Angiographic and Clinical Characteristics of Thoracolumbar Spinal Epidural and Dural Arteriovenous Fistulas
Background and Purpose—The purpose of this study is to compare the angiographic and clinical characteristics of spinal epidural arteriovenous fistulas (SEAVFs) and spinal dural arteriovenous fistulas (SDAVFs) of the thoracolumbar spine.
Methods—A total of 168 cases diagnosed as spinal dural or extradural arteriovenous fistulas of the thoracolumbar spine were collected from 31 centers. Angiography and clinical findings, including symptoms, sex, and history of spinal surgery/trauma, were retrospectively reviewed. Angiographic images were evaluated, with a special interest in spinal levels, feeders, shunt points, a shunted epidural pouch and its location, and drainage pattern, by 6 readers to reach a consensus.
Results—The consensus diagnoses by the 6 readers were SDAVFs in 108 cases, SEAVFs in 59 cases, and paravertebral arteriovenous fistulas in 1 case. Twenty-nine of 59 cases (49%) of SEAVFs were incorrectly diagnosed as SDAVFs at the individual centers. The thoracic spine was involved in SDAVFs (87%) more often than SEAVFs (17%). Both types of arteriovenous fistulas were predominant in men (82% and 73%) and frequently showed progressive myelopathy (97% and 92%). A history of spinal injury/surgery was more frequently found in SEAVFs (36%) than in SDAVFs (12%; P=0.001). The shunt points of SDAVFs were medial to the medial interpedicle line in 77%, suggesting that SDAVFs commonly shunt to the bridging vein. All SEAVFs formed an epidural shunted pouch, which was frequently located in the ventral epidural space (88%) and drained into the perimedullary vein (75%), the paravertebral veins (10%), or both (15%).
Conclusions—SDAVFs and SEAVFs showed similar symptoms and male predominance. SDAVFs frequently involve the thoracic spine and shunt into the bridging vein. SEAVFs frequently involve the lumbar spine and form a shunted pouch in the ventral epidural space draining into the perimedullary vein.
Spinal dural arteriovenous fistulas (SDAVFs) are the most common vascular shunts of the spine. They occur predominantly in men (80%), commonly involve the thoracolumbar spine, and usually cause progressive myelopathy because of venous congestion of the spinal cord.1,2 It is widely accepted that SDAVFs is fed by the radiculomeningeal artery, located on the dura mater of the spinal nerve root sleeve, and drains into the radiculomedullary vein.2,3 In contrast, spinal epidural arteriovenous fistulas (SEAVFs) are rare vascular shunts that have been thought to present with benign symptoms, such as radiculopathy. The shunt is located in the epidural space and drains into epidural veins.3–5 However, their angiographic and clinical features have not been well investigated in a large number of cases. Some investigators regard SDAVFs as lateral SEAVFs, in which the arteriovenous shunts develop in the lateral epidural space, although the arteriovenous shunt of an SDAVF is located on the dura mater.3,5 Therefore, SEAVFs are poorly understood and may be misdiagnosed as SDAVFs.
The aim of this study was to compare the angiographic and clinical characteristics of SEAVFs and SDAVFs of the thoracolumbar spine based on a retrospective multicenter cohort study.
This study is approved by the institutional review boards of all collaborative institutions.
Materials and Methods
We will not make our data and study materials available to other researchers.
A total of 207 consecutive patients with a diagnosis of SDAVFs, SEAVFs, or paravertebral arteriovenous fistulas (AVFs) of the thoracolumbar spine between 2005 and 2016 were enrolled from 31 neurovascular centers in Japan. All participating centers were neurosurgical or neuroendovascular training centers certified by the Japan Neurosurgical Society and the Japanese Society of Neuroendovascular Therapy and expressed interest in study participation (Table I in the online-only Data Supplement).
