Non-contrast MRI methods as a tool for the pre-operative assessment and surveillance of the arterio-venous fistula for haemodialysis

End-stage renal disease (ESRD) affects an estimated two million people worldwide and is increasing at a rate of around 3% per year [1]. Patients with ESRD require some form of renal replacement therapy (RRT). Haemodialysis (HD) is the most common form of RRT; however, the provision of optimal vascular access, essential for successful HD, has long remained a challenge. The arteriovenous fistula (AVF) first described by Brescia and Cimino in 1966 [2], revolutionised vascular access, allowing repeated access to high volume blood flow by forming a surgical anastomosis between the radial artery and cephalic vein. In response, blood flow through the venous vessel instantly increases. Ideally, the venous and arterial segments of a recently created AVF should dilate in response to the increased blood flow through the newly established low resistance pathway [3].

Unfortunately, vascular access dysfunction is a major cause of morbidity in the dialysis population—most commonly due to thrombosis, secondary to stenosis. Studies have shown that after 1 year, only around 60% of AVFs are still patent [4, 5] and have not required surgical intervention to remain so. Controversy exists as to the role of US in the pre-operative evaluation of the blood vessels for AVF creation. Pre-operative vessel mapping using US has been reported to reduce the immediate failure rate of AVFs and potentially increase patency rates [6‐9]. However, meta-analyses by Georgiadis and Wong [10, 11] back up positive short-term results but fail to show an increase in long term AVF patency.

Magnetic resonance imaging (MRI) is an attractive diagnostic tool as it can provide excellent angiographic images, with or without the use of contrast agent [12, 13]. Planken et al. [14] established that contrast-enhanced (CE) MRI was an accurate modality for providing pre-operative vessel diameters in patients requiring AVF creation. Further work by the same group [15] reported that CE-MRI was superior to US in detecting venous pathologies (most commonly upper arm cephalic vein occlusions) in 73 patients prior to AVF creation.

However, since 2006 there has been a marked decline in MRI based imaging involving patients with end stage renal disease and AVFs. This is likely due to the association between certain Gd-based contrast agents and nephrogenic systemic fibrosis (NSF), a fibrotic disease of the major organs [16]. Non-contrast enhanced MRI (NCE-MRI) methods are currently the safest option for this cohort. The gradient echo time of flight (ToF) sequence is commonly used in MR angiography, and relies on the inflow of blood to produce a hyper-intense signal. Various studies have reported that this sequence can visualise, assess failure and locate stenosis in AVFs [13, 17‐19].

Pre-operatively, very few studies have described the use of NCE-MRI to image the vessels of patients indicated for AVF creation. Menegazzo et al. (1998) [20] compared ToF imaging with venography to assess the depiction of vein diameters prior to AVF creation. ToF measurements were better correlated with subsequent surgical findings. In 2012, Bode et al. [21] used CE-MRI and a NCE balanced turbo field echo sequence to detect stenoses prior to AVF creation. The NCE-MRI detected 66% of (non-significant) stenosis found on CE-MRI, but importantly these were not seen on previous US imaging. More recently a T2* gradient echo “multi-echo data image combination” (MEDIC) sequence [22] has been used to highlight lower limb vessels, indicating possible future uses for this in pre-surgical mapping or surveillance of AVFs. The MEDIC sequence is a flow compensated 3D T2* weighted GRE technique where multiple echoes are acquired during the TR period via frequency encoding gradient reversals. The images acquired from each echo are then combined to form a “combination image”—benefitting from improved SNR. For non-contrast measurement of blood flow velocity, the widely used technique of phase contrast MRA (PC-MRA) is available [23]—providing the opportunity for pre-operative MR assessment of vessel structure and blood flow velocity to be used within a single modality.

The purpose of this pilot study therefore was to compare 3T NCE-MRI methods against the reference standard of US for the assessment of vessel geometry and blood flow velocity in the arms of both patients and healthy volunteers at a single time-point, and also in patient volunteers before and after AVF creation. Specifically, the objective was to examine the performance of three (non contrast) MRI pulse sequences (namely MEDIC, ToF and PC MRI) for these purposes, in order to establish whether they may prove useful the pre and post-surgical assessment of patients listed for AVF creation.

