The diagnostic role of capnography in pulmonary embolism
Original Contribution
The diagnostic role of capnography in pulmonary embolism
Ozlem Kar Kurt MD a,?, Sibel Alpar MD a, Tugrul Sipit MD a, Selma Firat Guven MD a, Hakan Erturk MD b, Mikail Koray Demirel MD c, Meliha Korkmaz MD c,
Mutlu Hayran MD d, Bahar Kurt MD e
aDepartment of Chest Diseases, Ataturk Chest Disease and Thoracic Surgery Teaching and Research Hospital, 06280 Ankara, Turkey
bDepartment of Radiology, Ataturk Chest Disease and Thoracic Surgery Teaching and Research Hospital, 06280 Ankara, Turkey
cDivision of nuclear medicine, Ankara Teaching and Research Hospital, 06340 Ankara, Turkey
dInstitute of Oncology, Department of Preventive Oncology, Hacettepe University, Faculty of Medicine, 06100 Ankara, Turkey
eDepartment of Chest Diseases, Abant Izzet Baysal University Medical School, 14100 Bolu, Turkey
Received 17 December 2008; revised 24 January 2009; accepted 25 January 2009
Abstract The aim of this study was to evaluate the diagnostic contribution of alveolar dead space fraction (AVDSf) measured using capnography in patients admitted with suspected pulmonary embolism (PE). A total of 58 patients who were admitted to our hospital with suspected PE between October 2006 and January 2008 were included in this study. All patients were assessed using the Wells Clinical score, capnography, computed tomographic pulmonary angiography, D-dimer measurement, lower-extremity venous Doppler ultrasonography, and V/Q scintigraphy. Forty patients (69%) had PE based on computed tomographic pulmonary angiography findings. The AVDSf value with the highest sensitivity and specificity, which was at the same time statistically significant, was 0.09. This value was consistent with the AVDSf value obtained using receiver operating characteristic analysis. In our study, the sensitivity of capnography was 70%, with a specificity of 61.1%, positive predictive value of 80%, and negative predictive value of 47.8%. The use of AVDSf in combination with any of the several scoring systems that evaluate clinical likelihood of PE and D-dimer levels resulted in higher sensitivity and specificity rates for the diagnosis of PE.
(C) 2010
Introduction
Pulmonary embolism (PE) is part of a group of thromboembolic disorders with a clinical spectrum ranging from mild symptoms to severe life-threatening Right heart failure. The occlusion of the pulmonary artery or its branches
* Corresponding author. Tel.: +90 505 2762811; fax: +90 312 3552135.
E-mail address: [email protected] (O.K. Kurt).
by a thrombus originating from the Venous system results in a clinical picture with high mortality and morbidity. According to several reports, the incidence of PE has been reported between 21 and 69/100 000 [1]. The co-presence of deep vein thrombosis and PE is referred to as Venous thromboembolic disease and is a condition that is serious and difficult to diagnose. Several studies have shown that the diagnosis of PE is successfully made in less than half of patients before their death [2,3]. In the International Cooperative Pulmonary
0735-6757/$ – see front matter (C) 2010 doi:10.1016/j.ajem.2009.01.031
Embolism Registry study, the 3-month crude Death rate of 2454 patients with acute PE was 17.5% [4]. In reality, the actual prevalence of PE in autopsy series has not changed much in the past 3 decades (12%-15% of hospitalized patients) [5]. Although the mortality rate is 30% in untreated cases, the advent of anticoagulation therapy has seen this rate fall to 2% to 8%. It is for his reason that research has focused on developing techniques for early and accurate diagnosis of this condition. In recent years, the use of capnography as a simple, noninvasive, fast, and practical test for the diagnosis of PE has been under investigation as an intriguing modality based on simple physiopathology [6-11]. Several studies have postu- lated cutoff levels for alveolar dead space fraction (AVDSf), and when used in combination with other parameters such as clinical probability and D-dimer, has yielded very impressive results in helping to rule out PE [8,9,11].
In this study on patients with suspected PE who presented to the emergency department (ED) or the chest diseases clinic for further evaluation, we aimed to evaluate the benefit of alveolar dead space measurement using capnography, alone or when used in combination with D-dimer and clinical probability, in the diagnosis of PE.