Data were collected retrospectively using a standardized form, including age, sex, symptoms, history of spinal surgery or trauma, treatment methods, treatment results, and angiography and magnetic resonance imaging in DICOM (digital imaging and communications in medicine) data format. Symptoms included myelopathy because of venous congestion, radiculopathy, intramedullary hemorrhage, subarachnoid hemorrhage, and others. All datasets were organized by 3 researchers (S.Y., K.T., and T.T.) at the coordinating center.
All images collected were reviewed by 6 reviewers (H.K., Y.M., Y.N., K.T., T.I., and M.H.) on a viewer using DICOM viewer software (Osirix) at the coordinating center. Each reviewer had >10 years of experience in spinal angiography. Five are neurointerventionalists certified by the Japanese Board of Neuroendovascular Therapy, and one is a spine neurosurgeon certified by the Neurospinal Society of Japan. After the preliminary review, 38 patients were excluded because of poor image quality (low spatial or temporal resolution) precluding analysis of the details of the angioarchitecture (n=20), sacral or cervical levels (n=12), intradural AVFs (n=5), and AVFs associated with spinal tumors (n=2). Ultimately, 168 lesions in 168 patients were analyzed in detail.
An anteroposterior view of selective angiography of each segmental artery was reviewed. In addition, lateral and oblique views were also interpreted when the AVF was detected in the anteroposterior view. Three-dimensional angiography, including cone beam computed tomography and multiplannar reconstruction images of rotational angiography, superselective angiography of the feeding arteries, video, and photographs taken during the surgical procedure were also reviewed when available.
All images were reviewed by the 6 reviewers to reach a consensus about the feeding artery, the location of the AVFs, the presence and location of a shunted venous pouch in the epidural space, and the drainage vein. At the end of each case review, the final diagnosis of AVF was made by a consensus among all the 6 readers, and it was compared with the initial diagnosis at the participating center.
An SDAVF was defined as an arteriovenous shunt located on the dura and draining to the intradural vein alone. An SEAVF was defined as an arteriovenous shunt located outside the dura, draining to the epidural vein with the formation of an epidural venous pouch, and then into the paravertebral vein or the intradural vein (radiculomedullary vein).
For the cases of SDAVF, the location of the shunt point in relation to the thecal sac on the anterior view of the spinal angiography was evaluated using the medial interpedicle line as the estimated lateral margin of the thecal sac. The medial interpedicle line is the tangent line connecting the most medial point of the pedicle circles between the upper and lower spinal levels (Figure 1). The location of the shunt point was classified into 3 types: lateral, on line, and medial. Lateral type suggests that the AVF is located on the dural sleeve and drains into the radiculomedullary vein, whereas medial type suggests an AVF located on the spinal dura mater (thecal sac) and draining into the bridging vein. The location of the shunt point is undetermined for the on line type.
For the cases of SEAVF, the location of the shunted epidural venous pouch was classified as ventral, lateral, or dorsal based on the anterior and lateral views of the spinal angiography. A lateral epidural AVF was defined as a shunted pouch lateral to the medial interpedicle line on the anterior view. The shunted pouches of the ventral and dorsal epidural AVFs are medial to the medial interpedicle line and are located on the ventral side and dorsal side in the spinal canal on lateral view, respectively.
Clinical and angiographic data, which included types of AVFs, angiographic findings (spinal levels, laterality, main feeding arteries, and drainage types of AVFs [perimedullary drainage, paravertebral drainage, or both]), and clinical findings (symptoms, age, sex, and history of spinal injury or surgery), were summarized using descriptive statistics.
All statistical analyses were performed using XLSTAT software. Correlations of types of AVFs with spinal levels, laterality, bilateral supply, symptoms, sex, and history of spinal surgery/trauma were statistically analyzed using the χ2 test. The statistical level of significance was set at P=0.05. Vascular anatomy related to SDAVFs and SEAVFs and anatomic terms used in this article are demonstrated in Figures I and II in the online-only Data Supplement.
