An Alternative Approach to Treating Spinal Epidural Arteriovenous Fistula: A Case Report of Direct Puncture
Article information
Abstract
We present a case of a patient in their 50s with a spinal epidural arteriovenous fistula (SEDAVF) at L2 level with intradural venous reflux. Initial transarterial embolization was attempted but failed due to vessel tortuosity and vasospasm. The second embolization was carried out with percutaneous puncture of the epidural venous sac under cone-beam computed tomography angiography (CBCTA) guidance. Following the complete obliteration of the fistula, resolution of the venous congestion and significant improvement of the patient’s symptoms were achieved. This case highlights the utility of the percutaneous approach as an alternative treatment strategy for SEDAVFs when traditional endovascular routes are not feasible. Advanced imaging techniques, such as CBCTA, facilitate precise navigation and successful embolization.
INTRODUCTION
Spinal epidural arteriovenous fistulas (SEDAVFs) are uncommon vascular anomalies that occur in the spinal epidural space, with an estimated incidence of less than 2% of all spinal arteriovenous malformations [1]. However, when intradural venous reflux occurs in SEDAVFs through a bridging or radiculomedullary vein, venous hypertension of the perimedullary veins may result in congestive myelopathy [2].
Endovascular treatment (EVT) is a first-line treatment modality in SEDAVFs [3]. The arterial approach is typically the first-line option. However, in cases of SEDAVFs with multiple feeders and fistula points, the transarterial approach may pose significant challenges. In such situations, a transvenous approach may be considered as an alternative [4]. Unlike spinal dural arteriovenous fistulas (SDAVFs), the draining veins of epidural AVFs are frequently situated in the azygous venous system. These veins often contain valves that impede retrograde flow and form metameric connections with paraspinal longitudinal veins, resulting in a complex and tortuous venous network that complicates catheter navigation and embolization procedures [5]. The percutaneous approach is a viable alternative when both arterial and venous approaches are unfeasible, especially in cases with multiple feeders or inaccessible draining veins.
However, limited literature is available on treating SEDAVFs using a percutaneous approach [6]. In this paper, we aim to share our clinical experience managing SEDAVFs that could not be effectively treated with a transarterial approach, highlighting the role of the percutaneous approach as a viable alternative.
CASE REPORT
A patient in their 50s visited the emergency room presenting with leg weakness and sacral pain that had progressively worsened over the previous 10 days. The patient had previously undergone posterior interbody fusion at the T11-L2 level and corpectomy at the L1 vertebral body at another hospital due to hip pain 4 years earlier.
Upon presentation, a whole spine magnetic resonance imaging (MRI) revealed the co-existence of multiple signal voids in the dorsal perimedullary space spanning from T7 to T12, and diffuse T2 hyperintensity in the spinal cord from T5 to L2, indicative of congestive myelopathy (Fig. 1A). These findings strongly suggested the presence of a spinal AVF, necessitating further investigation through spinal angiography.
(A) Sagittal T2WI of thoraco-lumbar region demonstrating long-segmental T2 high signal in the spinal cord with multiple flow voids at dorsal aspect of the cord. (B) Early phase of right L2 segmental artery angiography showing epidural venous pouch. (C) Late phase of the same injection showing extensive venous reflux to perimedullary veins up to T7 level. (D) A maximum intensity projection reconstructed image of cone-beam computed tomography angiography (CBCTA) demonstrating angioarchitecture of the spinal epidural arteriovenous fistula: main feeder of the shunt from dorsal somatic branch (long black arrow) goes to the shunt point (white arrow) and arterialized vein superiorly extending to become a triangular shaped venous pouch (black star). The venous pouch then drained inferiorly via paraspinal venous plexus until it refluxes into intradural segment. Initial part of intradural drainage marked with white arrowheads. (E–G) Multiplanar reconstruction images (E: coronal, F: sagittal, G: axial) show that the venous pouch (arrowhead on axial image) is located near the intervertebral foramen, clear from bony structures, making an advantageous route for percutaneous access (transverse line on coronal image refers to the level of the axial image, and vertical line on sagittal image refers to the level of the coronal image). (H) Left L2 segmental angiography at the beginning of the second treatment session. Note that the shunt amount has diminished compared to the diagnostic angiography. (I) Axial reformatted CBCTA image demonstrating the needle advancing toward the venous pouch (black arrow). (J) Hand injection angiography through the needle tip positioned within the venous pouch (black arrowhead). (K) Appearance of final Onyx 18 (Medtronic) cast after removal of the needle.
Selective angiography of the right L2 segmental artery revealed a SEDAVF with an epidural venous pouch (Fig. 1B) resulting in reflux to perimedullary veins up to the T7 level (Fig. 1C). The shunt and venous pouch were also opacified with the left L2 and right L1 segmental artery injection (not shown). Surgical implants from the previous posterior fusion and corpectomy procedure severely interfered with visualization of the angioarchitecture of the shunt.
