Spinal dural fistula and anterior spinal artery supply from the same segmental artery: Case report of volumetric T2 MRI diagnosis and rational endovascular treatment

Abstract

Spinal dural fistulas (SDAVFs) occasionally arise from the same segmental artery as the radiculomedullary branch to the anterior spinal artery. In such cases, selective fistula embolization that does not endanger the anterior spinal artery is not possible, and surgical fistula disconnection is recommended. We present an exceptional case in which rational embolization strategy of SDAVF was feasible because of separate origins from a common segmental artery pedicle of the ventral radiculomedullary artery and the dorsal radicular artery branch supplying the fistula.

Introduction

Spinal dural arteriovenous fistula (SDAVF) is a rare cause of progressive myelopathy.^1,2 It is more common in men than in women, and peak age of onset is between the fifth to seventh decade of life. Anatomically, SDAVF usually represents an arteriovenous shunt within the nerve root sleeve that originates from the radiculodural artery with retrograde drainage via a radicular or bridging vein into spinal cord surface veins. Cord dysfunction is thought to be due to venous congestion and resultant ischemia. The degree of congestion depends on the size of the fistula and, more important, on the extent of available venous outflow from the congested cord surface veins. Stepwise progression of symptoms may be related to successive closure of these venous outlets, which exacerbates cord congestion.

Most SDAVFs arise from mid- to-lower thoracic or lumbar vertebral levels,^1,3 and typically present with progressive lower back and leg pain, sensory disturbances, which may involve the perianal region or legs, bladder dysfunction and, ultimately, paraparesis, which may progress to paraplegia. When motor symptoms are present, a mixture of upper and lower motor neuron signs is a common finding due to the involvement both of cord and roots. The classic magnetic resonance (MR) finding is longitudinally extensive cord edema and multiple dilated veins along the surface of the cord and cauda equina. The extent of edema typically correlates with the degree of vascular congestion and the severity of symptoms.

Dynamic MR angiography and volumetric T2 MR imaging (MRI) are evolving techniques^4,5 that serve to better identify the site of the fistula; however, catheter angiography remains the gold standard for diagnosis.^3,6 Spinal angiography is used not only to diagnose and identify the site of the fistula, but also to establish the supply to the anterior spinal axis to ensure the safety of fistula embolization, to understand the degree of angiographic spinal vascular congestion, and to document the efficacy of the endovascular intervention. In experienced hands, endovascular treatments are highly safe and effective^1,2; however, not all dural fistulas are amenable to this approach. When dural fistulas arise at the same level as the radiculomedullary supply to the anterior spinal artery, surgical fistula disconnection is required to ensure preservation of the radiculomedullary artery.^2,7 We present here an instructive case in which a unusual anatomical variation allowed for a rational embolization strategy of a fistula that arose at the same level as the radiculomedullary supply to the anterior spinal artery.

Case presentation

A 64-year-old African American man re-presented to our Multiple Sclerosis (MS) Care Center with a complaint of worsening severe back pain and progressive paraparesis. The patient was seen in the center several years previously for possible MS following an episode of visual disturbance. However, in view of the history of a mild brainstem stroke, poorly controlled hypertension, hyperlipidemia, and brain MRI findings that were more consistent with vascular disease, the diagnosis of MS was rescinded and disease-modifying therapy for MS was discontinued. Past medical history was also significant for prostate cancer, for which he received radiation and hormone therapy, and subsequent diagnosis of clear cell bladder carcinoma, which was treated with total resection of the bladder and prostate and placement of a urine ostomy. Several months following bladder resection, he noted numbness in his thighs that was followed by severe, sometimes excruciating pain in his back and thighs that led to several visits to the emergency department and high-dose opioids. The pain was ongoing for several months when he noted increasing difficulty walking. MRI of the spinal cord at an outside hospital revealed longitudinally extensive intramedullary T2-hyperintense signal extending from the mid-thoracic cord to the conus, but no evidence of flow voids surrounding the cord. The patient was referred back to our MS center. His examination, approximately two weeks following the onset of walking difficulties, was significant for bilateral leg weakness (iliopsoas 3/5 bilaterally, tibialis anterior 4+/5 bilaterally; upper body strength fully preserved); absent reflexes at knees and ankles; and mild vibratory loss in the toes, worse on the left than the right side.

