22 Endovascular Treatment of Dural Arteriovenous Fistulas of the Brain
The endovascular treatment of dural arteriovenous fistulas (DAVFs) of the brain has greatly improved with the advent of precipitating liquid embolics based on the solvent dimethyl sulfoxide (DMSO). The success of the procedure is directly related to the occlusion of the (first segment of the) draining vein; a thorough understanding of the anatomical details is therefore mandatory. The treatment techniques may differ depending on whether the draining vein is represented by a dural sinus or a pial vein, and different anatomical features will suggest an arterial or venous approach, knowing that catheterization of cerebral veins needs a somewhat different technical approach compared to arterial catheterization. The introduction and development of new materials (liquid embolics, detachable tip microcatheters, performing guidewires, large and long compliant balloons for sinus temporary occlusion) allow today a very high rate of cure with a limited procedural risk. Most frequently the relatively rare periprocedural complications are hemorrhagic, related to venous thrombosis or to the possible rupture of small pial feeders. Surgery remains an excellent therapeutic alternative especially in DAVFs draining in a pial vein.
Keywords: dural arteriovenous fistula, embolization, dimethyl sulfoxide, sinus occlusion, sinus coiling, venous catheterization
- For practical purposes and for better understanding the different treatment approaches, dural arteriovenous fistulas (DAVFs) are classified in two main groups: (1) the ones draining first in a sinus and (2) the ones draining directly in a pial vein.
- Brain vessels (arteries or veins) as everywhere else in the body constitute one single giant network, where everything is potentially connected with anything else.
- Catheterization of the veins is quite more challenging than catheterization of arteries. It requires a learning of the complexity of venous anatomy and a specific training.
- The pretreatment analysis of the DAVF is aimed at finding the exact point of fistula, that is the very first segment of the draining vein: “the foot of the vein,” which will represent the target of the treatment.
- The advent of liquid embolic materials, based on the solvent dimethyl sulfoxide, has greatly improved the success rate of the endovascular approach to DAVFs.
Over the years, a few hundred dural arteriovenous fistulas (DAVFs) have been diagnosed, treated, and followed by our multi-disciplinary team consisting of neurologists, neurointensivists, neurosurgeons, and neuroradiologists at the authors’ institution. The material for this chapter comes from direct experience of the authors, treating DAVFs of the brain in the neuroradiology department of the Niguarda Hospital in Milan for the last 30 years.
A few introductory considerations are required to understand our philosophy in the endovascular treatment (EVT) of DAVFs:
Over the past 40 years, several classifications have been proposed for DAVFs of the brain. Djindjian and Merland,1 Borden et al,2 Cognard et al,3 and more recently Geibprasert et al4 have all contributed with important works. However, we will not follow any specific classification because we feel that in the context of treatment each one has limitations. For the purpose of this chapter, we will divide intracranial DAVFs in two groups: DAVFs located directly on a vein and those located on a sinus.(► Fig. 22.1)
- In most DAVFs, the bulk of the arterial flow is coming from the meningeal arteries, but it is not unusual to find feeders coming from the pial brain network (► Fig. 22.2) both on sinuses and on pial veins. The more you look for the more you find. It is usually inconsequential, but it may become an important issue in a few cases. At times, the small shunts are located on the vein a little more distal than the dura emergence of the vein. We have seen cases where a surgical disconnection of the vein achieved complete elimination of the meningeal shunts, but a little pial shunt on the brain surface was maintained. Such pial feeders may have been responsible for delayed hemorrhages after treatment (sinus occlusion) of a DAVF, much like rupture of residual nidus of a brain arteriovenous malformation (AVM) after the occlusion of the draining vein (see below). It is important to analyze the presence of possible pial feeders in DAVFs.
- Finally, a lesson learned treating DAVFs and AVMs is that the brain and head arterial vessels constitute an immense network very much connected at all levels. It is not only the well-known communicating arteries or the more discrete leptomeningeal collaterals that are able to connect different arterial territories: these connections are almost everywhere and given time and flow demand any link is possible: superficial brain arteries with deep ones, meningeal arteries with brain vessels, and vice versa. Not only the bone, the diploe, but also the muscles and skin may take part in the process. These are not new vessels that form in response to angiogenetic stimuli, but rather dilatations of preexisting very small channels, which increase in size only because of flow demand.
Fig. 22.1 (a) Dural arteriovenous fistula (DAVF) on a pial vein (Labbé vein): external carotid artery injection. The drainage is retrograde in veins reaching the superior sagittal sinus. (b) DAVF of the sigmoid sinus: common carotid artery injection. The drainage is orthograde toward the jugular vein.
