Classification Schemes of Cranial Dural Arteriovenous Fistulas

The clinical presentation of dural arteriovenous fistulas (DAVFs), in particular the associated risk of intracranial hemorrhage, shows a strong correlation with their pattern of venous drainage. The two most commonly used and clinically accepted DAVF classifications are the Merland-Cognard classification and the Borden classification, both based on the morphology of the venous drainage. A revised classification that grades DAVFs through a combination of angiographic and clinical features has also been proposed. This article offers a review of these various classification schemes, and discusses their application to treatment decision making.

Dural arteriovenous fistulas (DAVFs) are abnormal arteriovenous communications developed within a venous space contained between the two layers of the dura mater and typically vascularized by meningeal arteries. Rizzoli has been credited with the first anatomic description of a DAVF in 1881, although the lesion seems to have been a scalp arteriovenous fistula, draining intracranially through a torcular emissary vein. Sachs is widely quoted as having offered the first angiographic description of a DAVF in 1931; the origin of this claim is unclear, as angiography is neither mentioned nor illustrated in his monograph on brain tumors. A brief discussion of angiography as an ancillary test and an example of cerebral aneurysm were added to a second edition in 1949. However, one of Sachs’ patients may be the earliest reported case of a DAVF, likely a high-grade lesion in which branches of the middle meningeal artery were “found to connect with the vessels of the cortex” ( Fig. 1 ). To the authors’ knowledge, the first two angiographic demonstrations of DAVFs were published in 1936 by Bergstrand and colleagues. Of particular interest is that one of their cases was also a high-grade DAVF with significant (and possibly exclusive) cortical drainage documented both by angiography and during surgical exposure ( Fig. 2 ). These early observations of DAVFs, made before the era of noninvasive neuroimaging and involving cortical venous drainage, anticipate the link between pial reflux and clinical conspicuity that lies at the base of the DAVF classifications discussed in this article.

DAVFs are said to represent approximately 10% to 15% of all intracranial vascular malformations, most commonly involving the transverse, sigmoid, and cavernous sinuses. While the etiology of DAVFs remains controversial, it is thought that they represent acquired conditions developing in a context of dural sinus thrombosis or venous hypertension. Other associations have been documented, including trauma, prior neurosurgical procedures (eg, craniotomy), and ear infection. The clinical impact of DAVFs ranges from a benign course, including incidentally discovered, asymptomatic lesions, to malignant presentations resulting in profound cognitive impairment, intracranial hemorrhage, and death.

The natural history of DAVFs strongly correlates with their pattern of venous drainage, in particular the presence of reflux into pial veins. The venous anatomy of the lesion is therefore a central component of most DAVF classification schemes, starting with the first system proposed by Djindjian and colleagues in 1978. Since then, the link between cortical venous drainage and aggressive symptomatology has been confirmed, leading to continuous refinements in our understanding of DAVFs and their classification schemes. In particular, the mode of presentation of a DAVF has recently been shown to influence the risk of subsequent significant clinical event. This new information has led to the proposal of a revised classification that grades DAVFs according to a combination of angiographic and clinical features. This article offers a review of the various classification schemes currently in use, and discusses their application to treatment decision making.

Classification schemes of cranial dural arteriovenous fistulas

The Merland-Cognard Classification

Cognard and colleagues proposed a modification of the original Djindjian-Merland classification scheme based on the clinical presentation of 205 consecutive patients seen over an 18-year period. The Merland-Cognard classification divides DAVFs in 5 types (I–V), with Type II lesions further subdivided into Type IIa, IIb, and IIa+b ( Fig. 3 ). Type I and IIa DAVFs drain directly into dural venous sinuses, without pial reflux; in Type I, the venous drainage is antegrade whereas in Type IIa, it is partially or completely retrograde. Retrograde drainage is generally related to outflow impairment, that is, a dural sinus stenosis or occlusion, but can at times also result from high-volume arteriovenous shunting that exceeds the drainage capacity of a normal or even enlarged dural sinus. Type IIb and IIa+b DAVFs have dural and pial venous drainage; in Type IIb, retrograde drainage occurs into pial vein(s) only, whereas in Type IIa+b both dural and pial retrograde drainage is seen. Type III and IV DAVFs drain only into the pial venous system, either directly or via an isolated dural sinus segment. Note that the pial veins constituting the primary drainage of the DAVF can themselves drain into a dural sinus not directly connected to the lesion. In Type IV, the draining vein is enlarged (venous ectasia). Cognard and colleagues define venous ectasia as a segment larger than 5 mm in diameter or 3 times larger than the diameter of the draining vein. Finally, Type V DAVFs are cranial lesions draining into spinal perimedullary veins, potentially resulting in cervical myelopathy.

The strength of the Merland-Cognard classification lies in the correlation established between the angiographic characteristics of the DAVFs and their clinical presentation, which offers a management tool based on an objective assessment of the risk of adverse clinical event associated with a specific lesion. In their series, 83 of 84 patients with a Type I DAVF presented with nonaggressive symptoms, which included headache, bruit, minimal vertigo, and ocular symptoms not related to intracranial hypertension. Aggressive symptoms, such as intracranial hemorrhage, intracranial hypertension, focal neurologic deficits, and seizures, were found more frequently with higher grades, that is, in 37% of patients with Type IIa, 30% with Type IIb, 66% with Type IIa+b, 76% with Type III, and 97% with Type IV. All 12 patients with a Type V DAVF showed aggressive symptoms. Six of them had progressive myelopathy due to spinal cord venous hypertension. The perimedullary venous drainage extended down to the thoracic or lumbar levels in patients with myelopathy, whereas it was limited to the cervical region in those without myelopathy. Type V DAVFs without myelopathy were associated with subarachnoid hemorrhage in 5 cases and focal neurologic deficit in 1 case. DAVFs without cortical drainage (Type I or IIa) did not present with hemorrhage. The role of pial reflux and venous flow dynamics was further emphasized by the progressive increase in the hemorrhagic risk associated with less favorable venous configurations: 11% in Type IIb and IIa+b (combined pial and dural drainage), 40% in Type III (exclusive pial drainage without ectasia), and 65.5% in Type IV (exclusive pial drainage with ectasia).

The Borden Classification

The Borden classification scheme also takes inspiration from the morphologic characteristics of the venous drainage. It recognizes 3 principal types of DAVFs (I, II, III) and 2 subtypes (a, b) (see Fig. 3 ). Borden Type I DAVFs drain into dural venous sinuses without pial reflux; they typically have a benign course. Type II DAVFs drain into venous sinuses, but show retrograde flow into a pial vein, with an increased risk of neurologic events from venous hypertension or hemorrhage. Type III DAVFs drain directly into a pial vein (or an isolated dural sinus segment), are associated with significant venous hypertension, and typically present with hemorrhage or severe neurologic symptoms. Each type is further characterized by the complexity of the architecture of the arteriovenous shunt into Subtype A, a simple fistula with a direct, single connection between the feeding artery and the draining vein or sinus, and Subtype B, a complex lesion with multiple fistulous connections ( Fig. 4 ).