Article Text

Download PDFPDF

Transarterial embolization with ONYX for treatment of intracranial non-cavernous dural arteriovenous fistula with or without cortical venous reflux
  1. Katrien De Keukeleire1,
  2. Peter Vanlangenhove1,
  3. Jean-Pierre Kalala Okito2,
  4. Giorgio Hallaert2,
  5. Dirk Van Roost2,
  6. Luc Defreyne1
  1. 1Department of Interventional Radiology, Ghent University Hospital, Ghent, Belgium
  2. 2Department of Neurosurgery, Ghent University Hospital, Ghent, Belgium
  1. Correspondence to Dr K De Keukeleire, Department of Interventional Radiology, Ghent University Hospital, De Pintelaan 185, Ghent B-9000, Belgium; katriendekeukeleire{at}


Background and purpose To report our experience with transarterial ONYX embolization of intracranial non-cavernous dural arteriovenous fistulas (DAVFs) with or without cortical venous reflux.

Materials and methods Retrospective analysis of transarterial ONYX embolization in 20 patients with 21 DAVFs, presenting with intracranial hemorrhage (n=7), pulsatile bruit (n=7), vertigo (n=3), non-pulsatile bruit (n=1), headache (n=1) and epilepsy (n=1). Risk grading of DAVFs was Borden type I (n=6), type II (n=4) and type III (n=11).

Results 18 of 21 (85.7%) DAVFs were angiographically occluded immediately after embolization, with ONYX embolization only, in either one (n=12) or two sessions (n=2); with a combination of ONYX and glue or transvenous coiling in a single session (n=2) or in two sessions (n=1); or after previous transvenous coiling/glue embolization (n=1). At the 6 (4–14) month control digital subtraction angiography (DSA), available in 14 of 18 occluded DAVFs, one patient showed a small residual fistula (17/21 or 81% occluded). Mid-term DSA was not available because of early death (n=2) or patients were awaiting the examination (n=2). In three cases, treatment was incomplete. Of six Borden type I DAVFs, four were cured and two partially occluded with resolution of symptoms. In two DAVFs, neurosurgical access to the feeding artery allowed distal microcatheterization and successful embolization.

Conclusion Transarterial ONYX embolization offers an effective and safe treatment for all non-cavernous DAVFs, whether with or without cortical venous reflux.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Dural arteriovenous fistulas (DAVFs) represent 10–15% of all intracranial arteriovenous malformations.1 Although DAVFs may be found in any dural structures, they occur most frequently in the region of the transverse, sigmoid and cavernous sinuses.2 Patients may be asymptomatic or may experience symptoms ranging from non-aggressive (eg, headache, pulsatile bruit) to aggressive (eg, hemorrhage, seizures, neurologic deficit). Fistula behavior is determined by its venous drainage. Borden et al described three types of DAVFs.3 Borden type I DAVFs are supplied by meningeal arteries and drain anterograde into a meningeal vein or dural venous sinus. Type I DAVFs are often asymptomatic or present with pulsatile tinnitus or headache. They have a low bleeding risk and therefore intervention is only indicated if symptoms are intolerable. In type II DAVFs, the sinus shows outflow obstruction, causing retrograde flow into cortical veins and venous hypertension. Cortical venous reflux (CVR) is associated with neurological dysfunction and hemorrhage. Type III DAVFs show only drainage into cortical veins ensuing an even higher bleeding risk. Both Borden type II and III DAVFs are ‘aggressive’, requiring complete treatment.

