Accuracy or Outcome in Suspected Pulmonary Embolism
http://www.100md.com
《新英格兰医药杂志》
Sixteen years ago, the results of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study1 were published. That large-scale, multicenter trial definitively established the diagnostic characteristics of ventilation–perfusion scintigraphy of the lung, as compared with pulmonary angiography, for the diagnosis of pulmonary embolism. It was also the first trial to provide direct evidence that clinical assessment has a role beyond merely evoking pulmonary embolism as a diagnostic possibility and allows for the stratification of patients into three pretest categories of probability corresponding to an increasing prevalence of pulmonary embolism.
As reported by Stein et al. in this issue of the Journal,2 the investigators of the PIOPED II study again incorporated clinical assessment and determination of the performance of a test against a reference criterion in the evaluation of suspected pulmonary embolism. But in the past 15 years, much has changed in the field of pulmonary-embolism diagnosis. The question now is whether to use multidetector computed tomographic (CT) angiography (CTA), with or without CT venography (CTV) of the pelvic and thigh veins, to search for a proximal deep venous thrombosis. Clinical probability is no longer assessed implicitly on the basis of the clinician's global judgment but explicitly on the basis of a clinical prediction rule.3
So how does multidetector CTA of the chest perform in this population, which is predominantly outpatient (90 percent)? In the PIOPED II study, the specificity of chest CTA was excellent (96 percent), and the 83 percent sensitivity was certainly higher than that of single-detector imaging, which was previously reported to be approximately 70 percent.4,5 Nevertheless, the 17 percent false negative rate comes as a disappointment, as compared with other similar but less rigorously designed trials.6 It is also a surprise, considering the very low rate of thromboembolic events (below 2 percent) in patients who did not receive anticoagulant therapy because they had normal findings on multidetector CTA in two recent, large-scale outcome studies.7,8 How can one explain these apparently conflicting findings?
Patients in the PIOPED II study were, on average, 13 years younger than those in the outcome studies. There is no evidence that older age makes the interpretation of CTA with respect to pulmonary embolism more difficult,9 but even if it did, this finding should improve, not reduce, the accuracy of CTA in younger patients. Approximately a quarter of the patients in the PIOPED II study (266 of 1090) were not included in the sensitivity and specificity calculations since they did not complete all the required testing. In particular, 175 patients did not undergo pulmonary angiography despite having inconclusive results on noninvasive testing. However, these patients were unlikely to have had a substantial effect on the reported sensitivity of CTA. The diagnostic criteria for pulmonary embolism were sound, and the absence of pulmonary embolism in patients whose results on ventilation–perfusion scanning indicated a low or very low probability and who had a low clinical probability was verified by an uneventful follow-up, a finding that minimizes the risk of misclassification bias. Therefore, the most likely explanation for the findings of the PIOPED II study is that although multidetector (mainly four-slice) CTA is more sensitive than single-detector imaging, it still misses small, peripheral subsegmental clots that are better detected by ventilation–perfusion scintigraphy or classic pulmonary angiography. However, the clinical significance of such thrombi is the subject of much controversy,10 and the results of outcome studies in which many such patients were probably not treated on the basis of false negative findings on CTA suggest that most such thrombi do not need to be treated — and therefore do not need to be detected. Moreover, the advent of 16-slice and even 64-slice CTA in the near future may further improve the sensitivity of CTA in the periphery.
The PIOPED II study also provides a direct empirical demonstration of Bayes' theorem as applied to diagnosis, which states that the probability of disease after testing depends not only on the sensitivity and specificity of the test, but also on the clinical probability before testing.11 Indeed, in patients with a low or intermediate clinical probability of pulmonary embolism (as assessed by the Wells clinical prediction rule), normal findings on CTA had a high negative predictive value for pulmonary embolism (96 percent for patients with a low probability and 89 percent for patients with an intermediate probability), whereas the negative predictive value was only 60 percent in patients with a high probability before testing. Conversely, the positive predictive value of abnormal findings on CTA was high (92 to 96 percent) in patients with an intermediate or high clinical probability but much lower (58 percent) in patients with a low likelihood of pulmonary embolism. Hence, clinical assessment should be systematically included in the diagnostic workup of pulmonary embolism to allow for an adequate interpretation of test results. Normal results on CTA should be interpreted as ruling out pulmonary embolism only in patients without a high clinical probability, and clinicians should probably think twice before accepting the diagnosis of pulmonary embolism in a patient in whom the disease is thought to be clinically unlikely and CTA shows a small, isolated clot.