The initial diagnoses of the AVFs at the 31 individual centers were SDAVFs in 135, SEAVFs in 31, and paravertebral AVFs in 2 patients. The final diagnoses after precise review of the imaging data by 6 readers were SDAVFs in 108, SEAVFs in 59, and paravertebral AVFs in 1 patient. Twenty-nine cases (49%) of SEAVFs were misdiagnosed as SDAVFs at the individual centers. After exclusion of 1 case of a paravertebral AVF, the remaining 168 cases were further analyzed.
The median ages at presentation in patients with SDAVFs (SDAVF group) and SEAVFs (SEAVF group) were 64.4 and 66.6 years, respectively. Both groups predominantly contained men (89 of 108 in SDAVF group and 43 of 59 in SEAVF group). The most frequent symptom of the SDAVFs was progressive myelopathy (105 patients; 97.2%), followed by radiculopathy (5 patients) and back pain (4 patients). Subarachnoid hemorrhage and intramedullary hemorrhage were found in 1 patient each. The most frequent symptom of the SEAVFs was also progressive myelopathy (54 patients; 91.5%), followed by radiculopathy (7 patients) and back pain (1 patient). Mean value of the modified Rankin Scale at presentation was 2.9 (1–5) in SDAVFs group and 3.1 (0–5) in SEAVF group, respectively. There were no significant differences in age, sex, or symptoms between the 2 groups. A history of spinal surgery or trauma was more frequently observed in the SDAVF group (13 patients; 12%) than the SEAVF group (21 patients; 36%) with statistical significance (P=0.001; Table II in the online-only Data Supplement).
SDAVFs were much more frequent at the thoracic spine levels (94 of 108 cases; 87%) than at the lumbar spinal levels, whereas SEAVFs were frequently located at the lumbar spinal levels (49 of 59 cases; 83%). There were significant differences in frequency between SDAVFs and SEAVFs at the T5 to T8 levels (P<0.001) and L2 to L5 levels (P<0.001; Figure III in the online-only Data Supplement; Table).
There was no overall laterality for SDAVFs or SEAVFs. SDAVFs were located on the right side in 56 patients and the left side in 52 patients. SEAVFs were located on the right side in 22 patients and the left side in 31 patients. Six of 59 cases of SEAVFs showing diffuse epidural AVFs were located bilaterally.
Main Feeding Arteries
Meningeal branches of the radiculomeningeal artery were the main feeding arteries in 99 of 108 cases (91.7%) of SDAVFs. Meningeal branches from the prelaminar artery often fed the SDAVFs as an additional feeder (n=28; 26%) or the main feeder (n=9; 8.3%). The feeding arteries arose from 2 ipsilateral segmental arteries at serial spinal levels in 14 (13%) or from the bilateral segmental arteries in 5 cases (4.6%). These feeding arteries ran medially on the surface of the dural sleeve and turned longitudinally along the thecal sac. Then they gathered and joined the single vein on the inner dural surface. In typical cases, these longitudinal meningeal feeders and a drainage vein showed horizontal T sign in an anteroposterior view of angiography (Figure 2).
SEAVFs were mainly supplied from the dorsal somatic branches in 54 patients (91.5%), the prelaminar artery in 4, and the radiculomeningeal artery in 1. The ventral somatic branches also fed the SEAVF as accessory feeders in 14 patients. Multiple feeders joined to an epidural venous pouch at the multiple fistulous points (Figure 3). The feeders arose from ipsilateral multisegmental arteries at serial spinal levels in 15 (25.4%) cases. The SEAVFs were more frequently supplied from bilateral segmental arteries (n=29; 49.1%) than the SDAVFs with statistical significance (P<0.0001).
Locations of the Shunt Point in SDAVFs
Twenty cases were excluded from this analysis because of lack of the images depicting the pedicle circle. In the remaining 88 cases, the shunt point of the SDAVF was more frequently located medial to the medial interpedicle line (n=68; 77%) than any other position (on line in 16 cases, 18%; lateral in 4 cases, 5%; Figures 1B and 4).