To better understand the shunt structure, cone-beam computed tomography angiography (CBCTA) was performed (14s DCT Head Micro, ARTIS Icono; Siemens Healthineers) with injection of contrast media (Visipaque 320; GE Healthcare) at a rate of 0.8 mL/s and 2-second scan delay. A 3-dimensional maximum intensity projection image from the CBCTA is described in detail in Fig. 1D, with all metallic implants carefully eliminated manually on the workstation. Multiplanar reconstruction (MPR) of the CBCTA confirmed the shunt point’s location in the ventral epidural space.
Among the multiple feeders from both L2 and right L1 segmental arteries, the dorsal somatic branch (DSB) of the right L2 segmental artery carried the largest amount of shunt flow (marked with long black arrow in Fig. 1D). Transarterial embolization (TAE) through the DSB feeder was initially planned. The procedure was performed under general anesthesia (GA), and a 6F Envoy (Cerenovus; Johnson & Johnson) was placed at the proximal part of the right L2 segmental artery. However, this was not successful despite multiple attempts with various microcatheters and wires owing to the small caliber and acute angle of the proper feeder. We decided to try a type of modified pressure cooker technique. However, during the process, severe vasospasm with or without dissection occurred in the proximal segment of the segmental artery, which resulted in total occlusion of the artery. The SEDAVF was still visible on left L2 angiography, nevertheless, we could not proceed any further and the procedure was terminated with an early retreatment plan.
As a plan for the second session EVT, percutaneous access was chosen prior to a venous approach for several reasons: the epidural venous pouch was located near the right intervertebral foramen, and the entry route seemed to be clear of bony structures and arteries based on CBCTA MPR images (Fig. 1E–G). A transvenous approach could be considered; however, transvenous embolization (TVE) was put aside as the next option because we were reluctant to navigate into the ascending lumbar vein not knowing where the connection is between the ascending lumbar vein and the arterialized paraspinal veins. It may take too much time to navigate the devices near the shunt, and we may fail to reach the target no matter how close we can get.
Twelve days after the unsuccessful TAE, the second treatment procedure was performed under GA. For the percutaneous approach, the patient was placed in the prone position. For angiography, a 6F sheath was introduced through the left popliteal artery and a 5F Mickelson diagnostic catheter (Merit Medical) was used to select the left L2 segmental artery (right L2 segmental artery was still occluded; not shown). Left L2 injection demonstrated a decreased size of the epidural venous pouch probably resulted from a shunt reduction due to right L2 segmental artery occlusion (Fig. 1H). However, despite the decrease in volume, the shunt remained, with persistent venous reflux to the perimedullary veins.
The percutaneous access was performed with CBCTA guidance (5s Body DCT, ARTIS Icono; Siemens Healthineers), similar to a CT-guided biopsy technique. Skin markers were placed on the patient’s back from spinous process to right lateral side, centered at the L1–2 intervertebral disc level. An 18G biopsy needle (Cook Medical) was gradually advanced toward the target under CBCTA guidance. Contrast medium was injected from the left L2 segmental artery via a 5F diagnostic catheter (advanced through the left popliteal artery) at 1 mL/s for 7 seconds with a 2-second scan delay for every CBCTA acquisition. A total of 6 5s DCT scans were acquired before the needle reached the epidural venous pouch (black arrow, Fig. 1I). Hand-injection angiography through the needle confirmed the location of the needle tip within the pouch (Fig. 1J). And then, we injected Onyx 18 (Medtronic) directly through the needle. A total of 1.6 mL of Onyx was injected (Fig. 1K), and final angiography showed complete obliteration of the SEDAVF (not shown). The patient received steroids to manage edema and was discharged after a rehabilitation program, with no peri-procedural complications. The patient exhibited significant clinical improvement at discharge. The patient’s symptoms improved from P2S2M3S1 (moderate pain, moderate sensory change, wheelchair walking, difficult voiding) before the procedure to P1S1M1S0 (mild pain, mild sensory change, walking difficulty, normal voiding) after the procedure and before discharge [7]. On the follow-up spine MRI taken 2 months later, the congestive myelopathy and the flow voids completely disappeared.
DISCUSSION
In cases of SEDAVFs where progressive myelopathy is present, the importance of urgent and definitive treatment cannot be overstated. The primary goal of embolization is disconnecting the arteriovenous shunt and venous reflux to prevent further neurological deterioration. In this case, percutaneous access to the epidural venous sac aided by repeated CBCTA made it possible to occlude the shunt completely. Although neither prone positioning nor popliteal artery access are commonly used settings in neurointerventional procedures, both proved useful in this unique case and yielded good clinical outcomes.
