Materials and methods

Nine patients (seven men, two women; aged 17–72 years) with 10 arteriovenous fistulas were treated with Onyx in our department over a period of 19 months. The fistulas included four direct and indirect CCFs, Barrow types A–D, and six Cognard types IIa+b, III, IV and V DAVFs. It is our experience that the use of these two separate classification systems for CCFs and DAVFs provides a more discriminating set of assessment metrics for each type of lesion.11 12 Table 1 includes the characteristics of the Barrow and Cognard types and the numbers of each type seen in this study.

Each patient was treated with Onyx. Onyx is a non-adhesive liquid embolic agent made up of an ethylene vinyl alcohol (EVOH) co-polymer dissolved in dimethyl sulfoxide (DMSO) with suspended micronized tantalum powder to provide contrast for visualization under imaging. Onyx is delivered by slow controlled injection through a microcatheter into the arteriovenous fistula (AVF) under image guidance.13 Each Onyx kit contains an additional 1.5 ml vial of DMSO that is used only to fill the deadspace of the microcatheter (which varies depending on the selected microcatheter) and which must be cleared at the start of injection. DMSO is also pre-mixed into each vial of Onyx, which may contain 6% EVOH (Onyx 18) or 8% EVOH (Onyx 34). The DMSO solvent dissipates into blood, causing the EVOH co-polymer and suspended tantalum to precipitate in situ into a spongy coherent embolus. Onyx immediately forms a skin as the polymeric embolus solidifies from the outside to the inside while traveling more distally in the lesion.

Early reports on this liquid embolic technique raised concerns about DMSO and Onyx angiotoxicity as well as about other potential complications.14 15 However, the lower total dose and dose rates of superselective infusions has resulted in safe injections with low rates of complications.16 17 18 The injection technique and responses to complications are complex and, for this reason, physicians who wish to use Onyx are encouraged to complete a short training course and follow-up.

Results

Complete occlusion of 10 fistulas was obtained with 12 sessions (1.2 sessions per fistula). One patient developed caudal cranial nerve group palsy that recovered over the next several weeks. No recurrences at follow-up angiography after 6 months were noted, and permanent complications were not observed in our small series.

Our first experience with hemodynamic instability was during transarterial embolization of a type D CCF, feeding by multiple branches of the bilateral internal (ICA) and external carotid arteries (figure 1A,B). Embolization of the CCF was performed under general anesthesia with Onyx 18. Initial injection of Onyx through the distal branches of the middle meningeal artery (MMA) and sphenopalatine artery on the right resulted in decreased CCF flow. As we began embolization through the third pedicle (MMA branch), the patient experienced a rapid decrease in heart rate followed by asystole, despite immediate termination of Onyx injection. Atropine (1 mg) was administered, and the patient recovered normal sinus rhythm within 15–20 s. Cautious slow Onyx embolization was restarted without additional episodes of hemodynamic instability. Complete occlusion of CCF was achieved (figure 1C). The patient was extubated without focal neurologic symptoms, and cardiac monitoring for 24 h did not demonstrate instability.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Barrow type D carotid–cavernous fistula (CCF) and Cognard type V dural arteriovenous fistula (DAVF) at the craniocervical junction. (A) Anteroposterior view of the left external carotid artery (ECA) and (B) lateral view of the right ECA angiogram outlines CCF filling from multiple branches with outflow into the right inferior petrosal sinus. (C) Complete occlusion of CCF. (D) Right vertebral artery angiogram demonstrates a second fistula at the level of the hypoglossal canal, filling from the vertebral artery branch and venous outflow into the anterior condylar and spinal veins. Guide catheter within the right inferior petrosal vein. (E) Transvenous occlusion of the fistula. (F) MR angiography (time-of-flight technique, source images) demonstrates multiple flow signals along the right trigeminal nerve (arrows).

