Noncontrastenhanced MR angiography for detecting arteriovenous fistula dysfunction in haemodialysis patients.pdf

non-contrast-enhanced mr angiography for detecting arteriovenous fi stula dysfunction in haemodialysis patients w.t. jin a, g.f. zhanga,*, h.c. liub, h. zhanga, b. lib, x.q. zhub adepartment of radiology, obstetrics and gynecology hospital of fudan university, shanghai, china bdepartment of radiology, shanghai changzheng hospital, shanghai, china article information article history: received 9 november 2014 received in revised form 18 march 2015 accepted 9 april 2015 aim: to assess the diagnostic value of non-contrast-enhanced magnetic resonance angiog- raphy (nce-mra), using time-of-fl ight and black-blood mra, in the evaluation of arteriove- nous fi stulas in haemodialysis patients in comparison to multidetector computed tomography angiography (mdcta). material and methods: nce-mra and mdcta were performed on the same day in 21 patients on maintenance haemodialysis with dysfunctional arteriovenous fi stulas. the fi stulas included three segments: arterial infl ow, anastomosis, and venous outfl ow. two experienced observers, who were blinded to the results of the nce-mra, recorded in consensus the sig- nifi cant stenoses (?50%) seen on cta. two other experienced observers, unaware of the results of cta, independently recorded signifi cant stenoses (?50%) in the nce-mra. the sensitivity, specifi city, positive predictive value, negative predictive value, and accuracy of nce-mra were calculated, with mdcta as the standard reference. results: sixty-three vascular segments in the 21 patients were clearly displayed. for the two observers of nce-mra, the accuracy was 98% and 95.4%; sensitivity 96.4% and 96.4%; specifi city 97.1% and 94.3%; positive predictive value 96.4% and 93.1%; and, negative predictive value 97.1% and 97.1%. inter-/intra-observer agreement for detecting stenosis was excellent for nce-mra, with a weighted kappa of 0.968 (95% confi dence interval ci, 0.874e1) and 0.936 (95% ci, 0.848e1). conclusion: non-contrast-enhanced mra, using time-of-fl ight and black-blood mra, is a reproducible and reliable imaging technique for detecting ?50% stenosis in dysfunctional haemodialysis arteriovenous fi stulas. ? 2015 the royal college of radiologists. published by elsevier ltd. all rights reserved. introduction arteriovenous fi stula(avf),syntheticarteriovenous graft, and tunnelled central dialysis catheter are the major vascular access alternatives for long-term haemodialysis. avf, especially the radiocephalic fi stula at the wrist (bres- ciaecimino avf), is the preferred access of choice for haemodialysis.1e4vascularstenosisisthemajor * guarantor and correspondent: g.f. zhang, department of radiology, obstetrics and gynecology hospital of fudan university, huangpu district, no. 419 fangxie road, shanghai, china. tel.: 86 e-mail address: 344092668 (g.f. zhang). contents lists available at sciencedirect clinical radiology journal homepage: /10.1016/j.crad.2015.04.005 0009-9260/? 2015 the royal college of radiologists. published by elsevier ltd. all rights reserved. clinical radiology 70 (2015) 852e857 complication of avfs and can lead to reduced blood fl ow and shortened lifespan of the fi stulas. thus, early diagnosis and treatment of stenosis in avfs is very important.5 several surveillance techniques can be used for detecting impending vascular access stenosis: venous pressure mea- surements can trace distally located stenosis, but they do not refl ect venous outfl ow resistance.6 colour doppler fl ow ultrasonography is convenient and non-invasive for evalu- ating vascular access function,7but it has drawbacks: overestimation of stenosis at the arterial anastomosis, in- accuracy in the detection of non-thrombosed obstruction of venous outfl ow in the upper extremity, and absence of an angiographic map, which is necessary for surgical or percutaneous therapy. in addition, the quality of colour doppler ultrasound images depends on the skill of the operator. currently, digital subtraction angiography (dsa) is the reference standard for evaluating vascular access. dsa, however, also has drawbacks: radiation exposure of the patient and radiology personnel, the use of potentially nephrotoxic iodinated contrast agents, limited spatial in- formation,8and the frequent need tohospitalise patients for their safety. multidetectorcomputedtomographyangiography (mdcta) and magnetic resonance angiography (mra) can increase spatial