Classification and Historical Considerations

Historically, spinal dural fistulas were first referred to as angioma racemosum venosum by Wyburn-Mason in his 1943 monograph. This was later revised to angioma racemosum by Bergstrand and colleagues and Krayenbuhl and associates. Malis later described these lesions as long dorsal arteriovenous malformation (AVM).

Many classification schemes exist, contributing to the challenges posed by these lesions. The most common classification system used in the literature divides spinal vascular lesions into types I to IV. Type I lesions are by far the most common faced by most spine surgeons and the main focus of this chapter. Type I lesions represent a true spinal dural AVF where a radicular feeding artery communicates abnormally with the venous system of the spinal cord. Type II lesions represent glomus or intramedullary lesions, which are akin to brain AVMs occurring in the spinal cord. Type III lesions are juvenile or metameric AVMs associated with both extradural and intradural extension of the spinal AVM. Type IV lesions, originally described by Djindjian and more recently classified by Heros, are similar to type I lesions in their fistulous nature; however, these are true perimedullary fistulas, usually fed by the anterior spinal artery, and do not represent true dural AVFs. Spetzler has proposed a more clinically useful anatomic basis for classification of these lesions, which can help to guide treatment. A full discussion of spinal AVMs is beyond the scope of this chapter.

Demographics and Clinical Manifestations

Spinal DAVFs represent 80% to 85% of spinal vascular shunts, generally presenting in the fourth or fifth decade of life with a heavy male predominance (5 : 1). This is in contrast to spinal AVMs, which often present prior to age 40. For DAVFs, no genetic or familial tendencies have been identified, which again points to these lesions being acquired rather than congenital.

The typical pattern of clinical symptoms and natural history was first described by Aminoff and Logue in 1974, and their functional status grading scheme ( Table 124-1 ) is still frequently used to describe these lesions. The clinical presentation is characterized by an often-lagging pattern of progressive motor and sensory symptoms. Fifty percent of untreated patients will be disabled 3 years after the diagnosis. The most common location of these shunts is in the midthoracic region, so symptoms are primarily localized to the lower extremities and bowel and bladder; however, symptoms can vary depending on the particular anatomy of the lesion. In one review of the literature, lower extremity weakness was present in 88% of patients at the time of diagnosis, followed by lower extremity paresthesias (78%), back pain, (39%), urinary and bowel symptoms (84%), and impotence (24%). Pain may be local, radicular, or nonspecific, further confusing this entity with more common degenerative disc disease or musculoskeletal pain syndromes. In one retrospective review, a misdiagnosis was found in 81% of patients, leading to mistreatment in nearly 20% of the original population.

Given the pathophysiology of these lesions, most patients experience gradual onset of symptoms with slowly progressive clinical deterioration. Rapid onset of symptoms and deterioration has also been described, and this is thought to be related to acute thrombosis of the venous plexus leading to venous infarction of the spinal cord. Spontaneous recovery is almost never encountered. The time between the onset of symptoms and diagnosis has been reported to be between 12 and 44 months, with the mean duration of symptoms before diagnosis at 22.9 months. Symptoms with DAVF may be more prominent in the upright position or with exercise, likely secondary to the increased venous hydrostatic pressure when standing. This is another clinical feature that may help to differentiate DAVF from spinal AVM.

In contrast to cranial dural fistulas, hemorrhage is extremely uncommon in the spinal variety, likely secondary to their low-flow nature. Cases of spinal subarachnoid hemorrhage should be thoroughly investigated with spinal angiography, and the index of suspicion should be high for a spinal AVM as opposed to fistula.

Pathophysiology

It is important to understand that symptoms from type I spinal DAVFs are related to venous hypertension in the spinal cord, as opposed to true spinal AVMs, which cause symptoms from hemorrhage or mass effect. The actual dural pathology of fistula or AVM is inconsequential. Of utmost importance, however, is that the venous outflow of the dural lesion is into the coronal venous plexus of the spinal cord. This leads to venous congestion of the plexus, stagnation of arterial flow through the spinal cord itself, decreased perfusion pressure, edema formation, and ischemia.