Materials and methods
This study was undertaken in the Ataturk Chest Diseases and Thoracic Surgery Teaching and Research Hospital (Ankara, Turkey). The aim was to evaluate the diagnostic value of AVDSf measurement using capnography in patients who were hospitalized in the emergency and chest diseases wards with suspected PE. Inclusion criteria included age between 18 and 75 years, with no history of treated and/or recurrent PE. Written consent was obtained for each enrolled patient as per the Helsinki declaration. Those with confirmed or suspected pregnancy, known allergies to contrast solution, in poor general condition (comatose, disoriented and uncooperative, intubated or meeting criteria for shock), or were morbid obese (>=140 kg) were excluded from the study. After obtaining approval by the local ethic committee, a total of 58 patients who presented to our hospital with a suspicion of PE between October 2006 and January 2008 were included in this study. A detailed medical history and demographics were recorded for each patient on admission, and patients were questioned for risk factors for PE. The Wells score was completed for each patient, and subjects were divided into 3 groups according to their score (b2, low probability; 2-6, intermediate probability; N6, high probability).
Quantitative measurement of D-dimer was performed in
56 patients using an AMAX AUTO D-Dimer kit, Trinity Biotech, Bray, Ireland (polystyrene, microparticle agglutina- tion assay). Patients were grouped based on D-dimer levels of b500 and >=500 ug/L. Fifty-six patients also underwent bilateral lower-extremity venous Doppler ultrasonography (VDUSG) at the local radiology department. Patients were
|
PE present |
% |
PE absent |
% |
P |
|
|
Risk factor
|
|||||
|
n |
40 |
69 |
18 |
31 |
|
|
Age (y) |
64 |
54 |
.12 |
||
|
(22-81) |
(22-77) |
||||
|
Sex |
|||||
|
Male |
26 |
65 |
9 |
50 |
.28 |
|
Female |
14 |
35 |
9 |
50 |
|
|
Comorbidities |
|||||
|
COPD |
5 |
12.5 |
2 |
11.1 |
.88 |
|
DM |
2 |
5 |
3 |
16.7 |
.14 |
|
ASCD |
3 |
7.5 |
3 |
16.7 |
.36 |
|
HT |
14 |
35 |
4 |
22.2 |
.33 |
|
History of smoking |
20 |
50 |
8 |
44.4 |
.695 |
|
History of OC use |
1 |
7.1 |
0 |
1.0 |
|
|
History of abortion |
1 |
7.1 |
0 |
1.0 |
|
|
Immobile for more |
20 |
50 |
8 |
44.4 |
.695 |
|
than 3 d |
|||||
|
Operation within the |
10 |
25 |
3 |
16.7 |
.481 |
|
previous 4 wk |
|||||
|
Known history of |
0 |
0 |
|||
|
thrombophilia |
|||||
|
History of DVT |
7 |
17.5 |
0 |
.058 |
|
|
Malignancy (received |
2 |
5 |
1 |
5.6 |
1.0 |
|
treatment within the |
|||||
|
previous 6 mo) |
|||||
|
COPD indicates chronic obstructive pulmonary disease; DM, diabetes mellitus; ASCD, Atherosclerotic coronary disease; HT, hypertension; OC, oral contraceptive; DVT, deep vein thrombosis. |
|||||
grouped by the presence or absence of a thrombus. Ventilation perfusion (V/Q) scintigraphy was done for 50 patients, and consistent with the Prospective Investigation of Pulmonary Embolism Diagnosis study, patients were classi- fied as normal, low, intermediate, or high risk. All patients were evaluated with computed tomographic pulmonary angiography . Films were read by the same radiologist establishing the presence or absence of emboli. Finally, bedside levels of partial end-tidal carbon dioxide pressure (PETCO2) were measured for all patients using the TIDAL WAVE Sp Novametrix Handheld Capnography model 615, Novametrix Medical Systems Inc., Wallingford, Connecticut, USA, capnography device. With the help of arterial blood gas analysis, the AVDSf was calculated using the PaCO2-PETCO2 gradient and the Bohr formula (Formula 1). The alveolar-arterial oxygen gradient [P(A-a)] was calculated using Formula 2.
AVDSf(mm Hg) = (PaCO2 — PETCO2)/PaCO2 Formula1 P(A-a)O2 = (140 — PaCO2/0.8) — PaO2 Formula2
Table 1 The characteristics of the patients with suspected PE
Age-adjusted P(A-a)O2 predicted levels were obtained using Formula 3.