In the past, DAVFs have been treated with a variety of approaches, such as conservative treatment,4 stereotactic radiosurgery,5 transcatheter embolization,6–9 open surgery10 11 or a combination of these treatments.8

Recent developments allow many patients to be cured with transcatheter embolization. In transvenous coil embolization, the purpose is curative, aiming at venous outflow occlusion.12 This technique is a safe approach, with complete occlusion in 80–100% of cases.6 However, serious complications such as intracranial hemorrhage associated with vessel injury and incomplete embolization have been reported.7 Incomplete embolization may lead to worsening of symptoms. Transarterial embolization (TAE) with n-butyl cyanoacrylate (n-BCA) has been used for complex DAVFs inaccessible to a transvenous approach. Although results are relatively good,13 TAE with n-BCA requires extensive experience.8 ONYX (ev3 Neurovascular, Irvine, California, USA) is a liquid embolic agent whose lower viscosity and delayed precipitation makes deep fistula penetration possible. ONYX can be delivered with better control than n-BCA without the necessity of rapid withdrawal of the microcatheter. Although reports describing the use of ONYX for TAE have been very promising in complex DAVFs (Borden types II–III),14–26 it is unclear whether ONYX is effective and safe in benign type I DAVFs. In this paper, we report our technical, angiographic and clinical results of transarterial ONYX embolization for intracranial non-cavernous DAVFs with and without CVR.

Materials and methods

Patient population

From July 2005 to June 2010, we performed TAE with ONYX in 20 patients (10 women and 10 men; mean age 57.2 years, range 15–83 years) presenting with 21 non-cavernous intracranial DAVFs. In seven of 20 patients (35%), clinical presentation was intracranial hemorrhage. Seven patients (35%) presented with pulsatile bruit, three (15%) with vertigo, one (5%) with non-pulsatile bruit, one (5%) with headache and one (5%) with epilepsy.

DAVF characteristics

Risk grading of DAVFs was Borden type I (n=6) (figure 1), type II (n=4) and type III (n=11) (table 1). In 17 DAVFs the blood supply originated from the middle meningeal artery (MMA). Additional blood supply came from branches of the occipital artery (OA) (n=18), ascending pharyngeal artery (n=8), posterior auricular artery (n=6), tentorial arteries from the carotid siphon (n=6)/vertebral artery (VA) (n=5), posterior meningeal artery (PMA) (n=4), superficial temporal artery (n=2), posterior choroideal artery (PChA) (n=1) and aberrantly from native cerebral vessels (n=7). Mean number of arterial feeders was 4.5 (range 2–16).

Figure 1

(A) Right external carotid digital subtraction angiography (DSA) in lateral view shows a dural arteriovenous fistula (DAVF) type I with blood supply from the middle meningeal artery (arrow), petrosal branch of the middle meningeal artery (arrowhead), occipital artery (white arrow) and posterior auricular artery (white arrowhead); and venous drainage to the sigmoid sinus (asterisk). (B, C) Control right common carotid DSA (anteroposterior (AP) view) (B) at 18 months, 12 months after the second embolization session, shows complete occlusion of the DAVF, with an open sigmoid sinus on the corresponding venous phase (AP view) (C).

Table 1

Type of venous drainage, procedural data, and early and long term angiographic and clinical outcome of the 20 patients (21 dural arteriovenous fistulas)



All patients were treated under general anesthesia. Unilateral or bilateral femoral access involving the use of 5F–6F introducing sheaths permitted placement of the guiding and control catheters, positioned in the external carotid or vertebral arteries. A DMSO compatible microcatheter/microguidewire combination was placed in an arterial feeder close by to the fistula, if possible in the wedge position. Saline was used to flush the microcatheter and 0.25 ml of DMSO to fill the microcatheter's ‘dead space’. Afterwards, ONYX was injected using the subtraction roadmapping technique and fluoroscopic guidance, and continued until filling of the venous side of the DAVF or arterial reflux was observed. In the case of arterial reflux, a waiting time of at least 1 min was respected before ONYX was injected again. The length of reflux tolerated was between 1.5 and 3 cm, depending on the vascular territory.

If ONYX penetrated the network of arterial feeders, then the injection was continued until we suspected filling of the venous side of the DAVF. In the case of shunt penetration, we aborted the injection for 30 s to allow solidification of ONYX in an attempt to change the direction of penetration in the network of interconnecting feeders. This technique was particularly applied in type I DAVFs where we wanted to preserve sinus patency (figure 1). The procedure was terminated when reflux recurred and persisted and no ONYX was entering the fistula feeding network or when angiography showed complete fistula occlusion. Microcatheters were then pulled gently until they detached from their ONYX cast. Ultraflow microcatheters (ev3) could be stretched until rupture in their distal flexible end, if they did not move. Marathon microcatheters (ev3) are braided and stiffer, allowing much more force for retrieval before rupture could occur.


