Discussion

The development of dural AVF after sinus thrombosis, surgery, trauma and infection of adjacent air sinuses has been well documented,1–3 but developments after endovascular embolization of high-flow vascular malformations such as PAVFs have not been previously reported. The exact mechanism of development of such a dural AVF is unclear, and the theories are controversial.

To understand the formation of dural fistulas, a thorough knowledge of the anatomy of the structure and vasculature of the coverings of the brain is essential. The brain has a rich arterial and venous network on the surface but the capillaries themselves are not present in the pia arachnoid.4 The dura mater is thick, fibrous and inelastic with a complex vascular network, far in excess of the expected metabolic needs of a membrane furnishing only mechanical support. The main meningeal vessels are on the outer or the periosteal layer of the dura. They give rise to primary anastomotic arteries that are of uniform size and anastomose freely and frequently with each other and with major meningeal arteries. The primary anastomotic arteries give rise to four different arterial units: the arteries to the skull; the secondary anastomotic arteries lying in the outer layer; the arteriovenous shunts in the mid portion of the dura; and the penetrating vessels that traverse obliquely onto the juxta arachnoidal surface of the dura to form an extremely rich capillary network. The capillary network is present throughout the dura, including the falx and tentorium, and is extremely rich in the parasagittal region.5 The arachnoid is an avascular thin metabolically active covering which functions as a membrane to contain cerebrospinal fluid. Normally, the arachnoid is in apposition with the dura and there is no potential subdural space except in the pathological situations of subdural hematoma or hygroma. Electron microscopically, the junction consists of two parallel membranes approximately 200×10⁻⁸ cm apart, and the space is filled with dense granular material.6

Clinical experience and experimental studies in rats have shown that sinus thrombosis and venous or intrasinus hypertension are important factors related to the pathogenesis of dural AVF.2 ,7–9 Although either alone can produce a dural fistula, the combination leads to a higher chance. Endovascular embolization of high-flow PAVF involves a rapid reduction in the volume and velocity of blood flowing into the veins. The rapid reduction in blood flow causes relatively slow flow in the sinus. These hemodynamic changes in the sinus are not uniform, with some regions affected more than others possibly due to the septation and segmentation of the venous sinus channels. The slow flow induces thrombosis and secondarily localized hypertension to initiate the pathogenesis of the dural AVF.

Following embolization, the sinus thrombosis and localized hypertension necessary to initiate the event is usually subclinical. The slow flow and thrombosis in the venous channels may produce local ischemia of the endothelial cells and segmental venous hypertension. The hypoxia of the endothelial cells results in intracellular stabilization of hypoxia inducible factor (HIF1) which, in turn, induces vascular endothelial growth factor (VEGF) production.10–12 Increased shear stress due to localized hypertension promotes localized inflammation, thereby inducing production of VEGF and other angiogenic factors.13 The combination of the two effects results in a higher level of VEGF than either factor alone.14 In experimental dural AVFs, Uranishii et al found the VEGF staining to be stronger in the endothelium of the dural sinus, small arteries and veins in the sinus wall, along with basic fibroblast growth factor (bFGF) strongly staining in the subendothelial layer and hypertrophied media of the arteries in the sinus wall and in fibrous connective tissues around the sinus.15 Furthermore, the levels of VEGF are maximum 1 week following the induction of thrombosis and venous hypertension and decrease over the following 2 weeks.14 Similarly, Lawton et al have shown that, after inducing venous hypertension, the angiogenesis recorded was maximum at 1 week.16 As the dura is a structure with much more vascularity than its apparent needs, it readily yields to the angiogenic stimuli.

The angiogenic factors produce local angiogenesis either by sprouting or splitting of the pre-existing vascular structure.17 Sprouting angiogenesis is initiated by vascular leakage and local degradation of the basement membrane of the mother vessel followed by proteolytic degradation of the extracellular matrix. The leaking plasma proteins form a scaffold for the migrating endothelial cells. The initiation, cell division and migration are all orchestrated by the variable intracellular signaling of VEGF and other angiogenic factors such as fibroblast growth factor. Splitting angiogenesis occurs by intussusceptive growth resulting in multiplication of vessels by splitting their lumina through insertion of tissue pillars. Once formed, maturation takes place by recruiting mural cells consisting of pericytes and smooth muscle cells which subsequently lay the matrix and lamina,12 together with a gradual increase in flow in the vessel, and by recruiting more vessels proximally through the existing numerous artery to artery anastomosis on the dura. As most fistulas develop shortly after embolization, they take time to mature and become more obvious. The interval between embolization of the pial fistula and identification of a dural fistula ranged from 3 to 14 months in our patients, but this does not mean that the fistula did not occur earlier.

We frequently observe gross venous sinus remodeling after embolization of high-flow lesions and prefer to heparinize patients for 24–48 h to prevent acute total occlusion of the sinus. Although similar changes occur in all treated patients, only 25% of our patients developed a dural AVF. The post-embolization pressure changes are significant at the immediate draining veins and sinuses, and five out the 10 fistulas in our patients were located in this vicinity. However, the changes can occur anywhere along the sinus, and dural fistulas can occur infrequently at a distant location. At present we do not have any definite specificity to predict the group of patients who will develop a fistula in the future except pre-existing gross venous sinus occlusion, as was the case in two of our four patients (table 1).

Three different patterns of dural fistulae are possible by a similar mechanism: (1) dural fistula development into the dural venous sinus along the drainage of the fistula was the most common pattern and was seen in five of the 10 fistulas (50%); (2) dural fistula development locally into the thrombosed venous sac draining through the pre-existing venous drainage was seen in four of the 10 fistulas (40%) (figures 2 and 3) and was associated with local reactive angiogenesis from the pial vessels; and (3) dural fistula development locally into the thrombosed venous sac of the treated pial fistula and draining through a newly developed venous pathway was the least common pattern and was seen in one of the 10 fistulas (10%) (figure 1).

Of the 10 fistulas, five of them developed into the thrombosed venous sac. The thrombosed venous pouches are located in the subarachnoid space and are in almost direct contact with the dura baring the thin arachnoid. The angiogenic stimulus from the thrombosed veins induces growth of vessels from the vasculature-rich dura. These thrombosed venous sacs, in addition to the induction of dural fistulous communication, also induce reactive angiogenesis from surrounding parenchymal vessels possibly by the same mechanism.

The management of the de novo dural fistula follows the same principle as other dural fistulas. The basic principle is to occlude the shunting point; proximal occlusions will invariably result in recanalization as the dura is rich in vascular channels. The material used is based on individual choice and experience. Glue embolization is done with distal location of the catheter and with dilute glue to achieve maximum penetration. We used NBCA for four fistulas in two patients with s concentration ranging from 25% to 33%. Onyx 18 injections were used in one patient to occlude two feeders where multiple en passage dural feeders formed a fistula into the thrombosed venous pouch. In both locations we distally occluded the meningeal vessel using liquid coils and brought the microcatheter proximal to inject Onyx 18 by maximizing its flow into the fistula location. We successfully embolized one fistula from the posterior meningeal artery with absolute alcohol as we could not go distal and did not have a pedicle to allow reflux (figure 1).

None of the seven embolized fistulas recanalized, but one developed a new fistula proximal to the embolized location which was treated with radiosurgery as the same venous sac also had extensive reactive angiogenesis. At 3 years after radiosurgery the dural shunt was significantly reduced in size and did not warrant any further treatment (figure 2).

Not all fistulas need treatment; small fistulas with a minimal flow can safely be observed. If the fistula is enlarged during follow-up it becomes easier to place the catheter distally close to the fistula and embolize the fistulous location.