Fig. 22.2 (a) Dural arteriovenous fistula (DAVF) at the falcotentorial junction: right internal carotid artery injection. Feeders coming from tentorial branches of the meningohypophyseal trunk. (b) Right external carotid artery injection: feeders from the middle meningeal artery. (c) Feeders coming from the posterior meningeal artery. (d) Pial feeders from the transmesencephalic and retromesencephalic arteries (arrows).
The venous system has obviously the same properties, even though it is usually not quite easy to understand, or possibly the whole venous system is not well understood and these patterns may be there but not easy to recognize.
22.2 Endovascular Treatment of Intracranial DAVFs
As with all arteriovenous fistulas, EV cure is obtained only with occlusion of the proximal segment of the draining vein (the “foot” of the vein), where the arteriovenous shunts (A-V shunts) are actually located. This vein may be (1) a pial vein (running in the subarachnoid space on the surface of the cortex or otherwise) or (2) a sinus. There are profound clinical and methodological differences between the two forms. From an EV approach, there are also important technical issues that separate these two types. In general, it is better to avoid sacrifice of a sinus, except in selected cases, which will be discussed later. On the other hand, it is usually safe to occlude the foot of a pial vein (► Fig. 22.3).
Fig. 22.3 (a-c) The “foot” of the vein (arrows) at progressive enlargements.
22.2.1 Occlusion of the Pial Vein
Endovascularly the occlusion of the foot of the pial vein may be achieved in two ways: accessing it through the venous route or filling it through an arterial approach. The common use of the venous approach dates back to the early 1990s, when the detachable coils became available. It was preferred to the arterial approach because it offered high rates of cure, compared to the frequent complete or partial failures with intra-arterial injections of particles or glue.
The Venous Approach
It is always essential to have a complete and thorough study of the angiographic anatomy, but this is especially true in case one decides to reach the DAVF navigating the veins. First of all, the location and morphology of the foot of the vein, which is the procedure’s target point, has to be well understood. Then one has to define the possible venous routes to reach it. It may be very hard to find an accessible route, because of the very complex venous network on the surface of the brain and because of the change in morphology of the draining veins, due to the increased amount and speed of arterialized flow. It is not infrequent to see tortuous dilated veins, with local varicosities, which may prevent any useful catheterization.
Catheter navigation in the venous tree has multiple technical differences with arterial navigation. First and most important is the fragility of the venous compared to the arterial structure. It is commonly held to be true that veins are easily occluded, dissected, or even ruptured when forceful pressures are utilized. While it is a risk that must always be kept in mind, it is remarkable how at times cortical veins may well tolerate very aggressive maneuvers.
Secondly one very important aspect of venous catheterization is that the road map is much less intelligible: that is, it is difficult to have all the veins completely and fully injected with contrast medium. If one injects from the arteries, most veins will be only faintly opacified and some will be all but invisible. If one injects from the vein, the vessels distal to the catheter (the ones to navigate forwardly) will not be seen: only those in the direction one is coming from will be shown. This is a major difference with arterial catheterization where the use of an angiographic “roadmap” is always very useful and well depicts the way to go. It is therefore better to keep in mind that there are always many more veins than one has actually seen.
One third point is the variability of venous branching, where there are a lot more collaterals, connections, turnarounds, creating a multipotential network much more developed than on the arterial side. When navigating the veins, it is then easy to miss the main road and to catheterize secondary vessels (not known, not visible, not readily recognized), losing time and confidence.
Finally veins offer much less support to navigation of catheters and guide wires. Veins are more easily deformed and displaced, and therefore cannot counter the push of our devices. But more importantly their size is potentially much bigger than thought, that is, once inside a vein, one understands that the contrast injection has shown only a minimal part of the real caliber of the vessel. A vessel that looked a couple of millimeters in size may accept a wire loop three or four times that, or more. The precise position and direction of wires and catheters become therefore much more difficult.
All these negative features become evident when trying to catheterize a cortical vein backward from a sinus: the large size of the sinus offers no support, the outlet of the vein is often more complex than thought, and there are many unperceived venous branches in the proximity, both on the sinus wall and on the first segment of the vein. The problem has recently been partly solved by the introduction of longer intermediate catheters that provide the support veins do not give.
Once we have reached the correct position with a microcatheter—the correct position being the very initial part of the draining vein, the “foot”—we should be clear in our mind about the target of our treatment. The goal should be to fill and occlude the foot of the vein for a limited segment, not too long in order to avoid the occlusion of branching veins, possibly causing consequences on the normal drainage of the brain (► Fig. 22.4). Occluding a segment of 1 to 2 cm is usually more than enough.
Fig. 22.4 (a,b) Lateral and oblique view: dural arteriovenous fistula of the crista galli. (c) Cure post-transvenous treatment with coils. (d,e) Microcatheter in the frontal vein. (f) Coils in the foot of the vein.