Patients were heparinized over 24–48 h and eventually kept on low molecular weight heparins for 1–2 weeks when we observed slowing down of territorial venous outflow or contrast stasis in various cortical veins (n=10). Anticoagulation was indicated in type I DAVFs to maintain the patency of the sinus.

Technical and clinical success

Endovascular treatment was evaluated immediately by digital subtraction angiography (DSA) and subsequently by clinical improvement of symptoms. Long term clinical outcome was combined with a 6 month control DSA.


Microcatheterization and embolic agents

In all patients, microcatheterization and embolization with ONYX was initially performed by a transfemoral approach. However, in two patients, direct surgical access to the main feeding artery was requested because we assessed the level of distal microcatheterization inadequate for successful embolization. In the first patient, the MMA was opened via neurosurgical trepanation to allow direct cannulation and catheterization.27 In the other patient, the OA was transected neurosurgically to perform direct cannulation and transosseus microcatheterization of dural feeders. In two patients, additional transvenous coil embolization was attempted in the same session because TAE via the MMA was incomplete and other transarterial routes assessed were too delicate and risky (feeders branching from pial arteries or with potential blood supply to cranial nerves). In a patient with a DAVF at the foramen magnum, additional glue embolization during protective balloon occlusion of the VA8 in the same session led to near complete occlusion. The control angiogram the next day confirmed complete occlusion. In a patient with a type I sigmoid sinus DAVF, a minimal residual shunt was occluded in a second session by combined injection of ONYX and glue.

Embolization was performed with Ultraflow microcatheters in 18 (90.0%) patients and later with Marathon's in two (10.0%) patients. We used ONYX 18 in 16 (76.2%) DAVFs, ONYX 20 in eight (38.0%) and ONYX 34 in six (29.0%) DAVFs. The volume of ONYX administered per DAVF ranged from 0.4 to 18.3 ml (mean 3.59 ml) at an injection duration ranging from 12 to 208 min (mean 56.5 min). Thirty-six feeders in 21 DAVFs were embolized with ONYX. In the DAVFs occluded with ONYX embolization only, 21 feeders were used in 14 DAVFs, or 1.5 feeder/fistula (table 1).


Eighteen of 21 (85.7%) DAVFs were considered as occluded at the final control angiogram after embolization (table 1). Long term DSA follow-up however, available in 14 patients, confirmed complete occlusion in 13 (92%) patients (table 1). In one patient with a large type II DAVF of the confluence of sinuses, the control DSA after 6 months showed a very small residual fistula. In two patients (with three DAVFs, two of them occluded), long term follow-up was not available due to early death. Two other patients were awaiting control DSA to confirm complete obliteration while writing the manuscript. If we assume that in these four patients long term DSA would not be different from the post-embolization result, then 17 (81%) of 21 DAVFs were occluded. In 13 of 17 DAVFs, the occlusion was obtained with ONYX embolization only, in either a single (n=12) or two sessions (n=1).

In four of 21 (19.0%) fistulas with an incomplete obliteration, symptoms improved significantly. One patient was treated palliatively because access to the type I fistula was impaired by previous surgical ligation of feeders. In two of four incompletely cured DAVFs, additional treatment (radiosurgery n=1, embolization n=1) is scheduled. One patient (with one DAVF occluded and one partially occluded) died before the next treatment at day 37 in hospital because of severe aspiration pneumonia (he had only one functional lung, the other being severely destroyed by tuberculosis).


