Chest Imaging
Thromboembolic Disease
Comparison of Combined CT Pulmonary Angiography and Venography with Bilateral Leg Sonography in 70 Patients
Audio Available | 
Share
OBJECTIVE. The purpose of this study was to compare combined CT pulmonary angiography and venography with leg sonography for accuracy and relative efficacy in diagnosis of deep venous thrombosis from the popliteal vein to the common femoral vein.
SUBJECTS AND METHODS. Seventy consecutive patients with clinically suspected pulmonary embolism underwent both combined CT pulmonary angiography and venography and bilateral leg sonography within 24 hr. CT venograms were analyzed independently in a blinded fashion for quality of venous opacification and patency by two observers. CT venography was compared with sonography for femoropopliteal vein thrombosis, and the final assessment based on multiple subjective and objective clinical and imaging criteria was recorded in three categories: 1, CT venography better than sonography; 2, CT venography equivalent to sonography; and 3, sonography better than CT venography.
RESULTS. Sixty-eight patients (97%) had a satisfactory or good quality CT venography examination. Two CT venography studies had false-positive findings due to flow artifacts. Both CT venography and sonography had positive findings for deep venous thrombosis in five patients, and both had negative findings in 63 patients (100% sensitivity, 97% specificity, 100% negative predictive value, and 71% positive predictive value). CT venography was better and more efficacious than sonography (category 1) in 25 patients (36%). CT venography was equivalent to sonography (category 2) in 26 patients (37%), and sonography was better than CT venography (category 3) in 19 patients (27%).
CONCLUSION. Compared with sonography, CT venography in addition to CT pulmonary angiography is a relatively accurate method for evaluation of femoropopliteal venous thrombosis. Combined CT pulmonary angiography and CT venography may be more efficacious than sonography or two separate examinations in selected patients.
| Introduction |   |
|---|
Deep venous thrombosis (DVT) and pulmonary embolism, both a form of thromboembolic disease, constitute major health problems that result in significant morbidity and mortality. It is estimated that DVT and pulmonary embolism are associated with 300,000-600,000 hospitalizations a year in the United States and that as many as 50,000 individuals die each year as a result of pulmonary embolism [1].
Both DVT and pulmonary embolism are often difficult to detect on clinical examination. Although ventilation—perfusion lung scanning remains a commonly performed screening test for evaluation of pulmonary embolism, helical CT has emerged as a more accurate diagnostic test [2, 3] and is commonly used as the first and only screening test in patients with abnormal findings on chest radiography [4, 5]. Sonography of the legs is the most commonly used and is usually the only screening test for evaluation of DVT. Multiple tests are often performed including leg sonography, ventilation—perfusion scanning, helical CT, and, rarely, CT venography to evaluate for thromboembolic disease. Depending on the availability of equipment and personnel, these tests may take many hours to complete.
Helical CT has the potential to be an efficacious and reliable single examination for evaluation of both pulmonary embolism and DVT when indicated in selected patients. Although the technique of combined CT venography and CT pulmonary angiography has recently been studied [6, 7] in a small number of patients, to our knowledge a large prospective study comparing CT venography and pulmonary angiography with bilateral leg sonography has not been reported. We performed a prospective study in 70 patients comparing CT venography performed in addition to CT pulmonary angiography with bilateral leg sonography for femoropopliteal (infrainguinal) DVT. The purpose of our study was twofold: first, to assess the accuracy of CT venography using leg sonography as the standard of reference, and second, to determine the subset of patients in whom CT venography in addition to CT pulmonary angiography may be considered instead of performing two separate tests.
| Subjects and Methods | |
|---|
Between March 3, 1999, and September 27, 1999, 72 consecutive patients referred for helical CT to evaluate clinically suspected pulmonary embolism (48 acute, 21 chronic, and three acutely chronic) were given an opportunity to enter the study. All patients signed a written consent form approved by the hospital human subjects committee and institutional review board. One patient referred from the emergency department at night did not undergo leg sonography and was discharged after the CT venography and pulmonary angiography both had negative findings. Another patient had left leg sonography before CT pulmonary angiography and the observers were aware of the sonographic results. These two patients were excluded from the study. The remaining 70 patients, who underwent bilateral leg sonography within 24 hr (mean, 4 hr) after helical CT, formed the study group. No patient underwent conventional venography. The study patients included 65 men and five women, who were 34-81 years old (mean, 60.74 years). Medical charts were reviewed to record each patient's body weight, whether the patient was an inpatient or an outpatient at the time of study, and any significant comorbid conditions.