The patient was discharged on day 2 but developed caudal cranial nerve palsy over the following 2–3 days. He was placed on steroids and gradually recovered over several weeks. The patient had a second DAVF at the level of the craniocervical junction, supplied by the ascending pharyngeal artery, vertebral artery and posterior–inferior cerebellar artery, with venous drainage into the anterior condylar and spinal veins (Cognard type V). Because of the location of the fistula at the level of the hypoglossal canal, embolization via the transvenous approach was chosen to decrease the chance of injury to the cranial nerves. This was successfully performed with Onyx 34, injected through a microcatheter placed retrograde through the right internal jugular vein and anterior condylar vein (figure 1D). Complete occlusion of the fistula with no angiographic or clinical complications was achieved (figure 1E). Interestingly, no hemodynamic instability, in particular anticipated heart rhythm change, was observed in this patient during injection of DMSO and Onyx in close proximity to the vagus nerve. Review of his pre-embolization MR angiogram demonstrated multiple flow signals along the cisternal portion of the right trigeminal nerve as well as at the level of the ganglion (figure 1F).

Severe bradycardia (heart rate < 20 bpm) was seen during Onyx embolization of a traumatic CCF in a 17-year-old patient with complex skull base fractures. He had a long segment dissection of the right ICA with multiple pseudoaneurysms at the level of the neck and a high-flow direct (type A) CCF. Occlusion of the ICA below the ophthalmic artery was achieved with coils. A control angiogram demonstrated type C CCF filling from the right MMA branches that was treated with Onyx 18. Bradycardia was noted during Onyx injection and responded to atropine.

As a result of these experiences, we began to monitor hemodynamics more closely in cases involving Onyx use. Several cases with Onyx embolization of fistulas, including superior sagittal sinus type IV DAVF through the MMA branch, transverse–sigmoid sinus type IIa+b DAVF involving the occipital artery and MMA branches, two posterior fossa type III and type IV DAVFs supplied by the vertebral artery and ascending pharyngeal artery branches with transvenous and transarterial approaches as well as direct CCF transvenous embolization (with balloon inflated in the ICA to prevent reflux) were uneventful. In some cases of high-flow fistulas, we used a combination of hydrocoils and Onyx as an initial step to decrease flow within the fistula.19 No significant heart rate changes were observed.

Our next observed asystole event in this setting followed Onyx 18 injection during treatment of a type B CCF with venous drainage into the right superior ophthalmic vein. We used a direct trans-superior ophthalmic fissure approach. No heart rate change was noted during placement of the needle through the orbit or microcatheter into the cavernous sinus. As Onyx began filling the sinus and extended into the superior ophthalmic vein, the patient developed asystole, which responded to termination of Onyx injection and atropine administration with no additional deterioration during continuation of embolization.

In another patient, a heart rate below 15 bpm was observed during anterior fossa type IV DAVF embolization. The fistula was supplied by bilateral ophthalmic artery branches (ethmoidal arteries (EA)), with additional inflow from branches of the internal maxillary and anterior cerebral arteries, and venous outflow into the superior sagittal sinus via the leptomeningeal vein with venous ectasia (figure 2A). The patient refused surgery. Bradycardia was noted during the first session of Onyx 18 injection into the right EA and responded to atropine (figure 2B). A second session of embolization through the left EA and transvenous through a microcatheter positioned into the leptomeningeal vein (figure 2C,D) was performed after pretreatment with atropine (0.5 mg), and no significant change in heart rate was observed. Complete occlusion was achieved (figure 2E,F).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Type IV dural arteriovenous fistula (DAVF) between the bilateral anterior ethmoidal arteries and the frontal leptomeningeal vein. (A, B) Bilateral carotid artery angiogram demonstrates midline fistula; venous outflow into the superior sagittal sinus via the leptomeningeal–cortical vein with venous ectasia. (C) Angiogram through a microcatheter positioned in the right distal ophthalmic artery, pre-Onyx embolization. (D) Angiogram through a microcatheter positioned in the left distal ophthalmic artery guided placement of the second microcatheter via a transvenous approach into the leptomeningeal vein along the right frontal convexity (E). (F) Left carotid artery angiogram. Complete occlusion of the fistula. (G) Onyx cast.

