- Email: [email protected]
Magnetic resonance angiography of aorto-iliac disease
Magnetic Resonance Angiography of Aorto-Iliac Disease J. Shannon Swan, MD, Todd W. Kennell, MD, Charles W. Acher, MD, Dennis M. Heisey, PhD, Thomas M. Grist, MD, Frank R. Korosec, PhD, Mary E. Hagenauer, BA, Madison, Wisconsin
BACKGROUND: Four different techniques for aortoiliac magnetic resonance angiography (MRA) were assessed for accuracy using a digital subtraction angiography (DSA) gold standard. Surgeons’ confidence in their ability to generate treatment plans with MRA and DSA was assessed, in consultation with a radiologist. METHODS: Two different two-dimensional (2D) time-of-flight (TOF) sequences, a phase-contrast sequence, and a contrast-enhanced (CE) MRA sequence were used. Receiver operating characteristic (ROC) curves were plotted and areas (Az) calculated from radiologists’ readings. Surgeons’ confidence in their ability to utilize the images for treatment planning was assessed with a 5-point Likert scale. Thirty-six patients were evaluated. RESULTS: CE MRA had a sensitivity, specificity, and Az of .92, .93, and .96, respectively, for stenoses 50% or greater. CE MRA performed better than other sequences, but the improvement compared with gated 2D TOF was not statistically significant. Interobserver agreement for CE MRA and DSA yielded identical Kappa values. Surgeons were most confident in DSA, followed by CE MRA, which was significantly preferred to other techniques. CONCLUSIONS: CE MRA closely approximates DSA in terms of diagnostic accuracy. Surgeons considering treatment plans are confident in the CE MRA technique, relative to other MRA methods. Am J Surg. 2000;180:6 –12. © 2000 by Excerpta Medica, Inc.
S
ince early work by many investigators,1–3 it has been clear that two-dimensional (2D) time-of-flight (TOF) magnetic resonance angiography (MRA) is capable of imaging peripheral vascular disease (PVD). Roadblocks to previous implementation of MRA for eval-
uating PVD on a wide scale have been flow-dependent artifacts, the length of scan times, the need for dedicated MR receiver coils, and the inherent quality of the resulting images. With respect to artifacts, the pelvis has been a particular problem, owing to the orientation of the vessels and movement related to respiration and normal bowel peristalsis. Recent work is more encouraging. Studies now strongly indicate that gadolinium contrast-enhanced (CE) MRA is a good solution.4,5 CE MRA does not suffer as significantly from the above limitations. Thus, flow dependent artifacts are not as severe, and the vessel lumen detail is imaged wherever contrast is present. The resolution of CE MRA images can be considerably enhanced over previous techniques, and the inherent contrast characteristics of CE MRA images make the need for dedicated coils less of a factor. The result is much more like a conventional X-ray angiographic image, rendering easier interpretation for the many physicians previously trained with conventional angiographic methods. Of considerable importance to the patient is the less invasive character of MRA, thus suggesting a mechanism of improving short-term health outcomes related to PVD testing. Recent work has sought to quantify this improvement.6,7 This study is a multilevel technology assessment, according to the hierarchy proposed by Fryback and Thornbury.8 First, we explored the diagnostic accuracy of MRA in the pelvis using four different pulse sequences (Fryback and Thornbury level 2: “diagnostic accuracy efficacy”). We evaluated patients with techniques reflecting the varied solutions that have been attempted to optimize the imaging of the pelvis vasculature in recent years. Secondly, we wished to evaluate the effects of MRA imaging on the confidence of the ultimate end user of these images, the vascular surgeon (Fryback and Thornbury level 3: “diagnostic thinking efficacy”). This was done in a manner similar to what would be found in any typical healthcare scenario, where the radiologist and surgeon confer on the interpretative and clinical aspects of a case.
PATIENTS AND METHODS From the Departments of Radiology (JSS, TWK, TMG, FRK, MEH), Surgery (CWA, DMH), and Medical Physics (TMG, FRK), University of Wisconsin-Madison, Madison, Wisconsin. Dr. Swan has received support from the General Electric-Association of University Radiologists Fellowship, NIH R01 HL51370, and Nycomed Corporation. Requests for reprints should be addressed to J. Shannon Swan, MD, Indiana University Medical Center, Department of Radiology, 550 N. University Boulevard, Indianapolis, Indiana 46202. Manuscript submitted February 18, 2000, and accepted in revised form May 16, 2000.
6
© 2000 by Excerpta Medica, Inc. All rights reserved.
Patients Thirty-six patients were utilized for data analysis in this study. Thirty-one of them were male. The age range was 33 to 79 years, with a mean of 62. All patients were recruited from our community or surrounding areas, having been referred from surgical practitioners who were intending a vascular intervention for the patient, possibly involving the aorto-iliac area. Patients were enrolled in our study prospectively when they were scheduled to have a conventional angiogram by the referring physician. MRA images eventually acquired were not used to decide on whether a 0002-9610/00/$–see front matter PII S0002-9610(00)00412-8
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
patient would be scheduled for conventional angiography, if the MRA examination was obtained first. Informed consent was obtained using forms approved by our institutional review board. Angiographic Technique The conventional x-ray angiographic evaluation was done with digital subtraction angiography technique (DSA), with at least two opposite 30-degree oblique runs being acquired in each case when imaging the pelvis. All cases were intraarterial injections done with Seldinger technique at a groin entrance site, with the exception of two cases where total aortic occlusion necessitated a translumbar approach. Nonionic contrast was used in all patients, and preprocedure sedation was utilized. MRA Techniques The MRA techniques included cardiac gated and nongated two-dimensional (2D TOF), a cardiac-gated three dimensional (3D) phase-contrast (PC) approach, and a modified 3D CE MRA pulse sequence. All studies were performed using a General Electric (Signa Advantage, Milwaukee, WI) 1.5 Tesla unit with the body coil. The 2D TOF sequence used was the product General Electric method in clinical use at the time of the study. Scan time was 5 to 7 minutes depending on the number of slices acquired. The cardiac-gated 2D TOF sequence was written by modifying a basic 2D TOF sequence to allow gating and segmentation. Scan time was equal to 800 to 1,000 heartbeats (approximately 13 to 17 minutes). Scan time was shortened by using a rectangular field of view (FOV) when possible. In the cardiac-gated 3D PC sequence, scan time was generally 514 heartbeats (approximately 8.5 minutes at a heart rate of 60 beats per minute, approximately 6.5 minutes at 80 beats per minute). A series of three scans was performed to complete the CE MRA technique. First, a contrast transit time scan was obtained that consisted of a sagittal acquisition set-up to image a single slice location through the abdominal aorta at a rate of one image per second for 60 seconds. Then, 2 cc gadodiamide (Omniscan, Nycomed, Princeton, NJ) was injected intravenously via a small-bore peripheral catheter over 0.5 seconds, followed by a 15 cc saline flush at a rate of 1.5 cc/sec. This series of images showed the transit time of contrast agent from the arm to the aorta. Second, a coronal 3D scan was then obtained but no contrast was injected, so the data could be used for subsequent subtraction of nonvascular tissue signal. For the third scan, an injection of a gadodiamide bolus (0.2 to 0.3 mmol/kg body weight) at 1.5 cc/sec began the process, according to the transit time results. Scan times for the second and third scans were 30 to 45 seconds. Postprocessing of all of the MRA images was done to generate projections at 10-degree increments. Source images were available for confirmation of difficult readings. Image subtraction was done on an Advantage Windows workstation (General Electric) followed by the generation of reprojected images. Blinded Readings by Radiologists In each patient, the evaluation ideally included the infrarenal aorta, the right and left common iliac arteries, the
