- Email: [email protected]
Venous duplex imaging: Should it replace hemodynamic tests for deep venous thrombosis?
Venous duplex imaging: Should it replace hemodynamic tests for deep venous thrombosis? A n t h o n y J. Comerota, M D , FACS, Mira L. Katz, BS, RVT, L o r i L. Greenwald, M D , Eric Leefmans, M D , Michael Czeredarczuk, BA, RVT, and J o h n V. White, M_D, FACS, Philadelphia, Pa. Noninvasive diagnosis of deep venous thrombosis has traditionally relied on detection of alterations in venous hemodynamics. Although phleborheography is among the most sensitive tests, it is inadequate for diagnosing infrapopliteal and nonocclusive proximal thrombi and for surveillance of patients at high risk for deep venous thrombosis. Venous duplex imaging is a new technique being rapidly accepted, however, without the same critical analysis given to previous diagnostic modalities. The purpose of this study is to evaluate the diagnostic acumen of venous duplex imaging compared to phleborheography and ascending phlebography in two distinct patient groups, and to determine whether patient selection, and thus the location or magnitude of thrombi have significant influence on these diagnostic tests. One hundred ten extremities in 103 patients were prospectively evaluated with venous duplex imaging, phleborheography, and ascending phlebography within the same 24-hour period. Patients were categorized into one of two groups: Diagnostic--patients evaluated because of clinical suspicion of acute deep venous thrombosis; and Surveillance--patients at high risk of postoperative deep venous thrombosis after total joint replacement, but not symptomatic. Patients in the diagnostic group had a greater frequency of deep venous thrombosis (p < 0.001) and significantly more occluding above-knee thrombi (p ,= 0.054) compared to those in the surveillance group. Phleborheography detected 73% (27/37) of above-knee thrombi in the diagnostic group compared to 29% (2/7) in the surveillance group (p = 0.036). This difference was not noted with venous duplex imaging, which detected 100% of above-knee thrombi in both diagnostic and surveiUance groups and 78% (7/9) of aU below-knee thrombi. Venous duplex imaging was significantly more sensitive than phleborheography for detecting both above-knee and below-knee thrombi (p < 0.001) and importantly, the negative predictive value of venous duplex imaging was significantly higher than phleborheography (p < 0.001). Based on these observations it appears that venous duplex imaging offers superior noninvasive, diagnostic acumen without the limitations of traditional hemodynamic techniques. Venous duplex imaging reliably diagnosed nonoeclusive and infrapopliteal thrombi, thereby offering superior epidemiologic and natural history capability in patients with acute deep venous thrombosis. In experienced vascular laboratories venous duplex imaging may be more sensitive than ascending phlebography and may become the new diagnostic standard for deep venous thrombosis. (J VASC StSRG 1990;11:53-61.)
The noninvasive diagnosis of deep venous thrombosis (DVT) has assumed an important place in the clinical evaluation of patients with venous thromFrom the Sectionof VascularSurgery,VascularLaboratory,Temple UniversitySchool of Medicine. Supported in part by grants from the U.S. Public Health Service, research grant No. 2 MO1 RR00349, National Institutes of Health, GeneralClinicalResearchCenters Branch. Presented at the Thirty-seventhScientificMeeting of the North American Chapter, International Society for Cardiovascular Surgery, New York, N.Y., June 19-20, 1989. Reprint requests: AnthonyJ. Comerota,MD, FACS, Sectionof Vascular Surgery, Temple University School of Medicine, s3road & Ontario St., Philadelphia,PA 19140. 24/6/16342
boembolic disorders. Many investigators have reported excellent clinical utility for hemodynamic tests for DVT. ls Others have documented high falsenegative and false-positive rates, 68 casting doubt on the uniformity of the diagnostic reliability of these tests. A prospective evaluation was subsequently performed, which demonstrated that patient selection was critically important when assessing the diagnostic capability ofhemodynamic tests for DVT. 9 In that study patients evaluated because of a clinical suspicion of DVT had high diagnostic sensitivity for proximal vein thrombosis, 83% and 92% for impedance plethysmography (IPG) and phleborheography 53
54
Journal of VASCULAR SURGERY
Comerota et al.
Table I. Venous duplex imaging: Diagnostic criteria for DVT B-mode image
Table II. Phlebographic details of thrombi according to patient group
Doppler signal
Intralttminal echogenic material Absent intraluminal signal Continuous signal (no respiNoncompressibility of vein ratory phasisity) Blood flow around a lucent fill- Reduced or absent augmentaing defect tion maneuvers Loss of venodilation with Valsalva maneuver
(PRG), respectively. However, when patients were studied because of a high risk for postoperative DVT without overt clinical suspicion, the sensitivity for diagnosing proximal vein thrombosis dropped to 32% and 33%. Venous duplex imaging (VDI) is rapidly becoming accepted as an important noninvasive diagnostic test acute DVT. Although high ~diagnostic sen. . . . for :• slnvmes and specificities have been reported, 1°1z a prospective analysis evaluating whether patient selection affects the results of VDI has not been performed. Therefore the question remains whether patient selection, and thus the location and/or magnitude of thrombi, has significant influence on the diagnostic capability of VDI. The purpose of this study is to determine the diagnostic reliability of VDI compared to PRG and to evaluate the influence of patient selection on both techniques. METHODS
One hundred ten extremities in 103 patients were prospectively evaluated with VDI, PRG, and ascending phlebography within the same 24-hour period. Phleborheography was performed with the Cranley-Grass phleborheograph (Grass, Inc., Quincy, Mass.), and VDI was performed with the Biosound 2000IISA (Biosound, Inc., Indianapolis, Ind.) duplex imaging system. The technique of examination and method of interpretation for each test has been previously described. 2a° Table I lists the diagnostic criteria used for VDI. The vena cava and iliac veins were not visualized during this examination with the transducers used, unless the patient had very little body fat. Both noninvasive tests were performed at the same time, and with few exceptions, before phlebography. Ascending phlebography was performed according to the modified technique of Rabinov and Paulin. TM Patients who had unsuccessful or inadequate phlebography were excluded. Additionally, patients
Diagnostic (N = 72) Incidence Normal DVT Distribution AK BK Magnitude Occluding Nonocduding
Surveillance (N = 38)
p-value
39% (28•72) 61% (44172)
vs 76% (29/38) vs 24% (9•38)
<0.001 --
84% (37/44) 16% (7/44)
vs 78% (7/9) vs 22% (2/9)
NS --
81% (30/37) 19% (7/37)
vs 43% (3/7) vs 57% (4/7)
0.054
having phlebography during weekends, holidays, and those evaluated as outpatients were frequently unavailable for noninvasive testing. Also, as previously mentioned, patients having their noninvasive tests more than 24 hours before phlebography were~xeluded from analysis. Overall, somewhat more than twice the number ofphlebograms were required durhag the period of this study to capture the patients reported herein. Phlebograms were interpreted as normal, below-knee (BK)--clot isolated to the infrapopliteal veins, or above-knee (AK)--clot involving the popliteal and/or more proximal veins. The methods of patient selection for noninvasive testing and ascending phlebography varied between the two following distinct patient populations: (1) Diagnostic group: patients evaluated because of a clinical suspicion of acute DVT. Phlebography was selectively obtained at the discretion of the referring physician and was not necessarily predicated on the results of a noninvasive test. (2) Surveillance group: patients undergoing orthopedic surgery who were at high risk for postoperative DVT after total joint~eplacement. These patients had routine postoperative ascending phlebography as part of an ongoing surveillance program. Noninvasive testing was not'routinely requested, but was performed as an adjunct to the surveillance program after patient consent. Noninvasive studies were performed during the morning of the sixth or seventh postoperative day, with the phlebogram obtained during the afternoon of the same day. Statistical analysis Ascending phlebography was used as the reference standard for both PRG and VDI, and 2 × 2 contingency tables were made with the results of each of the noninvasive tests. Statistical differences between diagnostic tests were compared by X2 with the Fisher exact test.
