Kevin M. Duwe 1 Maria Shiau Nancy E. Budorick John H. M. Austin Yahya M. Berkmen

Transcription

1 2000 RRS Executive Council ward I Kevin M. Duwe 1 Maria Shiau Nancy E. udorick John H. M. ustin Yahya M. erkmen Received February 24, 2000; accepted after revision June 5, ll authors: Department of Radiology, New York Presbyterian Hospital, Columbia University College of Physicians and Surgeons, Milstein Hospital ldg., 2nd Fl., Fort Washington ve., New York, NY ddress correspondence to K. M. Duwe. JR 2000;175: X/00/ merican Roentgen Ray Society Evaluation of the Lower Extremity Veins in Patients with Suspected Pulmonary Embolism: Retrospective Comparison of Helical CT Venography and Sonography OJECTIVE. In patients undergoing a combined CT angiographic and CT venographic protocol, the accuracy of helical CT venography for the detection of deep venous thrombosis was compared with that of lower extremity sonography. MTERILS ND METHODS. Patients who had undergone a combined CT angiographic and CT venographic protocol and sonography of the lower extremities within 1 week were identified. The final reports were evaluated for the presence or absence of deep venous thrombosis. Statistical measures for the identification of deep venous thrombosis with helical CT venography were calculated. In each true-positive case, the location of the thrombus identified with both techniques was compared. ll false-positive and false-negative cases were reviewed to identify the reasons for the discrepancies. RESULTS. Seventy-four patients were included. There were eight patients (11%) with true-positive findings, 61 patients (82%) with true-negative findings, four patients (5%) with false-positive findings, and one patient (1%) with a false-negative finding. When comparing helical CT venography with sonography for the detection of lower extremity deep venous thrombosis, the sensitivity measured 89%; specificity, 94%; positive predictive value, 67%; negative predictive value, 98%; and accuracy, 93%. Of the eight true-positive cases, five had sites of thrombus that were in agreement on both CT venography and sonography. Of the five discordant cases, four were false-positives and one was a false-negative. Possible explanations for all discrepancies were identified. CONCLUSION. Compared with sonography, CT venography had a 93% accuracy in identifying deep venous thrombosis. However, the positive predictive value of only 67% for CT venography suggests that sonography should be used to confirm the presence of isolated deep venous thrombosis before anticoagulation is initiated. In addition, interpretation of CT venography should be performed with knowledge of certain pitfalls. V enous thromboembolic disease, which includes both deep venous thrombosis and resultant pulmonary embolism, is one of the most common preventable causes of death in hospitalized patients [1]. Each year, between 300,000 and 600,000 patients in the United States are hospitalized for deep venous thrombosis, pulmonary embolism, or both [2]. Estimates suggest that between 50,000 to 100,000 of these patients die as a result [2]. ecause thromboembolic disease is associated with a major risk of mortality, manifests often with nonspecific symptoms and signs, and can often be treated effectively, the use of diagnostic imaging to assist in the rapid and accurate diagnosis of pulmonary embolism and deep venous thrombosis has undergone continual evaluation and revision. In the 1960s and 1970s, the diagnosis of pulmonary embolism was based on the catheter angiogram. Subsequently, the less invasive ventilation-perfusion scan was often obtained without catheter angiography [3]. Most recently, helical CT angiography is threatening to usurp this role, especially in patients with abnormal findings on chest radiographs [4 10]. Likewise, the diagnosis of deep venous thrombosis traditionally required conventional catheter venography [11]. Currently, noninvasive sonography has largely taken the place of contrast-enhanced venography [12, 13]. fter the description of delayed helical CT venography after CT angiography in 1998, the optimal diagnostic evaluation of thromboembolic disease is in question once again [14]. JR:175, December

2 Duwe et al. Studies have already validated the accuracy of CT angiography for revealing pulmonary embolism above the subsegmental level [4 10]. Two additional studies have also shown that contrast-enhanced CT venography, with injection into the pedal veins, is comparable with contrast-enhanced venography for the diagnosis of deep venous thrombosis [15, 16]. However, to our knowledge, only two published studies have described the use of combined CT venography and CT angiography with a single upper extremity contrast injection. In the first study, Loud et al. [14] showed perfect correlation between the CT venographic results and sonographic results in five patients after CT angiography. In a recent larger prospective study, Loud et al. [17] again reported 100% correlation between CT venography and lower extremity sonography performed within 12 hr in 71 