In the United States, estimates suggest that 60,000–100,000 patients die of venous thromboembolism, including deep venous thrombosis (DVT) and pulmonary emboli, annually [
1]. Furthermore, one third of patients who have had a DVT will suffer long-term complications of postthrombotic syndrome, such as pain, swelling, discoloration, and scaling in the affected limb, which affect quality of life [
1]. Furthermore, some patients require long-term anticoagulation, subjecting them to increased risk of bleeding. The exact cost of venous thromboembolism is difficult to pinpoint but has been estimated at $2 billion to $10 billion annually [
1]. Although conventional venography has historically been the reference standard for the diagnosis of DVT in the lower extremities, duplex ultrasound has proven to be extremely sensitive and specific and is widely used because of its convenience, accessibility, and noninvasiveness [
2,
3]. However, ultrasound is limited in patients with casts, obesity, or edema. Ultrasound is also limited in diagnosing abdominopelvic DVT [
4,
5].
MRI has proven to be a valuable modality for DVT imaging using numerous techniques [
2]. The most frequently reported technique is a 2D time-of-flight technique [
6–
11]. Studies have found this technique to be highly sensitive and specific for the detection of DVT compared with conventional venography [
11,
12]. However, image acquisition is time consuming, and flow-related artifacts can be difficult to distinguish from true filling defects.
Contrast-enhanced (CE) MR angiography is considered superior to unenhanced MR angiography for most vascular distributions [
13–
17]. However, traditional IV gadolinium-based agents are not routinely used for lower extremity venous imaging. One of several limitations of CE MR venography (MRV) relative to MR arteriography is the dilution of the contrast bolus during the arterial phase despite the use of large quantities of contrast agent [
18]. To avoid this limitation, some authors have advocated direct MRV imaging, which has required the use of meticulous technique with pedal access, tourniquets, or complex software [
13,
18,
19] and, thus, is not routinely performed. Gadofosveset is a blood-pool contrast agent that reversibly binds to human serum albumin, resulting in a markedly extended intravascular retention time, prolonged imaging half-life, and improved magnetic resonance relaxivity (package insert, Ablavar, Lantheus Medical Imaging) [
20–
22]. These characteristics may theoretically overcome the limitations of traditional gadolinium-based contrast agents for imaging the lower extremity venous system. Therefore, the purpose of this study is to assess the diagnostic accuracy and performance characteristics (e.g., qualitative assessment of vessel visualization, contrast-to-noise ratio [CNR], and image acquisition time) of gadofosveset-enhanced MRV for the detection of DVT in the abdominopelvic and lower extremity venous system in comparison with a well-established and proven unenhanced gradient-recalled echo (GRE) MRV sequence.
Materials and Methods
This retrospective study was HIPAA compliant and approved by our institutional review board. A waiver of consent was obtained for review of patient records and images.
Patient Characteristics
Between March 23, 2010, and November 15, 2010, 30 consecutive patients were referred for an abdominopelvic or lower extremity MRV at our institution for evaluation of suspected DVT in the abdomen, pelvis, and lower extremities. The study population consisted of 15 women (mean [± SD] age, 57.7 ± 17.9 years; range, 20–77 years) and 15 men (mean age, 47.6 ± 20.3 years; range, 18–74 years). Specific indications for the examinations included lower extremity swelling (n = 10), stroke with concern for paradoxical emboli (n = 9), prior lower extremity DVT (n = 6), venous reflux disease with concern for iliac venous occlusion (n = 2), pulmonary hypertension (n = 2), and calf pain (n = 1). Patients with a glomerular filtration rate less than 60 mL/min/1.73 m2 or severe contrast agent allergy did not receive gadolinium-based contrast agent.
MRV Techniques
All MRV examinations were performed on a 1.5-T scanner (Avanto, Siemens Healthcare [n = 5]; or Signa HDx, GE Healthcare [n = 25]). Patients were imaged from the diaphragm to the proximal calf. The FOV ranged from 32 to 36 cm2. For the Avanto system, a 12- or 16-channel phased-array body coil was used. For the Signa HDx system, two six-channel body matrix coils were used in conjunction with a peripheral vascular coil. Respiratory gating was not used.
