|
|
| |||||||||||||||||||||||||||||||||||||||
Background Impedance plethysmography performed serially over a one-week period has been shown to be an effective diagnostic strategy for patients with clinically suspected acute deep-vein thrombosis. Compression ultrasonography has a high sensitivity and specificity for the detection of proximal-vein thrombosis. The clinical value of repeated ultrasonography in the management of symptomatic deep-vein thrombosis is unknown.
Methods We conducted a randomized trial in 985 consecutive outpatients with clinically suspected deep-vein thrombosis to compare the diagnostic value of serial impedance plethysmography (494 patients) and serial compression ultrasonography (491 patients). We compared the positive predictive values of both tests for the diagnosis of venous thrombosis, using contrast venography as a reference. The frequencies of venous thromboembolism during a six-month follow-up period were also compared in patients with repeatedly normal results in order to evaluate the safety of withholding anticoagulant therapy from such patients.
Results The positive predictive value of an abnormal ultrasonogram was 94 percent (95 percent confidence interval, 87 to 98 percent), whereas the predictive value of impedance plethysmography was 83 percent (95 percent confidence interval, 75 to 90 percent) (P = 0.02). In patients with repeatedly normal results, the incidence of venous thromboembolism during the six-month follow-up period was 1.5 percent (95 percent confidence interval, 0.5 to 3.3 percent) for serial compression ultrasonography, as compared with 2.5 percent (95 percent confidence interval, 1.2 to 4.6 percent) for serial impedance plethysmography.
Conclusions In making the diagnosis of deep-vein thrombosis in symptomatic outpatients, serial compression ultrasonography is preferable to impedance plethysmography, in view of its superior performance in detecting venous thrombosis.
Recently, there has been a substantial increase in the use of ultrasonography. Ultrasound scanners are available in most hospitals and are simple to operate, and the single criterion of compressibility of a venous segment (compression ultrasonography) is highly accurate9,10,11. However, because of its limited spatial resolution, ultrasonography cannot always provide adequate evaluation of the smaller deep veins of the calf. Thus far, a limited number of studies12,13,14 have suggested that compression ultrasonography may be useful in symptomatic patients, but the clinical validity of a normal compression ultrasonogram has not yet been evaluated by long-term clinical follow-up of large series of consecutive patients in order to document the safety of withholding anticoagulant treatment in patients with normal results. It is also controversial whether the investigation could be limited to a single examination of the entire deep-vein system of the leg (including the deep veins of the calf), or whether serial examinations (over a one-week period) of the popliteal and common femoral veins are indicated. In view of the lower sensitivity of compression ultrasonography for the detection of isolated calf-vein thrombi, we considered it necessary to perform compression ultrasonography serially10,11,12. We hypothesized that serial testing of the common femoral and popliteal veins with ultrasonography would be at least as accurate for the detection and exclusion of deep-vein thrombosis as impedance plethysmography. To test this hypothesis, consecutive symptomatic outpatients were randomly assigned to undergo either serial impedance plethysmography or serial compression ultrasonography, and the predictive value of an abnormal result (relative to the results of contrast venography) was compared with the incidence of venous thromboembolism during long-term follow-up in the patients with serially normal results who were not treated with anticoagulants.
Methods
Patients and Study Design
From November 1989 through October 1991, consecutive outpatients with clinically suspected acute deep-vein thrombosis of the leg who were referred by their general practitioners to the thrombosis unit of the Academic Medical Center in Amsterdam, the Thrombosis Service in Amsterdam, or the thrombosis unit of the Hamilton General Hospital in Hamilton, Canada, were included in the study. The referral patterns and diagnostic equipment of the three study centers were comparable. The study was approved by the institutional review boards.
Patients were considered ineligible if they were younger than 18 years, had received full-dose anticoagulant treatment for more than 48 hours before referral, had a history of a documented deep-vein thrombosis in the symptomatic leg, had symptoms suggestive of pulmonary embolism, had a known allergy to contrast material, were pregnant, or lived too far from the study center to return for serial testing. Informed consent was requested from all eligible patients.
