Hemorrhagic risk factors of endovascular onyx embolization of intracranial dural arteriovenous fistulas

Abstract

Background and purpose

Hemorrhagic complication is a disastrous complication of intracranial dural arteriovenous fistulas (DAVFs) embolization. This study was to analyze the possible risk factors for the hemorrhagic complication caused by endovascular embolization of DAVFs.

Methods

From January 2012 to July 2016, a total of 267 patients with intracranial DAVFs received endovascular Onyx embolization at our hospital. The demographic information, clinical presentation, angiographic features, endovascular treatment and hemorrhagic complications were reviewed. Univariate and multivariate logistic regression analyses were performed to evaluate the risk factors contributing to the post-procedural hemorrhagic complications.

Results

In 267 patients of DAVF treated with endovascular embolization, procedure-related hemorrhagic complication occurred in 12 (4.5%) patients. Univariate and multivariate logistic regression analyses showed that the pial arterial supplier (OR 13.630; 95% CI, 1.556–119.368; P = 0.018), giant venous aneurysm (OR 15.196; 95% CI, 2.505–92.183; P = 0.003) and Onyx volume ≥ 6 ml (OR 1.138; 95% CI, 1.006–1.288; P = 0.040) were significant factors associated with these hemorrhagic complications.

Conclusions

Hemorrhagic complications associated with endovascular DAVF embolization are not negligible. The pial arterial supplier, giant venous aneurysm and higher Onyx volume in one session may be risk factors for endovascular DAVF embolization.

Introduction

Dural arteriovenous fistulas (DAVFs) account for 10–15% of all intracranial arteriovenous malformations and represent anomalous shunts among arterial branches and dural venous sinuses, meningeal or cortical veins.¹ The natural history and clinical manifestations are determined by their location and angioarchitecture. DAVFs with pial venous drainage or venous aneurysm are risk factors for hemorrhagic presentation in patients with DAVF.^2–4 Management of DAVFs includes open surgery, radiosurgery, endovascular embolization, or a combination of these modalities. Endovascular Onyx embolization has now been accepted as the primary treatment strategy to manage DAVFs.^5–10 There is still a lack of detailed information concerning the incidence and risk factors of hemorrhagic complications related to endovascular procedures.^2–4,⁶ In this article, we analyzed the possible risk factors for the hemorrhagic complication caused by endovascular embolization of DAVFs.

Materials and methods

Study population

We retrospectively reviewed 267 patients with intracranial DAVFs treated by endovascular embolization in our hospital from January 2012 to July 2016. Spinal DAVFs (n = 2), children with dural sinus malformations (n = 1), or patients managed with surgery or radiosurgery as the initial treatment approach (n = 2) were all excluded from this study. Approval for this retrospective case series was obtained from the institutional review board in our hospital.

Demographic and symptoms at clinical presentation were collected as summarized in Table 1. The mean age was 47.4 ± 13.2 years (mean ±SD), and there was a male predominance (166:101; 62.2% males). The presentations prior to treatment were pulsatile tinnitus in 112 (41.9%) patients, ocular symptoms in 77 (28.8%), intracranial hemorrhage in 53 (19.9%) and incidental in 31 (11.6%). The fistula locations included tentorial in 82 (30.7%) patients, cavernous sinus in 71 (26.6%), lateral sinus in 48 (18.0%), anterior cranial fossa in 25 (8.4%), superior sagittal sinus in 21 (7.9%), marginal sinus in 13 (4.9%), torcular sinus in 5 (1.9%), sphenoparietal sinus in 2 (0.7%). According to the Cognard type classification¹¹: Cognard type I in 16 (6.0%) patients, type II in 123 (46.1%), and type III-IV in 128 (47.9%).

Table 1.

Clinical features of 267 patients.

Variable	Patients (n = 267)
Mean patient age (mean ± SD), years	47.4 ± 13.2
Male/female (%male)	166:101 (62.2%)
Presenting symptoms
Incidental	31 (11.6%)
Tinnitus	112 (41.9%)
Hemorrhage	53 (19.9%)
Ocular symptoms	77 (28.8%)
DAVF location
Tentorial	82 (30.7%)
Lateral sinus	48 (18.0%)
Anterior cranial fossa	25 (8.4%)
Superior sagittal sinus	21 (7.9%)
Cavernous sinus	71 (26.6%)
Torcular	5 (1.9%)
Marginal sinus	13 (4.9%)
Sphenoparietal sinus	2 (0.7%)
Cognard classification
I	16 (6.0%)
II	123 (46.1%)
III and IV	128 (47.9%)

In this study, patients were divided into hemorrhagic group and nonhemorrhagic group in accordance with whether the patients suffered intracranial hemorrhage or not at the diagnosis of DAVF. Intracranial hemorrhage was confirmed by CT scan. For all patients of the hemorrhagic group in this series, the source of intracranial hemorrhage was attributed to the DAVF.