P(A — a)O2 = 10 + 0.43(age — 20) Formula3
Statistical analysis was done using the Statistical Package for Social Sciences 16.0 for Windows (SPSS, Chicago, Ill), and a P value of less than .05 was considered statistically significant.
Results
The characteristics of the 58 patients with suspected PE are summarized in Table 1. According to CTPA findings, 40 patients (69%) had confirmed PE, with a median age of 64 years compared to the median age for those without PE which was 54 years. In the PE group, the most common symptoms were chest pain (82.5%), shortness of breath (72.5%), redness and swelling of the legs (50%), hemoptysis (40%), cough (35%), palpitations (10%), and syncope (5%). Rales on auscultation (40%), Homans sign (35%), tachy- cardia (25%), and an accentuated pulmonary component of S2 (2%) were the most commonly encountered physical examination findings.
The patients’ laboratory results, Wells scores, ABG analysis, P(A-a), PETCO2, and AVDSf are summarized in Table 2. Although there was no significant difference between the 2 groups with regard to complete blood count,
|
PE present PE absent P |
|
n 40 18
Leukocytes 8.5 (3.7-28.9) 9.2 (5.1-14.2) .77 (x103/mm3) Hemoglobin 13.2 (9.8-15.2) 13.4 (12.8-16.1) .66 (g/dL) Hematocrit (%) 37.7 +- 4.3 39.8 +- 2.1 .86 Platelet 254 (110-422) 226 (143-649) .45 (x103/mm3) ESR (/h) 44 (4-84) 60 (23-85) .71 Glucose 126 (60-308) 131 (110-137) .52 (mg/dL) BUN (mg/dL) 22 (10-42) 21 (11-54) .94 Creatinine 1.0 (0.7-1.5) 1.1 (0.5-1.8) .88 (mg/dL) SO2 (%) 92 (70-97) 96 (85-97) .06 D-dimer 1061 (285-6500) 735 (263-5995) .99 (ng/mL) pH 7.42 (7.34-7.52) 7.40 (7.38-7.48) .68 PCO2 (mm Hg) 34.5 (28.1-48.6) 38.4 (35.6-52.1) .057 PO2 (mm Hg) 63.0 (32.4-85) 64.6 (49.3-132) .86 P(A-a)O2 33.7 +- 2.3 32.0 +- 3.7 .71 PETCO2 28 (19-40) 35 (30-40) .003 Wells score 7.0 (3.0-10.5) 2.5 (0-5.5) b.001 AVDSf 0.174 +- 0.12 0.136 +- 0.14 .047 pCO2-PETCO2 4.7 [-2.8- (+18.6)] 5.6 [-1.0- (+17.1)] .094 |
|
ESR indicates erythrocyte sedimentation rate; BUN, blood urea nitrogen. |
Fig. 1 Receiver operating characteristic curve for the optimal cutoff point of the predictive value of AVDSf in the diagnosis of PE.
Serum glucose, and renal function tests, patients with PE had significant hypoxia and hypocapnia. For the patients with PE, the mean AVDSf was 0.174 compared with 0.136 in the Normal group (P = .047). The cutoff value for AVDSf in the detection of PE was calculated as 0.083 or higher using the receiver operating characteristic (ROC) analysis (sensitivity, 72.5%; specificity, 61.1%; area under the curve, 0.664 [0.516-0.812]; P = .047) (Fig. 1). The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and ? agreement value were calculated for every AVDSf value between 0.03 and 0.17. An AVDSf value of
Table 2 Laboratory results, Wells scores, ABG, PETCO2, P(A-a)O2, and AVDSf values of patients with suspected PE
0.09 was found to have the most statistically significance, with the highest sensitivity, specificity, and ? agreement values. This value was consistent with that obtained from the ROC analysis.
Table 3 shows a comparison of CTPA findings with D- dimer levels, bilateral lower-extremity VDUSG findings, Wells scores, and V/Q scintigraphy alone or in combination with AVDSF in patients with suspected PE.