On the day of referral, all patients underwent aphysical examination and were then randomly assigned to undergo either serial impedance plethysmography or serial real-time compression ultrasonography. The sealed-envelope method was used for randomization.
Impedance plethysmography was performed and interpreted as described previously15. Repeated testing and prolonged cuff inflation were employed with an impedance plethysmograph (model 200, Codman, Randolph, Mass.). The results were scored as normal, abnormal, or inadequate for interpretation. Real-time compression ultrasonography was performed with an Ultramark 4 scanner (Advanced Technology Laboratories, Dordrecht, the Netherlands, and Toronto) equipped with a 7.5-MHz linear-array transducer. Both the common femoral and popliteal veins were examined; the compressibility of these veins was assessed only in the transverse planes11. The outcomes were categorized as normal, abnormal, or inadequate for interpretation.
If the first result was normal, no anticoagulant treatment was instituted, and the test was repeated on the next day (day 2) and one week after the first visit (day 8)7. If the first result was abnormal or became abnormal on serial testing, ascending contrast venography was done to confirm the diagnosis and to assess the positive predictive value of an abnormal noninvasive test. Venography was performed according to the method of Rabinov and Paulin16. The criteria for a diagnosis of acute deep-vein thrombosis were the presence of an intraluminal filling defect on two different projections or nonfilling of a venous segment after repeated injection of contrast material. Venographic results were categorized as normal, indicative of proximal-vein thrombosis (thrombosis in the popliteal or more proximal veins or both, with or without concurrent thrombi in the veins of the calf), indicative of isolated calf-vein thrombosis, or inadequate for interpretation. All venograms were interpreted by a panel of experts without knowledge of the noninvasive test that the patient had undergone.
If the venogram was abnormal, anticoagulant therapy was instituted. If the venogram was normal, the patient was not treated and the abnormal noninvasive-test result was considered falsely abnormal. If the venogram was inadequate for interpretation, the patient was treated on the basis of the abnormal noninvasive-test result.
Patients with serially normal noninvasive-test results were considered not to have deep-vein thrombosis and were not treated with anticoagulants. To assess the accuracy of this conclusion, we determined the incidence of subsequent venous thromboembolism in these patients during long-term follow-up. For this reason, all these patients returned to the study centers at three and six months for clinical follow-up; no routine imaging tests were performed. Patients were asked to report to the study center immediately if signs or symptoms of venous thrombosis, pulmonary embolism, or both occurred. Venography was performed in all patients in whom new signs and symptoms of deep-vein thrombosis developed during the initial study period or during follow-up. Patients with clinically suspected pulmonary embolism were evaluated with perfusion-ventilation lung scanning. Lung scans were classified as normal or as being associated with either a high or any other probability of pulmonary embolism by a panel of three experts without knowledge of the patient's clinical condition or the type of noninvasive test that the patient had undergone.
An autopsy report was obtained, if available, for patients with repeatedly normal noninvasive tests who died during follow-up. If no autopsy was performed, the cause of death was determined by an independent physician.
Statistical Analysis
In four previous studies of symptomatic patients with normal serial impedance plethysmograms, the mean overall frequency of venous thromboembolism during long-term follow-up was approximately 1.5 percent4,5,6,7,8,17. Since it has been shown that the sensitivity of compression ultrasonography for the detection of venous thrombosis is at least as high as that of impedance plethysmography,9,11 we hypothesized that the incidence of subsequent venous thromboembolism in patients with normal serial ultrasonograms would at least equal that observed in patients with normal impedance plethysmograms. Therefore, we calculated that approximately 450 patients would be required in each group to provide sufficiently narrow confidence limits around the observed difference in the frequency of recurrent venous thromboembolism to conclude that the two methods have an essentially equal ability to exclude the presence of venous thrombosis.