patients. In the present study, we attempted to replicate the results of Loud et al. [17] using a retrospective study design. Helical CT venographic results of combined CT angiographic and CT venographic examinations were correlated with lower extremity sonographic findings for the identification of deep venous thrombosis when both examinations were performed within a 1- week interval. Materials and Methods retrospective review of all adult patients at our medical center who underwent combined CT angiographic and CT venographic evaluation for suspected pulmonary embolism from March 1999 to January 2000 was performed. Patients who underwent sonographic evaluations of the lower extremities within 1 week of the CT angiographic and CT venographic examination were included in the study group. Technically nondiagnostic examinations were excluded. CT ngiographic and CT Venographic Protocol Each examination was performed on a Somatom Plus 4 helical CT scanner (Siemens Medical Systems, Munich, Germany). n IV injection of 120 ml of iohexol (Omnipaque [300 mg organically bound I/mL]; Nycomed, Princeton, NJ) was administered at 3 ml/ sec via an 18- to 20-gauge antecubital catheter. reath-hold helical CT images were obtained with a 3-mm collimation and a 5-mm table feed (pitch of 1.6) from the dome of the diaphragm to the aortic arch 28 sec after the start of the injection. fter acquisition of the chest images (usually sec) and an additional 120-sec delay, helical CT images were obtained with a 10-mm collimation and a 10-mm table feed (pitch of 1) from the iliac crests to the knees. Images were routinely interpreted by subspecialty trained thoracic radiologists in cine format on a Siemens workstation or picture archiving and communication system (PCS). The primary criterion for deep IV thrombosis included IV filling defects (Figs. 1 1C). dditional secondary signs included venous dilatation, segmental nonvisualization, wall enhancement, and perivenous and subcutaneous infiltration. Sonographic Protocol Each examination was performed on either a 128XP/10 (cuson, Mountain View, C) or a Sonoline Elegra (Siemens Medical Systems) sonography unit and 5- or 7-MHz linear transducers. Transverse gray-scale imaging with and without compression and longitudinal color and spectral Doppler imaging with augmentation maneuvers were routinely performed from the upper common femoral to the popliteal veins bilaterally. Images were obtained specifically of the upper and lower common femoral vein; greater saphenous confluence; bifurcation of the superficial and deep femoral veins; proximal, mid, and distal superficial femoral and popliteal veins. If thrombus was identified in the uppermost veins, additional evaluation was performed of the external iliac and common iliac veins and inferior vena cava. If thrombus was seen in the popliteal veins, additional evaluation of the calf veins was undertaken. Examinations were either performed or checked by sonography-trained radiologists. Criteria for deep venous thrombosis included noncompressible veins, absence of vessel filling on color Doppler sonography, and either absent or abnormal spectral Doppler waveforms (Figs. 1D 1F). We reviewed the final dictated reports to determine the presence or absence of deep venous thrombosis on both helical CT venography and sonography. For the purposes of comparison, sonography was considered the gold standard [18]. Measures of sensitivity, specificity, positive and negative predictive values, and accuracy for helical CT venography versus sonography were calculated. For each true-positive finding, the site of thrombus on both helical CT venography and sonography was recorded and compared for agreement. Falsepositive and false-negative helical CT venographic results were further analyzed. Tabulation of the time delay between these studies was performed. The helical CT venographic images were obtained and reviewed by the authors in an attempt to reach a consensus opinion regarding the potential cause or causes for the discrepancies. Results From March 1999 to January 2000, 77 patients (47 women and 27 men; age range, years; mean age, 78.5 years) underwent both CT angiography and CT venography and lower extremity sonography within 1 week of each other. Of these 77 patients, three were excluded. The first patient was excluded because of a technically limited sonographic examination as a result of severe leg edema; the second, because of inadequate venous enhancement on helical CT venography; and the third, because of metallic artifacts on helical CT venography as a result of a proximal femoral prosthesis. The final reports of the remaining 74 patients were evaluated for the presence or absence of deep venous thrombosis. On sonography, two patients had clots in the calf that had not been shown on CT venography; these cases were tabulated as true-negatives because the areas included for both techniques were free of thrombus. Sonography revealed deep venous thrombosis in nine (12%) of the 74 patients and no deep venous thrombosis in 65 (88%). Helical CT venography