Volume 11 Number 1 flanuary 1990
Venous duplex imaging 55
Table HI. Reliability of noninvasive tests compared to phlebography Total PRG Sensitivity All AK BK Positive predictive value Specificity Negative predictive value VDI Sensitivity All AK BK Positive predictive value Specificity Negative predictive value
Diagnostic
55% 66% 0% 83% 89% 68%
(29/53) (29/44) (0/9) (29/35) (51/57) (51/75)
61% 73% 0% 82% 79% 56%
(27/44) (27/37)
96% 100% 78% 96% 93% 96%
(51/53) (44/44) (7/9) (51/53) (53/57) (53/55)
98% 100% 86% 96% 86% 96%
_RESULTS
C~' 110 extremities undergoing ascending phlebography, 48% (53/110) had demonstrable thrombi, and 52% (57/110) had normal phlebography results. Eighty-three percent (44/53) of patients with thrombi had clot extending into the pop,~titealor more proximal veins (AK), and two patients had isolated iliac vein occlusion. Seventeen percent (9/53) had isolated infrapopliteal venous thrombosis ~BK). The phlebographic details of thrombi by patient group are listed in Table II. The incidence and magnitude of thrombi were significantly different between the diagnostic and surveillance groups, whereas the distribution of thrombi was similar. The ~verall incidence of DVT was 61% (44/72) in the diagnostic group and 24% (9/38) in the surveillance group (p < 0,001). Although the distribution of ~,hrtmbi was similar between groups the magnitude was different, with 81% (30/37) of AK thrombi occluding veins of diagnostic patients compared to ¢~3% (3/7) of surveillance patients (p = 0.054). The reliability of the PRG in the diagnostic and surveillance groups is listed in Table III. There is a ,significant difference in the ability of the PRG to detect AK thrombi in diagnostic compared to patients in the surveillance group (p = 0.036). There #s a strong trend toward similar findings in all patients with DVT in the diagnostic compared to the surveillance group (p = 0.06). The specificity and negative predictive value are significantly better in patients in the surveillance group than in patients in the diagnostic group. - Table III summarizes the reliability of VDI in patients in the diagnostic group compared to patients m,,lle surveillance group. Unlike hemodynamlc stud-
Surveillance
p-value
(27/33) (22/28) (22/39)
vs vs vs vs vs vs
22% 29% 0% 100% 100% 81%
(2/9) (2/7) (0/2) (2/2) (29/29) (29/36)
0.06 0.036 NS NS 0.01 0.029
(43/44) (37/37) (6/7) (43/45) (24/28) (24/25)
vs vs vs vs vs vs
89% 100% 50% 100% 100% 97%
(8/9) (7/7) (1/2) (8/8) (29/29) (29/30)
NS NS NS NS 0.051 NS
(0/7)
ies, there is no difference in its sensitivity to detect overall, AK, or BK thrombi. There is no difference between positive and negative predictive value, however, the specificity is improved in the surveillance compared to the diagnostic group, and this difference appears statistically significant (p = 0.051). The comparative results of VDI and PRG are listed in Table IV. The overall sensitivity of VDI is 96% (51/53) compared to 55% (29/53) for PRG, (p < 0.001). Venous duplex imaging detected 100% (44/44) of AK thrombi compared to 66% (29/44) detected by PRG (p < 0.001). It is interesting to note that VDI detected 78% (7/9) of BK thrombi compared to 0% (0/9) detected by PRG (p < 0.001). The positive predictive value and specificity between the two noninvasive tests were not different, although the negative predictive value was significandy better for VDI. The comparative results of VDI and PRG for patients in the diagnostic group are listed in Table III. The sensitivities are significantly better for VDI for all patients with venous thrombosis, those with AK thrombi, and those with BK thrombi. The negative predictive value is also significantly better for VDI compared to PKG. Venous duplex imaging and PRG compared to ascending phlebography for patients in the surveillance group are summarized in Table IV. The overall sensitivity and sensitivity for detecting AK thrombi are significantly different between the two noninvasive studies. Below-knee thrombi were infrequent, and therefore conclusions cannot be drawn about the diagnostic sensitivity of VDI for BK thrombi in patients in the surveillance group. The positive predictive value, specificity, and negative prcdictive value were not different between the two tests.
Journal of VASCULAR SURGERY
56 Comerota et al.
Table IV. Venous duplex imaging and phleborheography compared to ascending phlebography VDI All patients Sensitivity All AK BK J~ Positive predictive value Specificity Negative predictive value Diagnostic patients Sensitivity All AK BK Positive predictive value Specificity Negative predictive value Surveillance patients Sensitivity All AK BK Positive predictive value Specificity Negative predictive value
PR G
p-value
96% 100% 78% 93% 93% 96%
(51/53) (44/44) (7/9) (51/55) (53/57) (53/55)
55% 66% 0% 83% 89% 68%
(29/53) (29/44) (0/9) (29/35) (51/57) (51/75)
p < 0.001 p < 0.001 NS NS NS p < 0.001
98% 100% 86% 91% 86% 96%
(43/44) (37/37) (6/7) (43/47) (24/28) (24/25)
61% 73% 0% 82% 79% 56%
(27/44) (27/37) (0/7) (27/33) (22/28) (22/39)
p < 0.001 p < 0.01 p < 0.01 NS NS p < 0.001
89% 100% 50% 100% 100% 97%
(8/9) (7•7) (1/2) (8/8) (29/29) (29/30)
22% 29% 0% 100% 100% 81%
(2/9) (2/7) (0/2) (2/2) (29/29) (29/36)
p < 0.05 p < 0.05 NS NS NS NS
Both PRG and VDI appropriately identified the two patients with iliac vein occlusion. Although results of the B-mode image were negative, results of the common femoral Doppler examination were abnormal, indicating a proximal obstruction. DISCUSSION
Patients at high risk for DVT (surveillance group) have previously presented a greater diagnostic challenge compared to those with signs and symptoms of acute DVT. 9 Ascending phlebography was considered the only accurate diagnostic method for evaluating these patients~ Since these patients had a high incidence of isolated infrapopliteal thrombi and nonocclusive thrombi, hemodynamic studies failed to identify clot in these patients. The PRG was shown to be a more sensitive test than IPG for detecting deep venous obstruction in a previous comparison, 9 therefore we chose the PRG as the best hemodynamic test (in our laboratory) for comparison to VDI. The PRG results reported herein are similar to our previous findingsg; however, there appears to be a trend toward a lower sensitivity for AK thrombi in patients in the diagnostic group. This is most likely the result of identification of patients with nonocclusive thrombi by VDI but who had normal PRGs, who were then evaluated by phlebography to resolve the discrepancy. In the past, most patients with symptoms in their legs and negative hemodynamic test results would not have had phlebography performed,
which seems to be the practice in many reputable laboratories, s-s Although we frequently use venous Doppler to complement other noninvasive tests to obtain addi4 tional information, we rarely use venous Doppler alone. Results of venous Doppler used alone vary widely. 1922 It is entirely dependent on the skill and,experience of the examiner, and excellent results obtained by one individual cannot be easily duplicated by other laboratories. We found that when combine4 with the B-mode image, venous Doppler was appropriately positive when patients had occlusive AK thrombi, whereas those with BK and infrapop~eak thrombi usually did not have abnormal Doppler examination results. When results of a common femoral or superficial femoral Doppler examination areclearly abnormal without any visible thrombi or unequivocally abnormal compression test, then suprainguinal thrombus must be suspected. Unlike hemodynamic studies, VDI has similar diagnostic capability for patients in the diagnostic and surveillance groups. It is interesting to note tha~ VDI has a high diagnostic sensitivity for BK as well as AK thrombi. Venous duplex imaging diagnosed 78% (7/9) of all BK thrombi and 86% (6/7) of BK thrombi in the diagnostic group. Although there were too few isolated infrapopliteal thrombi in the surveillance group to draw conclusions about the sen~ sitivity of VDI, it is our opinion that it would be equal to the diagnostic group. It has previously b ~ n
Volume 11 Number 1
Venous duplex imaging 57
lanuary1990
Fig. 1. B-mode image of paired popliteal veins adjacent to popliteal artery without probe pressure.