resolution and reconstruct the three- dimensional structures of a vascular access, thus further increasing sensitivity and specifi city of the technique.9e11 however, mdcta has some of the same drawbacks as dsa: risk of iodinated contrast reaction and radiation exposure. in addition, gadolinium-based contrast material- enhanced mra may be nephrotoxic in patients who have impaired renal function.12e14 non-enhanced time-of-fl ight (tof)-mra has been used for imaging avfs,15,16but disturbed fl ow near the stenosis may result in an inaccurate assessment of the fi stula geometry. black-blood (bb)-mra can clearly display the vascular geometry with the use of fl ow suppression techniques. the aim of the present study was to determine the use- fulness of nce-mra, using tof-mra and bb-mra, in evaluating dysfunctional avfs in haemodialysis patients in comparison with conventional mdcta as the standard of reference. material and methods patients twenty-one patients (16 male and fi ve female) on maintenancehaemodialysiswithbresciaeciminoavf dysfunction were enrolled. exclusion criteria were the routine mr imaging (mri) patient contraindications, such as the presence of pacemakers; claustrophobia; metal in the eyes; poor health requiring continual monitoring; and known allergic reactions to iodinated contrast agents. the mean age of the patients was 57 ?12.9 years (sd). the main causes of end-stage renal disease were primary glomeru- lonephritis (15 patients); benign nephrosclerosis (three patients);diabeticnephropathy(twopatients);and polycystic kidney disease (one patient). all patients under- went both nce-mra and mdcta examination within 1 day. the study was approved by the institutional ethics com- mittee, and written informed consent was obtained from all patients. mri all mra examinations were performed using a 3 t mri unit (achieva; philips healthcare, best, the netherlands). the patient was positioned supine in the magnet. the arm containing the fi stula was placed in the anatomical position at the patients side, as close to the centre of the magnet as possible. a surface coil (sense-flex-torso; philips medical systems) was placed over the site of the haemodialysis fi stula and fi xed with tape. mra sequences were acquired. the fi rst sequence was a two-dimensional (2d) tof-mra acquisition covering a transverse slab from approximately 10 cm above to 4 cm below the anastomosis, with acquisi- tion parameters of 11 ms/3.8 ms repetition time/echo time; 70? fl ip angle; 120 ? 120 ? 140 mm3volume of interest; 0.5 ? 0.5 mm/pixel inplane resolution; 2 mm section thickness; 300 ? 300 acquisition matrix; one signal ac- quired; and 4 minutes 31 seconds scan duration. based on the 2d-tof image, one radiologist, with extensive experi- ence in avf diagnosis, preliminarily determined the posi- tion of stenosis. the second sequence was obliqueesagittal t1-weighted bb-mra, double-inversion recovery turbo spin-echo (tse) sequences, with acquisition parameters of 800 ms/13 ms repetition time/echo time; 180? fl ip angle; 1.5 mm slice thickness; 256 ? 256 matrix size; two signals acquired; 120 mm ? 120 mm fi eld of view (fov), centred on the anastomosis; and 6 minutes 12 seconds scan duration. this coverage is usually suffi cient to image the vascular tree of avf at the wrist. the bb-mra sequence was performed using the mri three-point planscan technique, with sus- pected vascular stenosis shown by tof-mra images. three points were selected on the transverse images obtained with tof at the following locations relative to the anasto- mosis: proximal 2 cm (arterial); 0 cm (anastomosis); and distal 2 cm (venous). these three points defi ned the centre plane. mdcta mdcta was performed using a 256-mdct scanner (brilliance ict, philips healthcare, cleveland, oh, usa). the protocol used a pitch of 0.99, acquisition of 128 ? 