right and left external iliac arteries, the right and left common femoral arteries, the right and left superficial femoral arterial origins, and the right and left profunda femoral arterial origins. For the ungated 2D TOF acquisitions only, imaging continued distally, including the thighs to the popliteal trifurcations. This was done to facilitate the surgeons’ diagnostic thinking accuracy exercise that is further explained below. The vascular segments distal to the proximal superficial femoral and profunda femoral arteries were not included in the diagnostic accuracy analysis. Two experienced radiologists were given the images to evaluate in blinded fashion. All MRA images of the patients were scored by each reader independently, one pulse sequence at a time over a period of months. The DSA images were evaluated as the final group. The following scale was used for rating the disease in each vessel segment: 0 ⫽ normal, 1 ⫽ minimal disease, all stenoses being ⬍50% of the vessel diameter, 2 ⫽ one ⱖ50% stenosis, 3 ⫽ more than one ⱖ50% stenosis, 4 ⫽ complete occlusion of the affected vessel segment. This scheme is similar to the grading used in widely cited trials.1,3 Diagnostic accuracy was computed by first designating two cut points. We evaluated for significant disease by designating ratings greater than “1” from the disease scale mentioned above as being abnormal. Secondly, we assessed for complete vessel segment occlusions by designating readings greater than “3” as being abnormal. Receiver operating characteristic (ROC) curve areas were calculated for both significant disease and occlusion by pulse sequence, using pooled data from two readers. DSA was the gold standard. ROC data were computed with a nonparametric approach utilizing the trapezoid rule.9 Confidence intervals were constructed about the estimates of the area under the ROC curves, sensitivities and specificities by bootstrapping.10 Pairwise significance tests were performed by first constructing bootstrap confidence intervals about the differences of the parameter estimates of interest. The estimates were declared significantly different if the confidence interval did not include zero. We assessed interobserver agreement by computing Kappa11 and weighted Kappa12 statistics. Agreement weights were defined as in work by Cohen12 so that full agreement between readers was given a weight of 1, and partial agreement was given a weight of (1 ⫺ [r1 ⫺ r2] / 4). Asymptotic variances were calculated as in Fleiss et al.13 Surgeons’ Confidence in Imaging Techniques We evaluated surgeons’ perceived confidence in their ability to generate a treatment plan with the MRA images plus clinical information, in consultation with a radiologist. Clinical information consisted of all information that might be available prior to angiographic evaluation, including all vascular laboratory data, as well as physical examination data and clinical history. Although hemodynamic measurements obtained at angiography are obviously useful in aortoiliac evaluation, we felt they could serve as a possible source of bias against MRA, given that this measurement is typically obtained at, or associated with, angiography. Also, the decision to obtain gradient measurements and entertain other treatment, possibly in
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
7
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
TABLE I Clinically Significant Disease (≥50% stenosis): Diagnostic Accuracy with Four Different Pulse Sequences* MRA Method
ROC Az
95% CI
Sensitivity
95% CI
Specificity
95% CI
Gated 2D TOF Gated 3D PC 3D CE MRA 2D TOF
0.94 0.90† 0.96 0.90†
0.90–0.97 0.84–0.94 0.93–0.98 0.84–0.96
0.85 0.87 0.92 0.79†
0.73–0.93 0.78–0.94 0.85–0.97 0.65–0.89
0.93 0.85† 0.93 0.90
0.87–0.97 0.79–0.90 0.86–0.97 0.85–0.94
* Pooled data from two readers. † Significantly different (P ⬍.05) than 3D CE MRA. ROC Az ⫽ receiver operating characteristic curve area; CI ⫽ confidence interval; 2D TOF ⫽ two-dimensional time-of-flight; 3D CE MRA ⫽ three-dimensional contrast-enhanced magnetic resonance angiography; 3D PC ⫽ gated 3D phase contrast.
the angiography suite, is subsequent to reading the angiographic image. Therefore, to focus the evaluation on the images themselves, hemodynamic measurements at angiography were not utilized. The ungated 2D TOF images were shown first, because coverage was obtained from the pelvis through the popliteal trifurcations with this technique only. This was done so the surgeon would have a realistic view of the inflow and outflow vasculature with MRA alone. Ungated 2D TOF was selected for this purpose because of its ease of use and its widespread availability in private practice settings. Subsequently, the surgeon was shown the other MRA technique images in random order at one sitting. DSA images from the pelvis thru the popliteal trifurcations were viewed last. At the viewing of each set of images, input from a radiologist about the interpretation of findings was given. Then, with the patient’s clinical information in hand the surgeon ranked the image set appropriate to his confidence in his ability to generate a treatment plan given the images and the radiologist’s input. Surgeons were blinded to which pulse sequence they were evaluating at all times. The rankings were as follows: 1 ⫽ very confident, 2 ⫽ confident, 3 ⫽ equivocal, 4 ⫽ less confident, 5 ⫽ not confident. To minimize bias concerning the perceived usefulness of each pulse sequence, the radiologist gave interpretive information about the images presented, but made no value judgments to the surgeon about how well depicted the disease was in each set of images compared with other techniques. The Wilcoxon signed rank test was used to evaluate the significance of differences between surgeons’ confidence in their ability to generate treatment plans using the various pulse sequences and DSA with the radiologist’s input.
RESULTS Diagnostic Accuracy We found that MRA diagnostic accuracy for significant disease was best overall with CE MRA (Table I). ROC area, sensitivity, and specificity were seen to be consistently above .90. The final CE MRA images for our patients were of a consistent high quality (Figures 1 and 2). We had no allergic reactions to contrast administrations in our patients. Gated 3D PC was less accurate, with relatively low numbers seen in ROC area, sensitivity, and specificity (Table I). Besides not yielding optimal diagnostic accuracy, gated 3D PC was also relatively difficult to use. Because scan 8
times were heart-rate limited and gating was sometimes inconsistent in quality, some scans were 12 or more minutes in length. The relatively long echo time made for prominent signal voids at tight stenoses, making the grading of disease more difficult (Figure 1A). Because velocities of blood flow can vary a great deal in some patients with PVD, the image quality with gated 3D PC was sometimes quite suboptimal (Figure 2A). Gated 2D TOF gave results close to CE MRA in terms of ROC area and identical specificity. TOF related flow artifacts were still present with the gated sequence, as compared with standard 2D TOF (Figure 2B, C). However, pulsatility artifacts were considerably reduced with gating. Conventional 2D TOF gave less than optimal results, with an ROC area similar to gated 3D PC. This was expected, given that the stimulus for the development of the other techniques used here was the well-known inadequacy of 2D TOF in the pelvis. All pulse sequences were able to diagnose vascular occlusions satisfactorily, although gated 2D TOF and 3D PC performed significantly less well than CE MRA in terms of specificity (Table II). Only 3D PC had an ROC area significantly lower than CE MRA, but even so it was still high at .98. 2D TOF was essentially error free in evaluating occlusions, but was not significantly different than CE MRA. Good interobserver agreement was seen in this study. Kappa and weighted Kappa values indicate at least moderate or better agreement (Table III). Interobserver agreement for CE MRA and DSA was essentially identical and was substantial. The surgeons’ confidence ratings showed clear and significant differences in their comfort levels when viewing the various image sets. Surgeons were most confident when viewing DSA images. Confidence was then followed by CE MRA, gated 2D TOF, ungated 2D TOF, and gated 3D PC (Table IV). Thus, the confidence ratings ran in similar fashion to the diagnostic accuracy results when comparing the most accurate techniques.