Fig. 2. Compression of both veins is complete without compressionof artery, indicating that both veins are patent.
Table V. Studies comparing noninvasive hemodynamic tests to venous imaging for acute DVT Sensitivity Author
No.
Rosner et al. 15 1988 O'Leary et al. 16 1988 Patterson et al} 7 1989 Present series
25 37 49 110
IPG / P R G
100% 78% 75% 55%
(6/6) (14/18) (15/20) (29/53)
Specificity VDI
90% 88% 89% 96%
(9/10) (22/25) (24/27) (51/53)
IPG / P R G
63% 84% 45% 89%
(12/19) (16/19) (13/29) (51/57)
VDI
100% 96% 91% 93%
(22/22) (24/25) (31/34) (53/57)
In each previously published series, there were 31% to 32% more VDI studies reported than hemodynamic tests.
reported that real-time B-mode ultrasonography was inaccurate for detecting infrapopliteal venous thrombin; however, those investigators used the compression test (Figs. 1 and 2) as the sole criterion for "~agnosis. The additional information provided by the B-mode image itself has been most helpful in our experience. Although routinely the Doppler signal was not as useful as the image for the diagnosis ofBK thrombi, its use in equivocal examinations may be helpful, particularly the augmentation maneuver. The Doppler signal was most important for diagnosing the two patients with isolated iliac vein obstruction. Perhaps more importantly, the diligence and expertise of the technologist performing the test and the physician interpreting the test must be considered. Our data fit well with those published by Elias et al}4 who reported 854 limbs studied with realtime B-mode ultrasonography and phlebography for acute DVT. Real-time B-mode imaging detected '98% (325/333) of thrombi with 100% of AK and 9"% of BK thrombi accurately diagnosed.
Three previously published reports compared noninvasive hemodynamic tests for acute DVT to duplex imaging of the deep venous system (Table V).15-17 In general, the previously reported data agree with the data reported herein; however, several differences must be addressed. Rosner et al2s reported I00% sensitivity for IPG; whereas VDI missed a thrombus in a duplicated popliteal vein. The patient responsible for the false-negative VDI was not studied with IPG. O'Leary et al}6 report a 78% (14/18) sensitivity for hemodynamic tests; however, they excluded 15% (eight) of their patients because of indeterminant results. If these eight patients were ineluded as errors, the overall accuracy of their hemodynamic tests would drop to 68%. Although each of these three studies evaluates hemodynamic tests compared to VDI, each of these authors reports at least 30% more patients studied with VDI than hemodynamic studies. Therefore a true comparison is not possible. None of the three previous reports clarified the diagnostic accuracy of VDI for infrapopliteal
58
Comerota et al.
Journal of VASCULAR SURGERY
Fig. 3. B-mode image indicating partially occluding thrombus of proximal superficial femoral vein (double arrow). Thrombus is located within a valve cusp; whereas the opposite valve cusp fimctioned normally (single arrow). Fig. 5. B-mode image during compression shows ~[[~ldl¥ echogenic material within the superficial femoral vein (IT) and that it does not compress the probe pressure. Partially occluding thrombus of the superficial femoral vein was diagnosed. The superficial femoral artery (A) appears normal.
Fig. 4. Ascending phlebogram shows partially occluding thrombus in proximal superficial femoral vein within a valve cusp (double arrow). Doppler examination results were normal. thrombi. Two reports 16'17 identified isolated infrapopliteal thrombi as causes of false-negative examination results; however, both failed to differentiate the sensitivity of VDI for proximal versus infrapopliteal thrombi. Whether nonocclusive or infrapopliteal venous thrombi need to be diagnosed is a separate question.
Fig. 6. Ascending phlebogram of same patient shows contrast defect in the same location (arrow); however, the phlebogram was officially interpreted as normal, with the area in question read as a "flow defect from a side branch." However, it is important, to understand that in experienced hands, VDI, in contrast to hemodynamic tests, has the capability o f diagnosing nonocclus~e (Figs. 3 and 4) as well as infrapopliteal thrombi. Another question that might arise is why the pos-
Volume 11 Number 1 January 19.90
itive predictive value is not better than 9 3 % (51 / 5 5) for VDI. It is our opinion that VDI detected nonocclusive thrombi that were missed by ascending phlebography (Figs. 5 and 6). This raises yet another issue, which test, VDI or phlebography, is the true standard for the diagnosis of acute DVT. We believe it is premature to answer this question, and it probably should not be answered inasmuch as VDI and phlebography will agree in most patients. Discrepancies that arise between test results should evoke a critical evaluation of each test, with particular attention given to the quality of the images available. Combining the information at hand with sound judgement will lead to the proper conclusions in most cases. I n light o f t h e d a t a available, t h e answer to t h e q u e s t i o n o f w h e t h e r V D I s h o u l d replace h e m o d y n a m i c tests for D V T s h o u l d be yes.
RI~ERENCES 1. Hull RD, Hirsh J, Carter CJ, et al. Diagnostic efficacy of impedance plethysmography for clinicallysuspected deep vein thrombosis. A randomized trial. Ann Intern Meal 1985; 102:21-8. 2. Comerota AJ, Cranley JJ, Cook SE, Sipple P. Phleborheography--results of a 10-year experience. Surgery 1982;91: 573-81. 3. Wheeler HB, Anderson FA Jr. The diagnosis of venous thrombosis by impedance plethysmography. In: Bernstein EF, ed. Noninvasive diagnostic techniques in vascular disease, 3rd ed. St. Louis: CV Mosby Co, 1985:755. 4. Stallworth JM, Plonk GW, Home JB. Negative phleborheography, clinical follow-up in 593 patients. Arch Surg 1981;116:795-7. 5. Huisman MV, Buller HR, Wouter Ten Care JV, Vreeken J. Serial impedance plethysmography for suspected deep venous thrombosis in outpatients: the Amsterdam general ,practitioner study. N Engl J Med 1986;314:823-8. 6. Ramchandani P, Soulen RL, Fedullo LM, Gaines VD. Deep vein thrombosis: significant limitations of noninvasive tests. Radiology 1985;156:47-9. 7. Vaccaro P, Van Aman M, Miller S, Fochman J, Smead WL. Shortcomings of physical examination and impedance plethysmography in the diagnosis of lower extremity deep venous thrombosis. Angiology 1987;38:232-5.
DISCUSSION Dr. David S. S u m n e r (Springfield, Ill.). Once again, Dr. Comerota and his associates have demonstrated that plethysmographic tests have insufficient sensitivity to detect venous thrombi in surveillance groups. This comes as no surprise, since many o f these developing clots are nonocclusive or infrapopliteal and offer little or no impediment j venous blood flow. However, I am puzzled by the diminishing accuracy of the PKG in the hands of the same investigators, u s i n g - -
Venous duplex imaging 59 8. Young AE, Henderson BS, Phillips DA, Couch NP. Impedance plethysmography: its limitations as a substitute for phlebography. Cardiovasc Radiol 1978;1:233-9. 9. Comerota AJ, Katz ML, Grossi RJ, et al. The comparative value of noninvasive testing for diagnosis and surveillance of deep vein thrombosis. J VAsc SUWG1988;7:40-9. 10. Comerota AJ, Katz ML. The diagnosis of acute deep venous thrombosis by duplex venous imaging. Semin Vase Surg 1988;1:32-9. 11. Lensing AWA, Prandoni P, Brandjes D, et al. Detection of deep vein thrombosis by real-time B-mode ultrasonography. N Engl J Med 1989;320:342-5. 12. Langsfeld M, Hershey FB, Thorpe L, et al. Duplex B-mode imaging for the diagnosis of deep venous thrombosis. Arch Surg 1987;122:587-91. 13. Karkow WS, RuoffBA, Cranley JJ. B-mode venous imaging. In: Kempczinski RF, Yao JST, eds. Practical noninvasive vascular diagnosis. Chicago: Year Book Medical, 1987:464-85. 14. Elias A, Le Corff G, Bouvier JL, Benichou M, Serradimigni A. Value of real-time b-mode ultrasound imaging in the diagnosis of deep vein thrombosis of the lower limbs. Int Angiol 1987;6:175-82. 15. Rosner Nit, Doris PE. Diagnosis of femoropopliteal venous thrombosis: comparison of duplex sonography and plethysmography. AJR 1988;150:623-7. 16. O'Leary DH, Kane RA, Chase BM. A prospective study of the efficacy of B-scan sonography in the detection of deep venous thrombosis of the lower extremities. J Clin Ukrasound 1988; 16:1-8. 17. Patterson RB, Fowl RJ, Keller JD, Shomaker W, Kempczinski RF. The limitations of impedance plethysmography in the diagnosis of acute deep venous thrombosis. J VASe SURG 1989;9:725-30. 18. Rabinov D, Paulin S. Roentgen diagnosis of venous thrombosis of the leg. Arch Surg 1972;110:1336-42: 19. Bolton JP, Hoffman VJ. Incidence of early postoperative iliofemoral thrombosis. Br Med J 1975;1:247-9. 20. Milne RM, Gunn AA, Griffiths JMT, Ruckley DV. Postoperative deep venous thrombosis. A comparison of diagnostic techniques. Lancet 1971;2:445-7. 21. Barnes RW i Russell HE, Wilson MR. Doppler ultrasonic evaluation of venous disease, a programmed audiovisual instruction. Iowa City: University of Iowa: 1975. 22. Sumner D. Diagnosis of deep venous thrombosis by Doppler ultrasound. In: Nicolaides AN, Yao JST, eds. Investigation of vascular disorders. New York: Churchill Livingstone, 1981:377-402.
