| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
* Department of Radiology, Lasarettet Trelleborg, University of Lund, Hedvägen 46, SE-231 85 Trelleborg, Sweden
e-mail: ulf.nyman{at}skane.se
Department of Radiology, Institute of Selected Clinical Sciences, The Sahlgrenska Academy at Göteborg University, Göteborg, Sweden
In the February 2005 issue of Radiology, Dr Cham and colleagues (1) recommended the routine use of indirect computed tomographic (CT) venography in connection to CT pulmonary angiography for acute pulmonary embolism (PE). The diagnosis of deep venous thrombosis (DVT) by means of the addition of CT venography resulted in a 20% incremental increase (48 cases of DVT besides the 243 cases of PE) in venous thromboembolism detection in comparison with the use of CT pulmonary angiography alone. However, the absolute increase was only 3.0% (48 of 1590 cases), and among patients with either a negative (n = 1330) or inconclusive (n = 17) result at CT pulmonary angiography, 48 patients (3.6%) were diagnosed with DVT. The authors' positive test result is the sum of true-positive and false-positive diagnoses. With such a low positive test result, the false-positive rate (here defined as 1 – positive predictive value) grows rapidly when specificity decreases below the clinically unlikely level of 100%. This inevitable consequence and the authors' recommendation to routinely use indirect CT venography following CT pulmonary angiography raise some issues:
First, is the frequency of diagnosed DVT any different following normal conventional pulmonary angiography or ventilation-perfusion (V/Q) scintigraphy?
Second, what is the false-positive rate among those 3.6% patients with positive results for DVT when specificity drops below 100%?
Third, could another diagnostic strategy reduce the unnecessary risk of anticoagulating false-positive cases without jeopardizing patients who need treatment for venous thromboembolism?
DVT frequency following negative studies for acute PE.—In a meta-analysis, 3% and 7% of cases of DVT were diagnosed following negative conventional pulmonary angiography or V/Q scan findings, respectively (2). Pooled data of studies on CT venography following negative CT pulmonary angiography findings revealed a 4.5% (50 of 1116 patients) frequency of diagnosed DVT (3). Thus, the frequency of diagnosed DVT seems to be in the same range irrespective of which primary method was used to diagnose PE. Since it is generally held that patients can be safely left without anticoagulation following negative conventional pulmonary angiography (4,5) or V/Q (5) scan findings, routine venous studies following negative findings at CT pulmonary angiography may be questioned.
Specificity, prevalence, and false-positive rate.—The sum of the true-positive rate and false-positive rate, the positive test rate, is dependent on sensitivity and specificity, as well as prevalence of disease. Zero false-positive cases require a specificity of 100%. On the basis of the present 3.6% positive test rate of DVT, we calculated the false-positive rate (1 – positive predictive value) if specificity would drop below 100%. Already at 99% specificity, 25% of the DVT diagnoses would have been false, and that would increase to more than 80% if specificity drops to 97%, which is the highest specificity for CT venography cited by Cham et al (1) and reported in systematic reviews of ultrasonography (US) (6,7). Thus, there may be a delicate balance between the commendable ambition to improve the diagnostics of venous thromboembolism and the risk of hemorrhagic complications from anticoagulation in false-positive cases. Sensitivity, when decreased from 100% to 60%, has a negligible effect on the false-positive rate.
Stratification to clinical probability of PE a possible solution?—A clinically low probability of PE may indicate a 10% prevalence of PE, whereas a high probability may indicate prevalence as high as 70% (8). Combining a 10% prevalence of PE with 86% sensitivity and 94% specificity of CT pulmonary angiography for the diagnosis of PE (9) results in a false-negative rate (here defined as 1 – negative predictive value) of PE of only 1.6%. However, at 70% prevalence, the false-negative rate of PE increases to almost 25%, which would motivate further diagnostic tests (eg, venous studies). It seems reasonable to assume that patients with a high clinical probability of PE also have a higher prevalence of DVT. Anticipating a positive CT venography rate of 20% instead of 3.6% following negative results at CT pulmonary angiography, the false-positive rate will drop to about 4% and 11% at a specificity of 99% and 97%, respectively. Thus, it may be prudent to limit the CT venography resources to patients assessed to have a high clinical probability of PE.
In the present article, 97% of the CT venography examinations ([1590 – 48]/1590) may only have added cost, workload, radiation exposure, and risk of hemorrhages from anticoagulation in false-positive DVT if specificity drops a few percent below 100%.
If indirect CT venography is routinely used following CT pulmonary angiography (or US following negative CT pulmonary angiography findings), a positive diagnosis of DVT after negative CT pulmonary angiography findings should be confirmed with a second technique to minimize the risk of false-positive diagnoses.
Indirect CT venography (or US) may be a more valuable tool if only performed in patients assessed to have a high clinically probability of PE prior to CT pulmonary angiography, though the prevalence of DVT following negative CT pulmonary angiography findings in such a population needs to be evaluated in prospective studies.
 |
References |
|---|
 TOP References References |
|---|
Department of Radiology, New York Presbyterian Hospital-Weill Medical College of Cornell University, 525 East 68th Street, New York, NY 10021
e-mail: mattcham{at}hotmail.com
We disagree with the approach used by Dr Nyman in his analysis of our study. In his letter, Dr Nyman states that 80% of DVT diagnoses would be false-positive on CT venograms if the specificity drops to 97%. However, his analysis was limited to only a subgroup of our subjects who had DVT but no PE—that is, 48 of the total 1590. The combined tests, CT pulmonary angiography–CT venography, are performed in people suspected of having thromboembolic disease, either PE or DVT, and not only in those with no PE. The second test, CT venography, requires no additional contrast material injection.
To demonstrate this, we used the references and values specified by Dr Nyman: sensitivity and specificity of CT pulmonary angiography for PE of 86% and 94%, respectively (1); sensitivity and specificity of CT venography for DVT of 100% and 97%, respectively (2); and prevalence of DVT in proved PE of 40% (3).
When we apply the above values chosen by Dr Nyman to our data, the false-positive rate and false-negative rate of the combined study of CT pulmonary angiography–CT venography are lower than those of CT pulmonary angiography alone. For CT pulmonary angiography alone, the false-positive rate would be 35% (84 of 243) and the false-negative rate would be 2% (26 of 1347). For combined CT pulmonary angiography–CT venography, these rates would be 32% (92 of 291) and 1% (16 of 1299), respectively.
The discrepancy between our findings and those of Dr Nyman is due to his analysis of the subgroup. We hold that CT venography is not a test performed only when CT pulmonary angiography results are negative, but rather, the usefulness of CT venography must be evaluated together with CT pulmonary angiography for the diagnosis of thromboembolic disease.
|
References |
|---|
|
TOPReferencesReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