Patients were grouped based on their Wells score, with 7 (12.1%) patients falling into the low probability group, whereas the intermediate and high probability groups had 30 (51.7%) and 21 (36.2) patients, respectively. Pulmonary embolism detected by CTPA was observed in 1 patient (14.3%) in the low-probability group, in 18 patients (31%) from the intermediate-probability group, and in all 21 patients (100%) in the high-probability group. Patients in the high-probability group were considered as having PE, whereas those in the low- and intermediate-probability groups were accepted as not having PE (Table 3).
Ventilation perfusion scintigraphy was performed in 50 patients (86%). Findings were normal in 6 (12%), whereas for 10 patients (20%), the result was low probability. Eight patients (16%) were considered as having intermediate probability for PE, whereas 26 (52%) had high probability.
Computed tomographic pulmonary angiography detected PE in 3 (50%), 5 (50%), 5 (62.5%), and 24 (92.3%) of patients in the normal-, low-, intermediate-, and high-probability groups, respectively. Patients in the normal-, low-, and intermediate-probability groups were judged to be free of PE (Table 3).
Discussion
Pulmonary embolism is a condition that is serious and difficult to diagnose with high mortality and morbidity. The importance of early diagnosis and treatment highlights the
need for easier diagnostic modalities. Clinical assessment is the cornerstone of the diagnosis of PE in patients suspected of having the condition. In our study we used the likelihood algorithm proposed by Wells et al [12], which combines clinical findings with risk factors, electrocardiogram, and chest X-ray findings in patients suspected of having PE. In this study, patients with confirmed PE fell into the high- probability group (median Wells score, 7), whereas those without PE fell into the low-probability group (median score, 2.5), and the difference between both groups was statistically significant (P b .001) (Table 2).
Several noninvasive tests have been proposed to help decrease the dependency on invasive tests in patients with suspected PE. D-dimer is one such test. It is quite safe on
|
Table 4 Comparison of our study results on capnography for PE with other reports from the literature
|
||||||
|
Test |
Sn (%) |
Sp (%) |
PPV (%) |
NPV (%) |
||
|
Kar Kurt et al |
AVDSf (0.09) |
70 |
61.1 |
80 |
47.8 |
|
|
AVDSf + Wells scores |
80 |
61.1 |
82.1 |
57.9 |
||
|
AVDSf + D-dimer |
89.7 |
27.8 |
72.9 |
55.6 |
||
|
AVDSf + VDUSG |
82.5 |
55.6 |
80.5 |
58.8 |
||
|
V/Q + AVDSf |
82.5 |
52.9 |
80.5 |
56.2 |
||
|
Kline et al [8] |
AVDSf (0.20) |
67.2 |
76.3 |
|||
|
D-dimer |
93.8 |
67.1 |
||||
|
D-dimer + AVDSf |
98.4 |
51.6 |
||||
|
Rodger et al [9] |
AVDSf (0.15) |
79.5 |
70.3 |
|||
|
D-dimer |
83 |
57.6 |
||||
|
Hogg et al [10] |
AVDSf (0.32) |
95.3 |
20 |
6.2 |
98.7 |
|
|
Sanchez et al [11] |
AVDSf (0.15) |
68.5 |
81.5 |
71.1 |
79.5 |
|
|
AVDSf + Wicki scores |
70 |
61.1 |
80 |
47.8 |
||
ruling out PE, thus helping to avoid further unnecessary testing [13]. However, it has low specificity, particularly in the elderly, pregnant women, patients with cancer, and hospitalized patients where the D-dimer test correctly excludes PE in only 30% of the patients [14]. In our study, D-dimer had a sensitivity of 71.1% and specificity of 27.8%. In PE, the affected lung parenchyma has normal ventilation, but perfusion is affected. In the nonperfused alveolar space, the amount of carbon dioxide expressed is low. In summary, PE increases the amount of alveolar dead space, thus resulting in a decrease in carbon dioxide in expired air. However, several studies have shown that AVDSf values obtained using bedside capnography did not satisfactorily exclude PE [8,11]. In the same studies, it was demonstrated that the concomitant use of other tests such as D-dimer, Wells score, USG, and V/Q scintigraphy signifi- cantly increased the sensitivity and NPV of AVDSf in ruling out PE. For their ease, D-dimer and clinical probability
scores have been used more commonly (Table 4).