Locations of the Shunted Venous Pouch in SEAVFs
The shunted venous pouch of the SEAVFs was located much more frequently in the ventral epidural space (52 cases; 88.1%) than the lateral epidural space (6 cases; 10.2%) or the dorsal epidural space (1 case; 1.7%; Figures 3 and 4).
Drainage Types of SEAVFs
Every SEAVF formed an epidural venous pouch, and it drained into the perimedullary vein in 44 cases (75%), the paravertebral veins in 6 (10%), or both in 9 (15%; Figures 3, 4, and 5). In the cases of ventral SEAVFs, the drainage route from the epidural venous pouch to the radiculomedullary vein typically formed an acute angle, characteristic of J turn sign.
In 108 patients with SDAVFs, 107 patients were treated by surgical interruption of the intradural draining vein (n=39), transarterial embolization (TAE) using n-btyl 2 cyanoacrylate (n=47), or combination of open surgery and TAE (n=19). One patient was observed without treatment. Disappearance of SDAVF was obtained in 102 patients (95%), which includes 44 of 49 patients treated by TAE alone and all 58 patients treated by open surgery or combined treatment. Complications related to the procedure were encountered in 6 patients, which include arterial injury during catheterization (n=3), delayed spinal venous thrombosis after embolization (n=2), and wound infection after open surgery (n=1). Ninety-eight patients could be followed for 3 months, and modified Rankin Scale at 3 months after treatment was improved in 59 patients (60%), unchanged in 37 patients (38%), and worsened because of delayed spinal venous thrombosis in 2 patients (2%). Recurrence of SDAVF was observed in 3 patients (3%). All the 3 patients were initially treated by TAE alone, and they were retreated by open surgery.
Among the 59 patients with SEAVFs, 45 patients (76.3%) were treated by endovascular technique alone, including TAE, using n-btyl 2 cyanoacrylate (n=43) and transvenous embolization (n=2). Nine patients were treated by surgery of interruption of the intradural draining vein with or without surgical devascularization of the feeders, and 4 patients were treated by combination of open surgery and TAE. One patient was observed without treatment. Disappearance of SEAVF was obtained in 43 patients (74%), which includes 31 of 43 patients treated by TAE alone, 2 of 2 patients treated by transvenous embolization, and 10 of 13 patients treated by open surgery or combined treatment. Complications were encountered in 2 patients, which were arterial injury during catheterization. Fifty-one patients could be followed for 3 months. The modified Rankin Scale at 3 months after treatment was improved in 33 patients (65%), unchanged in 17 patients (33%), and worsened in 1 patient (2%). Recurrence of SDAVF was observed in 3 patients (6%), which were 2 cases initially treated by TAE alone and 1 treated by open surgery.
SEAVF is an entity consisting of a spinal arteriovenous shunt draining primarily into the epidural venous plexus, and it can also involve the bony structure adjacent to the epidural space.3,4,6 SEAVFs had been diagnosed as paravertebral AVFs or SDAVFs with epidural venous drainage before the recognition of this entity.7,8 SEAVFs are classified into 3 types according to their venous drainage, including intradural/perimedullary drainage, paravertebral drainage, and combined perimedullary and paravertebral drainage.9,10 SEAVFs with paravertebral drainage can be diagnosed because of their typical drainage to the outside of the spinal canal. However, even now, SEAVFs with perimedullary drainage alone are often incorrectly diagnosed as SDAVFs. In this study, 29 cases(49%) of SEAVFs were incorrectly diagnosed as SDAVFs at the individual centers.
Geibprasert et al5 classified SEAVFs and SDAVFs into 3 groups according to the location (ventral, lateral, and dorsal). According to their classification, SDAVF was classified as lateral epidural group, which frequently showed aggressive symptoms. They demonstrated that the ventral epidural group showed benign clinical presentations, a lower rate of spinal venous reflux, and predominantly female demographics.5 However, a few reports demonstrated high frequency of spinal venous reflux in ventral SEAVFs at thoracolumbar spine.9,10 The results of the present study showed no significant differences between the SDAVF and the SEAVF in sex, age, or clinical presentations. Both showed predominantly male demographics, frequent symptoms of progressive myelopathy, and frequent spinal venous drainage. This discrepancy may be because of the difficulty of diagnosing SEAVFs correctly.