GRE Pulse Sequence
All patients were scanned in the supine position in three or four different stations in the axial plane. Sections were 5 mm thick with a 5-mm intersection gap [
12]. For the Avanto system, the following imaging parameters were used: TR/TE, 50/2.3; slice thickness, 5 mm; distance between slices, 10 mm; number of slices, 120; acquisition matrix, 320 × 320; flip angle, 75°; and bandwidth, 320 Hz/pixel. For the Signa HDx system, the following imaging parameters were used: TR/TE, 34/13; slice thickness, 5 mm; distance between slices, 10 mm; number of slices, 120; acquisition matrix, 256 × 128; flip angle, 60°; and bandwidth, 31.25 Hz/pixel.
CE T1-Weighted Pulse Sequence
CE steady-state imaging was performed with contiguous 5-mm axial sections 5 minutes after the administration of 0.12 mL/kg (0.03 mmol/kg) of gadofosveset trisodium (Ablavar, Lantheus Medical Imaging) followed by a 20–30-mL saline flush through a peripheral upper extremity IV catheter. For the Avanto system, the following imaging parameters were used: TR/TE, 5.5/2.6; slice thickness, 4 mm; distance between slices, 0 mm; number of slices, 240; acquisition matrix, 384 × 307; flip angle, 10°; bandwidth, 305 Hz/pixel; and parallel imaging acceleration factor, 2. For the Signa HDx system, the following imaging parameters were used: TR/TE, 3.5/1.7; slice thickness, 6 mm (interpolated to 4 mm); distance between slices, 0 mm; number of slices, 240; acquisition matrix, 288 × 192; flip angle, 12°; bandwidth, 71.2 Hz/pixel; and parallel imaging acceleration factor, 2 (performed only for the abdomen station).
Image Interpretation
The imaging studies were retrospectively reviewed by six radiologists: two attending radiologists who routinely interpret MR angiography, two board-certified radiologists in fellowship training, and two junior radiology residents. The two attending radiologists have over 13 and 3 years of experience, respectively, interpreting MRV for detection of DVT in a large academic setting. The two radiologists in fellowship were enrolled in vascular-interventional and body-imaging fellowships, respectively. At our institution, responsibility for interpreting MRV studies is shared equally between the sections of vascular-interventional radiology and body imaging. Each radiologist evaluated the image datasets individually, and all the radiologists were blinded to the clinical history and were randomized as to which image dataset (GRE or CE MRV) would be reviewed first. Two weeks elapsed before review of the second image dataset to minimize recall bias. The venous system was divided into 13 segments: inferior vena cava (IVC; n = 25), right common iliac vein (n = 29), left common iliac vein (n = 29), right external iliac vein (n = 27), left external iliac vein (n = 28), right internal iliac vein (n = 30), left internal iliac vein (n = 30), right common femoral vein (n = 29), left common femoral vein (n = 29), right femoral vein (n = 29), left femoral vein (n = 28), right popliteal vein (n = 29), and left popliteal vein (n = 28). Segments containing an IVC filter, stent, or metallic prostheses were excluded (n = 20).
Qualitative Analysis
Each venous segment was assessed for thrombus, which was characterized as acute or chronic. For the purposes of this study, an acute DVT was defined as an occlusive or nearly occlusive filling defect within a vein of equal or greater caliber than the unaffected contralateral vessel, proximal segment, or distal segment, in that order of preference [
23,
24] (
Fig. 1). Chronic DVT was defined as nonocclusive, eccentrically adherent thrombus, web, or occluded vein with diminished caliber [
23] (
Fig. 2). The presence of collateral veins favored chronicity. In the event that segments contained both acute and chronic DVT, radiologists were instructed to rate a segment as simply acute DVT, because this was most pertinent to treatment algorithms. The segments were also assigned a score on a scale of 1–4 for vein visualization (1 = poor, 2 = fair, 3 = good, and 4 = excellent), luminal signal homogeneity (1 = heterogeneous, 2 = somewhat homogeneous, 3 = fairly homogeneous, and 4 = very homogeneous), and confidence pertaining to the presence or absence of thrombus (1 = unsure, 2 = mildly confident, 3 = moderately confident, and 4 = very confident).