as far as I can determine--similar techniques on a similar "diagnostiC' population. For example, in 1982 Dr. Comerota reported a sensitivity o f 92%, a specificity o f 95%, a positive predictive value of91%, and a negative predictive value o f 95%, for the P R G when used to diagnose DVT in 709 patients. His current dismal results stand in stark contrast to these previous values. Moreover, in the 1982 study the PRG detected 45 o f 57 isolated calf vein thrombi (a sensitivity o f 79%). N o w it detects none o f seven! What has changed? Is it the disease, the testing procedure, or the
60
Comerota et al.
method of reporting? Now we hear that the duplex scan is capable of accuracies similar to those previously reported with the PRG. Seven years from now will the duplex scan also fall from grace? In the not too distant past, numerous reputable investigators established the safety of using both the IPG and the PRG for ruling out DVT in patients suspected of having this disease. Are these laboratories now obligated to discard these tests? Despite the growing enthusiasm for VDI, the approach and the criteria used are so different that it can hardly be assumed that results obtained in one laboratory are equivalent to those from another. Some rely on compression alone and do not bother with the infrapopliteal veins; others emphasize the visualization of clots; and still others depend most heavily on the Doppler signal. Do we have enough information to standardize the test? I think not; yet, we are now hearing that duplex scanning should replace phlebography as the gold standard. Can a technologist realistically survey inch by inch the yards of veins in each leg? On the other hand, is it really important in the clinical setting that every isolated thrombus be detected? I agree with Dr. Comerota that duplex scanning is the current method of choice; but I predict that it will soon be replaced by color-flow imaging, which has several distinct advantages. It facilitates identification of veins, especially below the knee, provides simultaneous Doppler and B-mode information, and reduces the time required for each examination. Dr. Anthony Comerota. I thank Dr. Sumner for his comments. Dr. Sumner appropriately pointed to differences in the results. The differences he quoted from 1982 were essentially the results obtained in Cincinnati. When we moved to Philadelphia we noticed that our results with PRG were different from those we previously reported in Cincinnati and the IPG results reported from Worcester and McMaster. Therefore, the study we presented 2 years ago was performed, which paralleled the current data; however, Dr. Sumner failed to refer to those corroborating data. Clearly, it was patient selection that was responsible for those differences, and it was patient selection for phlebography that was most important. I think we made those points 2 years ago. Since our previous presentation, the audience might notice a difference in sensitivity for proximal thrombi in the diagnostic group. That also appears to be due to the selection of patients for phlebography. Both V D I and PRG were performed in patients suspected of DVT. When the venous duplex image identified a dot and the PRG was negative, ascending phlebography was strongly encouraged and often performed, thereby selecting patients in whom the PRG is likely to be falsely negative and reducing the sensitivity of the PRG in the diagnostic group. I do not believe that vascular laboratories should discard hemodynamic tests, at least not until they are facile with VDI. I would agree with Dr. Sumner that there is a
Journal of VASCULAR SURGERY
marked variability in technique as well as criteria for interpretation, and these criteria have not been standardized for venous duplex. However, V D I is more than just a compression test. We believe the image is important. The Doppler gives an added dimension and is particularly useful for proximal venous thrombi. I believe color incorporates anatomic information with physiologic information. I also believe it will speed the examination for the uninitiated and the less experienced. However, for those who are experienced with VDI, it is my impression that color will probably be less important. Dr. H. B r o w n e r Wheeler (Worcester, Mass.). Our hospital has experienced a dramatic increase in B-mode scanning, but we do more hemodynamic screening tests than ever. Logistically, we simply could not screen all patients suspected of DVT with B-mode scans. There remains an important role for simple low-cost bedside procedures. Dr. Comerota, how do you explain your low sensitivity with PRG for symptomatic proximal D V T when previous studies have reported nearly 100% sensitivity? Our,~wn accuracy for IPG would also be much less than earlier studies, if based on current venograms, simply because the original validation studies included venograms on all patients suspected of DVT, whereas now venograms are done on only about 5%--primarily problem patients. May similar bias in the selection of patients for venography affect your results? Also, if PRG fails to detect significant thrombi, how do you explain the excellent results of clinical management with PRG alone? Stallworth reported 593 patients with negative PRGs who were not given anticoagulants and had only a 0.2% incidence of nonfatal pulmonary embolism. Similar outcomes in untreated patients with negative IPGs have been reported by six different medical centers. If IPG and PRG fail to detect many thrombi, most of these thrombi must be relatively innocuous, since patient outcome is the ultimate gold standard for the value of any procedure, j% Dr. John J. Cranley (Cincinnati, Ohio). From 1971 to 1981 phleborheography was the main diagnostic test for deep venous thrombosis at Cincinnati's Good Samaritan Hospital. The data on which Dr. Comerota's 1982 report was based consisted of 709 consecutive phlebograms on extremities also studied by phleborheography. No patient having both tests was excluded. Dr. Comerota's recognition of the difference between patient populations studied for diagnosis or surveillance is correct and a significant step forward. All of our phlebograms were obtained for diagnosis, none for surveillance. As Dr. Porter has emphasized, comparison of results obtained in different eras is hazardous. Our conceptual changes alone are of significance here. When we initiated our studies with the PRG in 1971 we conceptualized D V T as a thrombus occluding mainstream veins. We were not seeking, and stressed the fact that we could not detect, clots outside the mainstream. Thus, partially occluding thrombi, thrombi in valve cusps, soleal veins, or small tributaries-~"
Volume 11 Number 1 January 1990
Venous duplex imaging 61
the popliteal or femoral vein systems detectable today, were ~ndetectable then. The phiebographic methods used in the two centers differed. At Temple University, the method of Rabinov ~ n d Paulin is used. This method permits flooding of the leg veins passively in the motionless limb. It provides more complete visualization of the muscular veins of the leg than ~he method we use, but as Comerota reported, it is less accurate in the groin and pelvis. At Good Samaritan Hospital, we used five14 × 51 inch cassettes, each containing three 14 × 17 inch films. This permitted visualization of ~11 the mainstream veins from the ankle to the vena cava, which was what I demanded. Finally, although we employed many technicians, we [~ad one senior technician who devoted her entire career to performing and teaching the PRG. The amount of time and maneuvers performed did not matter, as long as the result was clearcut. Such concentration on one test would T
be difficult today in a busy laboratory performing many tests. Indeed, without this devoted technician we might not be able today to repeat the results obtained a decade ago. But I am certain that the data Dr. Comerota presented in 1982 were correct. Dr. Comerota. In answer to Dr. Wheeler's second question, I believe the false-negative results of hemodynamic tests will certainly be diluted with the many truenegative examinations performed by most laboratories, and therefore the poor outcome in some of those will result in a relatively small percentage. I am sure that many of the false-negative tests might well remain silent over the long term. I do not necessarily agree that the ultimate outcome is the definitive evaluation of a diagnostic test. I ask Dr. Wheeler how often his IPG is negative and a venous duplex image is positive? Certainly Dr. Wheeler and his group have an excellent opportunity to provide those data in the future.