COMMENTS Basic MRA techniques have been relatively difficult to utilize in an easy, time-efficient fashion. The CE MRA technique represents a large paradigm shift. The scans can be set up quickly, and each CE 3D MRA scan (the precontrast and postcontrast scans) in this study is only a breath-hold in duration. Postprocessing is also quickly
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
Figure 1. Patient with a single significant stenosis of the left external iliac artery. A. Gated 3D PC sequence: note venous signal (arrowheads). A flow void is noted at the site of stenosis (arrow). B. Ungated 2D TOF: slightly more luminal detail is seen compared with gated 3D PC (arrow). C. Gated 2D TOF: a larger flow void is seen at the level of the stenosis (arrow) as compared with ungated 2D TOF or 3D PC. D. 3D CE MRA: the minimal lumen is seen that compares favorably with the DSA result. E. DSA result (see arrows). 3D PC ⫽ gated three-dimensional phase contrast; 2D TOF ⫽ two-dimensional time of flight; CE MRA ⫽ contrast-enhanced magnetic resonance angiography; DSA ⫽ digital subtraction angiography.
done at a workstation. CE MRA is also extremely robust in most circumstances. For all of the above reasons, CE MRA is considerably easier to use than other MRA methods. Because there are few flow-related artifacts and resolution can be enhanced relative to other MRA methods, the detail of the vessel lumen is much greater. CE MRA techniques continue to improve. A method of time-resolved contrast-enhanced 3D MR angiography,14,15 called 3D MR DSA, has recently been implemented. With this approach, the entire passage of a contrast bolus is imaged, much like a typical DSA run. One can then select any portion of the time-resolved sequence of images and create a 3D set of images. An alternative method that involves less data handling16 uses a moving table apparatus com-
bined with a low-rate intravenous contrast infusion. This approach results in short examination times with images of relatively high quality. Both of these CE MRA methods have recently been reported as feasible for multistation run-off examinations from the abdominal aorta to the feet in a half hour or less.15,16 As seen above, gated 2D TOF yielded diagnostic accuracy relatively close to CE MRA. This result could be further confirmed with a larger sample size, but our other work comparing DSA and gated 2D TOF treatment plans with the eventual surgical decisions17 suggests that the accuracy of this technique is substantial. However, a gated 2D TOF examination takes twice as long to perform as an examination done with CE MRA.
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
9
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
Figure 2. Patient with aortic aneurysm, extremely tortuous iliac arteries, and bilateral superficial femoral artery (SFA) occlusions. A. Gated 3D PC sequence returns very little vascular signal in this case. B. Ungated 2D TOF: in-plane and retrograde flow artifacts remove proximal iliac signal (arrows). Note bilateral SFA occlusions in this field of view (white outline arrows). Overlying bowel signal is evident (long arrows). C. Gated 2D TOF: flow artifacts (arrows) and bilateral SFA occlusions (white outline arrows) are noted as in B. Improved nonvascular tissue signal suppression is seen compared with B, and the aorta is more clearly displayed. D. 3D CE MRA shows tortuous iliac vessels with no loss of vascular signal. E. DSA correlation for distal aorta and iliac arteries. 3D PC ⫽ gated three-dimensional phase contrast; 2D TOF ⫽ two-dimensional time of flight; CE MRA ⫽ contrast-enhanced magnetic resonance angiography; DSA ⫽ digital subtraction angiography.
TABLE II Arterial Segment Occlusion: Diagnostic Accuracy with Four Pulse Sequences* MRA Method
ROC Az
95% CI
Sensitivity
95% CI
Specificity
95% CI
Gated 2D TOF Gated 3D PC 3D CE MRA 2D TOF
0.99 0.98† 0.99 1.00
0.98–1.00 0.96–0.99 0.98–1.00 0.99–1.00
1.00 0.98 0.98 1.00
1.00–1.00 0.94–1.00 0.94–1.00 1.00–1.00
0.98† 0.97† 1.00 1.00
0.97–0.99 0.95–0.99 0.99–1.00 0.99–1.00
* Pooled data from two readers. † Significantly different than 3D CE MRA. ROC Az ⫽ receiver operating characteristic curve area; CI ⫽ confidence interval; 2D TOF ⫽ two-dimensional time-of-flight; 3D CE MRA ⫽ threedimensional contrast-enhanced magnetic resonance angiography; 3D PC: gated 3D phase contrast.
The effects on diagnostic/surgical “thinking efficacy” in our study were limited to evaluating the surgeon’s confidence in various kinds of MRA and DSA for eventual decision making. One would expect that surgeons would be 10
more confident in DSA, given the entrenched tradition of conventional angiography as well as the quality imaging that DSA affords. With this as a standard of quality and perception, it could be hypothesized easily that the greater
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
TABLE III Interobserver Agreement: Kappa and Weighted Kappa Values
MRA Method
Observed Agreement
Kappa (95% CI)
Weighted Kappa* (95% CI)
Gated 2D TOF Gated 3D PC CE MRA 2D TOF DSA
0.66 0.64 0.72 0.63 0.73
0.51 (0.40–0.62) 0.51 (0.41–0.61) 0.61 (0.51–0.72) 0.47 (0.36–0.58) 0.61 (0.51–0.72)
0.75 (0.67–0.82) 0.68 (0.59–0.77) 0.78 (0.71–0.85) 0.72 (0.64–0.80) 0.78 (0.71–0.85)
* See text for agreement weights. MRA ⫽ magnetic resonance angiography; 2D TOF ⫽ two-dimensional timeof-flight; 3D PC ⫽ three-dimensional phase contrast; CE MRA ⫽ contrastenhanced MRA; DSA ⫽ digital subtraction angiography.
TABLE IV Surgeons’ Confidence in MRA Techniques for Treatment Planning Using Four Pulse Sequences and Conventional Angiography* MRA Method
x Value†
Gated 2D TOF
Gated 3D PC
3D CE MRA
2D TOF
Gated 2D TOF Gated 3D PC 3D CE MRA 2D TOF DSA
2.00 3.00 1.70 2.57 1.14
0.0034‡ 0.0269 0.0225 0.0001
0.0001 0.0420 0.0001
0.0002 0.0049
0.0001
* All images viewed with a radiologist, who provided interpretation of disease and explained artifacts present. † Mean value for ranks of each pulse sequence, using a scale as follows: 1 ⫽ very confident, 2 ⫽ somewhat confident, 3 ⫽ equivocal, 4 ⫽ less confident, 5 ⫽ not confident. ‡ Wilcoxon signed rank statistic (P value) for significance of difference between ranks for two compared angiographic techniques. MRA ⫽ magnetic resonance angiography; 2D TOF ⫽ two-dimensional timeof-flight; 3D PC ⫽ three-dimensional phase contrast; CE MRA ⫽ contrastenhanced MRA; DSA ⫽ digital subtraction angiography.