Reference Standard
If a duplex ultrasound was obtained within 3 days of the MRV, the interpretation served as the reference standard (
n = 16 patients). In our study, ultrasound detected eight positive segments and 79 negative segments. After review of our electronic medical records, no changes occurred between the physical examination and ultrasound to suggest the interval development of DVT in any of these patients. If no ultrasound was performed, the reference standard was determined in consensus by the two attending radiologists using all radiology studies and clinical information available (
Fig. 3).
Quantitative Analysis
CNR measurements were performed on each patent venous segment. The signal intensity of the venous segment (SIv) was determined by calculating the mean of three evenly distributed ovoid regions of interest (ROIs) within the vessel segment, encompassing as much lumen as possible. Background image noise was determined by calculating the average SD (SDbg) of three 500-pixel ROIs in artifact-free areas of air at the level of the second vascular ROI. The signal intensity of muscle (SIm) was obtained by placing three ovoid freehand 100-mm2 ROIs within muscles at the level of the venous ROIs. CNR was then calculated as follows: (SIv − SIm) / SDbg.
Image acquisition time was obtained from the DICOM headers. Each reader recorded the time for interpretation of each study in minutes.
Statistical Analysis
Statistical analyses were performed utilizing SPSS software (version 19.0.0, IBM). The mean sensitivity and specificity for the detection of acute and chronic DVT for GRE and CE MRV for acute DVT and chronic DVT were calculated and stratified according to reader experience level. The sensitivities and specificities between GRE and CE MRV were compared using the McNemar test. Comparison of the vein visualization score, venous homogeneity score, interpretation times, acquisition times, and CNR between GRE and CE MRV were calculated using the Wilcox-on signed rank test. The image acquisition times and interpretation times were compared using the paired Student t test. Interobserver variability between the two attending radiologists was calculated using the linear-weighted Cohen kappa statistic. Statistical significance was defined as p ≤ 0.05.
Discussion
Given the limitations of vascular ultrasound in the abdomen, alternative noninvasive cross-sectional imaging modalities are an important necessity for the detection of DVT in the IVC and iliac veins, as well as for patients who are unable to undergo adequate sonographic evaluation of the lower extremities because of post-surgical wounds, casts, excessive edema, or obesity. Although unenhanced MRV using a time-of-flight sequence has been shown to be highly sensitive and specific for the detection of DVT, the image acquisition time is lengthy, and the images are prone to flow-related artifact, which can require considerable experience to differentiate from filling defects [
5,
11]. Our study found that CE MRV using a blood-pool agent provides excellent sensitivity and specificity for detection of DVT that are higher than those of a conventional time-of-flight sequence. The qualitative assessments of vein visualization, venous signal homogeneity, and reader confidence were significantly improved with CE MRV compared with the GRE sequence. Finally, the radiologist interpretation time and image acquisition time were significantly improved with contrast enhancement.
MRV for the detection of abdominopelvic and lower extremity DVT has been shown with varying techniques to be highly sensitive and specific [
12,
25–
29]. Enden et al. [
30] compared DVT imaging with gadofosveset to a novel balanced turbo field-echo sequence in six patients and 15 healthy volunteers and found the sensitivity and specificity of the two techniques to be identical. However, the balanced turbo field-echo sequence has not been validated in comparison with a reference standard; thus, the accuracy of both techniques is uncertain.