B O U N D VOLUMES AVAILABLE TO SUBSCRIBERS
Bound volumes of the JOUV,NAL OF VASCULARSURGERY for 1990 are available to subscribers only. They may be purchased from the publisher at a cost of $54.00 ($68.00 international) for Vol. 11 (January to June) and Vol. 12 (July to December). Price includes shipping charges. Each bound volume contains a subject and author index, and all advertising is removed. Copies are shipped within 60 days after publication of the last issue in the volume. The binding is durable buckram with the journal name, volume number, and year stamped in gold on the spine. Payment must accompany all orders. Contact Circulation Fulfillment, The C.V. Mosby Company, 11830 Westline Industrial Drive, St. Louis, MO 63146-3318, USA. In the United States call toll free: (800)325-4177, ext. 351. In Missouri call collect: (314)872-8370, ext. 351. Subscriptions must be in force to qualify. Bound volumes are not available in place of a regular JOUrnAL subscription.
The noninvasive diagnosis of deep venous thrombosis (DVT) has assumed an important place in the clinical evaluation of patients with venous thromFrom the Sectionof VascularSurgery,VascularLaboratory,Temple UniversitySchool of Medicine. Supported in part by grants from the U.S. Public Health Service, research grant No. 2 MO1 RR00349, National Institutes of Health, GeneralClinicalResearchCenters Branch. Presented at the Thirty-seventhScientificMeeting of the North American Chapter, International Society for Cardiovascular Surgery, New York, N.Y., June 19-20, 1989. Reprint requests: AnthonyJ. Comerota,MD, FACS, Sectionof Vascular Surgery, Temple University School of Medicine, s3road & Ontario St., Philadelphia,PA 19140. 24/6/16342
boembolic disorders. Many investigators have reported excellent clinical utility for hemodynamic tests for DVT. ls Others have documented high falsenegative and false-positive rates, 68 casting doubt on the uniformity of the diagnostic reliability of these tests. A prospective evaluation was subsequently performed, which demonstrated that patient selection was critically important when assessing the diagnostic capability ofhemodynamic tests for DVT. 9 In that study patients evaluated because of a clinical suspicion of DVT had high diagnostic sensitivity for proximal vein thrombosis, 83% and 92% for impedance plethysmography (IPG) and phleborheography 53
54
Journal of VASCULAR SURGERY
Comerota et al.
Table I. Venous duplex imaging: Diagnostic criteria for DVT B-mode image
Table II. Phlebographic details of thrombi according to patient group
Doppler signal
Intralttminal echogenic material Absent intraluminal signal Continuous signal (no respiNoncompressibility of vein ratory phasisity) Blood flow around a lucent fill- Reduced or absent augmentaing defect tion maneuvers Loss of venodilation with Valsalva maneuver
(PRG), respectively. However, when patients were studied because of a high risk for postoperative DVT without overt clinical suspicion, the sensitivity for diagnosing proximal vein thrombosis dropped to 32% and 33%. Venous duplex imaging (VDI) is rapidly becoming accepted as an important noninvasive diagnostic test acute DVT. Although high ~diagnostic sen. . . . for :• slnvmes and specificities have been reported, 1°1z a prospective analysis evaluating whether patient selection affects the results of VDI has not been performed. Therefore the question remains whether patient selection, and thus the location and/or magnitude of thrombi, has significant influence on the diagnostic capability of VDI. The purpose of this study is to determine the diagnostic reliability of VDI compared to PRG and to evaluate the influence of patient selection on both techniques. METHODS
One hundred ten extremities in 103 patients were prospectively evaluated with VDI, PRG, and ascending phlebography within the same 24-hour period. Phleborheography was performed with the Cranley-Grass phleborheograph (Grass, Inc., Quincy, Mass.), and VDI was performed with the Biosound 2000IISA (Biosound, Inc., Indianapolis, Ind.) duplex imaging system. The technique of examination and method of interpretation for each test has been previously described. 2a° Table I lists the diagnostic criteria used for VDI. The vena cava and iliac veins were not visualized during this examination with the transducers used, unless the patient had very little body fat. Both noninvasive tests were performed at the same time, and with few exceptions, before phlebography. Ascending phlebography was performed according to the modified technique of Rabinov and Paulin. TM Patients who had unsuccessful or inadequate phlebography were excluded. Additionally, patients
Diagnostic (N = 72) Incidence Normal DVT Distribution AK BK Magnitude Occluding Nonocduding
Surveillance (N = 38)
p-value
39% (28•72) 61% (44172)
vs 76% (29/38) vs 24% (9•38)
<0.001 --
84% (37/44) 16% (7/44)
vs 78% (7/9) vs 22% (2/9)
NS --
81% (30/37) 19% (7/37)
vs 43% (3/7) vs 57% (4/7)
0.054
having phlebography during weekends, holidays, and those evaluated as outpatients were frequently unavailable for noninvasive testing. Also, as previously mentioned, patients having their noninvasive tests more than 24 hours before phlebography were~xeluded from analysis. Overall, somewhat more than twice the number ofphlebograms were required durhag the period of this study to capture the patients reported herein. Phlebograms were interpreted as normal, below-knee (BK)--clot isolated to the infrapopliteal veins, or above-knee (AK)--clot involving the popliteal and/or more proximal veins. The methods of patient selection for noninvasive testing and ascending phlebography varied between the two following distinct patient populations: (1) Diagnostic group: patients evaluated because of a clinical suspicion of acute DVT. Phlebography was selectively obtained at the discretion of the referring physician and was not necessarily predicated on the results of a noninvasive test. (2) Surveillance group: patients undergoing orthopedic surgery who were at high risk for postoperative DVT after total joint~eplacement. These patients had routine postoperative ascending phlebography as part of an ongoing surveillance program. Noninvasive testing was not'routinely requested, but was performed as an adjunct to the surveillance program after patient consent. Noninvasive studies were performed during the morning of the sixth or seventh postoperative day, with the phlebogram obtained during the afternoon of the same day. Statistical analysis Ascending phlebography was used as the reference standard for both PRG and VDI, and 2 × 2 contingency tables were made with the results of each of the noninvasive tests. Statistical differences between diagnostic tests were compared by X2 with the Fisher exact test.