showed deep venous thrombosis in 12 patients (16%) and no deep venous thrombosis in 62 (84%). There were eight true-positive findings (11%), 61 true-negative (82%), four false-positives (5%), and one false-negative (1%). On the basis of these findings, the sensitivity of helical CT venography was 89%; specificity, 94%; positive predictive value, 67%; negative predictive value, 98%; and accuracy, 93%. Of the eight cases with true-positive findings, agreement on the location of thrombus was excellent. Findings from five of the eight cases agreed perfectly, with the delay between examinations ranging from 1 to 2 days. The remaining three cases for which agreement was less than perfect included the following patients. The first patient was a 70-year-old man who showed minimal extension of thrombus into an adjacent common femoral vein on CT venography 5 days later (Figs. 1 1C). The second patient was an 89-year-old man who showed minimal resolution of thrombus in a common femoral vein on a follow-up sonography 1 day later. The third patient was a 76-year-old woman who displayed a large amount of deep venous thrombosis growth over a 3-day interval; the deep venous thrombosis was initially identified in the popliteal vein (Fig. 2). fter an arterial puncture, the patient developed a large adductor compartment hematoma and propagation of thrombus into the superficial femoral vein (Fig. 2). Interestingly, pulmonary embolism was also detected in this patient on CT angiography, suggesting that the popliteal or superficial femoral vein thrombus may have even embolized to the lung (Fig. 2C). In addition, the other two patients with changes in the location of deep venous thrombosis also were found to have pulmonary embolism. Findings in four cases were four false-positive. One patient had a small nonocclusive clot on a single image of the common femoral vein that may have either developed or not been imaged at the time of sonography, 2 days earlier (Fig. 3). The other three patients had slight dilatation and vague filling defects seen on only a single image in two popliteal veins and one superficial femoral vein. On further review, the observers agreed that 1526 JR:175, December 2000

3 Helical CT Venography of the Lower Extremity Veins Fig year-old man with deep venous thrombosis and pulmonary embolism revealed on CT venography and CT angiography., CT venographic image shows filling defect surrounded by rim of contrast material in left popliteal vein (arrow)., djacent CT venographic image again shows filling defect surrounded by rim of contrast material in left popliteal vein (arrow ). C, CT angiographic image reveals partial filling defect within pulmonary artery to left lower lobe (arrow). D, Transverse gray-scale sonographic images of left popliteal vein obtained 5 days before C without (left image) and with (right image) compression reveal noncompressible left popliteal vein (arrows). E, Transverse color Doppler sonographic image (shown in black and white) obtained 5 days before C reveals nonfilling of left popliteal vein (arrow). F, Spectral Doppler sonographic image obtained 5 days before C reveals noise within left popliteal vein (arrow). C E D F JR:175, December

4 Duwe et al. C Fig year-old woman with propagation of deep venous thrombosis and possible embolization., CT venographic image shows thrombus in right popliteal vein (arrow). This was isolated finding on sonogram obtained 3 days earlier (not shown)., CT venographic image shows extension of deep venous thrombosis into superficial femoral vein (arrow) after right femoral artery puncture complicated by large adductor compartment hematoma. C, CT angiographic image shows partially occlusive thrombus in right lower lobe pulmonary artery (arrow). Fig year-old woman with tiny thrombus seen on CT venography but not on sonography. CT venographic image shows tiny, nonocclusive filling defect in right common femoral vein (arrow). It is possible that this thrombus had developed since or had been overlooked on prior sonographic examination performed 2 days earlier. these focal defects might have represented volume averaging of valves (Fig. 4). The one false-negative finding was a case of extensive bilateral deep venous thrombosis involving the right common iliac to superficial femoral vein and the left common iliac to the popliteal vein on follow-up sonography 4 days later. On further review, the observers agreed that secondary signs of thrombus were present. These signs included subtle enhancement of the common femoral venous walls bilaterally, venous dilatation at this level, and both perivenous and subcutaneous edema (Fig. 5). n additional finding that the observers thought might suggest extensive bilateral deep venous thrombosis was the prolonged arterial phase of enhancement (Fig. 5). The time delay between helical CT venography and sonography was compared among the cases (Fig. 6). Of the examinations performed on the same day, all 14 had false-negative findings. Of the 24 examinations separated by a 1- day interval, one had a false-positive findings, three had true-positives, and 20 had true-negatives. There were 12 examinations separated by 2 days, of which two had false-positive findings, three had true-positives, and seven had true-negatives. Of the eight examinations separated by JR:175, December 2000