Although MR angiography of nearly all arterial and venous distributions has been shown to be superior with contrast enhancement compared with various unenhanced techniques [
13–
17], CE MRV of the lower extremities is not commonly performed. Infusion of conventional extracellular gadolinium-based agents from an arm IV typically achieves a relatively low concentration of contrast in the lower extremity veins because of the slow venous blood flow while the patient is lying motion-less on the MRI scanner platform and because the contrast agent becomes progressively diluted in the blood pool and is rapidly cleared by the kidneys. The only reported study using an upper extremity peripheral IV catheter to perform CE MRV with a conventional extra-cellular gadolinium-based agent for the detection of lower extremity DVT relied on specialized image subtraction software and a vacuum fixation bean bag to minimize leg movement [
19]. Although direct CE MRV with conventional extracellular gadolinium-based agents has been shown to provide excellent opacification of the lower extremity venous system, this technique relies on insertion of an IV catheter into a pedal vein, with direct injection to image the lower extremity veins. Unfortunately, this practice is difficult in patients with a swollen extremity (a typical indication for evaluation of DVT) and incurs a significant infection risk [
31].
Although our results show that gadofosveset-enhanced MRV is superior to unenhanced MRV in terms of accuracy for detection of DVT and qualitative assessment, the differences, although statistically significant, are not very large. This finding supports the previous evidence that GRE MRV is an accurate and reliable technique. However, the difference in radiologist interpretation time was substantial. Interpretation of CE MRV studies took approximately half the time required for GRE MRV studies. Furthermore, image acquisition time for the CE MRV images required less than one third of the image acquisition time for the GRE MRV images. The sensitivity and specificity for detection of DVT and confidence in the interpretation improved for all readers, regardless of experience level, including the junior level residents, who may be interpreting MRV studies during overnight shifts.
CT venography has been reported to have a pooled sensitivity of 71–100% for detection of DVT in the pelvis and lower extremities, and, in some centers, it is commonly performed in combination with CT angiography of the pulmonary arteries for suspected venous thromboembolism [
32]. However, a substantial radiation dose to the pelvis and thighs in the range of 3.2–9.1 mSv has been reported [
33]. Furthermore, the diagnostic quality of CT venography was reported to be insufficient in 11% of patients [
34]. Although efficiencies of scale can be accomplished with concurrent performance of CT venography of the pelvis and lower extremities along with pulmonary CT angiography, these limitations should be considered when choosing between CT venography and MRV.
Several limitations of this study are notable. First, we did not have conventional venography for comparison as the reference standard. The reference standard for our study was formulated from all available imaging modalities and ultra-sound imaging when available. However, both ultrasound and GRE MRV have been shown in multiple studies to have extremely high sensitivities and specificities compared with conventional venography. Furthermore, at our institution, conventional venography is no longer performed for the diagnosis of DVT. Therefore, we think that, in aggregate, using all available imaging is reasonable to serve as a reference standard with which to calculate sensitivities and specificities. Furthermore, this study is retrospective in nature, with the incumbent limitations. We attempted to make a distinction between acute versus chronic DVT, because the level of acuity affects treatment. However, the definition of acute versus chronic DVT based on cross-sectional imaging is not well delineated in the literature [
23,
24]. Our study was designed to evaluate the diagnostic accuracy and imaging characteristics of gadofosveset with respect to DVT. Before it is accepted, further studies are needed to evaluate its cost effectiveness. Finally, the number of segments containing DVT was not robust. This can lead to a less reliable sensitivity calculation, although the specificity should be relatively reliable.
In conclusion, CE MRV using a gadolinium-based blood-pool contrast agent qualitatively improved vessel visualization, signal homogeneity, and confidence levels for detecting DVT. The sensitivity and specificity for the detection of acute and chronic DVT were also improved compared with a conventional unenhanced sequence. Finally, significant improvement in efficiency was gained, in terms of image acquisition and radiologist interpretation time. These numerous advantages suggest that CE MRV with a blood-pool agent is preferable over unenhanced GRE MRV for patients able to receive gadolinium.