0.625 mm, rotation time of 0.5 seconds,120 kv, and 310 mas. the scan range was set from the inferior border of the mandible to the inferior border of the involved limb. non-ionic iodin- ated contrast agent (80e100 ml) was injected at a rate of 3e4.5 ml/s into the antecubital vein of the contralateral limb. a contrast agent concentration-tracking technology was used, with the aortic arch regarded as the region of interestandthethresholdsetat150hu.imaging commenced after a delay of 18 seconds, when the region of interest arrived at its threshold. w.t. jin et al. / clinical radiology 70 (2015) 852e857853 theimageswerereconstructedusingastandardalgorithm withtheuseofa0.625mmsectionthicknessanda0.625mm section gap. various post-processing techniques were used, such as surface reconstruction, multiple planar reconstruc- tion, maximum intensity projection, and volume rendering on the extended brilliance workspace workstation. imaging analysis for both nce-mra and mdcta, the haemodialysis access was divided into three segments: arterial infl ow, anasto- mosis, and venous outfl ow. two independent observers determined, in a blinded fashion, whether a signifi cant (?50%) stenosis was present.5for any disagreement in image analysis on mdcta, a third observers review was used to establish a consensus interpretation. the degree of stenosis on both nce-mra and mdcta was measured with electronic callipers. the measured diameter of the residual lumen at the point of maximal narrowing in a segment was compared with the measured diameter at a normal point in that segment by use of the formula: 1?(narrowest anastomotic diameter/normal adjacent diameter) ?100%. one experienced observer assessed image quality with conspicuity scoring for each segment on nce-mra and mdcta. a numeric scoring system was used with a four- point scale: 1, poor quality and inadequate information for answering clinical queries and making decisions about further therapy; 2, fair quality but inadequate information concerning clinical queries and making decision about further therapy; 3, good quality and adequate information; and 4, excellent quality and adequate information. statistical analysis all analyses were performed with spss19.0 and medcalc (ver.11). accuracy, sensitivity, specifi city, positive predictive value (ppv), and negative predictive value (npv) of nce- mra for the detection of signifi cant (?50%) stenosis were calculated, with mdcta as reference standard, and are presented with exact 95% confi dence intervals (cis) derived from the binomial distribution. interobserver agreement for both readers and intra-observer for observer 1 on the detection and grading of stenosis on mra were evaluated with a linear weighted version of cohens kappa. perfect agreement resulted in a kappa value of 1; agreement ex- pected on the basis of chance alone resulted in a kappavalue of 0. the strength of agreement was interpreted as poor (k 0.20), fair (k 0.21e0.40), moderate (k 0.41e0.60), good (k 0.61e0.80), or excellent (k 0.81e1.0). wilcoxons signed-rank test was used to evaluate differences in image quality between nce-mra and mdcta. results sixty-three vascular segments in 21 patients were eval- uated using nce-mra and mdcta. afterconsensus reading, 28 signifi cant stenoses and 35 normal vessels or mild ste- noses (50%) were identifi ed with mdcta. with mdcta as the standard of reference, the accuracy, sensitivity, speci- fi city, ppv, and npv for detecting a signifi cant stenosis were calculated for 63 segments. for observer 1, 28 signifi cant stenoses present on mdcta and one signifi cant stenosis not detected on mdcta were identifi ed on nce-mra. one stenosis that was graded as mild on nce-mra was graded as signifi cant on mdcta. the resulting accuracy was 96.8% (95% ci, 91.7e100.0%); sensi- tivity 96.4% (95% ci, 81.7e99.9%); specifi city 97.1% (95% ci, 85.1e99.9%); ppv 96.4% (95% ci, 81.6e99.9%); and, npv 97.1% (95% ci, 85.1e99.9%), respectively. for observer 2, nce-mra detected 29 signifi cant steno- ses, including two signifi cant stenoses not present on mdcta; one stenosis that was graded as a mild stenosis on nce-mra was graded as a signifi cant stenosis on mdcta. the resulting accuracy was 95.4% (95% ci, 89.3e100%); sensitivity 96.4% (95% ci, 81.7e99.9%); specifi city 94.3% (95% ci, 80.8e99.3%); ppv 93.1% (95% ci, 77.2e99.1%); and, npv 97.1% (95% ci, 84.7e99.9%) of nce-mra for detecting signifi cantstenoses.interobserverandintra-observer agreement for detecting and grading stenoses with nce- mra was also excellent, with a weighted kappa of 0.968 (95% ci, 0.874e1) and 0.936 (95% ci, 0.848e1), respectively. image quality for nce-mra was graded fair in one segment (1.5%), good in 10 segments (15.9%), and excellent in 52 segments (82.6%). image quality for ct angiography was graded good in four segments (6.3%) and excellent in 59 segments (93.7%). no segment was graded as poor quality. the image quality was not signifi cantly different between mdcta and nce-mra (p 0.785). discussion the results of the present study of haemodialysis avfs evaluated using nce-mra documented that the technique provides excellent