Technical complications were observed in two patients (10%). Strangely, on two occasions we experienced an unexplained distal rupture of a 0.010 Transend guidewire (Boston Scientific, Natick, Massachusetts, USA), albeit without clinical consequences. A complaint was submitted to the company but was unattended. In one of these patients, the microcatheter was stuck in the ONYX and broke on removal. The broken part of the microcatheter was carefully pushed up with the guiding catheter, salving it in the internal maxillary artery (IMA) and OA. The patient had mandibular pain and cheek numbness which resolved spontaneously. Low molecular weight heparins for 1 week and aspirin were prescribed. Three entrapped microcatheters (on a total of 36 microcatheterizations) broke on removal in the internal maxillary artery without any clinical consequence.

Procedure related complications were: facial nerve palsy, hemithermoanesthesia syndrome and protracted nuchal rigidity in one (14.3%) patient each. The right-sided hemithermoanesthesia occurred in a patient with a large deep seated type II DAVF of the vein of Galen, after additional transarterial ONYX embolization using the PChA. In one patient with a subarachnoid hemorrhage, transarterial ONYX embolization of a type II DAVF of the sigmoid sinus using the PMA (arising from the right VA) resulted in a right-sided peripheral facial nerve palsy.

Despite successful occlusion of all DAVFs presenting with intracranial hemorrhage, one 82-year-old man remained in a comatose state and finally died in hospital. Other bleeding related complications were: hemianopsia (n=2, 28.6%), quadranopsia (n=1, 14.3%) and dysphasia (n=1, 14.3%) patients.


Therapeutic strategies in DAVFs aim at complete obliteration of the fistula site and the abnormal venous outflow. Although in Borden type I DAVFs palliative treatment may be an option, in other types of DAVFs the high risk of (re-)bleeding compels full eradication. Over the years, interventional neuroradiologists developed techniques to approach DAVFs through the arterial as well as the venous system. In type I DAVFs, TAE turned out palliative (reducing flow and thus symptoms) whereas the venous technique was more often curative. On the other hand, in type II and particularly type III DAVFs, retrograde venous access is most of the time impossible, hence a transarterial solution is called for.

Preliminary experience with ONYX embolization for DAVFs has been very promising.14–18 In their pioneering study of 30 patients, Cognard et al included only DAVFs with CVR.17 In this Borden type II–III subgroup, Cognard et al reported 20 of 30 complete obliterations after a single procedure with ONYX only (66.7%) while overall immediate complete angiographic cure was achieved in 24 of 30 (80%), which seems far better than 48% cured DAVFs with n-BCA only.8 Similar occlusion rates of 70–92% were reported and confirmed by our study results (81%).18 20–22 24 26 A French group reported 100% angiographic closure rate in two ‘aggressive’ and four ‘non-aggressive’ DAVFs with CVR.16 These results suggest that ONYX allows for a more controlled injection technique with better fistula penetration.18 Taking into account that in Nogueira et al's (n=1/1), Lv et al's (n=4/5) and our study (n=4/6), type I DAVFs were treated successfully, it might well turn out that many, if not all, non-cavernous DAVFs are suitable for transarterial ONYX embolization. Moreover, we demonstrated that Borden type I DAVFs can be cured with transarterial ONYX embolization without occluding the sinus itself (figure 1). Although we do not treat type I DAVFs routinely, in patients with unbearable pulsatile tinnitus, transarterial ONYX embolization might lower the threshold to intervene actively.

We were successful in obliterating the arteriovenous shunt in a single ONYX embolization session in 13 of 21 (61.9%) DAVFs, a primary success rate which is similar to the results reported by Cognard et al and Nogueira et al (both 66.7%).17 18 Cure with ONYX embolization is impaired and even impossible if the DAVF has previously been treated with other embolization techniques (n=2) or surgery (feeder ligation, n=1). However, in a considerable number of cases with intended ONYX treatment (n=4/18), complementary embolic agents are required to yield complete cure. A versatile embolization technique remains mandatory, even if the number of obliterations with ONYX as a single agent is still increasing.23