Even though TAE is considered a primary EVT modality in SDAVFs or SEDAVFs, sometimes it can be quite challenging when the vessels are too small or tortuous. In SEDAVFs, multiple feeders and the existence of an epidural venous sac may complicate EVT even further. The embolization strategy we adopted during the TAE is similar to the plug-and-push technique previously described [8]. There may be various tactics and technical tips to make the liquid embolic material successfully and sufficiently penetrate the shunt point to achieve complete obliteration. Unfortunately, in this case, surgical implants interfered with visualization of the angioarchitecture made the first TAE much difficult than the usual EVT of spinal shunts.
TVE is not as widely utilized as TAE in the treatment of spinal AV shunts. According to Huang et al. [9], only 11% of SEDAVFs were treated with TVE. However, in our case, the lesion was located at the right L1–2 level, requiring a long navigation path caudally via the azygos vein. Additionally, for a cranial approach from the right common iliac vein, there may be multiple interruptions, making it difficult to reach the level of interest [10].
Another option is the percutaneous approach. In the case of a SEDAVF, since a large venous pouch is characteristically present [9] if this venous pouch exists in a location where direct access is possible, embolization can be attempted by direct puncture. Ramanathan et al. [6] reported an example of a patient with SEDAVF successfully treated percutaneously under flat-panel CT guidance, in which the patient relapsed following transarterial, surgical, and transvenous treatment.
In our case, several factors influenced our decision to perform a second embolization procedure using a percutaneous approach after a failed first embolization session. First, in a situation where the main feeder was occluded, the access route to other feeders, the left L2 and right L1 feeders, was long and difficult to select. Second, the venous route was not clearly identified for a transvenous approach. Third, the patient already had a surgical history near the lesion, making it difficult to operate again. The fortuitous location of the epidural venous pouch, accessible for direct puncture, made the percutaneous approach possible. Despite concerns regarding potential nerve damage due to the proximity of the shunt to the nerve root sleeve, the calculated risks were outweighed by the potential benefits, making it the most suitable option under the circumstances.
Another fundamental aspect of the percutaneous approach is that the patient must be in the prone position. Accordingly, an access route for angiography other than the femoral artery is necessary. The popliteal artery can be an appropriate option in terms of vessel size and accessibility [11]. In our case, we chose the popliteal artery over the radial artery because it better suited the usual layout of our angiosuite and the level of interest was the lumbar spine. Our patient underwent the procedure via ultrasound-guided puncture of the left popliteal artery, and no access site complications were identified.
SEDAVF is a rare condition, and its etiology remains unclear. However, several reports suggest a potential association with trauma or surgery [12]. One study reported that among 45 patients with EDAVF, 4 had a history of laminectomy, while 40 had no history of trauma (1 patient had a history of neck trauma) [13]. Another case report described a patient who developed SEDAVF 2 years after laminectomy, suggesting that changes in spinal anatomy may delay symptom onset [14]. Accordingly, it is not unreasonable to assume that the SEDAVF in our case may also have been caused by surgery-related anatomical changes. In particular, the fact that the lesion occurred near the site of prior surgery further supports this hypothesis.
Although CBCTA-guided percutaneous access enabled a safe and precise puncture, it is not without costs: additional radiation exposure and increased contrast media usage. That is why employed the 5-second body protocol for needle guidance. Although the 14-second unbinned acquisition provides the best image quality, needle advancement may require multiple scans; thus, we compromised high spatial resolution in favor of lower contrast volume, shorter scan times, and a larger field of view. It’s essential to have a precise understanding of the scan protocols available and their respective characteristics, so one can select and apply the most appropriate one for any given situation.
SEDAVFs with neurological symptoms require active intervention. The selection of the optimal treatment modality should be guided by anatomical considerations and vascular accessibility. This case highlights the feasibility and efficacy of a percutaneous approach for SEDAVF embolization when transarterial and transvenous strategies are impractical. High-resolution CBCTA serves as a valuable tool for procedural planning and execution, contributing to successful patient outcomes.
Notes
Fund
This study was supported by a 2024 clinical research grant of Pusan National University Yangsan Hospital.
Ethics Statement
The Institutional Review Board (IRB) of Pusan National University Yangsan Hospital approved the exemption from ethical review of this case report (IRB no. 55-2025-055). We anonymized patient information in the case report, such as sex and age, to prevent the identification of individuals.
Conflicts of Interest
The authors have no conflicts to disclose.
Author Contributions
Concept and design: BC and SKB. Analysis and interpretation: SKB. Data collection: BC, SHL, and SKB. Writing the article: BC and JR. Critical revision of the article: SKB. Final approval of the article: SHL and SKB. Obtained funding: JR. Overall responsibility: JR and SKB.