5 Helical CT Venography of the Lower Extremity Veins Fig. 4. Patients with possible deep venous thrombosis seen on single CT venographic images but not visible on sonography., CT venographic image of 27-year-old woman shows vague central filling defect in right superficial femoral vein (arrow)., CT venographic image of 55-year-old woman shows vague central filling defect in left popliteal vein (arrow). days, one had true-positive findings and seven had true-negatives. There were five examinations separated by 4 days; one had a false-negative finding, one had a false-positive, and three had true-negatives. Five examinations were separated by 5 days: one had a true-positive finding and four had true-negatives. There were two and four additional false-negative findings for which examinations were separated by 6 and 7 days, respectively. The false-positive findings occurred after a delay of from 1 to 4 days. The false-negative finding occurred after a 4-day delay. No discordant findings were identified among examinations with a delay of more than 4 days between examinations. Discussion In the present study, helical CT venography revealed a 93% accurate correlation with Fig year-old man with deep venous thrombosis identified on sonography but missed on CT venography., CT venographic image shows common femoral veins with subtle wall enhancement, dilatation (arrows), and perivenous and subcutaneous edema., CT venographic image shows superficial femoral arteries with dense arterial enhancement (arrows). This phenomenon may represent a poorly timed bolus, severe bilateral arterial insufficiency, or as in this case, probable delayed enhancement from extensive bilateral deep venous thrombosis. Fig. 6. ar chart shows distribution of examinations on basis of number of days separating helical CT venographic and sonographic examinations. Stripes = truenegative findings, white dots = true-positive findings, black dots = falsepositive findings, solid = false-negative finding. JR:175, December

6 Duwe et al. sonography for identifying deep venous thrombosis in patients with suspected pulmonary embolism. lthough these results may be considered excellent, we were unable to duplicate the perfect accuracy reported by Loud et al. [17]. When trying to understand why our results differed, we focused on the four falsepositive and one false-negative cases. The cases with false-positive findings were divided into two groups depending on the appearance of the filling defect, well-defined and eccentric versus ill-defined and central. The first subgroup consisted of one patient with a well-defined eccentric 3-mm nonocclusive thrombus in a common femoral vein. aldt et al. [16] described a similar nonocclusive thrombus in an eternal iliac vein that was identified on CT venography, but not on prospective sonography. nother possibility was that this abnormality developed after the sonographic examination 2 days earlier. s we showed with one of the patients with a true-positive finding, extensive thrombus extension can occur within 3 days. Surely, a 3-mm clot could form within 2 days. This discrepancy can be avoided in future studies by allowing less time between the examinations. The second subgroup consisted of three patients who had an ill-defined central filling defect seen on only single images (2 popliteal and 1 superficial femoral veins). The observers reviewed these images and thought each focal defect likely represented volume averaging of a valve and not thrombus. To better evaluate this possibility, narrower collimation, immediate follow-up sonography, or in vitro studies may be helpful. The one case with a false-negative finding was misinterpreted for two reasons. The first reason was the large extent of the thrombus, which prevented any of the veins from enhancing normally. The second reason was its bilateral involvement, which caused the absence of a normal control vein for comparison. The prolonged arterial enhancement is interesting. This finding generally suggests two possible explanations. First, helical CT venography was performed too soon after CT angiography. Second, there is severe bilateral arterial insufficiency. This case illustrates a third possibility that is, extensive bilateral deep venous thrombosis. dditional cases of extensive bilateral deep venous thrombosis will need to be identified to prove this hypothesis. dditional methodologic differences were identified between our study and that of Loud et al. [17] that may have also contributed to the differing results. The largest of these differences was the use of prospective versus retrospective designs. The retrospective design has well-documented disadvantages, but this design also has certain advantages. The largest disadvantage in our study was the failure to identify a large number of patients