Complication rates are low and do not seem to increase with the number of pedicles treated per patient.17 18 However, two patients had a permanent neurological deficit. In one patient, transarterial ONYX embolization using the PChA resulted in a right-sided hemithermoanesthesia due to reflux of ONYX. Embolization via the PChA is well known to be risky but the patient suffered a subarachnoid hemorrhage from a small ‘nidus-like’ fistula remnant, became extremely anxious and urged us to treat her instead of awaiting the scheduled radiosurgery. In the other patient, TAE of a type II DAVF of the sigmoid sinus using the PMA (arising from the right VA) resulted in a right-sided peripheral facial nerve palsy. Cranial nerve palsies were also reported by Cognard et al and attributed to reflux of ONYX in the MMA back to the foramen spinosum.17 26 The mechanism of facial nerve palsy in embolization via the PMA, as in our case, remains unclear, particularly because no deep penetration of ONYX in the MMA territory was observed. We were more restrictive with reflux in the meningeal arteries and preferred to look for an additional feeder. Moreover, we cooperated with our neurosurgeons to get direct access to the main feeder when distal microcatheterization was impeded by vessel tortuosity.27 Technical complications involved glued catheters and ruptured microguidewires. The latter complication occurred twice, fortunately uneventfully, but remained unexplained and has never occurred again in our practice. Stretching of glued microcatheters (only Ultraflow microcatheters) during retrieval can cause damage to adjacent nervous structures, as was transiently observed in one case. Entrapment of microcatheters should be avoidable nowadays by utilizing microcatheters with a high distal tensile strength (such as the Marathon) or with a detachable distal part.

Inadvertent pushing of the ONYX into a clogged fistula can suddenly provoke unexpected reflux, filling of arterio-arterial anastomoses and venous obstruction. To occlude the type II–III DAVF, entering of ONYX into the fistula vein is required and should be critically observed as the endpoint of injection. In the initial phase of embolization, ONYX will escape through the fistula and spread along the venous wall, forming a non-obstructive thin layer. But once a stable ONYX-bubble or plug forms at the microcatheter tip, this kind of laminated spreading of ONYX along the venous wall ceases and ONYX penetration into the fistula pouch and draining vein becomes more dense. In type II–III fistulas, at this moment, fistula occlusion is imminent and further injection superfluous. In type I fistula however, there are many fistulating points which all have to be occluded. Therefore, there is a substantial risk that ONYX might enter the sinus as a compact mass at different points. With the knowledge that incomplete transvenous embolization may end in disaster by diverting outflow into the cortical veins, we wanted to keep open the sinus at all cost. In view of the low bleeding risk of type I fistulas, transarterial ONYX embolization of these DAVFs has been debated. By understanding the mechanism of fistulation in type I DAVFs, we showed that TAE with ONYX is a serious option for suffering patients. Moreover, according to Shi et al, the venous balloon assisted technique provides a novel technique to preserve the venous sinus and adjacent key cortical veins and protect the patency of critical venous outflows.28 We think it is sensible to heparinize patients after a type I DAVF embolization routinely to avoid sinus thrombosis. In those type II–III cases in which DSA showed slow flow or stasis in veins clearly visible before occlusion of the fistula, we also advise heparinization. In other cases, the decision is at the operator's discretion.

With increasing experience, we could not confirm a difference between the ONYX concentrations 18 and 20. Only when the microcatheter tip was flapping in the fistula site was the more viscous ONYX 34 used to form a more stable plug, as advised by the manufacturer.

Meningeal feeders seem the most promising for complete success in ONYX embolization of DAVFs.16 Although for some authors microcatheter tip position does not seem to be a critical point,18 we learned that in the MMA as well as other feeding arteries, ONYX injection is best controlled with the microcatheter in a distally wedged position.17 In all such cases, filling of the fistula and arterial feeder network with ONYX was achieved more easily due to superior flow control and ONYX progression. When we could not reach such a microcatheter position in the principal arterial feeder, we opted for direct surgical access. Collaboration with our neurosurgeons lead to successful occlusion of the fistula or part of the fistula in two patients, one by direct cannulation of the MAA and one of the OA.


TAE with ONYX offers an effective and safe treatment for non-cavernous DAVFs with or without CVR. In non-aggressive DAVFs without CVR, it might offer a cure for the imploring symptomatic patient.



  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.