the degree of closeness to DSA quality that an MRA method possesses, the closer both diagnostic accuracy and clinician confidence will be to the DSA level. This correspondence was strongly suggested by our results, and argues for the use of CE MRA. Our study was somewhat limited by the “moving target”18 effect so commonly present in clinical trials involving MRI pulse sequence development, hence the use of multiple techniques for aorto-iliac evaluation. Even though the smaller sample size is a limitation, we were still able to detect significant suboptimal performance of gated 3D PC and ungated 2D TOF, as well as good performance of CE MRA. The radiologist needs a simple solution that will meet his or her needs and make the referring clinician confident in the technology provided. This work suggests that in the pelvis, a simple 2D TOF or 3D PC approach should be used with caution. Also, given the lesser time and effort invested along with excellent accuracy, the CE MRA approach is reasonable. The utility of CE MRA in the pelvis and elsewhere is further suggested by other studies.5,16,19,20 MRA of peripheral vessels is now viewed favorably by
public policy makers as well.21 Surgeons in our study clearly have confidence in the best of MRA techniques. However, the use of MRA is still limited to circumstances where there is a concerted effort on the part of radiologists and surgeons to integrate it into their practice. For the radiologist, this means building MRA protocols that answer the clinical question consistently, as well as providing an environment where patients can be imaged in a timely manner. The surgeon has to provide information and communicate needs effectively to the radiologist. Important information consists of adequate clinical history and data, including data that facilitates the screening of patients for contraindications, and clear communication of preferred display of the final images. Although DSA and optimally done MRA can probably both answer most questions, there are circumstances where one or the other examination is preferred. DSA is usually preferred where one is dealing with a clinically unstable patient. Some patients have clear absolute contraindications to MRA, such as ferromagnetic aneurysm clips, some types of prosthetic heart valves, auditory implants, and other implantable devices. It is extremely important that a screening scheme is in force to adequately query patients for possible problems. Some radiology departments employ nurses to document and clarify possible problem situations. This sometimes involves calling manufacturers of devices, and seeking out definitive outside medical records. However, metallic devices do not have to create safety concerns to be problematic. If the patient has metallic implants that are not contraindicated, such as most orthopedic hardware, the artifacts from the hardware may be sufficient to obscure the anatomy of interest. MRA is preferred where renal insufficiency is of concern. Additionally, in patients with an anaphylactic allergy history relative to iodinated x-ray contrast, MRA is the logical alternative. Some physicians prefer to monitor the patency of bypass grafts with MRA. However, such practice requires that metallic clips have not been used in the vicinity of grafts, since artifacts may occur which will partially or fully obscure the anatomy. Some patients will present with total aortic occlusions. This may make introduction of a catheter or visualization of downstream vasculature more challenging with DSA. MRA can function well in this context, and this was observed in our study. Also, MRA has also been shown to work well in imaging of distal runoff vessels where DSA may fail.1,3 Some patients present with findings suggesting a lesion amenable to endovascular treatment. Whether the patient should have MRA or DSA is dictated by circumstances. In many cases, it is currently easier to have the intervention and diagnosis in one visit. Thus, DSA would be a logical choice in such circumstances. However, if the referring clinician or the patient would prefer a less-invasive study be done first, or the clinician is not in a typical vascular intervention referral loop, then MRA may be a reasonable initial alternative. On the other hand, imaging evaluation in such circumstances could change if interventional MR guidance for vascular procedures becomes commonplace.22,23 MRA pulse sequence innovation has probably advanced into a plateau phase. We would anticipate that CE MRA
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
11
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
approaches will continue to find significant acceptance and wider application throughout the imaging community, largely supplanting older flow-dependent techniques based on increased speed and accuracy. Continued outcomes and technology assessment will be important in supporting this assertion.
REFERENCES 1. Owen RS, Baum RS, Carpenter JP, et al. Symptomatic peripheral vascular disease: selection of imaging parameters and clinical evaluation with MR angiography. Radiology. 1993;187:627– 635. 2. Yucel EK, Kaufman JA, Geller SC, Waltman AC. Atherosclerotic occlusive disease of the lower extremity: prospective evaluation with two-dimensional time-of-flight MR angiography. Radiology. 1993;187:637– 641. 3. Baum RA, Rutter CM, Sunshine JH, et al. Multicenter trial to evaluate vascular magnetic resonance angiography of the lower extremity. JAMA. 1995;274:875– 880. 4. Prince MR, Yucel EK, Kaufman JA, et al. Dynamic gadoliniumenhanced three dimensional MR arteriography. J MRI. 1993;3:877– 881. 5. Hany TF, Debatin JF, Leung DA, Pfammatter T. Evaluation of the aorto-iliac and renal arteries: comparison of breath-hold, contrast-enhanced, three-dimensional MR angiography with conventional catheter angiography. Radiology. 1997;204:357–362. 6. Swan JS, Fryback DG, Lawrence WL, et al. MR and conventional angiography: work in progress toward assessing utility in radiology. Acad Radiol. 1997;4:475– 482. 7. Swan JS, Fryback DG, Lawrence WF, et al. A time trade-off method for cost-effectiveness models applied to radiology. Med Decis Making. 2000;20:79 – 88. 8. Fryback DG, Thornbury JR. The efficacy of diagnostic imaging. Med Decis Making. 1991;11:88 –94. 9. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982; 143:29 –36. 10. Efron B. The Jackknife, the Bootstrap and Other Resampling Plans.
12
Philadelphia: Society for Industrial and Applied Mathematics; 1982. 11. Cohen J. A coefficient of agreement for nominal scales. Educ Psycholog Meas. 1960;20:37– 46. 12. Cohen J. Weighted Kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psycholog Bull. 1968;70:213–220. 13. Fleiss JL, Cohen J, Everett BS. Large sample standard errors of kappa and weighted kappa. Psycholog Bull. 1969;72:323–327. 14. Korosec FR, Frayne R, Grist TM, Mistretta CA. Time-resolved contrast-enhanced 3D MR angiography. MRM. 1996;36:345–351. 15. Swan JS, Grist TM, Frayne R, et al. Time-Resolved MR Angiography of the Peripheral Vasculature. Book of abstracts. Vancouver, BC: Fifth ISMRM; 1997:131. 16. Ho KYJAM, Leiner T, de Haan MW, et al. Peripheral vascular tree stenoses: evaluation with moving-bed infusion-tracking MR angiography. Radiology. 1998;206:683– 692. 17. Hoch JR, Kennell TW, Hollister MS, et al. Comparison of treatment plans for lower extremity arterial occlusive disease made with electrocardiograph-triggered two-dimensional time-of-flight magnetic resonance angiography and digital subtraction angiography. Am J Surg. 1999;178:166 –172. 18. Flamm CR. Technology assessment of magnetic resonance angiography: why pretty pictures are not enough. An overview. Invest Radiol. 1998;33:547–552. 19. Quinn SF, Sheley RC, Semonsen KG, et al. Aortic and lowerextremity arterial disease: evaluation with MR angiography versus conventional angiography. Radiology. 1998;206:693–701. 20. Snidow JJ, Johnson MS, Harris VJ, et al. Three-dimensional gadolinium-enhanced MR angiography for aorto-iliac inflow assessment plus renal artery screening in a single breath-hold. Radiology. 1996;198:725–732. 21. HCFA Coverage Issues Manual. August 1999:section 50-14. URL: http://www.hcfa.gov/pubforms/06_cim/ci50.htm _1_14 22. Adam G, Bu¨cker A , Glowinski A. Interventional magnetic resonance angiography. Semin Intervent Radiol. 1999;16:31. 23. Omary RA, Frayne R, Unal O, et al. MR angioplasty of renal artery stenosis in a pig model: a feasibility study. J VIR. 2000;11.