Sanchez et al [11] studied 270 patients with suspected PE who had positive D-dimer tests. Capnography and ABG measurements were obtained for each patient before spiral CT and lower-extremity VDUSG was performed. In this study, an AVDSF value of less than .15 was considered normal. Patients with negative spiral CT and USG findings, and those with low Wicki scores falling into the low-and intermediate-probability groups were excluded from study as not having PE. Patients with normal USG and spiral CT findings, but high-probability Wicki scores, were included in the study; and within the first 48 hours they underwent either one or both of V/Q scintigraphy and pulmonary angiography. In 45 patients (16.6%), the combination of low clinical probability and normal capnography managed to exclude PE in 99.1% sensitivity. In the 34 patients with high clinical probability and positive capnography, this combination had been 100% specific and 31.5% sensitive in correctly diagnosing PE. Overall, the combination of capnography and Wicki scores had a sensitivity of 70% and specificity of 61.1%. In this study by Sanchez et al, D-dimer was measured using the rapid quantitative enzyme-linked immunosorbent assay test. By excluding patients with D-dimer levels of less than 500, they enrolled a select group. Therefore, the sensitivity and specificity values provided for AVDSf are actually representative of its combination with D-dimer, and with clinical probability a triple combination.
In our study, the AVDSf cutoff point was set at 0.09 after analysis of the ROC. This value differs from that obtained in previous studies (0.15-0.32) [8-11]. Our results with AVDSf alone or in combination with Wells score, D-dimer, and lower-extremity VDUSG have been compared with previous studies in Table 4. The sensitivity of capnography in our study was increased from 70% to 89.7% with combination of D-dimer.
In a prospective multicenter study by Kline et al [8], 380 patients who presented to the ED were enrolled. They aimed to demonstrate that D-dimer and AVDSf values could help in
ruling out PE. D-dimer was measured using whole-blood agglutination assay (Simpli-RED, Agen Inc, Brisbane, Australia) on blood samples obtained for ABG analysis necessary for calculating AVDSf. Pulmonary embolism was confirmed in 64 patients (16.8%) in this study. The high PE ratio in our study could be attributed to the fact that hospitalized patients were included in our study, whereas patients who presented to the ED who had normal D-dimer and CTPA findings were excluded. Kline et al considered an AVDSf value of <=0.20 as normal, with a sensitivity of 67.2% and specificity of 76.3%. In the same study, D-dimer had a sensitivity of 93.8% and specificity of 67.1%. Positivity of either one of D-dimer or AVDSf (63/64 patients) had a sensitivity of 98.4% and specificity of 51.6%. In our study, the sensitivity and specificity of D-dimer were 71.1% and 27.8%, respectively, whereas for AVDSf these values were 70% and 61.1%. Either one of them being positive (34/40) successfully predicted PE with sensitivity of 89.7% and specificity of 27.8% (P = .092, ? = 0.202).
The low specificity of this combination in our study may
have been because of the technique used in measuring D- dimer levels. Kline et al reported a mean AVDSf value of
0.33 in 8 patients who eventually died of PE, whereas in the remaining 56 patients with PE, the mean value was 0.23. In 6 patients who did not have PE, but died due from other causes, this value was 0.11. It was suggested that AVDSf may play an important role in determining the severity of PE. Because the patients in our study were not followed up, we could not evaluate the link between AVDSf and overall survival. In the study by Kline et al, patients with suspected PE with a negative D-dimer value and normal AVDSf were found to have a 1% probability for having PE.
In a similar study by Rodger et al, PE was detected in 49 of 246 patients with suspected PE. An AVDSf value of less than 0.15 was considered normal [9]. A negative D-dimer value excluded PE with a sensitivity of 83.0% and a specificity of 57.6%. A normal value for AVDSf ruled out PE with 79.5% sensitivity and 70.3% specificity. The combina- tion of negative D-dimer with a normal AVDSf value was 98.4% sensitive and 38% specific in ruling out PE. The mean AVDSf value for those with confirmed PE was 0.27 compared with 0.11 for those without PE.
In a study by Hogg et al on 425 patients who presented with pleurisy, 5% had detectable PE, and AVDSf had a sensitivity and specificity of 95.3% and 20%, respectively, with an NPV of 98.7% and a PPV of 6.2% [10]. The combination of a positive AVDSf value obtained using the Bohr equation and a positive D-dimer value had a sensitivity of 90.5% and a specificity of 72.3% with NPV and PPV values of 99.3% and 13.2%, respectively.