Several differences in angioarchitectural features are observed between the SDAVFs and SEAVFs.
First, ventral SEAVFs with perimedullary drainage show a characteristic J turn sign consisting of the epidural pouch and the radiculomedullary vein.
Second, several differences in other angioarchitectural features, including the main feeding artery and frequency of bilateral feeders, are also observed. The majority of the SDAVFs are mainly supplied by meningeal branches from the radiculomeningeal artery. These meningeal branches anastomose vertically and drain into the single intradural vein, which typically forms horizontal T sign on a frontal view of spinal angiography.11 On the contrary, SEAVFs are almost always supplied by epidural arteries, especially the dorsal somatic branch, often bilaterally.9
Third, the spinal levels involved are significantly different. SEAVFs should be primarily suspected for cases of spinal AVF at the lumbar spinal levels because SEAVFs were much more frequently located at lumbar spinal levels than SDAVFs.
These angiographic findings can help to differentiate SEAVFs from SDAVFs.9–11 Three-dimensional angiography, particularly cone beam computed tomography and multiplannar reconstruction images, is useful for evaluation of these angioarchitecture.9,12
Bridging Venous Drainage of SDAVFs
It is generally thought that SDAVFs are located at the dura mater of the spinal nerve root sleeve and drain into the radiculomedullary vein.2,3 In healthy anatomy, the drainage veins from the spinal cord, called radiculomedullary veins, run along the nerve root and pierce the dura at the spinal nerve root sleeve; radiculomedullary veins are found in 60% of healthy subjects.13 The remaining 40% shows venous drainage of the spinal cord via the bridging vein, which runs apart from the nerve root and pierces the dura mater of the spinal cord to join the epidural venous plexus. SDAVFs can involve either the radiculomedullary vein or the bridging vein.3 However, the bridging venous drainage has not been clearly noted. In our results, the shunt point of the SDAVF was lateral to the medial interpedicle line in only 5% of cases. This indicates SDAVFs commonly shunt to the bridging vein.
It is important to differentiate SEAVFs from SDAVFs in selection of treatment strategy. TAE using liquid embolic materials is an effective and less-invasive technique for the treatment of SDAVFs.14 SDAVFs can cure when n-btyl 2 cyanoacrylate or Onyx reaches into the proximal portion of the intradural draining vein. However, it is often difficult to navigate a microcatheter into a proper feeding artery because the main feeders of SDAVFs from the radiculomeningeal artery usually originate at acute angle and run tortuously. Furthermore, radiculomedullary artery or radiculopial artery often originates from the radiculomeningeal artery together with feeders of SDAVFs. In this study, less than half of cases were treated by TAE alone. Surgical interruption of the intradural draining vein is a promising method for the treatment of SDAVFs although it is more invasive technique. In our series, 54% of the cases of SDAVFs were treated by open surgery alone or combined with TAE, and all of them disappeared without recurrence. The majority of SDAVFs shunt to the bridging vein, which penetrates the dorsal spinal dura matter, and, therefore, shunted bridging vein can be easily identified during surgical procedure. On the contrary, primary surgical treatment is less frequently undergone for the SEAVFs.15 Interruption of the intradural draining vein can cause remnant of epidural AVF, and recruitment of the retrograde intradural drainage may occur via radiculomedullary vein at another spinal level.16 Surgical approach to the arterized venous pouch has a risk of massive bleeding, and it is difficult because the venous pouch is usually located at ventral epidural space. Therefore, endovascular technique is preferred to use for the treatment of SEAVFs in resent reports.9,10,15 Nasr et al15 recently reported a case series of SEAVF at a single center. In their series, 18 of 24 patients (75%) were treated by endovascular technique, 4 (16.7%) by open surgery, and 2 (8.3%) by combination of both. Similarly, our study shows open surgery was less frequently performed approximately in 13 cases (22%) of SEAVFs. Three of the 13 SEAVFs persisted after surgical treatment. Regarding endovascular treatment of SEAVFs, curative rates of SEAVFs by endovascular treatment varied from 59% to 100%.9,10,15 SEAVFs can be completely obliterated when a liquid embolic material fills in the venous pouch and partially in the drainage vein. Catheterization to the proper feeders of SEAVF is easier than that for SDAVF because the main feeder of the dorsal somatic branch usually runs straight, and, therefore, complete obliteration of the SEAVFs can be easily obtained when the SEAVFs have a small epidural venous pouch with few feeders. However, complete filling of the embolic materials in the entire venous pouch and draining veins is often difficult for the cases with a large epidural venous pouch fed by numerous feeders and with paravertebral drainage.9 Combined techniques or transvenous embolization should be applied for such cases.