Today we have ample choice of occluding devices or materials. The most popular are the detachable coils, normally used for aneurysms. They are very easy to use, reliable, well controllable, and well known. One should try to obtain a very dense tangle in a short segment of vein. There might be times though when the progressive coiling will displace the catheter more proximally, away from the correct location, with suboptimal results. In order to avoid that, one should maintain a continuous pressure on the catheter (not easy, because of the scarce support offered by the vein, as we discussed earlier) or place preventively a second catheter distal to the first that will stay in position during the coiling. This double catheter technique is quite interesting for the occlusion of major sinuses (see below), but it could be very hard to realize in small cortical veins, when navigation is usually very challenging.
Liquid materials, cyanoacrylates or dimethyl sulfoxide (DMSO) based (*), may also be used to occlude vessels. They are mostly used in the arterial compartment, but the injection at the venous side is becoming more and more popular. The disadvantage of the venous injection is of course that the material, behaving as contrast medium would, will run away from the fistula site, rather than into it, resulting in the occlusion of more veins than desired. It takes therefore quite a lot of expertise to obtain the correct placement of the embolic material. At times one can associate the liquids with coils: a few coils are deployed in the desired position forming a nest where the liquid will be subsequently trapped, completing the occlusion. The use of “gluing” liquids has one more potential danger, that is, gluing of the microcatheter in place. Retrieval could become impossible or at least very dangerous with the risk of tearing numerous different veins along the way. With the advent of the recent detachable tip microcatheters, this occurrence is less of a problem, but it is still a possibility. In case it gets stuck, the microcatheter could be left in place. For that reason, some would prefer to access the intracranial veins from the jugular vein approach, rather than from the femoral vein, which would leave the microcatheter in the inferior vena cava, with the possibility of it curling up into the right atrium. Other occluding devices like detachable balloons or plugs have not been used in pial veins to our knowledge, the major difficulty being to navigate such devices against the direction of flow. The procedure ends when one has the evidence that the DAVF is completely obliterated. It does not suffice not seeing it any more when injecting contrast medium through the guiding catheter. It is almost always necessary to inject also all the other arteries, on both the same and the other side, both the anterior and the posterior circulation, both the meningeal and the pial arteries. More important is to well assure of the presence of a dense plug in the foot of the draining vein.
*DMSO-based liquid materials (DMSO-LM) are chemical solutions where the solvent is DMSO, which rapidly evaporates in contact with blood. The remaining solute will precipitate and solidify, behaving similarly to lava: it forms a crust around a liquid core, which under pressure will be later able to break the crust and flow in a new direction. The commercially available products are, in order of appearance on the market, Onyx (EV3, now Medtronic), Squid (Balt), and Phil (Microvention). In Onyx and Squid, the precipitating agent is ethylene vinyl alcohol (EVOH), while in Phil the agent is hydroxyl ethyl methacrylate (HEMA).
The Arterial Approach
The treatment of DAVFs through the arterial route is probably the approach preferred by most operators, because it is usually safe and most successful. History tells us that it has not always been as successful as it is today. In the 1970s, 1980s, and 1990s, neurointerventionists could use particles (dura mater and polyvinyl alcohol) and glue (Histoacryl). Both proved to be ineffective for the same reason: they could not occlude the foot of the vein. Either they remained on the arterial side, trapped in the arterial network proximal to the A-V shunt, or they would pass into the vein, but then fly away with the high blood flow. In both cases, occasionally the cure could be randomly obtained by occluding most of the main arterial feeders, causing a major drop of the flow through the fistula and a spontaneous thrombosis of the vein. With glue, it was also possible that if some drops would remain lodged in the first segment of the draining vein, the inflammatory response of the vessel would go on to cause its complete occlusion. With the turn of the century, the advent of DMSO-LM allowed a different and more efficient approach. It is now possible to obtain a high rate of successful occlusions of DAFVs, whenever the catheter is located sufficiently close to the A-V shunt. The injection of the material obtains the progressive filling of most of the arterial network and of the foot of the vein (► Fig. 22.5). The great advantage compared to cyanoacrylates is the fact that today’s liquid embolic material does not fly away with flow but stays in the foot of the vein and it can be slowly and continuously injected so that it accumulates there. Most of the DAVFs draining in pial veins can today be cured in such a way (► Fig. 22.6).
Fig. 22.5 (a) Dural arteriovenous fistula (DAVF) draining in a pial vein. (b) Superselective injection through a microcatheter. (c) Initial filling of the arterial network and part of the vein. (d) Final injection of the foot of the vein (arrow). (e) Complete DAVF occlusion.