and elevated levels because the positive predictive value is not sufficiently high to reliably diagnose. The fact that the assay takes several hours to perform and is more suitable for multiple assays further limits the clinical utility of this test. Nevertheless, the ELISA has potential to be useful in selected patients who pose diagnostic prob- Clinically Suspected "' ELISA <3 nglml 2:3() nglml I I Excluded Not Excluded I Further Investigation Required Clinically Suspected Nondiagnostic Lung Scan \ I Latex Agglutination Assay* <5 nglml 2:5 ng/ml I I Excluded Not Excluded *This requires further evaluation in clinical trials. I Further Investigation Required FIGURE 2. Potential applications of -dimer testing. lems. The use of 5 nglml as a cutoff would have resulted in missing 1 patient with in this study and 4 of24 -positive patients in our previous study. 14 Although this would result in an increase in specificity, it is less acceptable clinically. The latex agglutination assay, which is more suitable for individual testing and can be performed rapidly, has a lower sensitivity than the ELISA, but the specificity is higher. We had to use a different cutoff (5 nglml) for the latex agglutination assay because this is the lowest level associated with an abnormal test result. Therefore, the lower sensitivity and higher specificity, compared with the ELISA, could be due to either the use of a different cutoff or the characteristics of the assay; the fact that 6 of 7 -positive patients with normal latex agglutination assays had ELISA results of greater than 5 nglml suggests that the latter is the explanation. The lower sensitivity of the latex agglutination assay limits its clinical utility as a screening test in all patients with clinically suspected since a normal test result does not reliably exclude in all patients. However, it has potential utility in patients with nondiagnostic lung scans (Fig 2) in whom the presence of a normal test result showed a negative predictive value of 9 percent. Although this is not perfect, it is comparable to the negative predictive value of pulmonary angiography and serial IPG, 5 17 and it is better than basing management decisions on clinical grounds. However, before the use of the latex agglutination assay is advocated, it should be validated in a large prospective clinical management study. This is because the number of patients with non-high-probability scans who were positive is relatively small and therefore, the 95 percent confidence intervals on the observed sensitivity of9 percent are relatively wide (56 to 1 percent). For example, if the true sensitivity of the assay is 7 percent, the negative predictive value would decrease to 95 percent. Further, the assay should be validated in patients with a high pretest likelihood and non-high-probability VIQ scans in whom the prevalence of is approximately 4 percent. 1 For example, in such patients, a sensitivity of 9 percent and a specificity of 55 percent would provide a negative predictive value of only 9 percent. It is important to address the validity of our conclusions, particularly with respect to the clinical classification of the patients. There are two ways of evaluating the assay in patients with clinically suspected. The first is to perform pulmonary angiography, the reference standard for the diagnosis of, either in all patients or in patients whose V/Q scans are abnormal. This is expensive and invasive. The second approach is to use a clinical management approach that is corroborated by long-term clinical outcome. CHEST I 14 I 6 I DECEMBER,