Confidence intervals for the true incidence of recurrent venous thromboembolism and the positive predictive value were calculated from the binomial distribution. Exact confidence intervals for the difference in positive predictive values and in the incidence of recurrent venous thromboembolism between the two diagnostic groups were calculated according to the method of Thomas and Gart18. Fisher's exact test was used when appropriate. Two-sided P values of less than 0.05 were considered to indicate statistical significance.
Results
During the study period, 1500 consecutive outpatients with clinically suspected acute deep-vein thrombosis were seen at the three study centers. Of these patients, 398 (26.5 percent) were excluded for the following reasons: a previous episode of venous thrombosis in the same leg (197 patients; 49.5 percent), preexisting anticoagulant treatment (110 patients; 27.6 percent), geographic inaccessibility (42 patients; 10.6 percent), pregnancy (28 patients; 7 percent), a history of allergy to contrast material (14 patients; 3.5 percent), or an age of less than 18 years (7 patients; 1.8 percent). Thus, a total of 1102 patients were considered eligible for the study, of whom 117 (10.6 percent) declined to participate. Therefore, 985 patients entered the study: 494 were assigned to undergo serial impedance plethysmography, and 491 were assigned to serial ultrasonography.
The base-line clinical characteristics of the two study groups are shown in Table 1.
|
Impedance Plethysmography
Impedance plethysmography was not feasible in 4 of the 494 patients (0.8 percent) randomly assigned to undergo this procedure, because of massive swelling of the leg or an inability to position the leg properly. In all other patients the test could be performed, and the results were adequate for interpretation. An abnormal result either at presentation or on repeated testing was obtained in 125 of the 490 patients. Venography was performed in 109 of the patients (87 percent) who had abnormal results. Venography was not performed in eight patients because of an inability to puncture a foot vein, and it was not performed in another eight patients because of logistic problems. Of the venograms obtained, 89 (82 percent) showed acute deep-vein thrombosis, 18 (16 percent) were completely normal, and 2 (2 percent) were inadequate for interpretation. Thus, 89 of the 107 patients with abnormal impedance plethysmograms and interpretable venograms had deep-vein thrombosis (positive predictive value, 83 percent; 95 percent confidence interval, 75 to 90 percent). The distribution of the abnormal impedance plethysmograms (whose results were confirmed by venography) during the initial study week is shown in Table 2. In 14.6 percent of the 89 patients with abnormal impedance plethysmograms and documented venous thrombosis, thrombi were detected on repeated investigation. Only two patients had isolated calf-vein thrombi.
|
Ultrasonography
Compression ultrasonography was performed adequately in all but 1 of the 491 patients (0.2 percent); this patient had a large radiation ulcer in the groin, which made investigation of the common femoral vein impossible. Of the 490 patients, 100 had an abnormal ultrasonogram during the first study week, and 90 (90 percent) of these underwent venography. Venography was not performed in the other 10 patients for technical or logistic reasons. Of the venograms obtained, 84 (93 percent) showed acute deep-vein thrombosis, 5 (6 percent) were normal, and 1 (1 percent) was uninterpretable. Consequently, 84 of 89 patients with abnormal ultrasonograms and interpretable venograms had deep-vein thrombosis, yielding a positive predictive value of 94 percent (95 percent confidence interval, 87 to 98 percent). Table 2 shows the distribution of the abnormal ultrasonograms and abnormal venograms in the 84 patients during the initial study week. All these patients had proximal-vein thrombosis.
Symptoms suggestive of pulmonary embolism developed in one patient on the sixth study day. Perfusion lung scanning and ultrasonography showed no abnormalities. The patient was not treated with anticoagulants, and she completed the long-term follow-up.