These fistulas are often thought of as actual dural AVMs. A structure between the two leaflets of dura resembling a glomerulus has been described, its role to maintain constant intraspinal venous pressure despite frequent changes in intra-abdominal and intrathoracic pressure. Although the exact mechanism of formation of the fistula is poorly understood, it is thought that the arterialization occurs in this region, making the glomerulus into a sort of small dural AVM. Most often, a single radiculomedullary artery enters the spine dorsolaterally at the dural nerve root sleeve ( Fig. 124-1 ). These fistulas can be supplied by multiple branches over several levels and are further subdivided based on their size (types A to C). The radiculomedullary arterial feeder to the fistula tends to be separate from the branch that normally penetrates the dura to anastomose with the anterior or posterior spinal arteries, but this is an obvious consideration for treatment. No valves are present within the radial veins or coronal venous plexus; therefore, the arterialized pressure is transmitted via these veins directly to the spinal cord parenchyma. Venous hypertension and the resulting venous ischemia and loss of autoregulation can result in the devastating necrotizing myelopathy described by Foix and Alajouanine.

A, Normal vascular anatomy of the spinal cord. The radiculomedullary artery enters the dura at the nerve root sleeve, anastomosing with the anterior and posterior spinal arteries. B, Spinal dural arteriovenous fistula. The nidus is located in the dura mater at the nerve root sleeve. This is usually supplied by a single branch of the radiculomedullary artery and drains into an enlarged intradural (medullary) vein running to the dorsal venous plexus.

(Copyright Cleveland Clinic Foundation. All rights reserved.)

Imaging

Although the first diagnoses of spinal AVM and fistulas were made with myelography, this modality has been replaced nearly completely by magnetic resonance imaging (MRI). MRI is already the modality in most widespread use for patients with myelopathy or lower extremity neurologic symptoms, and a high index of suspicion should prompt the physician to include imaging of the thoracic and cervical spine, as opposed to just the lumbar region as is necessary for most radiculopathy workups. Spinal cord edema, seen on T2-weighted images, has been found to be nearly 100% sensitive for spinal DAVF, but this finding is rather unspecific and can be caused by many other etiologies such as intramedullary tumor, demyelinating process, or trauma. Dilated intradural veins are a more specific finding; however, turbulent flow of spinal fluid on T2 weighted sequences may be confused for abnormal spinal vasculature. Further, some studies have shown that 50% of patients or more may not show any abnormal vasculature at all on conventional MRI sequences. Donghai and colleagues found T2 hyperintensity in the cord in 87.1% of patients and dilated perimedullary vessels in 77%, making these, by far the most common MRI finding ( Figs. 124-2 and 124-3 ). However, they did note that conventional MRI was normal in 6.4% of patients. MRA has also been shown to be useful in this population with a sensitivity and specificity of 91% and 78%, respectively, in one series. High-resolution, contrast enhanced magnetic resonance angiography (MRA) has also been shown to be highly reliable in showing the side and spinal level of the actual fistula and appears to be superior to time-resolved sequences. By providing improved localization, all of these techniques are at least useful in limiting the amount of contrast dye and radiation necessary during spinal angiography.

A, Sagittal T2-weighted MRI of the thoracic spine showing dilated venous channels dorsal to the spinal cord (single arrow) and increased T2 signal within the caudal spinal cord (double arrows). B, A-P digital subtraction angiography showing the radiculomedullary artery at T9 on the right entering into a fistula (arrow) and draining into the dorsal venous plexus of the spinal cord. C, Intraoperative picture showing the vein as it enters from the foramen and anastomoses with the arterialized dorsal venous plexus. D, Indocyanine green angiography showing the fistula, prior to ligation. E, Intraoperative photograph postligation of the entering vein. The dorsal venous plexus is no longer arterialized.

A, Sagittal T2-weighted MRI of the thoracic spine showing dilated venous channels dorsal to the spinal cord (double arrows) and a midthoracic syrinx. B, A-P digital subtraction angiogram shows a dural AVF fed from a branch of the right T7 radiculomedullary artery with drainage into the perimedullary venous plexus. C, A-P angiogram postembolization with onyx. The fistula is completely obliterated. D, Unsubtracted image shows the onyx case flowing into the draining vein. E, MRI 6 months posttreatment, the midthoracic syrinx is still present (arrow), but the patient’s symptoms did improve.

The gold standard imaging study for spinal DAVF by which all other modalities are compared is diagnostic spinal angiography (see Figs. 124-2 and 124-3 ). In general, if there is an abnormal vascular process affecting the spine, it will be found on a complete, high-quality, spinal angiogram. Spinal angiography has the disadvantages of being an invasive procedure with inherent risks, typically requires high doses of both contrast dye and radiation, and can be quite long and tedious for the inexperienced angiographer or with difficult anatomy. Nonetheless, spinal angiography is essential in these cases, both pre- and post-op, for diagnosis and confirmation of cure of the fistula. Angiography is also useful in identification of the artery of Adamkiewicz, which is rarely involved in the pathology of spinal fistulas (more so in AVMs) but can have catastrophic consequences if compromised during treatment. In general, the overall characteristics and location of the fistula can be investigated with MRI and computed tomography angiography (CTA). This then allows for a focused spinal angiogram covering a few levels above and below the lesion and minimizing risk to the patient.