In our study, the median PETCO2 obtained from capno- graphy measurements for patients with confirmed PE was 28 (19-40), whereas in those without PE, this value was 35 (30- 40), a statistically significant difference (P = .003). This result is consistent with physiologic PETCO2 variation. Studies have shown that a normal PaCO2-PETCO2 gradient
is not sensitive enough to safely exclude a diagnosis of PE [15]. Verschuren et al [15] reported that a cutoff value of 3 had a sensitivity of 77% and specificity of 70%. In our study, the median gradient in the PE group was 4.7 compared with
5.6 in those without PE (P = .094).
In recent years, the use of capnography as a simple, noninvasive, fast, and practical test for the diagnosis of PE has been under investigation as an intriguing modality based on simple physiopathology. Meaningful results have been reported in studies on the value of AVDSf (obtained from capnography measurements using the Bohr equation) combined with parameters such as clinical probability and D-dimer in ruling out PE [9]. Capnography may help avoid unnecessary further testing in up to 30% of patients [11].
In our study, as with other similar studies, we concluded that capnography has its shortcomings when used alone in the diagnosis of PE. However, the use of AVDSf in combination with any of the several scoring systems that evaluate the clinical likelihood of PE (eg, Wells, Wicki, and Pisa) and D-dimer levels resulted in higher sensitivity and specificity rates for the diagnosis of PE. In fact, combining all 3 modalities may alleviate the need for Imaging techniques in the diagnosis of PE. Of course there is dire need for large-scale studies to support our findings.
References
- Fennerty T. The diagnosis of pulmonary embolism. BMJ 1997;314: 425-9.
- Dalen J. Clinical diagnosis of acute pulmonary embolism. In: Morpugo M, editor. Pulmonary embolism. NewYork: Dekker; 1994. p. 55-66.
- Rubinstein I, Murray D, Haffsten V. Fatal pulmonary emboli in hospitalized patients. Arch Intern Med 1988;148:1425-6.
- Goldhaber SZ, Visani L, De Rosa M. Acute pulmonary embolism: clinical outcomes in the International Cooperative Pulmonary Embolism Registry (ICOPER). Lancet 1999;353:1386-9.
- Stein PD, Henry JW. Prevalence of acute pulmonary embolism among patients in a general hospital and at autopsy. Chest 1995;108:78-81.
- Nutter DO, Massumi RA. The arterial-alveolar carbon dioxide tension gradient in diagnosis of pulmonary embolus. Dis Chest 1966; 50:380-7.
- Hatle L, Rokseth R. The arterial to end-expiratory carbon dioxide tension gradient in acute pulmonary embolism and other cardiopul- monary diseases. Chest 1974;66:352-7.
- Kline JA, Israel EG, Michelson EA, et al. Diagnostic accuracy of a bedside D-dimer assay and alveolar dead-space measurement for rapid exclusion of pulmonary embolism: a multicenter study. JAMA 2001; 285:761-8.
- Rodger MA, Jones G, Rasuli P, et al. Steady-state end-tidal alveolar dead space fraction and D-dimer: bedside tests to exclude pulmonary embolism. Chest 2001;120:115-9.
- Hogg K, Dawson D, Tabor T, et al. respiratory dead space measurement in the investigation of pulmonary embolism in out- patients with pleuritic chest pain. Chest 2005;128:2195-202.
- Sanchez O, Wermert D, Faisy C, et al. Clinical probability and alveolar dead space measurement for suspected pulmonary embolism in patients with an abnormal D-dimer test result. J Thromb Haemost 2006;4:1517-22.
- Wells PS, Ginsberg JS, Anderson DR, et al. Use of a clinical model for safe management of patients with suspected pulmonary embolism. Ann Intern Med 1998;129:997-1005.
- Stein PD, Hull RD, Patel KC, et al. D-dimer for the exclusion of deep venous thrombosis and acute pulmonary embolism: a systematic review. Ann Intern Med 2004;140:589-602.
- Perrier A, Desmarais S, Miron MJ, et al. Noninvasive diagnosis of venous thromboembolism in outpatients. Lancet 1999;353:190-5.
- Verschuren F, Liistro G, Coffeng R, et al. Volumetric capnography as a screening test for pulmonary embolism in the emergency department. Chest 2004;125:841-50.