A dural arteriovenous fistula (DAVF) was first reported by Sachs in 1931 [1]. They were initially thought to develop from pre-existing dural arteriovenous microshunts by Kerber and Newton [2] or to develop in response to abnormal venous pressures by Brainin and Samec [3]. In the 1970s, it was realised that venous drainage influenced their behaviour, notably by Aminoff and Kendall [4] who differentiated fistulas draining to the cavernous sinus from those draining to the transverse/sigmoid sinuses and Castaigne et al. [5] and Djindjian et al. [6] who proposed classifications based on venous drainage patterns. The former divided lesions into three groups depending on whether venous drainage was to sinus or cortical veins, anticipating the seminal classification of Djindjian and Merland [7] published in 1973.

The current consensus is that they are acquired lesions, usually developing in response to thrombosis of a sinus or a lesion that causes abnormally high venous pressure. The arguments for a congenital aetiology are weak and based on the additional finding of aneurysms, brain arteriovenous malformations or other arteriovenous fistulas in some patients.

The evidence for their being acquired is more compelling. DAVFs have been reported in patients with:

A history of previous local trauma or hypercoagulability state was present in 66% of patients in a meta-analysis [11]. Trauma may rupture arteries adjacent to veins to create a fistula acutely or result in thrombosis of a sinus. It is assumed that sinus thrombosis may cause venous hypertension and local cerebral hypoperfusion which stimulates production of endothelial growth factors [e.g. vascular endothelial growth factor (VEGF) and hypoxia-inducible factor-1alpha]. The resulting angiogenesis leads to the development of arteriovenous shunts. An alternative, but not contradictory hypothesis, is that a change in arteriovenous pressure gradients opens arteriovenous microshunts between dural vessels. This latter process being triggered by raised arterial pressure, head injury, partition and sinus compression by tumour or surgical occlusion of a sinus, though these are also factors that might precipitate sinus thrombosis.

However, the relationship with sinus thrombosis is not clear, since sinus thrombosis does not always cause a DAVF and thrombosis may occur within them. Thus, lesions may evolve with changing degrees of venous thrombosis causing changes in venous drainage patterns, fluctuations in symptoms and signs and even spontaneous closure of fistulas [12]. Progressive restriction of venous outflow may cause retrograde cortical venous drainage and venous hypertension [13].

Age: Lesions have been reported in children but most present in adults [14]. In a literature review of 248 treated cases, Lucas et al. [11] found a mean age at presentation of 50.3 years (range 1–87 years).

Sex: There is a modest male predominance, which is most evident for higher-grade lesions (55% male in Lucas et al.’s review).

Incidence: It is difficult to establish an incidence for this diagnosis from the literature. Most authors quote an estimate of Newton and Conquist (made in 1969) that they represent 10–15% of all intracranial vascular malformations [14]. The incidence (detection rate) in the Scottish Intracranial Vascular Malformation Study (SIVMS) was 0.16/100,000 (95% CI 0.08–0.27) adults per annum and therefore somewhat lower at 7% of all intracranial vascular malformations in this cohort [15].

The dura is thickened with intense vascular proliferation within and around the sinus wall. A spongy mass of fibrous tissue is found inside the sinus, and primary and secondary arteriovenous shunts are seen on the venous side of the network. Stenosis or occlusion of sinuses is frequent but not always present.

Much has been written about the classification of DAVFs. This is not surprising since they compose a group of lesions with heterogeneous symptoms, signs and natural histories. One approach is to separate progressive from non-progressive lesions based on their symptoms and signs (which in turn is related to the location of the DAVF). Another is based on an analysis of the angioarchitecture (i.e. venous drainage pattern) and whether drainage involves cortical venous reflux (CVR). Both have been shown to correlate with prognosis and natural history (Figs. 10.1, 10.2, 10.3, 10.4 and 10.5).

The three important classifications based on angioarchitecture are presented in Tables 10.1, 10.2 and 10.3. The first generally accepted classification was by Djinjian and Merland [7]. It was enlarged by Cognard et al. [16] and, about the same time, a simplified system was proposed by Borden et al. [17]. Today, these last two systems are used almost equally frequently. The Borden system captures the important distinction between lesions with progressive or benign natural histories but lacks the detail of the French authors.