If spinal angiography is negative for any dural fistula or other pathology, then the clinician should proceed to cerebral angiography. A “Cognard type V” fistula has been described as an intracranial dural fistula with spinal venous drainage. These fistulas generally occur in the region of the foramen magnum and feature primarily anterior spinal venous drainage, which may not be apparent on MRI. Though not common, these lesions are important entities to remember, especially in patients where there is a high index of suspicion but relatively unremarkable imaging workup.

Management

Once a diagnosis of spinal DAVF has been established, expeditious treatment via a multidisciplinary approach is best. Although the time from diagnosis to treatment has not shown a correlation with outcome in some studies, the goal of treatment is to at least halt the progression of symptoms. Thus, treatment should proceed as soon as possible so that the patient does not have the opportunity for further decline. We have found that treatment of these complex lesions is best with a multidisciplinary team of neurosurgeons, neuroendovascular specialists, and ultimately rehabilitation physicians. Regardless of the modality of treatment chosen, the goal is to disconnect the fistula from the medullary venous drainage, thereby obliterating the venous hypertension. Occlusion of only the arterial inflow with failure to obliterate the draining vein results in a high likelihood of fistula recanalization and recurrence.

Endovascular therapy involves placement of a microcatheter in the radiculomeningeal artery as close as possible to the fistula, followed by the injection of embolic material (see Fig. 124-3 ). Embolization is safer and more effective the closer one can get the microcatheter to the fistula. Endovascular therapy has the advantage of being less invasive, can be attempted at the time of diagnosis with angiography, and may lead to shorter hospital stay and accelerated postdischarge recovery. However, endovascular techniques have traditionally had a higher recurrence rate than surgery, and once the arterial side is embolized, endovascular access to the fistula and draining vein may no longer be possible, necessitating surgical therapy. Newer microcatheters and microwires have improved ability to navigate the tortuosity of the vessels and get direct access to the fistula. If endovascular embolization is the chosen treatment, extreme care must be exercised to identify the radiculomedullary artery of Adamkiewicz. If this branch is inadvertently embolized, anterior spinal artery infarction and paraplegia can easily result.

For many years, the standard surgical treatment for DAVFs involved stripping the enlarged venous plexus from the spinal cord. Surgical therapy aims at disconnection of the fistula by placing a clip on the draining medullary vein. The patient is placed in the prone position and, under general anesthesia, a laminectomy or hemilaminectomy is performed at the level where the fistula has been localized. After opening the dura, usually engorged posterior spinal veins are visualized. The proximal draining vein is then localized as it enters through the foramen and a clip is placed at this point so that it is as proximal as possible. Indocyanine angiography with the surgical microscope is useful to confirm that the fistula has been obliterated and in some cases may obviate the need to perform postoperative formal spinal angiography (see Fig. 124-2 ). Care should be taken for watertight dural closure to decrease the risk of postoperative cerebrospinal fluid leak.

One meta-analysis shows that surgery is clearly superior to endovascular therapy in terms of obliteration of the fistula; 98% of those treated with surgery were cured at the initial treatment compared to only 46% of patients treated with endovascular therapy. Unfortunately, the literature is heterogeneous in terms of endovascular therapy, and the overall results are biased by older studies using embolic agents (i.e., polyvinyl alcohol [PVA] particles), which are no longer used in the treatment of DAVFs. Newer embolic agents such as N-butyl cyanoacrylate (NBCA) and Onyx have had success rates reported from 25% to 89% if embolic material can be seen entering the draining vein. Therefore, the first attempt of treatment by endovascular means is unsuccessful nearly 50% of the time. The complication rate is slightly lower for surgery compared to embolization (1.9% versus 3.7%), but the complication rate for both procedures is relatively favorable. Despite the lesser chance of initial cure and slightly higher complication rate, the less invasive nature of embolization has led some centers to use this method as their first-line therapy, reserving surgery for recurrence or for cases that are not amenable to endovascular embolization. In our experience, the difference in recovery and length of stay between patients treated surgically and those treated endovascularly has not been significant. Therefore, we tend to use the approach where we are most confident of cure at the initial treatment. This leads to surgery in most patients; however, if the vascular anatomy is favorable and we are confident we can obliterate the draining vein, we will use endovascular embolization as first-line therapy.