In this pilot study, the feasibility of using 3T MRI techniques for the assessment of upper and lower arm blood flow geometries and velocities at potential AVF sites is described. For morphological assessments, the MRI T2* MEDIC technique used compared favourably against the current reference standard of US, with very similar diameters obtained in patients and healthy volunteers. Further, the “combined” observation of hypo-intense vessel lumen with hyper-intense vessel edges in the vicinity of the anastomosis using MEDIC was a consistent finding in the four examples studied. Further studies are warranted to explore this pattern of contrast behaviour, but initial evidence suggests that it may have the potential to assist with the visualisation of AVF remodelling events.
Increased knowledge of the remodelling process of arteriovenous fistulae is required to help improve scientific understanding of AVF failure, and ultimately to facilitate the pre-operative identification of patients whose fistulae may fail to reach maturation. Current guidelines recommend physical examination or US, but these cannot provide the anatomical detail required to increase the knowledge of the remodelling process. The purpose of this work was to explore wider imaging opportunities as an alternative pre-operative assessment to gain extra anatomical information. We utilised two MRI sequences (ToF and MEDIC), alongside US to compare the geometries of arteries and veins in patients with renal failure (n = 6) and healthy volunteers (n = 10). In a subset of the patients, AVF remodelling was also monitored post-surgery.

The MEDIC and US morphology measures were generally in good agreement with each other. It is possible that the higher SNR of the 3D MEDIC sequence enabled clearer visualisation of the vessel diameters relative to those of the ToF images. Further, the presence of the higher signal “vessel edge” detail on MEDIC (clearly seen in the anastamosis regions on the patients) may additionally contribute to the “true vessel width” not seen on 2D-TOF. From the perspective of US, a possible concern is the fact that tourniquets are routinely used to instigate a level of venous congestion—potentially affecting the vein diameters. However, work presented by van Bemmelen et al. [24] determined that differences between US diameter measurement values obtained with and without a tourniquet were not significant for either the cephalic or the basilic vein. They concluded that vein diameters are affected by warm water immersion, but not the use of a tourniquet.

The lack of MRI contrast media used in this study completely removes any possibility of NSF or contrast induced nephropathy. The MEDIC sequence has the advantage of acquiring the imaging in 3D—thus allowing measurements to be made throughout the length of the vessel, unlike selective measurements from US. Further, the MEDIC does not appear to suffer from through-plane signal losses in quite the same way as ToF. An extension to consider the central vessels would also be feasible as part of a future study. Previous research by Planken et al. [14] has reported that contrast enhanced MR angiography (CE-MRA) is more accurate than US when comparing diameter measurements made during AVF surgery, but no other similar study has been performed using a NCE-sequence. In our study, the radial artery diameter in the PRF participants as measured by MEDIC was in agreement with literature values measured by US [25]. Vessels used for AVF creation were above the recommended minimum vessel diameters stated in the literature [9, 26, 27], but lower than that recommended by Lauvao et al. [28].

Our results show that, pre-operatively, the vessels can have a heterogeneous cross-sectional area (CSA). The implications of this for AVF maturation requires further study. It is possible that narrower (but non-stenotic) regions which exist pre-operatively, may be more prone to stenosis development post-AVF creation. Post-operatively, vessels had achieved or were approaching the recommended diameter of 6 mm for cannulation. This adds to the sentiment that pre-operative vessel diameters are not the major predictor of maturation that some studies would suggest. Heterogeneous changes in the venous segment of an AVF have been observed before [29, 30]. In this study, we included a pre-operative analysis, enabling us to determine if any areas had undergone a decrease in area. It is possible that the areas of post-operative narrowing observed could be the result of neo-intimal hyperplasia, and that the areas of lower dilation are related to local endothelial dysfunction or abnormal flow patterns. However, this would need to be determined by further studies involving larger cohorts, multiple post-surgical analyses and/or haemodynamic simulations.

Few studies have focussed on the changes that occur in the arterial segment, post AVF creation. Corpateux et al. [3] found that the radial artery did not significantly change over a 3-month period. However, we noted large increases in the CSA of the radial arteries studied. The observations noted by us may corroborate the use of MRI for providing additional anatomical data, which may form part of the larger picture of AVF-remodelling.

The purpose of this study was not to demonstrate the diagnostic ability of the MRI sequences for AVF stenosis detection, although regions of narrowing were identified in the venous segment of two AVFs. One of these participants (PRF3) did develop a stenosis at a later date, which was detected by routine US imaging. Further work is needed to determine the pattern of stenosis formation post AVF, and whether MRI may have a role in targeted post-operative stenosis screening.