There are different protocols for spinal angiography using different angiographic machines because of the retrospective multicenter cohort study design. Therefore, it is difficult to analyze the angioarchitecture of numerous small feeders precisely based on conventional angiography in some cases. We excluded 20 cases from this study because of poor image quality. Three-dimensional rotational angiography with multiplanar reformatted image reconstruction, cone beam computed tomography, and superselective angiography are useful for precise evaluation by providing detailed angioarchitecture. Therefore, a prospective study with a large number of cases using the same angiographic protocol with 3-dimensional angiography would be required for further evaluation.
SDAVFs and SEAVFs showed similar symptoms of myelopathy and male predominance. SDAVFs frequently involve the thoracic spine and are commonly fed by the radiculomeningeal artery and drained into the bridging vein. SEAVFs frequently involve the lumbar spine and form a shunted pouch in the ventral epidural space, draining into the perimedullary vein or paravertebral vein. Recognition of these differences is helpful for correct diagnosis.
Sources of Funding
This study is financially supported by Japanese Society of Neuroendovascular Therapy.
Presented in part at the Annual Meeting of the World Federation of Interventional and Therapeutic Neuroradiology, Budapest, Hungary, October 16–19, 2017.
Guest Editor for this article was Giuseppe Lanzino, MD.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.019131/-/DC1.
- Received June 5, 2017.
- Revision received September 18, 2017.
- Accepted September 28, 2017.
- © 2017 The Authors.
Stroke is published on behalf of the American Heart Association, Inc., by Wolters Kluwer Health, Inc. This is an open access article under the terms of the Creative Commons Attribution Non-Commercial-NoDerivs License, which permits use, distribution, and reproduction in any medium, provided that the original work is properly cited, the use is noncommercial, and no modifications or adaptations are made.
- Krings T,
- Geibprasert S
- Patsalides A,
- Knopman J,
- Santillan A,
- Tsiouris AJ,
- Riina H,
- Gobin YP
- Geibprasert S,
- Pereira V,
- Krings T,
- Jiarakongmun P,
- Toulgoat F,
- Pongpech S,
- et al
- Chul Suh D,
- Gon Choi C,
- Bo Sung K,
- Kim KK,
- Chul Rhim S
- Heier LA,
- Lee BC
- Brinjikji W,
- Yin R,
- Nasr DM,
- Lanzino G
- Takai K,
- Komori T,
- Taniguchi M
- Aadland TD,
- Thielen KR,
- Kaufmann TJ,
- Morris JM,
- Lanzino G,
- Kallmes DF,
- et al
- Lasjaunias P,
- Berenstein A,
- Ter Brugge K
- Gemmete JJ,
- Chaudhary N,
- Elias AE,
- Toma AK,
- Pandey AS,
- Parker RA,
- et al
- Nasr DM,
- Brinjikji W,
- Clarke MJ,
- Lanzino G