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
BACKGROUND: Four different techniques for aortoiliac magnetic resonance angiography (MRA) were assessed for accuracy using a digital subtraction angiography (DSA) gold standard. Surgeons’ confidence in their ability to generate treatment plans with MRA and DSA was assessed, in consultation with a radiologist. METHODS: Two different two-dimensional (2D) time-of-flight (TOF) sequences, a phase-contrast sequence, and a contrast-enhanced (CE) MRA sequence were used. Receiver operating characteristic (ROC) curves were plotted and areas (Az) calculated from radiologists’ readings. Surgeons’ confidence in their ability to utilize the images for treatment planning was assessed with a 5-point Likert scale. Thirty-six patients were evaluated. RESULTS: CE MRA had a sensitivity, specificity, and Az of .92, .93, and .96, respectively, for stenoses 50% or greater. CE MRA performed better than other sequences, but the improvement compared with gated 2D TOF was not statistically significant. Interobserver agreement for CE MRA and DSA yielded identical Kappa values. Surgeons were most confident in DSA, followed by CE MRA, which was significantly preferred to other techniques. CONCLUSIONS: CE MRA closely approximates DSA in terms of diagnostic accuracy. Surgeons considering treatment plans are confident in the CE MRA technique, relative to other MRA methods. Am J Surg. 2000;180:6 –12. © 2000 by Excerpta Medica, Inc.
S
ince early work by many investigators,1–3 it has been clear that two-dimensional (2D) time-of-flight (TOF) magnetic resonance angiography (MRA) is capable of imaging peripheral vascular disease (PVD). Roadblocks to previous implementation of MRA for eval-
uating PVD on a wide scale have been flow-dependent artifacts, the length of scan times, the need for dedicated MR receiver coils, and the inherent quality of the resulting images. With respect to artifacts, the pelvis has been a particular problem, owing to the orientation of the vessels and movement related to respiration and normal bowel peristalsis. Recent work is more encouraging. Studies now strongly indicate that gadolinium contrast-enhanced (CE) MRA is a good solution.4,5 CE MRA does not suffer as significantly from the above limitations. Thus, flow dependent artifacts are not as severe, and the vessel lumen detail is imaged wherever contrast is present. The resolution of CE MRA images can be considerably enhanced over previous techniques, and the inherent contrast characteristics of CE MRA images make the need for dedicated coils less of a factor. The result is much more like a conventional X-ray angiographic image, rendering easier interpretation for the many physicians previously trained with conventional angiographic methods. Of considerable importance to the patient is the less invasive character of MRA, thus suggesting a mechanism of improving short-term health outcomes related to PVD testing. Recent work has sought to quantify this improvement.6,7 This study is a multilevel technology assessment, according to the hierarchy proposed by Fryback and Thornbury.8 First, we explored the diagnostic accuracy of MRA in the pelvis using four different pulse sequences (Fryback and Thornbury level 2: “diagnostic accuracy efficacy”). We evaluated patients with techniques reflecting the varied solutions that have been attempted to optimize the imaging of the pelvis vasculature in recent years. Secondly, we wished to evaluate the effects of MRA imaging on the confidence of the ultimate end user of these images, the vascular surgeon (Fryback and Thornbury level 3: “diagnostic thinking efficacy”). This was done in a manner similar to what would be found in any typical healthcare scenario, where the radiologist and surgeon confer on the interpretative and clinical aspects of a case.
PATIENTS AND METHODS From the Departments of Radiology (JSS, TWK, TMG, FRK, MEH), Surgery (CWA, DMH), and Medical Physics (TMG, FRK), University of Wisconsin-Madison, Madison, Wisconsin. Dr. Swan has received support from the General Electric-Association of University Radiologists Fellowship, NIH R01 HL51370, and Nycomed Corporation. Requests for reprints should be addressed to J. Shannon Swan, MD, Indiana University Medical Center, Department of Radiology, 550 N. University Boulevard, Indianapolis, Indiana 46202. Manuscript submitted February 18, 2000, and accepted in revised form May 16, 2000.
6
© 2000 by Excerpta Medica, Inc. All rights reserved.
Patients Thirty-six patients were utilized for data analysis in this study. Thirty-one of them were male. The age range was 33 to 79 years, with a mean of 62. All patients were recruited from our community or surrounding areas, having been referred from surgical practitioners who were intending a vascular intervention for the patient, possibly involving the aorto-iliac area. Patients were enrolled in our study prospectively when they were scheduled to have a conventional angiogram by the referring physician. MRA images eventually acquired were not used to decide on whether a 0002-9610/00/$–see front matter PII S0002-9610(00)00412-8
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
patient would be scheduled for conventional angiography, if the MRA examination was obtained first. Informed consent was obtained using forms approved by our institutional review board. Angiographic Technique The conventional x-ray angiographic evaluation was done with digital subtraction angiography technique (DSA), with at least two opposite 30-degree oblique runs being acquired in each case when imaging the pelvis. All cases were intraarterial injections done with Seldinger technique at a groin entrance site, with the exception of two cases where total aortic occlusion necessitated a translumbar approach. Nonionic contrast was used in all patients, and preprocedure sedation was utilized. MRA Techniques The MRA techniques included cardiac gated and nongated two-dimensional (2D TOF), a cardiac-gated three dimensional (3D) phase-contrast (PC) approach, and a modified 3D CE MRA pulse sequence. All studies were performed using a General Electric (Signa Advantage, Milwaukee, WI) 1.5 Tesla unit with the body coil. The 2D TOF sequence used was the product General Electric method in clinical use at the time of the study. Scan time was 5 to 7 minutes depending on the number of slices acquired. The cardiac-gated 2D TOF sequence was written by modifying a basic 2D TOF sequence to allow gating and segmentation. Scan time was equal to 800 to 1,000 heartbeats (approximately 13 to 17 minutes). Scan time was shortened by using a rectangular field of view (FOV) when possible. In the cardiac-gated 3D PC sequence, scan time was generally 514 heartbeats (approximately 8.5 minutes at a heart rate of 60 beats per minute, approximately 6.5 minutes at 80 beats per minute). A series of three scans was performed to complete the CE MRA technique. First, a contrast transit time scan was obtained that consisted of a sagittal acquisition set-up to image a single slice location through the abdominal aorta at a rate of one image per second for 60 seconds. Then, 2 cc gadodiamide (Omniscan, Nycomed, Princeton, NJ) was injected intravenously via a small-bore peripheral catheter over 0.5 seconds, followed by a 15 cc saline flush at a rate of 1.5 cc/sec. This series of images showed the transit time of contrast agent from the arm to the aorta. Second, a coronal 3D scan was then obtained but no contrast was injected, so the data could be used for subsequent subtraction of nonvascular tissue signal. For the third scan, an injection of a gadodiamide bolus (0.2 to 0.3 mmol/kg body weight) at 1.5 cc/sec began the process, according to the transit time results. Scan times for the second and third scans were 30 to 45 seconds. Postprocessing of all of the MRA images was done to generate projections at 10-degree increments. Source images were available for confirmation of difficult readings. Image subtraction was done on an Advantage Windows workstation (General Electric) followed by the generation of reprojected images. Blinded Readings by Radiologists In each patient, the evaluation ideally included the infrarenal aorta, the right and left common iliac arteries, the
