TO THE EDITOR:
Reinartz et al. have to be commended for reassessing ventilation–perfusion scintigraphy for diagnosis of pulmonary embolism in light of recent advances such as SPECT methodology and 99mTc-based ultrafine aerosols (1). They are correct in stating that comparisons with pulmonary CT angiography should incorporate these technological advances.
There are good reasons to believe that tomographic imaging could supersede the sensitivity of planar techniques, simply by avoiding the overlapping of small perfusion defects by normal tissue. For example, phantom experiments have shown that perfusion defects in the mediobasal segment of the lower lobe would go unnoticed on planar images (2). As expected, the data of Reinartz et al. (1) support the better sensitivity of SPECT over planar techniques. Moreover, pulmonary angiography, which was used as the reference method in the Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) study, may underdiagnose pulmonary embolism (3).
So, there is a need to reassess scintigraphy for diagnosis of pulmonary embolism both because technology has progressed and because weaknesses of prior assessments have been recognized. On the other hand, the limitations of the retrospective study by Reinartz et al. (1) should be acknowledged as well. In their study, the final diagnosis was subjectively based on all imaging data together with clinical data, including follow-up and D-dimer levels. Therefore, the final diagnosis may have been biased to various degrees by the techniques to be compared, as well as by the clinical decisions that had been made from them.
Therefore, one should exercise caution before definitively adopting the authors’ proposal to report all mismatches as embolisms. It may be hard to admit to the referring physician that, after all the trouble we and our patient took to perform a perfusion–ventilation study, still, we do not know whether the patient has pulmonary embolism. Yet, this may be the most honest answer and is preferable to guesswork. The intermittent-probability category is just a way of identifying those patterns that do not allow a final conclusion and that may require further diagnostic studies. Besides, management studies have convincingly shown precisely how to resolve these indeterminate cases by further diagnostic examination, if at all necessary (4).
This will always involve a further cost. Therefore, if such patterns can be avoided by technical improvements (e.g., because of additional mismatches identified on a tomographic study), all the better. Or if they are at error, because of limitations in the PIOPED studies that have defined them, one should of course eliminate them. But if such patterns are the result of conceptual limitations inherent in perfusion scintigraphy, we would probably do better to continue to label them as indeterminate readings.
There is reason to believe that such limitations are inherent in perfusion scintigraphy, a technique that reveals arterial or arteriolar obstruction, instead of clots. Indeed, any defect seen on a perfusion scan opens up the differential diagnosis of obstruction, which could be due to a lesion either in the lumen, in the vessel wall, or outside the vessel. Distinguishing between matched and mismatched perfusion defects does not completely resolve this differential diagnosis; for example, apart from embolism, mismatches may occur with vasculitis or with extrinsic tumors that spare the airways. Neither would a concurrent chest radiograph, as proposed by the authors (and also mandatory when using the PIOPED scheme) allow one to settle all diagnostic questions. Reinartz et al. (1) point out that “the PIOPED study gives no physiologic explanation of why large mismatch defects should be a sign for pulmonary embolism while small ones are not.” But the experimental finding from the PIOPED study that small mismatches do not always mean embolism may just signify that in small mismatches the other differential diagnoses are relatively more frequent or that the matched or mismatched nature is more difficult to certify for small defects. Alternatively, this finding could have been an error introduced by the limited sensitivity of pulmonary angiography. At this point, however, no definitive proof of this assumption exists. So, it seems likely that some scan patterns will never allow one to rule in or to rule out pulmonary embolism and would be most appropriately termed “indeterminate.” A further practical consideration is that small defects, even when due to pulmonary embolism, may not always be the harbinger of life-threatening pulmonary embolism and therefore may sometimes be left untreated.