Volume 11 Number 1 flanuary 1990
Venous duplex imaging 55
Table HI. Reliability of noninvasive tests compared to phlebography Total PRG Sensitivity All AK BK Positive predictive value Specificity Negative predictive value VDI Sensitivity All AK BK Positive predictive value Specificity Negative predictive value
Diagnostic
55% 66% 0% 83% 89% 68%
(29/53) (29/44) (0/9) (29/35) (51/57) (51/75)
61% 73% 0% 82% 79% 56%
(27/44) (27/37)
96% 100% 78% 96% 93% 96%
(51/53) (44/44) (7/9) (51/53) (53/57) (53/55)
98% 100% 86% 96% 86% 96%
_RESULTS
C~' 110 extremities undergoing ascending phlebography, 48% (53/110) had demonstrable thrombi, and 52% (57/110) had normal phlebography results. Eighty-three percent (44/53) of patients with thrombi had clot extending into the pop,~titealor more proximal veins (AK), and two patients had isolated iliac vein occlusion. Seventeen percent (9/53) had isolated infrapopliteal venous thrombosis ~BK). The phlebographic details of thrombi by patient group are listed in Table II. The incidence and magnitude of thrombi were significantly different between the diagnostic and surveillance groups, whereas the distribution of thrombi was similar. The ~verall incidence of DVT was 61% (44/72) in the diagnostic group and 24% (9/38) in the surveillance group (p < 0,001). Although the distribution of ~,hrtmbi was similar between groups the magnitude was different, with 81% (30/37) of AK thrombi occluding veins of diagnostic patients compared to ¢~3% (3/7) of surveillance patients (p = 0.054). The reliability of the PRG in the diagnostic and surveillance groups is listed in Table III. There is a ,significant difference in the ability of the PRG to detect AK thrombi in diagnostic compared to patients in the surveillance group (p = 0.036). There #s a strong trend toward similar findings in all patients with DVT in the diagnostic compared to the surveillance group (p = 0.06). The specificity and negative predictive value are significantly better in patients in the surveillance group than in patients in the diagnostic group. - Table III summarizes the reliability of VDI in patients in the diagnostic group compared to patients m,,lle surveillance group. Unlike hemodynamlc stud-
Surveillance
p-value
(27/33) (22/28) (22/39)
vs vs vs vs vs vs
22% 29% 0% 100% 100% 81%
(2/9) (2/7) (0/2) (2/2) (29/29) (29/36)
0.06 0.036 NS NS 0.01 0.029
(43/44) (37/37) (6/7) (43/45) (24/28) (24/25)
vs vs vs vs vs vs
89% 100% 50% 100% 100% 97%
(8/9) (7/7) (1/2) (8/8) (29/29) (29/30)
NS NS NS NS 0.051 NS
(0/7)
ies, there is no difference in its sensitivity to detect overall, AK, or BK thrombi. There is no difference between positive and negative predictive value, however, the specificity is improved in the surveillance compared to the diagnostic group, and this difference appears statistically significant (p = 0.051). The comparative results of VDI and PRG are listed in Table IV. The overall sensitivity of VDI is 96% (51/53) compared to 55% (29/53) for PRG, (p < 0.001). Venous duplex imaging detected 100% (44/44) of AK thrombi compared to 66% (29/44) detected by PRG (p < 0.001). It is interesting to note that VDI detected 78% (7/9) of BK thrombi compared to 0% (0/9) detected by PRG (p < 0.001). The positive predictive value and specificity between the two noninvasive tests were not different, although the negative predictive value was significandy better for VDI. The comparative results of VDI and PRG for patients in the diagnostic group are listed in Table III. The sensitivities are significantly better for VDI for all patients with venous thrombosis, those with AK thrombi, and those with BK thrombi. The negative predictive value is also significantly better for VDI compared to PKG. Venous duplex imaging and PRG compared to ascending phlebography for patients in the surveillance group are summarized in Table IV. The overall sensitivity and sensitivity for detecting AK thrombi are significantly different between the two noninvasive studies. Below-knee thrombi were infrequent, and therefore conclusions cannot be drawn about the diagnostic sensitivity of VDI for BK thrombi in patients in the surveillance group. The positive predictive value, specificity, and negative prcdictive value were not different between the two tests.
Journal of VASCULAR SURGERY
56 Comerota et al.
Table IV. Venous duplex imaging and phleborheography compared to ascending phlebography VDI All patients Sensitivity All AK BK J~ Positive predictive value Specificity Negative predictive value Diagnostic patients Sensitivity All AK BK Positive predictive value Specificity Negative predictive value Surveillance patients Sensitivity All AK BK Positive predictive value Specificity Negative predictive value
PR G
p-value
96% 100% 78% 93% 93% 96%
(51/53) (44/44) (7/9) (51/55) (53/57) (53/55)
55% 66% 0% 83% 89% 68%
(29/53) (29/44) (0/9) (29/35) (51/57) (51/75)
p < 0.001 p < 0.001 NS NS NS p < 0.001
98% 100% 86% 91% 86% 96%
(43/44) (37/37) (6/7) (43/47) (24/28) (24/25)
61% 73% 0% 82% 79% 56%
(27/44) (27/37) (0/7) (27/33) (22/28) (22/39)
p < 0.001 p < 0.01 p < 0.01 NS NS p < 0.001
89% 100% 50% 100% 100% 97%
(8/9) (7•7) (1/2) (8/8) (29/29) (29/30)
22% 29% 0% 100% 100% 81%
(2/9) (2/7) (0/2) (2/2) (29/29) (29/36)
p < 0.05 p < 0.05 NS NS NS NS
Both PRG and VDI appropriately identified the two patients with iliac vein occlusion. Although results of the B-mode image were negative, results of the common femoral Doppler examination were abnormal, indicating a proximal obstruction. DISCUSSION
Patients at high risk for DVT (surveillance group) have previously presented a greater diagnostic challenge compared to those with signs and symptoms of acute DVT. 9 Ascending phlebography was considered the only accurate diagnostic method for evaluating these patients~ Since these patients had a high incidence of isolated infrapopliteal thrombi and nonocclusive thrombi, hemodynamic studies failed to identify clot in these patients. The PRG was shown to be a more sensitive test than IPG for detecting deep venous obstruction in a previous comparison, 9 therefore we chose the PRG as the best hemodynamic test (in our laboratory) for comparison to VDI. The PRG results reported herein are similar to our previous findingsg; however, there appears to be a trend toward a lower sensitivity for AK thrombi in patients in the diagnostic group. This is most likely the result of identification of patients with nonocclusive thrombi by VDI but who had normal PRGs, who were then evaluated by phlebography to resolve the discrepancy. In the past, most patients with symptoms in their legs and negative hemodynamic test results would not have had phlebography performed,
which seems to be the practice in many reputable laboratories, s-s Although we frequently use venous Doppler to complement other noninvasive tests to obtain addi4 tional information, we rarely use venous Doppler alone. Results of venous Doppler used alone vary widely. 1922 It is entirely dependent on the skill and,experience of the examiner, and excellent results obtained by one individual cannot be easily duplicated by other laboratories. We found that when combine4 with the B-mode image, venous Doppler was appropriately positive when patients had occlusive AK thrombi, whereas those with BK and infrapop~eak thrombi usually did not have abnormal Doppler examination results. When results of a common femoral or superficial femoral Doppler examination areclearly abnormal without any visible thrombi or unequivocally abnormal compression test, then suprainguinal thrombus must be suspected. Unlike hemodynamic studies, VDI has similar diagnostic capability for patients in the diagnostic and surveillance groups. It is interesting to note tha~ VDI has a high diagnostic sensitivity for BK as well as AK thrombi. Venous duplex imaging diagnosed 78% (7/9) of all BK thrombi and 86% (6/7) of BK thrombi in the diagnostic group. Although there were too few isolated infrapopliteal thrombi in the surveillance group to draw conclusions about the sen~ sitivity of VDI, it is our opinion that it would be equal to the diagnostic group. It has previously b ~ n
Volume 11 Number 1
Venous duplex imaging 57
lanuary1990
Fig. 1. B-mode image of paired popliteal veins adjacent to popliteal artery without probe pressure.
Fig. 2. Compression of both veins is complete without compressionof artery, indicating that both veins are patent.
Table V. Studies comparing noninvasive hemodynamic tests to venous imaging for acute DVT Sensitivity Author
No.
Rosner et al. 15 1988 O'Leary et al. 16 1988 Patterson et al} 7 1989 Present series
25 37 49 110
IPG / P R G
100% 78% 75% 55%
(6/6) (14/18) (15/20) (29/53)
Specificity VDI
90% 88% 89% 96%
(9/10) (22/25) (24/27) (51/53)
IPG / P R G
63% 84% 45% 89%
(12/19) (16/19) (13/29) (51/57)
VDI
100% 96% 91% 93%
(22/22) (24/25) (31/34) (53/57)
In each previously published series, there were 31% to 32% more VDI studies reported than hemodynamic tests.
reported that real-time B-mode ultrasonography was inaccurate for detecting infrapopliteal venous thrombin; however, those investigators used the compression test (Figs. 1 and 2) as the sole criterion for "~agnosis. The additional information provided by the B-mode image itself has been most helpful in our experience. Although routinely the Doppler signal was not as useful as the image for the diagnosis ofBK thrombi, its use in equivocal examinations may be helpful, particularly the augmentation maneuver. The Doppler signal was most important for diagnosing the two patients with isolated iliac vein obstruction. Perhaps more importantly, the diligence and expertise of the technologist performing the test and the physician interpreting the test must be considered. Our data fit well with those published by Elias et al}4 who reported 854 limbs studied with realtime B-mode ultrasonography and phlebography for acute DVT. Real-time B-mode imaging detected '98% (325/333) of thrombi with 100% of AK and 9"% of BK thrombi accurately diagnosed.