right and left external iliac arteries, the right and left common femoral arteries, the right and left superficial femoral arterial origins, and the right and left profunda femoral arterial origins. For the ungated 2D TOF acquisitions only, imaging continued distally, including the thighs to the popliteal trifurcations. This was done to facilitate the surgeons’ diagnostic thinking accuracy exercise that is further explained below. The vascular segments distal to the proximal superficial femoral and profunda femoral arteries were not included in the diagnostic accuracy analysis. Two experienced radiologists were given the images to evaluate in blinded fashion. All MRA images of the patients were scored by each reader independently, one pulse sequence at a time over a period of months. The DSA images were evaluated as the final group. The following scale was used for rating the disease in each vessel segment: 0 ⫽ normal, 1 ⫽ minimal disease, all stenoses being ⬍50% of the vessel diameter, 2 ⫽ one ⱖ50% stenosis, 3 ⫽ more than one ⱖ50% stenosis, 4 ⫽ complete occlusion of the affected vessel segment. This scheme is similar to the grading used in widely cited trials.1,3 Diagnostic accuracy was computed by first designating two cut points. We evaluated for significant disease by designating ratings greater than “1” from the disease scale mentioned above as being abnormal. Secondly, we assessed for complete vessel segment occlusions by designating readings greater than “3” as being abnormal. Receiver operating characteristic (ROC) curve areas were calculated for both significant disease and occlusion by pulse sequence, using pooled data from two readers. DSA was the gold standard. ROC data were computed with a nonparametric approach utilizing the trapezoid rule.9 Confidence intervals were constructed about the estimates of the area under the ROC curves, sensitivities and specificities by bootstrapping.10 Pairwise significance tests were performed by first constructing bootstrap confidence intervals about the differences of the parameter estimates of interest. The estimates were declared significantly different if the confidence interval did not include zero. We assessed interobserver agreement by computing Kappa11 and weighted Kappa12 statistics. Agreement weights were defined as in work by Cohen12 so that full agreement between readers was given a weight of 1, and partial agreement was given a weight of (1 ⫺ [r1 ⫺ r2] / 4). Asymptotic variances were calculated as in Fleiss et al.13 Surgeons’ Confidence in Imaging Techniques We evaluated surgeons’ perceived confidence in their ability to generate a treatment plan with the MRA images plus clinical information, in consultation with a radiologist. Clinical information consisted of all information that might be available prior to angiographic evaluation, including all vascular laboratory data, as well as physical examination data and clinical history. Although hemodynamic measurements obtained at angiography are obviously useful in aortoiliac evaluation, we felt they could serve as a possible source of bias against MRA, given that this measurement is typically obtained at, or associated with, angiography. Also, the decision to obtain gradient measurements and entertain other treatment, possibly in
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
7
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
TABLE I Clinically Significant Disease (≥50% stenosis): Diagnostic Accuracy with Four Different Pulse Sequences* MRA Method
ROC Az
95% CI
Sensitivity
95% CI
Specificity
95% CI
Gated 2D TOF Gated 3D PC 3D CE MRA 2D TOF
0.94 0.90† 0.96 0.90†
0.90–0.97 0.84–0.94 0.93–0.98 0.84–0.96
0.85 0.87 0.92 0.79†
0.73–0.93 0.78–0.94 0.85–0.97 0.65–0.89
0.93 0.85† 0.93 0.90
0.87–0.97 0.79–0.90 0.86–0.97 0.85–0.94
* Pooled data from two readers. † Significantly different (P ⬍.05) than 3D CE MRA. ROC Az ⫽ receiver operating characteristic curve area; CI ⫽ confidence interval; 2D TOF ⫽ two-dimensional time-of-flight; 3D CE MRA ⫽ three-dimensional contrast-enhanced magnetic resonance angiography; 3D PC ⫽ gated 3D phase contrast.
the angiography suite, is subsequent to reading the angiographic image. Therefore, to focus the evaluation on the images themselves, hemodynamic measurements at angiography were not utilized. The ungated 2D TOF images were shown first, because coverage was obtained from the pelvis through the popliteal trifurcations with this technique only. This was done so the surgeon would have a realistic view of the inflow and outflow vasculature with MRA alone. Ungated 2D TOF was selected for this purpose because of its ease of use and its widespread availability in private practice settings. Subsequently, the surgeon was shown the other MRA technique images in random order at one sitting. DSA images from the pelvis thru the popliteal trifurcations were viewed last. At the viewing of each set of images, input from a radiologist about the interpretation of findings was given. Then, with the patient’s clinical information in hand the surgeon ranked the image set appropriate to his confidence in his ability to generate a treatment plan given the images and the radiologist’s input. Surgeons were blinded to which pulse sequence they were evaluating at all times. The rankings were as follows: 1 ⫽ very confident, 2 ⫽ confident, 3 ⫽ equivocal, 4 ⫽ less confident, 5 ⫽ not confident. To minimize bias concerning the perceived usefulness of each pulse sequence, the radiologist gave interpretive information about the images presented, but made no value judgments to the surgeon about how well depicted the disease was in each set of images compared with other techniques. The Wilcoxon signed rank test was used to evaluate the significance of differences between surgeons’ confidence in their ability to generate treatment plans using the various pulse sequences and DSA with the radiologist’s input.
RESULTS Diagnostic Accuracy We found that MRA diagnostic accuracy for significant disease was best overall with CE MRA (Table I). ROC area, sensitivity, and specificity were seen to be consistently above .90. The final CE MRA images for our patients were of a consistent high quality (Figures 1 and 2). We had no allergic reactions to contrast administrations in our patients. Gated 3D PC was less accurate, with relatively low numbers seen in ROC area, sensitivity, and specificity (Table I). Besides not yielding optimal diagnostic accuracy, gated 3D PC was also relatively difficult to use. Because scan 8
times were heart-rate limited and gating was sometimes inconsistent in quality, some scans were 12 or more minutes in length. The relatively long echo time made for prominent signal voids at tight stenoses, making the grading of disease more difficult (Figure 1A). Because velocities of blood flow can vary a great deal in some patients with PVD, the image quality with gated 3D PC was sometimes quite suboptimal (Figure 2A). Gated 2D TOF gave results close to CE MRA in terms of ROC area and identical specificity. TOF related flow artifacts were still present with the gated sequence, as compared with standard 2D TOF (Figure 2B, C). However, pulsatility artifacts were considerably reduced with gating. Conventional 2D TOF gave less than optimal results, with an ROC area similar to gated 3D PC. This was expected, given that the stimulus for the development of the other techniques used here was the well-known inadequacy of 2D TOF in the pelvis. All pulse sequences were able to diagnose vascular occlusions satisfactorily, although gated 2D TOF and 3D PC performed significantly less well than CE MRA in terms of specificity (Table II). Only 3D PC had an ROC area significantly lower than CE MRA, but even so it was still high at .98. 2D TOF was essentially error free in evaluating occlusions, but was not significantly different than CE MRA. Good interobserver agreement was seen in this study. Kappa and weighted Kappa values indicate at least moderate or better agreement (Table III). Interobserver agreement for CE MRA and DSA was essentially identical and was substantial. The surgeons’ confidence ratings showed clear and significant differences in their comfort levels when viewing the various image sets. Surgeons were most confident when viewing DSA images. Confidence was then followed by CE MRA, gated 2D TOF, ungated 2D TOF, and gated 3D PC (Table IV). Thus, the confidence ratings ran in similar fashion to the diagnostic accuracy results when comparing the most accurate techniques.