Finally, does searching for deep venous thrombosis by adding CTV of the proximal lower-limb veins improve the accuracy of CTA? Apparently so, since the sensitivity of chest CTA combined with CTV was 90 percent, as compared with 83 percent for chest CTA alone. However, the absolute gain owing to CTV was modest (the identification of 14 additional patients with pulmonary embolism among the 824 patients with a reference diagnosis), which is reflected by a mere 2 percentage point increase in the negative predictive value (97 percent, as compared with 95 percent). Also, CTV combined with clinical assessment did not yield significantly different predictive values from those with chest CTA alone. On the other hand, 105 of the 192 patients with pulmonary embolism (55 percent) had a thrombus in the lower-limb veins, confirming the value of looking for deep venous thrombosis in suspected pulmonary embolism. Although the proportion of patients with pulmonary embolism in whom compression ultrasonography detects a proximal deep venous clot (approximately 40 percent) is lower than that for CTV,12 performing this entirely noninvasive test to reduce the requirement for CTA may be cost-effective.
In summary, the PIOPED II study convincingly establishes the diagnostic performance of multidetector CTA, at least in outpatients. These data, along with those from recent outcome studies, support the use of multidetector CTA for suspected pulmonary embolism as a stand-alone imaging technique in most patients. However, clinicians should be wary of results that are discordant with their clinical judgment, especially in the rare case of a patient with a high likelihood of pulmonary embolism and normal findings on CTA. CTV does not appear to improve the diagnostic yield of CTA enough to justify the additional irradiation.
No potential conflict of interest relevant to this article was reported.
Source Information
From the Divisions of General Internal Medicine (A.P.) and Angiology and Hemostasis (H.B.), Department of Internal Medicine, Geneva University Hospital and Geneva Faculty of Medicine, Geneva.
References
The PIOPED Investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism: results of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED). JAMA 1990;263:2753-2759.
Stein PD, Fowler SE, Goodman LR, et al. Multidetector computed tomography for acute pulmonary embolism. N Engl J Med 2006;354:2317-2327.
Wells PS, Anderson DR, Rodger M, et al. Excluding pulmonary embolism at the bedside without diagnostic imaging: management of patients with suspected pulmonary embolism presenting to the emergency department by using a simple clinical model and d-dimer. Ann Intern Med 2001;135:98-107.
Perrier A, Howarth N, Didier D, et al. Performances of helical computed tomography in unselected outpatients with suspected pulmonary embolism. Ann Intern Med 2001;135:88-97.
Van Strijen MJ, De Monye W, Kieft GJ, Pattynama PM, Prins MH, Huisman MV. Accuracy of single-detector spiral CT in the diagnosis of pulmonary embolism: a prospective multicenter cohort study of consecutive patients with abnormal perfusion scintigraphy. J Thromb Haemost 2005;3:17-25.
Quiroz R, Kucher N, Schoepf UJ, et al. Right ventricular enlargement on chest computed tomography: prognostic role in acute pulmonary embolism. Circulation 2004;109:2401-2404.
Perrier A, Roy P-M, Sanchez O, et al. Multidetector-row computed tomography in suspected pulmonary embolism. N Engl J Med 2005;352:1760-1768.
van Belle A, Buller HR, Huisman MV, et al. Effectiveness of managing suspected pulmonary embolism using an algorithm combining clinical probability, D-dimer testing, and computed tomography. JAMA 2006;295:172-179.
Righini M, Bounameaux H, Perrier A. Effect of age on the performance of single detector helical computed tomography in suspected pulmonary embolism. Thromb Haemost 2004;91:296-299.
Le Gal G, Righini M, Parent F, van Strijen M, Couturaud F. Diagnosis and management of subsegmental pulmonary embolism. J Thromb Haemost 2006;4:724-731.
Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine. 2nd ed. Boston: Little, Brown, 1991.
Kearon C, Ginsberg JS, Hirsh J. The role of venous ultrasonography in the diagnosis of suspected deep venous thrombosis and pulmonary embolism. Ann Intern Med 1998;129:1044-1049.(Arnaud Perrier, M.D., and)
As reported by Stein et al. in this issue of the Journal,2 the investigators of the PIOPED II study again incorporated clinical assessment and determination of the performance of a test against a reference criterion in the evaluation of suspected pulmonary embolism. But in the past 15 years, much has changed in the field of pulmonary-embolism diagnosis. The question now is whether to use multidetector computed tomographic (CT) angiography (CTA), with or without CT venography (CTV) of the pelvic and thigh veins, to search for a proximal deep venous thrombosis. Clinical probability is no longer assessed implicitly on the basis of the clinician's global judgment but explicitly on the basis of a clinical prediction rule.3
So how does multidetector CTA of the chest perform in this population, which is predominantly outpatient (90 percent)? In the PIOPED II study, the specificity of chest CTA was excellent (96 percent), and the 83 percent sensitivity was certainly higher than that of single-detector imaging, which was previously reported to be approximately 70 percent.4,5 Nevertheless, the 17 percent false negative rate comes as a disappointment, as compared with other similar but less rigorously designed trials.6 It is also a surprise, considering the very low rate of thromboembolic events (below 2 percent) in patients who did not receive anticoagulant therapy because they had normal findings on multidetector CTA in two recent, large-scale outcome studies.7,8 How can one explain these apparently conflicting findings?
Patients in the PIOPED II study were, on average, 13 years younger than those in the outcome studies. There is no evidence that older age makes the interpretation of CTA with respect to pulmonary embolism more difficult,9 but even if it did, this finding should improve, not reduce, the accuracy of CTA in younger patients. Approximately a quarter of the patients in the PIOPED II study (266 of 1090) were not included in the sensitivity and specificity calculations since they did not complete all the required testing. In particular, 175 patients did not undergo pulmonary angiography despite having inconclusive results on noninvasive testing. However, these patients were unlikely to have had a substantial effect on the reported sensitivity of CTA. The diagnostic criteria for pulmonary embolism were sound, and the absence of pulmonary embolism in patients whose results on ventilation–perfusion scanning indicated a low or very low probability and who had a low clinical probability was verified by an uneventful follow-up, a finding that minimizes the risk of misclassification bias. Therefore, the most likely explanation for the findings of the PIOPED II study is that although multidetector (mainly four-slice) CTA is more sensitive than single-detector imaging, it still misses small, peripheral subsegmental clots that are better detected by ventilation–perfusion scintigraphy or classic pulmonary angiography. However, the clinical significance of such thrombi is the subject of much controversy,10 and the results of outcome studies in which many such patients were probably not treated on the basis of false negative findings on CTA suggest that most such thrombi do not need to be treated — and therefore do not need to be detected. Moreover, the advent of 16-slice and even 64-slice CTA in the near future may further improve the sensitivity of CTA in the periphery.
The PIOPED II study also provides a direct empirical demonstration of Bayes' theorem as applied to diagnosis, which states that the probability of disease after testing depends not only on the sensitivity and specificity of the test, but also on the clinical probability before testing.11 Indeed, in patients with a low or intermediate clinical probability of pulmonary embolism (as assessed by the Wells clinical prediction rule), normal findings on CTA had a high negative predictive value for pulmonary embolism (96 percent for patients with a low probability and 89 percent for patients with an intermediate probability), whereas the negative predictive value was only 60 percent in patients with a high probability before testing. Conversely, the positive predictive value of abnormal findings on CTA was high (92 to 96 percent) in patients with an intermediate or high clinical probability but much lower (58 percent) in patients with a low likelihood of pulmonary embolism. Hence, clinical assessment should be systematically included in the diagnostic workup of pulmonary embolism to allow for an adequate interpretation of test results. Normal results on CTA should be interpreted as ruling out pulmonary embolism only in patients without a high clinical probability, and clinicians should probably think twice before accepting the diagnosis of pulmonary embolism in a patient in whom the disease is thought to be clinically unlikely and CTA shows a small, isolated clot.