Repeat MRI was performed at our institution on a 3 T Skyra Siemens magnet. Postcontrast (10 ml Gadavist) fat saturation sequence (repetition time (TR) = 599 ms; echo time (TE) = 9.4 ms; matrix = 384 × 269; field of view (FOV) = 360 × 360; thickness = 3 mm) showed extensive cord edema without identifying any abnormal cord vasculature (Figure 1). However, volumetric isotropic three-dimensional (3D) sampling perfection with application-optimized contrast with different flip-angle evolutions (SPACE) sequence (TR = 1670 ms; TE = 144 ms; matrix = 320 × 320; FOV = 320 × 320; thickness = 1 mm; duration = 4 minutes) demonstrated presence of a few flow voids surrounding the conus, compatible with engorged veins (Figure 2). Notably, those flow voids were not visible on the simultaneously obtained conventional sagittal T2 sequence (Figure 2, TR = 3500 ms; TE = 99 ms; matrix = 448 × 448; thickness = 3 mm). Based on the findings suspicious for an SDAVF, a catheter angiogram was planned.

Figure 1.

(a) Conventional T2, (b) conventional gradient echo, and (c) conventional sagittal thoracic spine postcontrast T1 sequences demonstrating extensive cord edema without obvious vascular congestion.

Figure 2.

Three-dimensional SPACE T2 sequence sagittal views showing cord edema and abnormally dilated vessels dorsal to the cord not appreciated on conventional two-dimensional sequences.

Catheter angiography and embolization

The procedure was performed under general anesthesia, on a Siemens Q Biplane, with 5F short sheath femoral access, utilizing a 5F renal diagnostic catheter (Cordis, Fremont, CA), with fluoroscopic rates of 4 and 7.5 frames per second (fps), and digital subtraction angiography (DSA) rates between 0.5 and 2 fps. Both fistula and radiculomedullary artery supply to the anterior spinal system arose from the left T8 segmental artery (Figure 3). Venous-phase images showed no identifiable outflow for the congested cord surface veins, consistent with the extensive cord edema seen on MRI despite the fact that surface veins were not seen on conventional T2 sequences and other nonvolumetric MRI sequences.

Figure 3.

(a–e) (a) Left T8 segmental artery injection demonstrating the dual presence of spinal dural fistula (black arrow) and radiculomedullary artery (dashed arrow) supply to the anterior spinal artery (white arrows). (b) and (c) Early- (b) and late- (c) phase views of fistula venous runoff demonstrating complete lack of radicular venous outflow in the fistula venous system. (d) and (e) Early- and late-phase views of left T12 segmental artery injection showing artery of Adamkiewicz and lack of identifiable spinal/radicular veins in the venous phase (e) corresponding to marked vascular congestion.

The artery of Adamkiewicz⁸ was identified at the left T12 segmental artery. The ascending limb of the anterior spinal artery segment supplied by the Adamkiewicz was judged to be insufficient to ensure mid-thoracic cord perfusion should T8 supply be compromised. Venous-phase images showed marked congestion with no angiographically identifiable cord surface veins.

High-magnification, stereoscopic, and DynaCT⁹ (12-second rotation time, 2 ml/s 50% dilution contrast injection) images demonstrated separate origins of the radiculomedullary artery and of a radiculodural artery supplying the fistula from the common left T8 segmental artery (Figure 4). This anatomical disposition reflects the normal presence of two radiculodural arteries—dorsal and ventral—supplying their respective nerve roots, adjacent dura, and epidural spaces. Ventral radiculodural artery typically also supplies the posterior aspect of the vertebral body and, when developmentally selected, the anterior spinal axis, in which case it is called the radiculomedullary artery.¹⁰ The dorsal radicular artery supplies the dorsal epidural space, dura, and frequently the posterior spinal arterial system. Typically, dorsal and ventral radicular arteries alike arise from a common radiculodural trunk. However, the length of this common trunk is variable and on occasion, the two trunks may have separate origins (Figure 5). The separate origins of the two trunks in our case was definitively established by stereoscopic and DynaCT views. Subselective catheterization and DSA of the dorsal radicular artery supplying the fistula showed no communication with the anterior spinal system. Following establishment of adequate microcatheter position immediately proximal to the fistula, injection of 50:50 N-butyl-2-cyanoacrylate:Lipidol with added tantalum to enhance visualization was performed under live subtraction, achieving permeation of the fistula and its proximal draining vein (Figure 6). Postembolization views showed complete fistula closure with preserved anterior spinal arterial supply. Injection of the left T12 Adamkiewicz level showed significant improvement in cord venous congestion, as evidenced by visualization of cord surface veins (Figure 7). As is nearly always the case, venous drainage remains impaired because of the paucity of radicular/bridging draining veins.

Figure 4.