COMMENTS Basic MRA techniques have been relatively difficult to utilize in an easy, time-efficient fashion. The CE MRA technique represents a large paradigm shift. The scans can be set up quickly, and each CE 3D MRA scan (the precontrast and postcontrast scans) in this study is only a breath-hold in duration. Postprocessing is also quickly
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
Figure 1. Patient with a single significant stenosis of the left external iliac artery. A. Gated 3D PC sequence: note venous signal (arrowheads). A flow void is noted at the site of stenosis (arrow). B. Ungated 2D TOF: slightly more luminal detail is seen compared with gated 3D PC (arrow). C. Gated 2D TOF: a larger flow void is seen at the level of the stenosis (arrow) as compared with ungated 2D TOF or 3D PC. D. 3D CE MRA: the minimal lumen is seen that compares favorably with the DSA result. E. DSA result (see arrows). 3D PC ⫽ gated three-dimensional phase contrast; 2D TOF ⫽ two-dimensional time of flight; CE MRA ⫽ contrast-enhanced magnetic resonance angiography; DSA ⫽ digital subtraction angiography.
done at a workstation. CE MRA is also extremely robust in most circumstances. For all of the above reasons, CE MRA is considerably easier to use than other MRA methods. Because there are few flow-related artifacts and resolution can be enhanced relative to other MRA methods, the detail of the vessel lumen is much greater. CE MRA techniques continue to improve. A method of time-resolved contrast-enhanced 3D MR angiography,14,15 called 3D MR DSA, has recently been implemented. With this approach, the entire passage of a contrast bolus is imaged, much like a typical DSA run. One can then select any portion of the time-resolved sequence of images and create a 3D set of images. An alternative method that involves less data handling16 uses a moving table apparatus com-
bined with a low-rate intravenous contrast infusion. This approach results in short examination times with images of relatively high quality. Both of these CE MRA methods have recently been reported as feasible for multistation run-off examinations from the abdominal aorta to the feet in a half hour or less.15,16 As seen above, gated 2D TOF yielded diagnostic accuracy relatively close to CE MRA. This result could be further confirmed with a larger sample size, but our other work comparing DSA and gated 2D TOF treatment plans with the eventual surgical decisions17 suggests that the accuracy of this technique is substantial. However, a gated 2D TOF examination takes twice as long to perform as an examination done with CE MRA.
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
9
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
Figure 2. Patient with aortic aneurysm, extremely tortuous iliac arteries, and bilateral superficial femoral artery (SFA) occlusions. A. Gated 3D PC sequence returns very little vascular signal in this case. B. Ungated 2D TOF: in-plane and retrograde flow artifacts remove proximal iliac signal (arrows). Note bilateral SFA occlusions in this field of view (white outline arrows). Overlying bowel signal is evident (long arrows). C. Gated 2D TOF: flow artifacts (arrows) and bilateral SFA occlusions (white outline arrows) are noted as in B. Improved nonvascular tissue signal suppression is seen compared with B, and the aorta is more clearly displayed. D. 3D CE MRA shows tortuous iliac vessels with no loss of vascular signal. E. DSA correlation for distal aorta and iliac arteries. 3D PC ⫽ gated three-dimensional phase contrast; 2D TOF ⫽ two-dimensional time of flight; CE MRA ⫽ contrast-enhanced magnetic resonance angiography; DSA ⫽ digital subtraction angiography.
TABLE II Arterial Segment Occlusion: Diagnostic Accuracy with Four Pulse Sequences* MRA Method
ROC Az
95% CI
Sensitivity
95% CI
Specificity
95% CI
Gated 2D TOF Gated 3D PC 3D CE MRA 2D TOF
0.99 0.98† 0.99 1.00
0.98–1.00 0.96–0.99 0.98–1.00 0.99–1.00
1.00 0.98 0.98 1.00
1.00–1.00 0.94–1.00 0.94–1.00 1.00–1.00
0.98† 0.97† 1.00 1.00
0.97–0.99 0.95–0.99 0.99–1.00 0.99–1.00
* Pooled data from two readers. † Significantly different than 3D CE MRA. ROC Az ⫽ receiver operating characteristic curve area; CI ⫽ confidence interval; 2D TOF ⫽ two-dimensional time-of-flight; 3D CE MRA ⫽ threedimensional contrast-enhanced magnetic resonance angiography; 3D PC: gated 3D phase contrast.
The effects on diagnostic/surgical “thinking efficacy” in our study were limited to evaluating the surgeon’s confidence in various kinds of MRA and DSA for eventual decision making. One would expect that surgeons would be 10
more confident in DSA, given the entrenched tradition of conventional angiography as well as the quality imaging that DSA affords. With this as a standard of quality and perception, it could be hypothesized easily that the greater
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
TABLE III Interobserver Agreement: Kappa and Weighted Kappa Values
MRA Method
Observed Agreement
Kappa (95% CI)
Weighted Kappa* (95% CI)
Gated 2D TOF Gated 3D PC CE MRA 2D TOF DSA
0.66 0.64 0.72 0.63 0.73
0.51 (0.40–0.62) 0.51 (0.41–0.61) 0.61 (0.51–0.72) 0.47 (0.36–0.58) 0.61 (0.51–0.72)
0.75 (0.67–0.82) 0.68 (0.59–0.77) 0.78 (0.71–0.85) 0.72 (0.64–0.80) 0.78 (0.71–0.85)
* See text for agreement weights. MRA ⫽ magnetic resonance angiography; 2D TOF ⫽ two-dimensional timeof-flight; 3D PC ⫽ three-dimensional phase contrast; CE MRA ⫽ contrastenhanced MRA; DSA ⫽ digital subtraction angiography.
TABLE IV Surgeons’ Confidence in MRA Techniques for Treatment Planning Using Four Pulse Sequences and Conventional Angiography* MRA Method
x Value†
Gated 2D TOF
Gated 3D PC
3D CE MRA
2D TOF
Gated 2D TOF Gated 3D PC 3D CE MRA 2D TOF DSA
2.00 3.00 1.70 2.57 1.14
0.0034‡ 0.0269 0.0225 0.0001
0.0001 0.0420 0.0001
0.0002 0.0049
0.0001
* All images viewed with a radiologist, who provided interpretation of disease and explained artifacts present. † Mean value for ranks of each pulse sequence, using a scale as follows: 1 ⫽ very confident, 2 ⫽ somewhat confident, 3 ⫽ equivocal, 4 ⫽ less confident, 5 ⫽ not confident. ‡ Wilcoxon signed rank statistic (P value) for significance of difference between ranks for two compared angiographic techniques. MRA ⫽ magnetic resonance angiography; 2D TOF ⫽ two-dimensional timeof-flight; 3D PC ⫽ three-dimensional phase contrast; CE MRA ⫽ contrastenhanced MRA; DSA ⫽ digital subtraction angiography.