who had undergone both examinations on the same day. We were able to identify only 14 such patients. For these 14 patients, the CT venographic and sonographic results all agreed, but all revealed negative findings for deep venous thrombosis. It was for this reason that we chose to include patients with examinations separated by a larger time interval. We chose 1 week for three reasons. First, Holm et al. [19] suggested that even after adequate anticoagulation only 3% of the deep venous thrombosis cases resolve completely by 1 week. Second, Shah et al. [20] used a 1-week interval in assessing the accuracy of routine pelvic CT compared with lower extremity sonography for the detection of deep venous thrombosis. Lastly, if the delay between our studies allowed complete resolution or new development of thrombus, discrepancies should have been more apparent in the studies separated by longer intervals. This appeared to not be the case. Instead, discordant results were seen in cases separated by as few as 1 and as many as 4 days (Fig. 6). In fact, reducing the allowed delay between examinations to 4 days would have excluded 11 cases with true-negative findings. This would have only reduced the specificity from 94% to 93% and the accuracy from 93% to 92% without affecting the sensitivity or positive or negative predictive values. n advantage of our retrospective design relates to the interpretation of the CT venographic studies. Instead of having the images interpreted by a small number of expert observers who were involved in the study design, we used four thoracic radiologists with differing levels of experience. They were unbiased because they were unaware of any potential study on this subject. We believe that this design more accurately depicts the situation seen in daily radiology practice. Furthermore, this design permitted the identification of the potential interpretive pitfalls. second methodologic difference between our study and that of Loud et al. [17] was in the CT venographic protocols. Loud et al. performed CT venography using a 5-mm collimation (10-mm collimation in 20 of 71 patients) and obtained images every 5 cm from the upper calves to the diaphragm 3.5 min after the start of contrast injection. For our protocol, we used a contiguous 10-mm collimation from the knees to the iliac crests 3 min after the start of contrast injection. Perhaps the use of a 5-mm collimation in our study might have prevented the misdiagnosis of valves as thrombus. We suggest that prospective studies performed with different collimations, skip segments, time delays, and volumes of contrast material will help delineate an optimal CT venographic protocol. One potential limitation that applies to both our study and that of Loud et al. [17] is the choice of sonography as the gold standard. It is possible that contrast-enhanced venography should remain the gold standard imaging study for the diagnosis of deep venous thrombosis. However, two limitations of contrast-enhanced venography support the use of sonography as the standard. First, contrast-enhanced venography is technically inadequate in 5 10% of the studies [21 24] and interobserver variability is reportedly 10% [25]. Second, sonography has been proven to be highly sensitive (89 100%) and specific (99 100%) for deep venous thrombosis when compared with contrast-enhanced venography of the thigh [26 28]. The long-term use of helical CT venography will be defined by the limitations of CT angiography. These limitations include a large range of reported sensitivity (53 100%) and specificity (81 100%) [29] and, in particular, a small but significant number of examinations that reveal false-negative findings, are technically inadequate, or both [4 10]. Patients who would otherwise be denied appropriate anticoagulation or would require another test before undergoing anticoagulation are the ones who might truly benefit from helical CT venography. In multiple studies, false-negative results range from 0.5% to 20% and technically limited examinations range from 0.7% to 10% [4 10]. In our series, only one CT angiographic examination (1.3%) was reported as technically not interpretable. However, 20 of these CT angiography reports (27%) described a degree of limitation because of respiratory motion. Therefore, until CT angiography shows consistently improved sensitivity, specificity, and insensitivity to patient motion, helical CT venography will likely have some role in the evaluation of patients with suspected pulmonary embolism. lthough the recent study of Loud et al. [17] indicates perfect agreement between helical CT venography and sonography in the diagnosis of deep venous thrombosis when experts interpret studies, we suggest a conservative approach to the use of helical CT venographic results at this time. We recommend that in patients without pulmonary embolism a positive finding on helical CT venography for deep venous thrombosis be confirmed on sonography before anticoagulation is undertaken. Furthermore, helical CT venography interpretation should be performed 1530 JR:175, December 2000