These classifications emphasise the importance of increased pressure in cortical veins to the development of symptoms or spontaneous haemorrhage. Lesions with CVR are higher grades with angiographic evidence of retrograde drainage into leptomeningeal (i.e. veins of the arachnoid and pia mater) and cortical veins. They are more likely to cause symptoms. Type I lesions, in which drainage is confined to an adjacent sinus without backflow or reflux, are unlikely to bleed or cause symptoms other than bruit or headache (Fig. 10.1). Type IIa lesions are similarly unlikely to cause serious symptoms but DAVF of the cavernous sinus is an exception because if drainage is directed anteriorly to orbital veins it may overwhelm their drainage capacity and cause sight-threatening symptoms due to venous hypertension, i.e. proptosis, ophthalmoplegia and raised intraocular pressure. An additional classification for fistulas involving the cavernous sinus was proposed by Barrow et al. in 1985 [19], but this is confusing because it mixes carotid–cavernous fistulas with DAVFs of the cavernous sinus. I suggest it is best avoided.

Baltsavias et al. [20] have recently proposed a classification based on a more detailed analysis of venous drainage patterns. They use three angiographic criteria to define eight subgroups of DAVFs with leptomeningeal venous drainage. These are: (a) direct or non-direct drainage depending involvement of leptomeningeal veins alone or in addition to the sinus, respectively; (b) exclusive or non-exclusive drainage depending on this being only by leptomeningeal veins or involving sinus, dural or emissary veins, respectively; and (c) cortical veins showing features of venous hypertension or not. The classification emphasises the role of bridging veins (between sinus and cortical veins) in the angioarchitecture of type III lesions [21]. Time will tell how useful it is in practice.

Location and arterialisation of cortical veins are the primary features determining the symptoms and signs caused by DAVFs. Both influence the behaviour and natural history of lesions. The approximate proportions of DAVF at intracranial locations and frequency of most common presenting symptom are presented together in Table 10.4.

Symptoms and signs due to DAVFs include:

Spontaneous haemorrhage causes parenchymal haematoma and/or subarachnoid haemorrhage and it may be secondary to venous infarction. The risk is 2% per year for all types of DAVFs but consistently reported to be higher in the presence of CVR and therefore in type 11b – V lesions [24]. In a meta-analysis, Kobayashi et al. calculated an odds ratio of presentation with haemorrhage of 23.2 (95%CI 13.8–39.0) in patients with CVR [25]. The risk of future haemorrhage is also increased for patients after haemorrhage [26]. A follow-up study of patients with CVR reported annual haemorrhage rates of 3.7% for all and 1.5% and 7.4% for patients without or after haemorrhage, respectively [27]. This increased haemorrhage risk also applies to patients with symptoms of venous hypertension secondary to CVR. Strom et al. [28] reported the incidence of haemorrhage on follow-up was 5.9% in patients with no or minimal symptoms and 18.2% in those with symptoms of CVR. The highest risk of rebleeding is in the acute period after haemorrhage; a 35% rate of rebleeding within 2 weeks of presentation with haemorrhage has been reported [29].

It is well recognised that symptoms due to DVAFs may alter over time [24]. This is assumed to be due to evolution to a higher or lower grade lesion. Spontaneous resolution is well recognised to occur sometimes. This has been attributed to thrombosis of the sinus but cases with spontaneous resolution and patent sinuses have been observed [30].

The severity of symptoms at presentation is a factor in progression and Zipfel et al. suggested that it should be a criterion for the classification of DAVFs [31]. It is difficult to separate the effect of haemorrhage from other causes of symptoms in some reports. However, it is clear that location is important in determining symptoms and the risk of progression. Awad et al. [32] reviewed 377 case reports in the literature and calculated a ratio of those with an aggressive neurological course (defined as heamorrhage or progressive focal neurological deficit other than ophthalmoplegia) to those with benign serious symptoms (100 cases) according to location. They concluded that no location was immune from progression but this was least often seen in DAVFs involving the transverse-sigmoid sinuses and cavernous sinus and most often seen at the tentorial incisura and anterior fossa locations (see Table 10.5). Angiographic features associated with progression were CVR, venous aneurysm or varices, and Galenic drainage. Unlike other fistulas, the rate of shunting did not correlate with progression [32].

The diagnostic pathway depends on the patient’s clinical presentation and the likely location of the DAVF.

Though treatment should be considered for all lesions, spontaneous closure has been reported in about 10% of DAVFs. In a follow up study of DAVFs of the cavernous region, spontaneous complete regression of symptoms occurred in 19 of 26 patients followed but fistula closure was only confirmed by angiography in 4 patients [34]. Thus for type I lesions, intervention is only indicated if symptoms are distressing for the patient. Although evolution between types (due to opening or closing of the primary drainage) has been reported, it is unusual, and if it occurs, it is usually accompanied by an alteration in symptoms. Therefore, if symptoms are stable and tolerated by the patient, there is no indication for intervention but they should be monitored. For most type II lesions, intervention is usually recommended, and for type III and IV lesions, urgent surgical or endovascular treatment is indicated.