the degree of closeness to DSA quality that an MRA method possesses, the closer both diagnostic accuracy and clinician confidence will be to the DSA level. This correspondence was strongly suggested by our results, and argues for the use of CE MRA. Our study was somewhat limited by the “moving target”18 effect so commonly present in clinical trials involving MRI pulse sequence development, hence the use of multiple techniques for aorto-iliac evaluation. Even though the smaller sample size is a limitation, we were still able to detect significant suboptimal performance of gated 3D PC and ungated 2D TOF, as well as good performance of CE MRA. The radiologist needs a simple solution that will meet his or her needs and make the referring clinician confident in the technology provided. This work suggests that in the pelvis, a simple 2D TOF or 3D PC approach should be used with caution. Also, given the lesser time and effort invested along with excellent accuracy, the CE MRA approach is reasonable. The utility of CE MRA in the pelvis and elsewhere is further suggested by other studies.5,16,19,20 MRA of peripheral vessels is now viewed favorably by
public policy makers as well.21 Surgeons in our study clearly have confidence in the best of MRA techniques. However, the use of MRA is still limited to circumstances where there is a concerted effort on the part of radiologists and surgeons to integrate it into their practice. For the radiologist, this means building MRA protocols that answer the clinical question consistently, as well as providing an environment where patients can be imaged in a timely manner. The surgeon has to provide information and communicate needs effectively to the radiologist. Important information consists of adequate clinical history and data, including data that facilitates the screening of patients for contraindications, and clear communication of preferred display of the final images. Although DSA and optimally done MRA can probably both answer most questions, there are circumstances where one or the other examination is preferred. DSA is usually preferred where one is dealing with a clinically unstable patient. Some patients have clear absolute contraindications to MRA, such as ferromagnetic aneurysm clips, some types of prosthetic heart valves, auditory implants, and other implantable devices. It is extremely important that a screening scheme is in force to adequately query patients for possible problems. Some radiology departments employ nurses to document and clarify possible problem situations. This sometimes involves calling manufacturers of devices, and seeking out definitive outside medical records. However, metallic devices do not have to create safety concerns to be problematic. If the patient has metallic implants that are not contraindicated, such as most orthopedic hardware, the artifacts from the hardware may be sufficient to obscure the anatomy of interest. MRA is preferred where renal insufficiency is of concern. Additionally, in patients with an anaphylactic allergy history relative to iodinated x-ray contrast, MRA is the logical alternative. Some physicians prefer to monitor the patency of bypass grafts with MRA. However, such practice requires that metallic clips have not been used in the vicinity of grafts, since artifacts may occur which will partially or fully obscure the anatomy. Some patients will present with total aortic occlusions. This may make introduction of a catheter or visualization of downstream vasculature more challenging with DSA. MRA can function well in this context, and this was observed in our study. Also, MRA has also been shown to work well in imaging of distal runoff vessels where DSA may fail.1,3 Some patients present with findings suggesting a lesion amenable to endovascular treatment. Whether the patient should have MRA or DSA is dictated by circumstances. In many cases, it is currently easier to have the intervention and diagnosis in one visit. Thus, DSA would be a logical choice in such circumstances. However, if the referring clinician or the patient would prefer a less-invasive study be done first, or the clinician is not in a typical vascular intervention referral loop, then MRA may be a reasonable initial alternative. On the other hand, imaging evaluation in such circumstances could change if interventional MR guidance for vascular procedures becomes commonplace.22,23 MRA pulse sequence innovation has probably advanced into a plateau phase. We would anticipate that CE MRA
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000
11
MAGNETIC RESONANCE ANGIOGRAPHY OF AORTO-ILIAC DISEASE/SWAN ET AL
approaches will continue to find significant acceptance and wider application throughout the imaging community, largely supplanting older flow-dependent techniques based on increased speed and accuracy. Continued outcomes and technology assessment will be important in supporting this assertion.
REFERENCES 1. Owen RS, Baum RS, Carpenter JP, et al. Symptomatic peripheral vascular disease: selection of imaging parameters and clinical evaluation with MR angiography. Radiology. 1993;187:627– 635. 2. Yucel EK, Kaufman JA, Geller SC, Waltman AC. Atherosclerotic occlusive disease of the lower extremity: prospective evaluation with two-dimensional time-of-flight MR angiography. Radiology. 1993;187:637– 641. 3. Baum RA, Rutter CM, Sunshine JH, et al. Multicenter trial to evaluate vascular magnetic resonance angiography of the lower extremity. JAMA. 1995;274:875– 880. 4. Prince MR, Yucel EK, Kaufman JA, et al. Dynamic gadoliniumenhanced three dimensional MR arteriography. J MRI. 1993;3:877– 881. 5. Hany TF, Debatin JF, Leung DA, Pfammatter T. Evaluation of the aorto-iliac and renal arteries: comparison of breath-hold, contrast-enhanced, three-dimensional MR angiography with conventional catheter angiography. Radiology. 1997;204:357–362. 6. Swan JS, Fryback DG, Lawrence WL, et al. MR and conventional angiography: work in progress toward assessing utility in radiology. Acad Radiol. 1997;4:475– 482. 7. Swan JS, Fryback DG, Lawrence WF, et al. A time trade-off method for cost-effectiveness models applied to radiology. Med Decis Making. 2000;20:79 – 88. 8. Fryback DG, Thornbury JR. The efficacy of diagnostic imaging. Med Decis Making. 1991;11:88 –94. 9. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982; 143:29 –36. 10. Efron B. The Jackknife, the Bootstrap and Other Resampling Plans.
12
Philadelphia: Society for Industrial and Applied Mathematics; 1982. 11. Cohen J. A coefficient of agreement for nominal scales. Educ Psycholog Meas. 1960;20:37– 46. 12. Cohen J. Weighted Kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psycholog Bull. 1968;70:213–220. 13. Fleiss JL, Cohen J, Everett BS. Large sample standard errors of kappa and weighted kappa. Psycholog Bull. 1969;72:323–327. 14. Korosec FR, Frayne R, Grist TM, Mistretta CA. Time-resolved contrast-enhanced 3D MR angiography. MRM. 1996;36:345–351. 15. Swan JS, Grist TM, Frayne R, et al. Time-Resolved MR Angiography of the Peripheral Vasculature. Book of abstracts. Vancouver, BC: Fifth ISMRM; 1997:131. 16. Ho KYJAM, Leiner T, de Haan MW, et al. Peripheral vascular tree stenoses: evaluation with moving-bed infusion-tracking MR angiography. Radiology. 1998;206:683– 692. 17. Hoch JR, Kennell TW, Hollister MS, et al. Comparison of treatment plans for lower extremity arterial occlusive disease made with electrocardiograph-triggered two-dimensional time-of-flight magnetic resonance angiography and digital subtraction angiography. Am J Surg. 1999;178:166 –172. 18. Flamm CR. Technology assessment of magnetic resonance angiography: why pretty pictures are not enough. An overview. Invest Radiol. 1998;33:547–552. 19. Quinn SF, Sheley RC, Semonsen KG, et al. Aortic and lowerextremity arterial disease: evaluation with MR angiography versus conventional angiography. Radiology. 1998;206:693–701. 20. Snidow JJ, Johnson MS, Harris VJ, et al. Three-dimensional gadolinium-enhanced MR angiography for aorto-iliac inflow assessment plus renal artery screening in a single breath-hold. Radiology. 1996;198:725–732. 21. HCFA Coverage Issues Manual. August 1999:section 50-14. URL: http://www.hcfa.gov/pubforms/06_cim/ci50.htm _1_14 22. Adam G, Bu¨cker A , Glowinski A. Interventional magnetic resonance angiography. Semin Intervent Radiol. 1999;16:31. 23. Omary RA, Frayne R, Unal O, et al. MR angioplasty of renal artery stenosis in a pig model: a feasibility study. J VIR. 2000;11.
THE AMERICAN JOURNAL OF SURGERY® VOLUME 180 JULY 2000