MRI is rarely used in a clinical setting with for ESRD patients due to financial constraints, patient contraindications and concerns over the use of MRI contrast agents. Contraindication rates to MRI imaging of around 10% in patients with ESRD have been reported previously [12, 31]. Research involving MRI in patients on, or awaiting haemodialysis has also slowed considerably since the occurrence of NSF. Non-contrast enhanced sequences such as black-blood or ToF MRI have been studied previously [29, 30], and the relative limitations of each technique are reported. The MEDIC sequence used in this study (a gradient echo technique with T2* weighting) provides good agreement with US measures, but in common with ToF techniques it also suffers from inflow-related signal losses at the anastomosis. However, the MEDIC sequence does provide good vessel edge detail at the locations in and around the anastomosis, and it is possible that this may prove to be useful as a surveillance tool to characterise the development of stationary structures such as possible stenosis sites.

This study has several limitations. Firstly, detailed optimisation work was not performed on the ToF sequence to maximise the vessel/background contrast for use at the chosen anatomical location—the vendor default parameters were utilised. The flip angle was slightly lower than one might normally expect, although since we were observing relatively slow flow within the veins the use of a larger flip angle may have resulted in partial saturation of the venous signal. It is difficult to optimise the sequence when both arteries and veins are being observed, although we accept that the negative effect of using a smaller flip angle will be an elevated signal from stationary tissue. Lower flip angles were also necessary in order for the acquisition to stay within acceptable SAR limits on our 3T scanner for the chosen TR. Other experimental complexities included the fact that the anatomy was, by definition, positioned at the periphery of the scanner bore in an area of relatively heterogeneous magnetic field—leading to unavoidable regions of inhomogeneous signal intensity. Finally, we were not able to apply fat suppression because doing this would have resulted in a longer sequence TR—leading to unacceptably long scan times. The TR was also implemented with consideration of the expected venous flow velocities likely to be present in the healthy volunteers. We estimated that the slowest diastolic blood flow velocities present would be about 10 cm/s, and using an estimate that the TR would need to be equal or greater than the slice thickness (1.5 mm) divided by the minimum flow velocity (10 cm/s) to allow full in-flow of unsaturated blood—i.e., the TR would need to be a minimum of approximately 15 ms. Our choice of TR 14.5 ms was, therefore, “at the limit” in terms of the slowest blood flow velocities likely to be encountered. However, since the cardiac cycle was postulated to involve velocities mostly above 10 cm/s, we felt that using this short TR would be the best solution to ensure the shortest scan time—and lower chances of subject motion artefacts. We also used a TE of 5.8 ms (the shortest TE to acquire fat/water out of phase images at 3.0T), which was implemented to ensure that the signal from water and fat were out of phase—thus helping to minimise stationary background tissue signals a little further.

Other modern MRI pulse sequences are potentially available for non-contrast enhanced visualisation of the vascular system. These include, for example, “black blood” spin echo sequences (utilising double or triple inversion pulses to null the signal from the vessels), and cardiac gated gradient-echo based sequences that typically use inversion recovery along with a steady state free precession read-out after a systolic inflow period to allow movement of hyper-intense blood into the imaging plane. Common examples of these are Non-Contrast MRA of Arteries and Veins (NATIVE, Siemens), Quiescent Interval Steady State (QISS, Siemens); Balanced-Triggered Angiography Non-Contrast Enhanced (B-TRANCE, Philips); Inhance Inflow IR (IFIR, GE) and Time Spatial Labeling Inversion Pulse (TimeSLIP, Cannon). These sequences may be useful in the future when it becomes possible to optimise them easily for arteries and veins interchangeably.

In summary, our study has demonstrated that the MEDIC sequence can offer accurate depiction of both the arterial and venous dimensions in healthy volunteers and for patients requiring detailed vessel mapping of AVF sites. This information can be acquired in a single imaging session and does not rely on contrast agent injections, which makes it suitable for imaging wider patient groups with ESRD.

This pilot study has demonstrated that non-contrast enhanced MRI can be used to describe vessel morphology and blood flow at anatomical sites associated with AVF surgery—with results comparable to the reference standard of US. Additionally, the pattern of image contrast associated with the MEDIC sequence at the anastomosis site may assist with future studies designed to examine the AVF maturation process without the need for contrast agents.