| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
PLENARY SESSION |
1 From the Department of Radiology, Michigan State University, 184 Radiology Bldg, East Lansing, MI 48824-1303. From the Plenary Session, Friday Imaging Symposium: Algorithmic Controversies, at the 1999 RSNA scientific assembly. Received February 14, 2000; revision requested February 23 and received March 13; accepted March 13. Address correspondence to the author (e-mail: alex.gottschalk@radiology.msu.edu).
Index Terms: Embolism, pulmonary, 60.72 • Lung, CT, 60.12115 • Lung, diseases, 60.72 • Lung, perfusion, 60.12171 • Lung, ventilation, 60.1217
The purpose of this article is to suggest new algorithms for the interaction between ventilation-perfusion scanning and spiral computed tomography (CT). Much of the data that are discussed comes from the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study. The criteria for the original PIOPED ventilation-perfusion scan evaluation were developed in 1983 and initially tested in practice sessions in 1984. Once these criteria had been finalized, they were not changed throughout the course of the PIOPED study, for which patient recruitment began in January 1985 and ended in the fall of 1986. Thus, the PIOPED study is a relatively old study; nevertheless, the PIOPED database has been extremely useful because it contains precise descriptions of a large number of ventilation-perfusion scans and their corresponding angiograms. In addition, extensive clinical data were collected for each patient. The PIOPED database is very large; for example, there are over 250 possible data entry points for each ventilation-perfusion scan (1).
The original PIOPED report analyzed the main arm of the PIOPED study, which included 902 patients in whom pulmonary embolism was either diagnosed or excluded at pulmonary angiography (84% of cases) or at clinical follow-up that lasted up to 12 months after initial presentation (2). The primary conclusions of the PIOPED study were that high-probability ventilation-perfusion scans were reliable indicators for pulmonary embolism and that normal or near normal scans were reliable in the exclusion of the disease. In the PIOPED study, these findings were coupled with clinical probability assessment. A high-probability clinical assessment coupled with a high-probability ventilation-perfusion scan demonstrated an accuracy of 96% in diagnosing pulmonary embolism, whereas a low-probability clinical assessment coupled with a normal or near normal ventilation-perfusion scan demonstrated a similar accuracy in excluding pulmonary embolism. The main problem with the portion of the PIOPED study relating to ventilation-perfusion scanning was that a definitive diagnosis was obtained in less than 30% of cases.
Revisions of PIOPED Criteria
In an effort to improve the ability to make an accurate definitive diagnosis, the PIOPED nuclear medicine working group developed revised PIOPED criteria (3). In essence, the revised criteria indicated that (a) a single, moderate-sized mismatched segmental defect was inappropriate for low probability, (b) multiple matched defects could be considered low probability if they were the only perfusion defects present, and (c) the cutoff for high probability (ie, two segmental equivalent defects) was appropriate.
Using data from the PIOPED study and other studies, Hull and Raskob (4), acting on a suggestion by the late K. Moser, advocated that low- and intermediate-probability ventilation-perfusion scans be combined to form a "nondiagnostic" category. The authors argued that there were too many emboli present in the "low-probability" categories. In particular, many referring physicians were incorrectly interpreting "low probability" to mean "no probability." In the PIOPED study, 22% of the patients in the "nondiagnostic" category had pulmonary emboli.
Dalen (5) subsequently developed an algorithm stating that a high-probability perfusion scan was a reliable indicator for pulmonary embolism and that affected patients should undergo anticoagulant therapy. Similarly, a normal ventilation-perfusion scan was a contraindication for anticoagulant therapy provided there was no evidence of deep venous thrombosis on ultrasonographic or impedance plethysmographic studies of the legs. All other ventilation-perfusion scan findings would be considered nondiagnostic, and these patients should be followed up with either leg studies or pulmonary angiography. A variation of this general algorithm was subsequently proposed by Stein et al (6), who suggested that normal or near normal ventilation-perfusion scan findings are acceptable and that noninvasive leg tests or pulmonary angiography should not be recommended unless a strong clinical suspicion for pulmonary emboli is present.
This sequence of events indicates that ventilation-perfusion scanning must provide an increased number of accurate definitive findings if its usefulness is to be maximized. Two approaches have been taken to achieve this goal. The first is to increase the number of patients for whom a high-probability reading can be made. The second is to develop criteria to extract information from the "nondiagnostic" category that will be useful to pulmonologists and other referring physicians.