Long-Term Follow-up
Impedance Plethysmography
Three hundred sixty-four patients had normal serial impedance plethysmograms. Of these patients, three were lost to follow-up (one patient moved abroad three months after randomization, at which time he was in good health; no information could be obtained on the other two patients). During the six-month period, a total of 8 patients died (all of diseases other than venous thromboembolism) (Table 3), and 19 patients (5.2 percent) had signs and symptoms suggestive of venous thromboembolism. Thirteen patients had recurrent leg symptoms, and six patients had chest symptoms. The diagnosis of venous thromboembolism was confirmed by objective tests in eight patients (2.2 percent): seven patients had acute deep-vein thrombosis on venography and one patient had a lung scan that suggested a high probability of pulmonary embolism. These episodes occurred 3 weeks (two patients), 4 weeks (one patient), 5 weeks (two patients), 6 weeks (one patient), 8 weeks (one patient), and 10 weeks (one patient) after randomization. In all the other patients, the diagnosis of venous thromboembolism was excluded and follow-up was uneventful. Hence, the overall incidence of venous thromboembolism during follow-up (including the pulmonary embolism that occurred during the initial study week) was 2.5 percent (nine patients; 95 percent confidence interval, 1.2 to 4.6 percent).
|
During the initial study period, 390 patients had repeatedly normal compression ultrasonograms; 5 of these patients did not complete the six-month follow-up (2 patients moved from the area, and 3 declined to come for follow-up visits). Seven deaths were recorded, all of which were unrelated to venous thromboembolism (Table 3). Symptoms of recurrent venous thrombosis or pulmonary embolism were encountered in 21 patients (5.4 percent), 6 of whom received a confirmed diagnosis of thromboembolism (1.5 percent; 95 percent confidence interval, 0.5 to 3.3 percent). Five patients had evidence of acute deep-vein thrombosis on venography; one patient had a lung scan indicating a high probability of pulmonary embolism. These thromboembolic events occurred 3 weeks (one patient), 4 weeks (one patient), 6 weeks (one patient), and 12 weeks (three patients) after randomization. In the remaining 15 patients, objective testing indicated that venous thromboembolism was not present, and none of these patients had new symptoms of venous thromboembolism during follow-up.
Comparison of Impedance Plethysmography with Ultrasonography
The probability that venous thrombosis would be confirmed by venography in the event of an abnormal impedance plethysmogram or ultrasonogram was 83 percent and 94 percent, respectively; this difference in the positive predictive values was statistically significant (P = 0.02) (Table 4). During follow-up, 9 of 364 patients in the impedance-plethysmography group (2.5 percent), as compared with 6 of 390 patients in the ultrasonography group (1.5 percent), had documented venous thromboembolism. Thus, there was an observed difference of 1.0 percent in the incidence of venous thromboembolism between the two groups in favor of the ultrasonography group (P>0.2). Therefore, it is unlikely (P<0.05) that the true difference would be more than 1.3 percent in favor of the impedance-plethysmography group or more than 2.7 percent in favor of the ultrasonography group.
|
In the present study of a series of consecutive outpatients with clinically suspected deep-vein thrombosis, we randomly compared compression ultrasonography with the extensively evaluated noninvasive reference method, impedance plethysmography4,5,6,7,8,19. We used the positive predictive value (i.e., the confirmation by contrast venography of an abnormal noninvasive-test result) as an index of accuracy for the detection of deep-vein thrombosis. The direct comparison of these values revealed that serial ultrasonography is superior to serial impedance plethysmography in correctly diagnosing venous thrombi (positive predictive value, 94 and 83 percent, respectively; P = 0.02). Since anticoagulant therapy is usually instituted in all patients with an abnormal noninvasive-test result, this difference is clinically relevant, because the likelihood of unnecessary treatment is almost three times higher when a strategy involving serial impedance plethysmography is employed. It should be noted that the observed positive predictive value of impedance plethysmography in our study was 5 to 10 percent lower than that reported previously6,7. This difference may be explained by the slight decrease in the prevalence of deep-vein thrombosis in the study patients. On the other hand, the positive predictive value of ultrasonography in the patients studied is very high and consistent with the results of earlier studies9,10,11,12.