6 This approach has been used successfully to validate the use of serial lpg in patients with clinically suspected and non-high-probability V/Q scans, by demonstrating that, without anticoagulant therapy, patients who have normal serial lpg have a low rate of thromboembolic events in long-term follow-up. 17 Inherent in this approach is the fact that some patients with non-high-probability V/Q scans, whom we classified as negative, had as a cause of their presenting symptoms. However, only 1 of the 9 patients (1.1 percent) with non-high-probability lung scans and normal serial lpg developed a thromboembolic event in follow-up, corroborating the clinical utility of this clinical classification. Our classification of patients with normal perfusion scans or normal pulmonary angiography as negative is valid, based on the results of published studies On the other hand, two studies have now demonstrated that the vast majority of patients with high-probability lung scans have as a cause of their symptoms. 3 5 Finally, it is important to note that the results of our study are valid for the assays used in this study but may not apply to other latex agglutination and ELISA kits. ACKNOWLEDGMENTS: The authors would like to thank Joanne McGinnis, R.N., and Marilyn Johnston, A.R.T., for their contributions to this study. REFERENCES Hull RD, Raskob GE, Carter CJ, Coates J, Gill GJ, Sackett DL, et al. Pulmonary embolism in outpatients with pleuritic chest pain. Arch Intern Med 19; 14: Bell WR, Simon TL, DeMets DL. The clinical features of submassive and massive pulmonary embolism. Am J Med 1977; 62: Hull RD. Hirsh J, Carter CJ, Jay R, Dodd, Ockelford PA, et al. Pulmonary angiography, ventilation lung scanning and venography for clinically suspected pulmonary embolism with abnormal perfusion lung scan. Ann Intern Med 193; 9: Hull RD, Hirsh J, Carter CJ, Raskob G, Gill GJ, Jay R, et al. Diagnostic value of ventilation-perfusion lung scanning in pa tients with suspected pulmonary embolism. Chest 195; : The PIOD Investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism. JAMA 199; 263: Bounameaux H, Slosman D, de Moerloose P. Reber G. Laboratory diagnosis of pulmonary embolism: value of increased levels of plasma and thrombin-antithrombin III <.'lmplexes. Biomed Pharm:u:other 199; 43:35-7 Bridey F, Philipotteau C, Dreyfus M, Simonneau G. Plasma D dimer and pulmonary embolism. Lancet 199; 1: Speiser W, Leitha T, Dudczak R. Lechner K. Letter to the editor. Lancet 199; 1:792 9 Goldhaber SZ, Vaughan DE, Tumeh SS, Loscalzo J. Utility of cross-linked fibrin degradation products in the diagnosis of pulmonary embolism. Am Heart J 19; 116:55-1 Bounameaux H, Cirafici P. DeMoerloose P. Schneider PA, Slosman D, Reber G, et al. Measurement of as diagnostic aid in suspected pulmonary embolism. Lancet 1991; 337: Pelzer H, Schwarz A, Heimburger N. Determination of human thrombin-antithrombin Ill complex in plasma with an enzymelinked immunosorbent assay. Thromb Haemost 19; 59: Rowbotham BJ, Carrol P. Whitaker AN, Bunce IH, Cobcroli RG, Elms MJ, et al. Measurements of crosslinked fibrin derivates use in the diagnostic of venous thrombosis. Thromb Haemost 197; 57: Bounameaux H, Schneider PA, Reber G, De Moerloose P. Krahenbuhl B. Measurement of plasma D-Dimer for diagnosis of deep venous thrombosis. AJCP 199; 91 : Demers C, Ginsberg JS, Johnston M, Brill-Edwards P. Panju es in pa A. and thrombin-antithrombin Ill t ~ > m p l e x tients with clinically suspected p u l m ~ embolism m a r. Thromb Haemost 1992; 67:4-12 y 15 Heaton DC, Billings JD, Hickton CM. Assessment of assays for the diagnosis of deep venous thrombosis. J Lab Clin Med 197; 11: Hull R, Van Aken WG, Hirsh J, Gallus AS, Hoicka G, Turpie AGG, et al. Impedance plethysmography using the occlusive cuff technique in the diagnosis of venous thrombosis. Circulation 1976; 5: Hull RD, Raskob G, Coates G, Panju AA, Gill GJ. A new noninvasive management stratej.,'y for patients with suspected pulmonary embolism. Arch Intern Med 199; 149: Kelley MA, Carson JL, Palevsky HI, Schwartz JS. Diagnosing pulmonary embolism: new facts and strategies. Ann Intern Med 1991; 114: Hull RD, Raskob GE, Coates G, Panju AA. Clinical validity of a normal perfusion lung scan in patients with suspected pulmonary embolism. Chest 199; 97: D-Oimer in Clinically Suspected Pulmonary Embolism (Ginsberg et al)