7 Helical CT Venography of the Lower Extremity Veins with a knowledge of certain pitfalls. These include ill-defined central venous filling defects seen on single images, which may represent valves, and prolonged arterial enhancement, which may represent extensive bilateral deep venous thrombosis. Further prospective studies of CT angiography and helical CT venography, especially with the widespread implementation of multidetector helical CT, will be required to fully delineate the role of these techniques in the evaluation of patients with suspected pulmonary embolism. References 1. Dalen JE, lpert JS. Natural history of pulmonary embolism. Prog Cardiovasc Dis 1975;17: National Institutes of Health Consensus Conference. Prevention of venous thrombosis and pulmonary embolism. JM 1986;256: Sostman HD, Ravin CE, Sullivan DC, Mills SR, Glickman MG, Dorfman GS. Use of pulmonary angiography for suspected pulmonary embolism: influence of scintigraphic diagnosis. JR 1982; 139: Goodman LR, Curtin JJ, Mewissen MW, et al. Detection of pulmonary embolism in patients with unresolved clinical and scintigraphic diagnosis: helical CT versus angiography. JR 1995; 164: Remy-Jardin M, Remy J, Petyt L, et al. Diagnosis of acute pulmonary embolism with spiral CT: comparison with pulmonary angiography and scintigraphy. Radiology 1996;200: Teigen CL, Maus TP, Sheedy PF, et al. Pulmonary embolism: diagnosis with contrast-enhanced electron-beam CT and comparison with pulmonary angiography. Radiology 1995;194: VanRossum, Pattynama PM, Treurniet FE, et al. Spiral CT angiography for detection of pulmonary embolism: validation in 124 patients. Radiology 1996;201: Mayo JR, Remy-Jardin M, Muller NL, et al. Pulmonary embolism: prospective comparison of spiral CT with ventilation-perfusion scintigraphy. Radiology 1997;205: Ferretti GR, osson JL, uffaz PD, et al. cute pulmonary embolism: role of helical CT in 164 patients with intermediate probability at ventilation-perfusion scintigraphy and normal results at duplex US of the legs. Radiology 1997;205: Kim KI, Muller JL, Mayo JR. Clinically suspected pulmonary embolism: utility of spiral CT. Radiology 1999;210: Hull R, Hirsch J, Sackett DL, et al. Clinical validity of a negative venogram in patients with clinically suspected venous thrombosis. Circulation 1981;64: Naidich J, Feinberg W, Karo-Harman H, et al. Contrast venography: a re-assessment of its rate. Radiology 1988;168: Cronann JJ, Dorfman G, Grusmark J. Lower-extremity DVT: further experience with and refinements of US assessment. Radiology 1988;168: Loud P, Grossman ZD, Klippenstein DL, Ray CE. Combined CT venography and pulmonary angiography: a new technique for suspected thromboembolic disease. JR 1998;170: Stehling MK, Rosen MP, Weintraub J, Kim D, Raptopoulos V. Spiral CT venography of the lower extremity. JR 1994;163: aldt MM, Zontsich T, Stumpflen, et al. Deep venous thrombosis of the lower extremity: efficacy of spiral CT venography compared with conventional venography in diagnosis. Radiology 1996;200: Loud P, Katz DS, Klippenstein DL, Shah RD, Grossman ZD. Combined CT venography and pulmonary angiography in suspected thromboembolic disease: diagnostic accuracy for deep venous evaluation. JR 2000;174: Redman HC. Deep venous thrombosis: is contrast venography still the gold standard? Radiology 1988;168: Holm H, Finnanger, Hartmann, et al. Heparin treatment of deep venous thrombosis in 280 patients: symptoms related to dosage. cta Med Scand 1984;215: Shah, uckshee N, Yankelevitz DF, Henschke CI. ssessment of deep venous thrombosis using routine pelvic CT. JR 1999;173: enedict KT, Wheeler H, Patwardhan N. Impedance plethysmography: correlation with contrast venography. Radiology 1977;125: Comerota J, White JV, Katz MI. Diagnostic methods for deep venous thrombosis: venous Doppler examination, phleborheography, iodine- 125 fibrinogen uptake, and phlebography. m J Surg 1985;150 [suppl]: Huisman MV, uller HR, ten Cate JW, Vreeken J. Serial impedance plethysmography for suspected deep venous thrombosis in outpatients: the msterdam General Practitioner Study. N Engl J Med 1986;314: Hull R, Hirsh J, Sackett DL, Stoddard G. Cost effectiveness of clinical diagnosis, venography, and noninvasive testing in patients with symptomatic deep vein thrombosis. N Engl J Med 1981;304: McLachlan MSF, Thomson JG, Taylor DW, Kelly ME, Sackett DL. Observer variation in the interpretation of lower limb venograms. JR 1979; 132: Cronan JJ, Dorfman GS, Scola FH, et al. Deep venous thrombosis: US assessment using vein compression. Radiology 1987;162: Lensing W, Prandoni P, randjes D, et al. Detection of deep venous thrombosis by real time mode ultrasonography. N Engl J Med 1989;320: axter GM, McKechnie S, Duffy P. Colour Doppler ultrasound in deep venous thrombosis: a comparison with venography. Clin Radiol 1990; 42: Rathbun SW, Raskob GE, Whitsett TL, et al. Sensitivity and specificity of helical computed tomography in the diagnosis of pulmonary embolism: a systematic review. nn Intern Med 2000;132: JR:175, December