CT scans were obtained with a CT unit (Model 2000; Picker International, Cleveland, OH). The CT pulmonary angiography protocol that we used has been described [3, 5]. The CT pulmonary angiography parameters were 3-mm collimation, pitch of 2, mean z-axis coverage from aortic arch to the level of at least mid ventricles in a single breath-hold, 200 mA, and 130 kVp. Iodinated contrast material was administered as a bolus with an automated injector (MCT Plus; Medrad, Pittsburgh, PA). The injection was carefully monitored by a registered nurse or a physician. A total of 100-150 mL of undiluted iopramide solution (Ultravist 300; Berlex Laboratories, Wayne, NJ) or of iothalamate meglumine (Conray 60; Mallinckrodt Medical, St. Louis, MO) was injected at a rate of 4 mL/sec with a delay of 15-20 sec before scanning. A delay of 25 sec was used in one patient in whom IV access at the ankle was used. Three minutes after the start of injection, 10-mm-thick nonhelical axial images were acquired at 20-mm intervals from the knees to the renal veins. Depending on the height of the patient, 34-37 CT venograms were acquired. Each CT venogram took approximately 2-3 min.
Bilateral leg sonography was performed generally with a 5- or 7-MHz linear array transducer (Logic 700, General Electric Medical Systems, Milwaukee, WI; or HDI 3000, Advanced Technology Laboratories, Bothell, WA) by experienced sonographers. The main diagnostic criterion used for DVT was loss of venous compressibility. Change in venous spectral waveforms or color Doppler sonograms was used as supportive evidence of DVT. The sonography scan time was the same as the time to evaluate the deep veins with compression in patients in whom this technique was adequate (average, 5 min each leg). The scanning time included the time to assess the veins with positive criteria (spectral and color Doppler imaging) if these techniques had to be used to assess patency (average, 20 min each leg). At least one of the radiologists who interpreted the CT venography study prospectively was present during sonography for all patients except one in whom the sonography was performed by the on-call resident at night. Adequacy of depiction of the common femoral vein; proximal, middle, and distal superficial femoral vein; and the popliteal vein; and relative ease of performance of sonography versus CT venography were assessed subjectively and objectively. The CT venography and sonography were compared, and the final assessment was recorded in three categories: CT venography better than sonography (category 1), CT venography equivalent to sonography (category 2), and sonography better than CT venography (category 3).
CT venograms were interpreted prospectively at the time of examination after viewing on the monitor and the hard copies. All CT venograms were also reviewed independently at a later date by a second observer who was unaware of the sonography study results. The observers independently completed a standard grading sheet. CT venograms (window width, 250-300 H; window level, 40-80 H) were graded for quality of venous opacifiaction. Image quality was considered good when a high degree of venous opacification was shown, satisfactory when the images were sufficient for analysis of venous patency without a high degree of opacification, and poor when venous patency could not be assessed. Visualized portions of the inferior vena cava, renal veins, portal vein, and iliac veins were also analyzed for patency, but the results were not included for statistical analysis in this study.
The criterion used to diagnose acute DVT was the presence of a definite intraluminal filling defect (Fig. 1). Additional findings suggestive of DVT were venous distention with the filling defect and perivenous fat stranding. Findings of chronic DVT such as small, thick-walled, poorly enhancing veins; heterogeneously enhancing veins; and presence of collaterals were looked for but were not used for statistical analysis.