Discussion

Hemodynamic instability, including severe bradycardia and asystole, were observed in several patients during Onyx embolization of AVF. Based on our data, the most profound cardiovascular effect was observed at the beginning of Onyx injection. Previous work on DMSO safety suggests possible angiotoxicity in animal models and that this effect was dose and rate dependent.20 It was suggested that an acute cardiovascular effect and functional neurotoxicity may also occur.21 22

In all of our cases, the amount of Onyx injected was minimal to produce systemic effects. We adhered strictly to recommendations on injection rates of DMSO and Onyx during our procedures, not to exceed 0.1 ml/min.

Observed reactions were not patient specific because they occurred only during specific vessel injection, after uneventful injection into other branches in the same patient. The reactions did not appear to be volume dependent because significantly larger amounts of Onyx were injected in other patients treated for large arteriovenous malformations and AFVs. No significant mass effect was demonstrated. Concentration is likely to play a role, with low-flow and small-compartment fistulas having the strongest response. Patients with large high-flow fistulas (eg, direct CCFs) had no reaction, probably because of dilution of DMSO with the high-flow or large-space blood pool. It was suggested previously that the mechanism of trigeminocardiac reflex could be due to mechanical irritation of the trigeminal nerve branch by the catheter within the MMA during Onyx embolization.23 We have multiple cases of MMA embolization with particles and coils, both for hemostasis in trauma patients and for presurgical tumor embolization. We do not recall any similar reactions.

Our observations and evaluation of imaging and angiographic findings suggest that the observed responses could be the result of the direct effect of DMSO on the trigeminal nerve and its branches, especially the orbital division. The proposed reflex leading to this event is the trigeminal cardiac reflex. First detected in children undergoing strabismus surgery,24 the reflex has also been reported during skull base surgery, craniofacial surgery, balloon compression rhizolysis of the trigeminal ganglion and tumor resection in the cerebellopontine angle.25 26 27 The afferent tracts of the trigemenocardiac reflex derive from the ophthalmic division of the trigeminal nerve although tracts from the maxillary and mandibular division have been documented. These tracts synapse with the visceral motor nucleus of the vagus nerve, located in the reticular formation of the brainstem. The efferent portion is carried by the vagus nerve from the cardiovascular center of the medulla to the sinoatrial node. The reflex is more prominent in children; hypoxemia and hypercapnia contribute to more profound reflex. The nucleus of the tractus solitarius, located in the dorsomedial aspect of the caudal medulla, receives and integrates multiple visceral afferents regulating autonomic functions, including the cardiovascular and respiratory systems.28 29 The nucleus of the tractus solitarius also receives input from the ventral region of lamina I and II of the caudomedial region of the trigeminal nucleus. Apnea could be a part of this reflex.30

The reflex was responsive to atropine in all of our patients and was short in duration. Interestingly, rebound increases in heart rate above baseline were not noticed. All patients were monitored for 24 h postoperatively and remained stable. Hemodynamic fluctuations did not lead to adverse clinical consequences.

Conclusions

Treatment of DAVF and some types of CCFs with Onyx in our experience is an effective and relatively safe technique, resulting in durable results with a small number of sessions required. Hemodynamic changes, including asystole, may represent a direct effect of DMSO on the trigeminal nerve, resulting in a trigeminocardiac reflex.

Proposed measures to prevent this complication include pretreatment with atropine. Prophylactic placement of transvenous pacemakers may be necessary in patients with contraindications to atropine use or when atropine may not be effective because of underlying heart block. Hypoxemia and hypercapnia could aggravate the situation through the trigeminorespiratory reflex. Because apnea could be part of the reflex, mechanical ventilation should be considered in patients undergoing this treatment.

Additional studies to evaluate the direct effect of DMSO/Onyx on the trigeminal nerve in animals and the extent/duration of neurotoxicity may be warranted.