How should we proceed further, then? Because a favorable patient outcome does matter more than a correct diagnosis (only embolism that needs to be treated should be detected), outcome studies withholding anticoagulant treatment from patients with a low diagnostic probability are probably the best way to assess the sensitivity of diagnostic modalities for clinically relevant pulmonary embolism. Such data exist for some diagnostic strategies, including planar ventilation–perfusion scintigraphy, but they are lacking for tomographic scintigraphy. In stark contrast, a recent management study has shown that multidetector-row CT can even be used as the sole imaging study (5). If perfusion–ventilation scintigraphy is to survive in clinical practice, we will need that same level of evidence. More, if sensitivity is our strong point, as suggested by the data from Reinartz et al. (1), it would be in our interest to compare the 3-mo embolic risk in cohorts diagnosed by either ventilation–perfusion scintigraphy or CT.
Given the lack of a reference method with 100% sensitivity (which would be needed to identify all patients without pulmonary embolism), the specificity of a diagnostic modality for pulmonary embolism is even more difficult to judge, although this is an important issue in view of the hazards of anticoagulant treatment. Assessing the specificity will involve a comprehensive search for alternative diagnoses, but care should be taken to mask readers of scintigraphy for the results of this search.
In conclusion, the work of Reinartz et al. (1) provides an impetus for further prospective and more rigorous studies on ventilation–perfusion scintigraphy for diagnosing pulmonary embolism. As indicated by Reinartz et al., these will need to incorporate state-of-the-art techniques and revised interpretation criteria. Until those are proven that way, however, I am afraid that from time to time, it will be wise to admit that still, we do not know.
REFERENCES
REPLY:
We appreciate Dr. De Geeter’s interest in our article (1). Indeed, it is true that lung scintigraphy—like all other diagnostic procedures developed so far—is tainted with a certain probability of error. Although the same applies to multislice spiral CT, apparently the radiologic community has sufficient confidence in their method to give definitive diagnoses. Sure enough, scintigraphic mismatch defects can be caused by nonembolic diseases. On the other hand, it cannot be excluded that some of the segmental or subsegmental clots detected by CT are, rather, partial-volume artifacts or other phenomena instead of embolisms. Because no perfect gold standard exists for the diagnosis of pulmonary embolism, there is no possibility of verifying the scan results objectively. To sum it up, we could say of nuclear medicine physicians—still, they do not know, and of radiologists—they do not know either but continue to give definitive diagnoses anyway, and with considerable success.
In our opinion, it is well founded to diagnose embolism on lung scans when mismatch defects are detected. Apart from embolism, such mismatch defects are induced by only a few and, more important, rare nonembolic diseases or therapeutic interventions such as arteritis, vessel stenosis, lung cancer, nodal enlargement, and radiation therapy (2). Some of these conditions can be excluded by anamnesis or plain chest radiography so that the probability of a false diagnosis is further reduced. In this context, it appears reasonable and well balanced to use the diagnostic approach proposed by Howarth et al. (3), according to which embolism should be diagnosed in cases of mismatch defects of half-segment size or larger. By doing so, they achieved a sensitivity of 0.98 and a specificity of 0.96 in a study group of 924 patients. These data are indeed impressive and underline the diagnostic power of lung scintigraphy.
As far as concerns Dr. De Geeter’s notion that “only embolism that needs to be treated should be detected,” we strongly disagree. Although it is true that not all embolic clots are “the harbinger of life-threatening pulmonary embolism,” it is also true that at autopsy, 50% of the patients dying from pulmonary embolism show residuals of earlier embolic events (4). In our opinion, a diagnostic procedure should be as exact as possible in reflecting pathologic changes. It is not within the competence of the diagnostician to decide about therapeutic options. All the diagnostician should do is provide accurate information to the physician responsible for the patient. To withhold or disregard any findings because they may be irrelevant for treatment is, in our opinion, irresponsible, especially because the therapeutic regime for pulmonary embolism is under constant evolution. Who can predict what therapeutic impact subsegmental embolisms may have in the future?
If we answer with “maybe” once too often, there is a good possibility that no one is going to ask us anymore. In conclusion, we can only stress the importance of striving for and expressing definitive diagnoses.