 View larger version (110K) | Fig. 1. —47-year-old man with acute deep venous thrombosis. CT venogram obtained at distal thigh shows intraluminal filling defect (arrows) in distal superficial veins bilaterally. Note mild stranding of perivenous fat. Common femoral veins and right external iliac vein shown only on CT venography also were involved (not shown). |
The unit of analysis was patient rather than lower extremity. Sensitivity, specificity, and positive and negative predictive values were calculated by standard methods, and exact 95% confidence intervals (CIs) were calculated according to the binomial distribution for acute DVT from the common femoral vein to the popliteal vein. Fisher's exact test and the Mantel-Haenszel chi-square test for trend were used to assess association between patient's weight or volume of contrast material and the relative efficacy of CT venography versus sonography. A p value of less than 0.05 was considered statistically significant.
| Results | |
|---|
Twenty-one patients (31%) were inpatients at the time of study entry for various reasons including total hip or knee arthroplasty or other orthopedic surgery (n = 15) and coronary artery bypass graft or extremity vascular graft surgery (n = 6). Six patients had a history of pulmonary artery hypertension. Two patients had congestive cardiomyopathy. CT venography was better than sonography in 25 patients (36%), CT venography was equivalent to sonography in 26 (37%), and sonography was better than CT venography in 19 patients (27%). Compression sonography was difficult to perform, uncomfortable for the patient, or inadequate in 25 (36%) of 70 study patients because of various clinical reasons including obesity limiting visualization of distal superficial veins (n = 20), tenderness due to open wounds or recent saphenous vein harvesting (n = 5), tender purpuric lesions in one patient, and congested veins that were tender and difficult to compress in one patient with cardiomyopathy. CT venography was not as efficacious as sonography in 19 patients because of image degradation from flow artifacts (Fig. 2), beam hardening artifacts from orthopedic hardware, and suboptimal opacification due to inadequate delay before imaging. Three patients with below-knee amputation and atrophic musculature were better examined with sonography. No artifacts from orthopedic hardware were seen on sonography, but streak artifacts were seen on CT venography.
 View larger version (91K) | Fig. 2. —66-year-old man with flow artifact. CT venogram at mid thigh shows apparent central filling defect in right superficial vein (white arrow) that was seen on multiple contiguous images (not shown). Note more homogeneous and satisfactory opacification of left superficial vein (black arrow). |
Forty-four patients (63%) had satisfactory-quality and 26 patients (37%) had good-quality CT venography examinations. CT venography studies in two patients (2.9%) were false positive. Results are summarized in Tables 1,2,3.
Great saphenous vein thrombus was depicted on both CT venography and sonography in two patients and on CT venography alone in one patient in whom sonography was performed at night by the on-call resident and in whom this vein was not examined. Small saphenous vein thrombus was revealed in one patient on both CT venography and sonography. A small nonocclusive superficial DVT was seen only on CT venography (Fig. 3) in one patient who also had DVT in the left popliteal vein that was revealed on both examinations. Three patients had suprainguinal DVT revealed on CT in the right internal and common iliac vein (with negative findings for DVT on leg sonography), portal vein thrombosis (with positive DVT findings on leg sonography and CT venography), and left subclavian vein thrombosis (revealed indirectly by the presence of extensive collateral vessels in the left chest wall). None of the patients had isolated inferior vena cava DVT and none had positive findings on CT venography for DVT and negative findings on CT pulmonary angiography for pulmonary embolism. Twelve patients had positive findings on CT pulmonary angiography for pulmonary embolism; five of these patients had DVT revealed on both CT venography and sonography.
 View larger version (94K) | Fig. 3. —70-year-old man with tiny nonoccluding thrombus in left superficial vein. CT venogram at mid thigh shows nonoccluding filling defect (arrow) in posterior aspect of duplicated segment of left superficial vein. Note slower flow on right. |
The prospective sensitivity of CT venography compared with sonography from the popliteal to the common femoral vein was 100%; the 95% confidence interval was not estimable. The specificity was 97% (95% CI, 88-99%), positive predictive value was 71%, and the negative predictive value was 100%. The association between the patient's weight and the relative efficacy of CT venography versus sonography was not significant (Fisher's exact test, p = 0.63; Mantel-Haenszel chi-square test for trend, p = 0.25). The association between volume of contrast material and the relative efficacy of CT venography versus sonography was also not significant (Fisher's exact test, p = 0.68; Mantel-Haenszel chi-square test for trend, p = 0.47).