image quality while consistently and accurately demonstrating signifi cant stenosis with high sensitivity and specifi city and negligible interobserver and intra-observer variability. in previous studies, mdcta detected dysfunctional avfs in haemodialysis patients with good results,17,18but mdcta has two distinct disadvantages: exposure of the patient to ionising radiation and the use of iodinated contrast agents, which may cause severe allergic reaction and lead to reduced kidney function. contrast-enhanced mra, which does not involve radiation, has been used for evaluating dysfunctional avfs, with impressive results.16,19unfortu- nately, concerns about a possible relationship between gadolinium-based contrast agents and nephrogenic sys- temic fi brosis (nsf) have brought into question the use of these agents for mra.20,21although this relationship has not been conclusively proven, an association has been seen in several studies.22e26in 2007, the us food and drug administration issued a precautionary black-box warning on gadolinium agents, recommending special caution in their use, especially in renal-compromised patients.27 w.t. jin et al. / clinical radiology 70 (2015) 852e857854 nce-mra, including tof-mra and bb-mra, has been used in a wide variety of vascular studies, including those of the brain, neck, heart, lungs, hands, and lower extrem- ities.28e30varying results have been published concerning the ability of tof-mra to depict and detect stenosis in avfs. in those reports, tof resulted in overestimation of stenosis because of fl ow voids caused by post-stenotic intravoxel phase dispersion.15,16,31individually, tof-mra does not accurately depict the vascular geometry of the stenoses. in the present study, after the position of stenosis was deter- mined preliminarily with tof-mra, a bb-mra sequence wasperformedwiththemrithree-pointplanscan technique; suspected vascular stenosis was depicted by the tof-mra images. bb-mra is a non-invasive high-resolu- tion imaging, combined with fl ow-suppression techniques. bb-mra allows clear visualisation and gives high sensitivity for detection of vascular lumens. it also offers excellent spatial and contrast resolution of vessel walls, allowing better detection of thickened walls and narrowed lumens (fig 1). most importantly, bb-mra is not affected by disturbed fl ow near stenotic segments. with combined tof-mra and bb-mra, images not only can be reformatted to show the access from arbitrary projection angles, but also to show the vascular lumens and vessels walls in the figure 1 (a) tof-mra reveals a suspected stenosis at the anastomosis (arrowhead) and signal of turbulent fl ow at the venous outfl ow (arrow). (b) bb-mra reveals a ?50% stenosis at the anastomosis (arrowhead) and mild stenosis at the venous outfl ow (arrow). (ced) mdcta of the same avf reveals ?50% stenosis at (arrowhead), correlating with that revealed in nce-mra. w.t. jin et al. / clinical radiology 70 (2015) 852e857855 stenotic segments clearly. for example (fig 2), tof-mra revealsasuspectedvascularstenosisatthevenous outfl ow (fig 2a, arrow), and it is uncertain whether the stenosis was mild or signifi cant on tof-mra individually; however, on bb-mra (fig 2b, arrow), the vascular lumen and vessel walls are clearly depicted, and after measure- ment signifi cant stenosis was detected (50%). in the present study, both observers reported one ste- nosis not seen on nce-mra but seen on mdcta. the cause for this failure of nce-mra might be that 2d bb-mri se- quences were used, which are limited in depicting the complex orientations of vessels because section gaps with 2d techniques result in less spatial information. three- dimensional bb-mri sequences may provide better spatial anatomical defi nition of the fi stula, but they require more time (8e12 minutes) than most patients can tolerate, thus motion artefact and low image quality may result. the shorter time requirement of 2d bb-mri (6 minutes) is more tolerable. a clear drawback of nce-mra was the limited scanning span, resulting in imaging a targeted area but not the entire haemodialysis fi stula including the venous outfl ow tract and central veins. moreover, nce-mra requires a long scan duration (10e20 minutes) to display central veins. on the other hand, mdcta and ce-mra allows imaging of the complete vascular tree of haemodialysis access. in a future experiment, the feasibility of stenosis detection with nce- mra in central veins is planned. there are limitatio