DynaCT coronal maximum intensity projection images showing separate origins of dorsal radiculodural artery supplying the fistula and ventral radiculomedullary artery supplying the anterior spinal system. White open arrow: proximal ventral radiculomedullary artery. Black open arrow: retrocorporeal branch of ventral radiculomedullary artery. White arrowhead: ventral radiculomedullary artery in the neural foramen. White dashed arrow: ventral radiculomedullary artery in the spinal canal. Black arrowheads: dorsal radiculodural artery. Black dashed arrow: fistula. Black arrows: dorsal spinal vein draining the fistula.

Figure 5.

(a) and (b) Cranial view: White open arrow: proximal ventral radiculomedullary artery. Black open arrow: retro-corporeal branch of ventral radiculomedullary artery. White arrowhead: ventral radiculomedullary artery in the neural foramen. White dashed arrow: ventral radiculomedullary artery in the spinal canal. Black arrowheads: dorsal radiculodural artery. Black dashed arrow: fistula. Black arrows: dorsal spinal vein draining the fistula. Solid white arrows: anterior spinal artery. (b) Schematic showing typical disposition of dorsal and ventral radiculodural arteries arising from a common pedicle on the right. On the left, ventral radiculomedullary artery and dorsal radiculodural artery supplying the fistula originate separately from the segmental artery.

Figure 6.

(a) and (b) Unsubtracted and digital subtraction angiography views of microcatheter injection at the embolization position immediately proximal to the fistula. (c) N-butyl-2-cyanoacrylate cast.

Figure 7.

(a) Postembolization left T8 injection showing obliteration of fistula and preservation of the anterior spinal artery. (b) Early and (c) late views of left T12 injection demonstrating reduced cord venous congestion, allowing for visualization of previously obscured spinal cord surface veins (white arrows).

After fistula embolization, the patient was maintained on a heparin drip for two days and then on 325 mg daily aspirin for two weeks to minimize the possibility of thrombosis of engorged spinal veins. He was discharged to acute rehabilitation on postprocedure day 2. His subsequent course was complicated by kidney infection that resolved with intravenous antibiotics. Examination four weeks after the procedure showed remarkable neurologic improvement with full leg strength (5/5), ability to ambulate without an assistive device, and resolution of low back and lower extremity pain (opioids no longer necessary). His only residual symptom at this time is right lower extremity numbness below the ankle.

Discussion

Despite continued improvement in MR technology, noninvasive diagnosis of SDAVF remains a challenge. High degree of suspicion of SDAVF in a proper clinical context and clinical and radiographic expertise play key roles in diagnosis, particularly when imaging does not show telltale signs of SDAVF. A recent report by Kannath et al.⁴ highlighted the utility of 3D SPACE T2 in the determination of fistula site in patients with known fistulas, and an earlier report by Morris and colleagues⁵ discussed the use of phase-cycled fast imaging employing steady state acquisition and constructive interference steady state sequences for the same purpose. In our case, volumetric cord MRI provided important clues to diagnosis; however, volumetric imaging is not yet routinely available. Thus, when clinical suspicion is high and routine MRI fails to demonstrate flow voids, referral to a tertiary center with capacity for advanced MRI and catheter angiography should be considered. Timely and accurate diagnosis of SDAVF is essential, as SDAVF is one of very few potentially reversible causes of paraparesis, as in our patient. And, of course, further studies of the role of volumetric T2 in the diagnosis of SDAVF may be warranted.

Our practice for all dural fistulas arising from the same level as either anterior or posterior spinal artery supply is surgical treatment. This is in line with the worldwide standard of care.^1,2,7 In this setting, we usually place a temporary clip on the proximal intradural fistula-draining vein and then perform an intraoperative angiogram to confirm fistula closure and preservation of anterior spinal supply. This is followed by removal of the temporary clip, division and cautery of the fistula-draining radicular vein, and preservation of cord surface veins. We do not advocate a change in this well-established, safe, and effective practice, but wish to add a caveat that under certain circumstances surgery may not be necessary if the origin of the AFV supply and of the anterior spinal artery are separated in space even when they arise from the same segmental artery. We likewise prefer surgery for other complex dural fistulas with potential for endovascular cord injury, such as those at the C1 level. We continue to advise caution and underscore the importance of expertise in SDAVF treatment to maximize positive outcomes. Specifically, endovascular treatment should take into account the possibility liquid embolic reflux during injection that, if extensive, could result in occlusion of an anatomically nearby branch of separate origin. A related technical caveat is the possibility of release of a piece of polymerized embolic material during microcatheter removal, which could likewise result in unintended branch occlusion.

Our case demonstrates how appreciation of an individual’s neurovascular anatomy permits an elegant and responsible departure from standard treatment protocols. Additionally, our case illustrates the utility of the 3D SPACE T2 technique for diagnosis of SDAVF, which escaped detection on conventional nonvolumetric sequences.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.