| Discussion | |
|---|
The sensitivity and specificity of CT venography when combined with CT pulmonary angiography was relatively high. There were two cases of false-positive findings due to flow artifacts. Flow artifacts due to nonhomogeneous mixing of contrast-enhanced and unenhanced blood, especially in the inferior vena cava at the level of the renal veins, are commonly observed on abdominal CT. Although flow artifacts can usually be easily differentiated from true filling defects by experienced observers, rarely these artifacts may be confused with DVT. Flow artifacts were more commonly seen in patients with pulmonary artery hypertension, cardiomyopathy, congestive heart failure, and peripheral artery disease. Longer delay before scanning could have resulted in better opacification of veins in this subset of patients similar to the longer delay before scanning proposed for these patients for CT pulmonary angiography [8]. Currently we use a delay of 4 min for patients with suspected slow flow or abnormal hemodynamic status.
Although the difference was not statistically significant, in eight (53%) of the 15 patients in the greater-than-100-kg weight group CT venography was scored better than sonography (category 1) (Table 2) compared with only four patients (27%) in whom sonography was more efficacious (category 3). The patient's weight is an important criteria for contrast material dosage but we did not compare these two parameters directly. A relatively large volume of contrast material to fill the capacious venous system may help decrease flow artifacts. Sixteen (37%) of 43 CT venography studies were graded good and better than sonography (category 1) in the later part of our study when we used 150 mL of contrast material compared with nine (33%) of 27 patients in whom 100 mL of contrast material was administered in the earlier part of our study, which was not a statistically significant difference. In our study we used undiluted contrast material similar to what is used in routine CT of the abdomen and pelvis. It is not clear whether diluted contrast material would mix better and result in fewer flow artifacts. One can only speculate that flow artifacts may be fewer with a large volume of diluted contrast material. Future studies with larger numbers of patients may clarify what if any influence these factors have on flow artifacts.
Although none of the CT venography studies had false-negative findings, when compared with sonography, in one patient, two images, one showing the right iliac vein and the other showing the confluence of iliac veins, were interpreted as possible DVT by one observer and as negative for DVT by the second observer. This patient underwent abdominal and pelvic CT 3 days later. The routine pelvic CT scan clearly showed a filling defect at the confluence of the iliac veins (Fig. 4A,4B).
 View larger version (94K) | Fig. 4A. —66-year-old man with acute deep venous thrombosis (DVT). CT venogram at confluence of iliac veins shows vague filling defect (arrow) in left aspect of confluence. One observer interpreted this defect as suggestive of DVT, and another observer interpreted it as negative for DVT. |
 View larger version (93K) | Fig. 4B. —66-year-old man with acute deep venous thrombosis (DVT). CT scan of pelvis obtained 3 days after A at same level clearly shows intraluminal filling defect (arrow) at confluence. |
The CT venography technique in our study was different from the technique described by Loud et al. [6, 9]. The scan interval was 2 cm in our study compared with 5 cm, and we did not image the calf veins. The importance of isolated calf vein DVT as the cause of clinically important pulmonary embolism or persistent lower extremity symptoms has been a subject of debate in the literature [10]. Kakkar et al. [11] examined 132 consecutive patients during the postoperative period to determine the natural history of DVT. Of 40 patients who developed DVT, pulmonary embolism developed in four who had DVT in the femoropopliteal veins. Subsequent complications were not seen in any of the patients with clots limited to the calf veins [11]. In a retrospective study of 283 patients, Gottlieb et al. [12] concluded that sonography of the calf is unnecessary at initial examination to identify patients at risk of clinically important pulmonary embolism or propagation of DVT into the thigh. A study published in 1996 reported that a limited compression sonography study that examined only the popliteal and common femoral veins was sufficient to detect most significant DVT [13]. Small nonocclusive DVT involving short venous segments especially in asymptomatic patients has been reported, although the clinical significance is not clear [14, 15]. This information could translate into different scanning parameters for CT venography, for which additional studies may provide answers in the future.