Improving Ventilation-Perfusion Scan Criteria
Stein et al (7) suggested that patients be categorized on the basis of the results of clinical assessment for prior cardiopulmonary disease. This clinical evaluation was related to the number of segmental equivalent defects present on the ventilation-perfusion scan. These authors examined both arms of the PIOPED study (ie, patients who underwent obligatory angiography and those who underwent angiography only if their physician requested it) and showed that a single, moderate-sized mismatched segmental lesion (0.5 segmental equivalent defects) had a positive predictive value (PPV) of 80% for pulmonary embolism provided there was no history of cardiopulmonary disease. In contrast, in patients with a history of cardiopulmonary disease, a single, moderate-sized mismatched segmental lesion had a PPV of 68% for pulmonary embolism. In these patients, two mismatched segmental equivalent defects were needed to achieve a PPV of 80% for high probability (Table 1). Furthermore, the difference between patients with and without prior cardiopulmonary disease was statistically significant for 0.5, 1.0, and 1.5 segmental equivalent defects. In short, substituting a lower number of mismatched perfusion defects for high-probability diagnosis in patients without prior cardiopulmonary disease allowed an increase of up to 50% in correct high-probability diagnoses, thereby consistently improving sensitivity with very little change in specificity.
|
|
Although Stein et al (6) recognized the usefulness of coupling normal or near normal ventilation-perfusion scan findings with clinical findings, this approach does not provide nuclear medicine physicians with useful criteria for the "near normal" category. In the PIOPED study, a scan was considered near normal if (a) one of the PIOPED readers considered the study to be low-probability and the other considered it to be normal, (b) one reader considered the study to be very low probability and the other considered it to be normal, or (c) both readers considered the study to be very low probability. Note that in the PIOPED study, the "very low probability" category was quite restrictive, requiring three or fewer small segmental perfusion defects (ie, defects involving less than 25% of a segment) with a normal chest radiograph; the ventilation scan findings were irrelevant. After the revised PIOPED criteria were released, many authors developed additional criteria that were potentially useful in defining the "very low probability for pulmonary embolism" category. Stein and Gottschalk (10) recently collated this material, defining very low probability criteria as those demonstrating a PPV of less than 10% for pulmonary embolism. From their work and that of others, six criteria were identified (Table 3).
|
Proposed Algorithms
I believe that ventilation-perfusion scanning and spiral CT can be used interactively to diagnose pulmonary embolism. To do this successfully, the number of indeterminate ventilation-perfusion scans must be decreased. The PIOPED study clearly shows that patients in this category have about a 33% chance of having a pulmonary embolism. The typical clinical approach to these patients (ie, treating them according to best clinical judgment) is, in my opinion, unsatisfactory. One alternative approach would be to screen all patients with ventilation-perfusion scanning, use spiral CT to follow up any patients with intermediate-probability or indeterminate findings, and then perform pulmonary angiography or other studies depending on the spiral CT findings. I believe that clinical categorization of patients as those with and those without prior cardiopulmonary disease could make this process easier and more reliable because more correct high-probability readings could be made in patients with no prior cardiopulmonary disease, thereby increasing sensitivity without changing specificity.
Separating patients into those with and those without prior cardiopulmonary disease is not a difficult clinical task; it could be achieved in large part by a technologist with a relatively simple questionnaire. Patients are considered to have no prior cardiac disease if they have no history of valvular heart disease, coronary artery disease, or other heart disease and no history of left- or right-sided heart failure prior to the episode of suspected acute pulmonary edema. Patients are considered to have no prior pulmonary disease if they have no history of asthma, chronic obstructive pulmonary disease, interstitial lung disease, or other lung disease and no recognized acute pneumonia or acute respiratory distress syndrome at the time of evaluation for suspected acute pulmonary edema. In addition, these patients must have no history of pulmonary edema.
Another algorithm has been suggested by P. O. Alderson (presentation to the Fleischner Society, Tucson, Arizona, April 1999). In this approach, chest radiography is used for initial screening. If chest radiographic findings are "simple" (ie, normal or near normal), ventilation-perfusion scanning should be effective, whereas if the findings are "complicated," spiral CT should be effective. Both of these categories are tempered by the clinical assessment. In general, I believe that when either ventilation-perfusion scan findings or spiral CT findings do not match the clinical assessment (eg, high-probability clinical findings with a normal spiral CT scan), additional testing is required. The original PIOPED database has been used to evaluate ventilation-perfusion scans accompanied by normal chest radiographic findings. In brief, this approach was not helpful: Emboli were still seen in 17% of patients in the low-probability group (11). However, ventilation-perfusion scan criteria have been modified significantly since the original PIOPED study, and, as described later, this has resulted in a significant improvement in diagnostic accuracy.