Finally, does searching for deep venous thrombosis by adding CTV of the proximal lower-limb veins improve the accuracy of CTA? Apparently so, since the sensitivity of chest CTA combined with CTV was 90 percent, as compared with 83 percent for chest CTA alone. However, the absolute gain owing to CTV was modest (the identification of 14 additional patients with pulmonary embolism among the 824 patients with a reference diagnosis), which is reflected by a mere 2 percentage point increase in the negative predictive value (97 percent, as compared with 95 percent). Also, CTV combined with clinical assessment did not yield significantly different predictive values from those with chest CTA alone. On the other hand, 105 of the 192 patients with pulmonary embolism (55 percent) had a thrombus in the lower-limb veins, confirming the value of looking for deep venous thrombosis in suspected pulmonary embolism. Although the proportion of patients with pulmonary embolism in whom compression ultrasonography detects a proximal deep venous clot (approximately 40 percent) is lower than that for CTV,12 performing this entirely noninvasive test to reduce the requirement for CTA may be cost-effective.
In summary, the PIOPED II study convincingly establishes the diagnostic performance of multidetector CTA, at least in outpatients. These data, along with those from recent outcome studies, support the use of multidetector CTA for suspected pulmonary embolism as a stand-alone imaging technique in most patients. However, clinicians should be wary of results that are discordant with their clinical judgment, especially in the rare case of a patient with a high likelihood of pulmonary embolism and normal findings on CTA. CTV does not appear to improve the diagnostic yield of CTA enough to justify the additional irradiation.
No potential conflict of interest relevant to this article was reported.
Source Information
From the Divisions of General Internal Medicine (A.P.) and Angiology and Hemostasis (H.B.), Department of Internal Medicine, Geneva University Hospital and Geneva Faculty of Medicine, Geneva.
References
The PIOPED Investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism: results of the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED). JAMA 1990;263:2753-2759.
Stein PD, Fowler SE, Goodman LR, et al. Multidetector computed tomography for acute pulmonary embolism. N Engl J Med 2006;354:2317-2327.
Wells PS, Anderson DR, Rodger M, et al. Excluding pulmonary embolism at the bedside without diagnostic imaging: management of patients with suspected pulmonary embolism presenting to the emergency department by using a simple clinical model and d-dimer. Ann Intern Med 2001;135:98-107.
Perrier A, Howarth N, Didier D, et al. Performances of helical computed tomography in unselected outpatients with suspected pulmonary embolism. Ann Intern Med 2001;135:88-97.
Van Strijen MJ, De Monye W, Kieft GJ, Pattynama PM, Prins MH, Huisman MV. Accuracy of single-detector spiral CT in the diagnosis of pulmonary embolism: a prospective multicenter cohort study of consecutive patients with abnormal perfusion scintigraphy. J Thromb Haemost 2005;3:17-25.
Quiroz R, Kucher N, Schoepf UJ, et al. Right ventricular enlargement on chest computed tomography: prognostic role in acute pulmonary embolism. Circulation 2004;109:2401-2404.
Perrier A, Roy P-M, Sanchez O, et al. Multidetector-row computed tomography in suspected pulmonary embolism. N Engl J Med 2005;352:1760-1768.
van Belle A, Buller HR, Huisman MV, et al. Effectiveness of managing suspected pulmonary embolism using an algorithm combining clinical probability, D-dimer testing, and computed tomography. JAMA 2006;295:172-179.
Righini M, Bounameaux H, Perrier A. Effect of age on the performance of single detector helical computed tomography in suspected pulmonary embolism. Thromb Haemost 2004;91:296-299.
Le Gal G, Righini M, Parent F, van Strijen M, Couturaud F. Diagnosis and management of subsegmental pulmonary embolism. J Thromb Haemost 2006;4:724-731.
Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine. 2nd ed. Boston: Little, Brown, 1991.
Kearon C, Ginsberg JS, Hirsh J. The role of venous ultrasonography in the diagnosis of suspected deep venous thrombosis and pulmonary embolism. Ann Intern Med 1998;129:1044-1049.(Arnaud Perrier, M.D., and)