Three previously published reports compared noninvasive hemodynamic tests for acute DVT to duplex imaging of the deep venous system (Table V).15-17 In general, the previously reported data agree with the data reported herein; however, several differences must be addressed. Rosner et al2s reported I00% sensitivity for IPG; whereas VDI missed a thrombus in a duplicated popliteal vein. The patient responsible for the false-negative VDI was not studied with IPG. O'Leary et al}6 report a 78% (14/18) sensitivity for hemodynamic tests; however, they excluded 15% (eight) of their patients because of indeterminant results. If these eight patients were ineluded as errors, the overall accuracy of their hemodynamic tests would drop to 68%. Although each of these three studies evaluates hemodynamic tests compared to VDI, each of these authors reports at least 30% more patients studied with VDI than hemodynamic studies. Therefore a true comparison is not possible. None of the three previous reports clarified the diagnostic accuracy of VDI for infrapopliteal
58
Comerota et al.
Journal of VASCULAR SURGERY
Fig. 3. B-mode image indicating partially occluding thrombus of proximal superficial femoral vein (double arrow). Thrombus is located within a valve cusp; whereas the opposite valve cusp fimctioned normally (single arrow). Fig. 5. B-mode image during compression shows ~[[~ldl¥ echogenic material within the superficial femoral vein (IT) and that it does not compress the probe pressure. Partially occluding thrombus of the superficial femoral vein was diagnosed. The superficial femoral artery (A) appears normal.
Fig. 4. Ascending phlebogram shows partially occluding thrombus in proximal superficial femoral vein within a valve cusp (double arrow). Doppler examination results were normal. thrombi. Two reports 16'17 identified isolated infrapopliteal thrombi as causes of false-negative examination results; however, both failed to differentiate the sensitivity of VDI for proximal versus infrapopliteal thrombi. Whether nonocclusive or infrapopliteal venous thrombi need to be diagnosed is a separate question.
Fig. 6. Ascending phlebogram of same patient shows contrast defect in the same location (arrow); however, the phlebogram was officially interpreted as normal, with the area in question read as a "flow defect from a side branch." However, it is important, to understand that in experienced hands, VDI, in contrast to hemodynamic tests, has the capability o f diagnosing nonocclus~e (Figs. 3 and 4) as well as infrapopliteal thrombi. Another question that might arise is why the pos-
Volume 11 Number 1 January 19.90
itive predictive value is not better than 9 3 % (51 / 5 5) for VDI. It is our opinion that VDI detected nonocclusive thrombi that were missed by ascending phlebography (Figs. 5 and 6). This raises yet another issue, which test, VDI or phlebography, is the true standard for the diagnosis of acute DVT. We believe it is premature to answer this question, and it probably should not be answered inasmuch as VDI and phlebography will agree in most patients. Discrepancies that arise between test results should evoke a critical evaluation of each test, with particular attention given to the quality of the images available. Combining the information at hand with sound judgement will lead to the proper conclusions in most cases. I n light o f t h e d a t a available, t h e answer to t h e q u e s t i o n o f w h e t h e r V D I s h o u l d replace h e m o d y n a m i c tests for D V T s h o u l d be yes.
RI~ERENCES 1. Hull RD, Hirsh J, Carter CJ, et al. Diagnostic efficacy of impedance plethysmography for clinicallysuspected deep vein thrombosis. A randomized trial. Ann Intern Meal 1985; 102:21-8. 2. Comerota AJ, Cranley JJ, Cook SE, Sipple P. Phleborheography--results of a 10-year experience. Surgery 1982;91: 573-81. 3. Wheeler HB, Anderson FA Jr. The diagnosis of venous thrombosis by impedance plethysmography. In: Bernstein EF, ed. Noninvasive diagnostic techniques in vascular disease, 3rd ed. St. Louis: CV Mosby Co, 1985:755. 4. Stallworth JM, Plonk GW, Home JB. Negative phleborheography, clinical follow-up in 593 patients. Arch Surg 1981;116:795-7. 5. Huisman MV, Buller HR, Wouter Ten Care JV, Vreeken J. Serial impedance plethysmography for suspected deep venous thrombosis in outpatients: the Amsterdam general ,practitioner study. N Engl J Med 1986;314:823-8. 6. Ramchandani P, Soulen RL, Fedullo LM, Gaines VD. Deep vein thrombosis: significant limitations of noninvasive tests. Radiology 1985;156:47-9. 7. Vaccaro P, Van Aman M, Miller S, Fochman J, Smead WL. Shortcomings of physical examination and impedance plethysmography in the diagnosis of lower extremity deep venous thrombosis. Angiology 1987;38:232-5.
DISCUSSION Dr. David S. S u m n e r (Springfield, Ill.). Once again, Dr. Comerota and his associates have demonstrated that plethysmographic tests have insufficient sensitivity to detect venous thrombi in surveillance groups. This comes as no surprise, since many o f these developing clots are nonocclusive or infrapopliteal and offer little or no impediment j venous blood flow. However, I am puzzled by the diminishing accuracy of the PKG in the hands of the same investigators, u s i n g - -
Venous duplex imaging 59 8. Young AE, Henderson BS, Phillips DA, Couch NP. Impedance plethysmography: its limitations as a substitute for phlebography. Cardiovasc Radiol 1978;1:233-9. 9. Comerota AJ, Katz ML, Grossi RJ, et al. The comparative value of noninvasive testing for diagnosis and surveillance of deep vein thrombosis. J VAsc SUWG1988;7:40-9. 10. Comerota AJ, Katz ML. The diagnosis of acute deep venous thrombosis by duplex venous imaging. Semin Vase Surg 1988;1:32-9. 11. Lensing AWA, Prandoni P, Brandjes D, et al. Detection of deep vein thrombosis by real-time B-mode ultrasonography. N Engl J Med 1989;320:342-5. 12. Langsfeld M, Hershey FB, Thorpe L, et al. Duplex B-mode imaging for the diagnosis of deep venous thrombosis. Arch Surg 1987;122:587-91. 13. Karkow WS, RuoffBA, Cranley JJ. B-mode venous imaging. In: Kempczinski RF, Yao JST, eds. Practical noninvasive vascular diagnosis. Chicago: Year Book Medical, 1987:464-85. 14. Elias A, Le Corff G, Bouvier JL, Benichou M, Serradimigni A. Value of real-time b-mode ultrasound imaging in the diagnosis of deep vein thrombosis of the lower limbs. Int Angiol 1987;6:175-82. 15. Rosner Nit, Doris PE. Diagnosis of femoropopliteal venous thrombosis: comparison of duplex sonography and plethysmography. AJR 1988;150:623-7. 16. O'Leary DH, Kane RA, Chase BM. A prospective study of the efficacy of B-scan sonography in the detection of deep venous thrombosis of the lower extremities. J Clin Ukrasound 1988; 16:1-8. 17. Patterson RB, Fowl RJ, Keller JD, Shomaker W, Kempczinski RF. The limitations of impedance plethysmography in the diagnosis of acute deep venous thrombosis. J VASe SURG 1989;9:725-30. 18. Rabinov D, Paulin S. Roentgen diagnosis of venous thrombosis of the leg. Arch Surg 1972;110:1336-42: 19. Bolton JP, Hoffman VJ. Incidence of early postoperative iliofemoral thrombosis. Br Med J 1975;1:247-9. 20. Milne RM, Gunn AA, Griffiths JMT, Ruckley DV. Postoperative deep venous thrombosis. A comparison of diagnostic techniques. Lancet 1971;2:445-7. 21. Barnes RW i Russell HE, Wilson MR. Doppler ultrasonic evaluation of venous disease, a programmed audiovisual instruction. Iowa City: University of Iowa: 1975. 22. Sumner D. Diagnosis of deep venous thrombosis by Doppler ultrasound. In: Nicolaides AN, Yao JST, eds. Investigation of vascular disorders. New York: Churchill Livingstone, 1981:377-402.