I can now present new evidence to suggest that these algorithms are reasonable. In Europe, a prospective trial of spiral CT in the diagnosis of pulmonary embolism that also included nuclear medicine studies has been conducted by the European Society of Thoracic Imaging Prospective Evaluation of Pulmonary Embolism (ESTIPEP). This study was organized largely by Dr C. J. Herold, and the database for the study is housed with Dr Herold's department at the University of Vienna. It was my privilege to work with Dr Herold and his colleagues, Dr C. Schaefer-Prokop, Dr M. Prokop, and J. Hahne. The data that follow have been evaluated but at the time of this writing are as yet unpublished. The first series was obtained when I read ventilation-perfusion scans as a consensus reader for the ESTIPEP trial. Every time I encountered a scan that I considered to have one or more criteria for very low probability for pulmonary embolism, I made note of it. When my readings were finished, I was able to use the official reading of the spiral CT scan as a reference criterion to prospectively assess the usefulness of the very low probability criteria. My thesis was that if the very low probability criteria were correct, over 90% of these cases would demonstrate normal findings at spiral CT. In fact, 93% of over 40 cases demonstrated normal spiral CT findings. A second series of 115 cases was obtained to see how effectively these new criteria could be used in cases in which chest radiographic findings were normal or near normal. These cases were selected by Dr Schaefer-Prokop, who sent me the chest radiographs and ventilation-perfusion scans but provided no patient history. Because I did not have clinical information, I could not categorize patients as those with and those without prior cardiopulmonary disease. I was able to give a definitive reading (ie, pulmonary embolism definitely present or absent) in 83% of cases, with a sensitivity, specificity, PPV, negative predictive value, and accuracy of 90% or more. The reference criteria for this series consisted largely of definitive spiral CT readings in ESTIPEP, although it also included some patients who had undergone pulmonary angiography. In addition, 9% of the patients could not be evaluated, largely because the three central spiral CT readers all disagreed. Recently, a sensitivity of 87% and a specificity of 90% have been cited for the central spiral CT readings in ESTIPEP (12). These data clearly support using chest radiography to screen patients such that those with normal or near normal radiographic findings are sent to ventilation-perfusion scanning and the rest are sent to spiral CT.
Quality issues always arise when comparing two techniques. Since ESTIPEP, new mosaic detectors have become available and have improved the technique of spiral CT for pulmonary embolism. At the same time, nuclear medicine technique in ESTIPEP was often suboptimal. For example, ESTIPEP required only 300,000 counts for most of the images on the perfusion scan, and no collimator was specified. In contrast, in the much older PIOPED study, 750,000 counts were required for comparable images with a low-energy, all-purpose collimator. I suspect (but cannot prove) that current nuclear medicine technique would have permitted an even greater number of definitive ventilation-perfusion diagnoses to be made.
Conclusions
Ventilation-perfusion scanning can be a highly accurate and useful noninvasive test in diagnosing acute pulmonary embolism. Algorithms designed to correctly integrate findings at ventilation-perfusion scanning and spiral CT should be developed at any institution where both tests are available.
Footnotes
Abbreviations: ESTIPEP = European Society of Thoracic Imaging Prospective Evaluation of Pulmonary Embolism PIOPED = Prospective Investigation of Pulmonary Embolism Diagnosis PPV = positive predictive value
References
This article has been cited by other articles:
|
|
 |
|
  P. Reinartz, J. E. Wildberger, W. Schaefer, B. Nowak, A. H. Mahnken, and U. Buell Tomographic Imaging in the Diagnosis of Pulmonary Embolism: A Comparison Between V/Q Lung Scintigraphy in SPECT Technique and Multislice Spiral CT J. Nucl. Med., September 1, 2004; 45(9): 1501 - 1508. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Itti, S. Nguyen, F. Robin, S. Desarnaud, J. Rosso, A. Harf, and M. Meignan Distribution of Ventilation/Perfusion Ratios in Pulmonary Embolism: An Adjunct to the Interpretation of Ventilation/Perfusion Lung Scans J. Nucl. Med., December 1, 2002; 43(12): 1596 - 1602. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOGRAPHICS | RADIOLOGY | RSNA JOURNALS ONLINE |