We also assessed the outcome of long-term clinical follow-up in patients with serially normal noninvasive-test results (in whom anticoagulant treatment was withheld) as a measure of the accuracy with which the diagnosis was excluded. The incidences of subsequent venous thromboembolism in the two groups during six months of follow-up were similar (1.5 percent in the ultrasonography group and 2.5 percent in the impedance plethysmography group), and they agree with earlier observations in similar groups of patients who underwent serial impedance plethysmography5,6,7. All episodes were nonfatal and, in the great majority of patients in both groups, they consisted of venous thrombosis of the legs.
Which diagnostic strategies can be recommended for patients with clinically suspected acute venous thrombosis? As a matter of course, ascending contrast venography can be used,20 but this technique has the disadvantages of being invasive, being expensive, and requiring access to hospitals21. Alternatively, noninvasive methods can be employed, and the results of our randomized study indicate that both serial impedance plethysmography and serial ultrasonography are equally effective in excluding the possibility of venous thrombosis, but that ultrasonography is to be preferred in view of its higher clinical usefulness in confirming the presence of venous thrombosis. An advantage of ultrasonography is its universal availability in hospitals, although the costs of performing ultrasonography are approximately twice those of performing impedance plethysmography22.
Considering the documented high sensitivity and specificity of the simple technique of compression ultrasonography (of the popliteal and common femoral veins only) in symptomatic patients, it is questionable whether the addition of Doppler ultrasonography (duplex and color-coded) will improve the diagnostic accuracy. It is possible that these technical advances may be less cost effective and may decrease specificity, since flow artifacts may be mistaken for venous thrombi. It should be emphasized, as has been advocated before,12 that compression ultrasonography needs to be repeated over the course of a week, since our findings show that 6 percent of the cases of deep-vein thrombosis were detected during serial testing.
The results of this comparative study allow us to conclude that serial compression ultrasonography is preferable to serial impedance plethysmography for the diagnosis of clinically suspected deep-vein thrombosis in outpatients.
Supported by a grant (OG 89-116) from the Ziekenfondsraad, Amstelveen, the Netherlands. Dr. Buller is the recipient of a fellowship from the Royal Netherlands Academy of Arts and Sciences.
We are indebted to Mrs. H. Jagt and Mrs. J. Johnson for their assistance in collecting the clinical data; to Mrs. C. Ballard, Mrs. L. Degrow, Mrs. M. Glotzbach, Mrs. M. Met, Mrs. L. Newkirk, Mrs. S. Smith, Mrs. E. Spekschneider, and Mrs. M. Wieringa for their skilled technical assistance; and to Dr. A.J. Bakker, Dr. D. Batchelor, and the radiologists of the Hamilton General Hospital for their skilled performance of the venographic procedures.
Source Information
From the Center for Hemostasis, Thrombosis, Atherosclerosis and Inflammation Research, Academic Medical Center, Amsterdam (H.H., H.R.B., A.W.A.L., J.W.C.); the Hamilton Civic Hospitals General Division, Hamilton General Hospital-McMaster Clinic, Hamilton, Ont., Canada (A.W.A.L., A.G.G.T.); and the Thrombosis Service, Amsterdam (L.P.C.).
Address reprint requests to Dr. Buller at the Center for Hemostasis, Thrombosis, Atherosclerosis and Inflammation Research, Academic Medical Center, Meibergdreef 9, 1105 AZ Amsterdam Zuidoost, the Netherlands.
References
| |||||||||||||||||||||||||||||||||||||||
Related Letters:
The Diagnosis of Deep-Vein Thrombosis
Moser K. M., Fedullo P. F., Heijboer H., Buller H. R., ten Cate J. W.
Extract |
Full Text
N Engl J Med 1994;
330:863-864, Mar 24, 1994.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