as far as I can determine--similar techniques on a similar "diagnostiC' population. For example, in 1982 Dr. Comerota reported a sensitivity o f 92%, a specificity o f 95%, a positive predictive value of91%, and a negative predictive value o f 95%, for the P R G when used to diagnose DVT in 709 patients. His current dismal results stand in stark contrast to these previous values. Moreover, in the 1982 study the PRG detected 45 o f 57 isolated calf vein thrombi (a sensitivity o f 79%). N o w it detects none o f seven! What has changed? Is it the disease, the testing procedure, or the
60
Comerota et al.
method of reporting? Now we hear that the duplex scan is capable of accuracies similar to those previously reported with the PRG. Seven years from now will the duplex scan also fall from grace? In the not too distant past, numerous reputable investigators established the safety of using both the IPG and the PRG for ruling out DVT in patients suspected of having this disease. Are these laboratories now obligated to discard these tests? Despite the growing enthusiasm for VDI, the approach and the criteria used are so different that it can hardly be assumed that results obtained in one laboratory are equivalent to those from another. Some rely on compression alone and do not bother with the infrapopliteal veins; others emphasize the visualization of clots; and still others depend most heavily on the Doppler signal. Do we have enough information to standardize the test? I think not; yet, we are now hearing that duplex scanning should replace phlebography as the gold standard. Can a technologist realistically survey inch by inch the yards of veins in each leg? On the other hand, is it really important in the clinical setting that every isolated thrombus be detected? I agree with Dr. Comerota that duplex scanning is the current method of choice; but I predict that it will soon be replaced by color-flow imaging, which has several distinct advantages. It facilitates identification of veins, especially below the knee, provides simultaneous Doppler and B-mode information, and reduces the time required for each examination. Dr. Anthony Comerota. I thank Dr. Sumner for his comments. Dr. Sumner appropriately pointed to differences in the results. The differences he quoted from 1982 were essentially the results obtained in Cincinnati. When we moved to Philadelphia we noticed that our results with PRG were different from those we previously reported in Cincinnati and the IPG results reported from Worcester and McMaster. Therefore, the study we presented 2 years ago was performed, which paralleled the current data; however, Dr. Sumner failed to refer to those corroborating data. Clearly, it was patient selection that was responsible for those differences, and it was patient selection for phlebography that was most important. I think we made those points 2 years ago. Since our previous presentation, the audience might notice a difference in sensitivity for proximal thrombi in the diagnostic group. That also appears to be due to the selection of patients for phlebography. Both V D I and PRG were performed in patients suspected of DVT. When the venous duplex image identified a dot and the PRG was negative, ascending phlebography was strongly encouraged and often performed, thereby selecting patients in whom the PRG is likely to be falsely negative and reducing the sensitivity of the PRG in the diagnostic group. I do not believe that vascular laboratories should discard hemodynamic tests, at least not until they are facile with VDI. I would agree with Dr. Sumner that there is a
Journal of VASCULAR SURGERY
marked variability in technique as well as criteria for interpretation, and these criteria have not been standardized for venous duplex. However, V D I is more than just a compression test. We believe the image is important. The Doppler gives an added dimension and is particularly useful for proximal venous thrombi. I believe color incorporates anatomic information with physiologic information. I also believe it will speed the examination for the uninitiated and the less experienced. However, for those who are experienced with VDI, it is my impression that color will probably be less important. Dr. H. B r o w n e r Wheeler (Worcester, Mass.). Our hospital has experienced a dramatic increase in B-mode scanning, but we do more hemodynamic screening tests than ever. Logistically, we simply could not screen all patients suspected of DVT with B-mode scans. There remains an important role for simple low-cost bedside procedures. Dr. Comerota, how do you explain your low sensitivity with PRG for symptomatic proximal D V T when previous studies have reported nearly 100% sensitivity? Our,~wn accuracy for IPG would also be much less than earlier studies, if based on current venograms, simply because the original validation studies included venograms on all patients suspected of DVT, whereas now venograms are done on only about 5%--primarily problem patients. May similar bias in the selection of patients for venography affect your results? Also, if PRG fails to detect significant thrombi, how do you explain the excellent results of clinical management with PRG alone? Stallworth reported 593 patients with negative PRGs who were not given anticoagulants and had only a 0.2% incidence of nonfatal pulmonary embolism. Similar outcomes in untreated patients with negative IPGs have been reported by six different medical centers. If IPG and PRG fail to detect many thrombi, most of these thrombi must be relatively innocuous, since patient outcome is the ultimate gold standard for the value of any procedure, j% Dr. John J. Cranley (Cincinnati, Ohio). From 1971 to 1981 phleborheography was the main diagnostic test for deep venous thrombosis at Cincinnati's Good Samaritan Hospital. The data on which Dr. Comerota's 1982 report was based consisted of 709 consecutive phlebograms on extremities also studied by phleborheography. No patient having both tests was excluded. Dr. Comerota's recognition of the difference between patient populations studied for diagnosis or surveillance is correct and a significant step forward. All of our phlebograms were obtained for diagnosis, none for surveillance. As Dr. Porter has emphasized, comparison of results obtained in different eras is hazardous. Our conceptual changes alone are of significance here. When we initiated our studies with the PRG in 1971 we conceptualized D V T as a thrombus occluding mainstream veins. We were not seeking, and stressed the fact that we could not detect, clots outside the mainstream. Thus, partially occluding thrombi, thrombi in valve cusps, soleal veins, or small tributaries-~"
Volume 11 Number 1 January 1990
Venous duplex imaging 61
the popliteal or femoral vein systems detectable today, were ~ndetectable then. The phiebographic methods used in the two centers differed. At Temple University, the method of Rabinov ~ n d Paulin is used. This method permits flooding of the leg veins passively in the motionless limb. It provides more complete visualization of the muscular veins of the leg than ~he method we use, but as Comerota reported, it is less accurate in the groin and pelvis. At Good Samaritan Hospital, we used five14 × 51 inch cassettes, each containing three 14 × 17 inch films. This permitted visualization of ~11 the mainstream veins from the ankle to the vena cava, which was what I demanded. Finally, although we employed many technicians, we [~ad one senior technician who devoted her entire career to performing and teaching the PRG. The amount of time and maneuvers performed did not matter, as long as the result was clearcut. Such concentration on one test would T
be difficult today in a busy laboratory performing many tests. Indeed, without this devoted technician we might not be able today to repeat the results obtained a decade ago. But I am certain that the data Dr. Comerota presented in 1982 were correct. Dr. Comerota. In answer to Dr. Wheeler's second question, I believe the false-negative results of hemodynamic tests will certainly be diluted with the many truenegative examinations performed by most laboratories, and therefore the poor outcome in some of those will result in a relatively small percentage. I am sure that many of the false-negative tests might well remain silent over the long term. I do not necessarily agree that the ultimate outcome is the definitive evaluation of a diagnostic test. I ask Dr. Wheeler how often his IPG is negative and a venous duplex image is positive? Certainly Dr. Wheeler and his group have an excellent opportunity to provide those data in the future.
B O U N D VOLUMES AVAILABLE TO SUBSCRIBERS
Bound volumes of the JOUV,NAL OF VASCULARSURGERY for 1990 are available to subscribers only. They may be purchased from the publisher at a cost of $54.00 ($68.00 international) for Vol. 11 (January to June) and Vol. 12 (July to December). Price includes shipping charges. Each bound volume contains a subject and author index, and all advertising is removed. Copies are shipped within 60 days after publication of the last issue in the volume. The binding is durable buckram with the journal name, volume number, and year stamped in gold on the spine. Payment must accompany all orders. Contact Circulation Fulfillment, The C.V. Mosby Company, 11830 Westline Industrial Drive, St. Louis, MO 63146-3318, USA. In the United States call toll free: (800)325-4177, ext. 351. In Missouri call collect: (314)872-8370, ext. 351. Subscriptions must be in force to qualify. Bound volumes are not available in place of a regular JOUrnAL subscription.