The sensitivity for CT venography may not be reliable given the relatively small number of patients in whom both CT venography and sonography had positive findings for DVT. A large number of CT venography examinations were negative for DVT in our study. This is at least partly explained by the study entry criteria. One third of the study patients, especially those with a history of chronic thromboembolic disease, were being treated with anticoagulant therapy. DVT was not clinically suspected in most of our patients. This fact is in agreement with a prospective study reported recently in which none of the 89 patients in group 1 (no symptoms or risk factors) were found to have DVT, yielding a prevalence of 0% [16]. Although a large number of patients in our study, especially those with orthopedic surgery including total hip or knee arthroplasty (n = 15), were at increased risk for DVT, the prevalence of DVT in this group was 0% (0/15), which is lower than a rate of 5-12% reported in literature [15]. This may be because in our institution these patients receive routine prophylaxis against DVT, and it is possible that small nonocclusive DVT was not detected on either CT venography or sonography. One can only speculate whether CT venograms at 1-cm rather than 2-cm intervals would have detected more nonocclusive small DVTs; however, performing CT venography at 1-cm intervals would double the number of images to evaluate, increase scanning time, and possibly increase the radiation dose. Narrower scan intervals and thinner collimation, such as 5-mm thick sections at 5-mm intervals, may become routine with the newer multidetector scanners. Analysis of cost-effectiveness and assessment of radiation dose was not an objective of this study.
Sonography as the standard of reference is the main limitation of our study because sonography is less than 100% accurate. The sensitivity of compression sonographic screening for proximal DVT is reported to be only 62% after total hip or knee arthroplasty [17]. Lack of symptoms due to small nonocclusive thrombi in these patients probably explains the low sensitivity of sonography. In clinical practice, contrast-enhanced venography is nearly obsolete; therefore, a relatively new technique for leg vein evaluation has to be compared with sonography, which is the standard of care and is highly accurate in symptomatic patients. In one patient, a tiny nonocclusive filling defect in the left superficial femoral vein was seen prospectively only on CT venography (Fig. 3). Despite streak artifacts from orthopedic hardware on some images, patients with orthopedic hardware compose a subset of patients in whom CT venography in addition to CT pulmonary angiography may still be considered because asymptomatic small nonoccluding thrombus may be revealed only on CT venography.
Another subset of patients in whom CT venography may be indicated are the intensive care unit patients who are intubated and in whom one cannot assess leg symptoms. The quality of CT pulmonary angiography in these patients is often limited by breathing artifacts. Additional information about patency of leg veins may be helpful to evaluate thromboembolic disease in these patients because the treatment for either pulmonary embolism or DVT is generally the same. Sonographic examination requested for intensive care unit patients is often bilateral and can be not only time-consuming but also challenging because of dressings, trophic changes, and lack of mobility in these patients.
Sonography was as efficacious as CT or better in thin mobile patients referred from the emergency department or outpatient clinics. In three patients with below-knee amputation and atrophic musculature, sonography was found to show better the small poorly enhancing veins. Physical proximity of sonography facility and availability of equipment and personnel could influence whether sonography is performed before referring a patient for helical CT.
We conclude that CT venography is reasonably accurate and is useful to exclude a diagnosis of DVT. It is fast and easy to perform. CT venography in addition to CT pulmonary angiography may be efficacious as a single test (one-stop examination) instead of two separate tests (i.e., sonography and CT) in select patients. The following indications for CT venography in addition to CT pulmonary angiography are based on results of this study: suspected thromboembolic disease in obese patients with moderate to severe bilateral leg edema, no symptoms in high-risk patients, and nondiagnostic CT pulmonary angiography in intubated patients.
Presented at the annual meeting of the American Roentgen Ray Society, May 2000, Washington, DC.
Address correspondence to K. Garg.
| References | |
|---|