Neuroimaging review

For each DAVF, we extracted detailed angiographic information, including Cognard types, feeding arteries (dural and pial arteries), the location of fistula (sinus and non-sinus locations), and the diameter of venous aneurysm (non-giant < 2.5 cm and giant ≥ 2.5 cm). Sinus type was DAVFs locating in the cavernous sinus, the transverse-sigmoid sinus, superior sagittal sinus, marginal sinus, as well as the torcular sinus. The non-sinus type included tentorial, petrosal and ethmoidal DAVFs.

Endovascular treatment

General anesthesia was utilized in all cases. Intra-procedure heparinization was achieved by dripping 3000 IU heparin/500 mL of 0.9% saline through the guiding catheter continuously during the procedure. No additional pre or post-procedural anticoagulation was utilized.

Among the 267 patients, one embolization procedure was performed in 232 patients, two procedures in 29 patients and more than two procedures in 6 patients. Transarterial Onyx(Onyx-18, Medtronic, USA) embolization was performed in 209 patients, transvenous Onyx embolization was in 53 patients, and combined two approaches in 5 patients. Additional radiosurgery was performed in 8 patients. Onyx volume of each session was recorded. The procedure-related hemorrhagic complication was according to the CT examination within 30 days after procedure. Patients were divided into the hemorrhagic complication group and non-hemorrhagic complication group.

Follow-up protocol

Clinical and angiographic follow-up were obtained. MR angiography was regarded as an alternative in cases where DSA was not available. The treatment was considered to be curative when complete fistula obliteration achieved on follow-up DSA. Neurological outcome was assessed using modified Rankin Score scale (mRS).

Statistical analysis

SPSS version 19.0 (SPSS, Inc., Chicago, IL, USA) was used for statistical analysis.

Except for patient age at diagnosis, all the variables were categorical and reported as count and percentage. Patients’ ages were calculated as mean ± SD and paired t tests were used to examine their differences. Cognard types were analyzed using Mann-Whitney U test. Factors of gender, location of the fistula, prior intracranial hemorrhage prior to the endovascular treatment, embolization outcome (complete or partial), the pial arterial supplier, the giant venous aneurysm ≥ 2.5 cm, the high Onyx volume in one session ≥ 6 ml were analyzed using paired Pearson χ² tests. If P value of < 0.05 in univariate analysis, then multivariate logistic regression analysis was performed. P value of < 0.05 was considered statistically significant.

Results

One hundred and sixty (59.9%) DAVFs were at sinus location and 107 (40.1%) DAVFs were at non-sinus location. Pial arterial supplier was found in 97 (36.3%) patients. Giant venous aneurysm was found in 21 (7.9%) patients. The mean volume of Onyx per session was 3.8 ± 3.0 ml. The Onyx volume in one session more than 6 ml was recorded in 36 (13.5%) patients. The overall initial angiography showed complete occlusion was achieved in 187 (70.0%) patients, near-complete occlusion in 20 (7.5%) and partial occlusion in 60 (22.5%) patients.

Hemorrhagic complication

Hemorrhagic complication occurred in 12 (4.5%) patients (Figures 1 to 5): 9 (75%) non-sinus and 3 (25%) sinus DAVFs. Nine (75%) patients were male and 3 (25%) were female. The mean age was 45.3 ± 15.6 years (range 18–69 years). Cognard I was found in no patient. Cognard II was found in 3 (25.0%) patients, and type III-IV in 9 (75.0%) patients. Eleven (91.7%) patients had pial arterial supplier. Giant venous aneurysm was found in 8 (66.7%) patients. Onyx volume more than 6 ml in one session was found in 9 (75%) patients. The interval between the end of the endovascular treatment and the onset of intracranial hemorrhage ranged from 30 minutes to 7 hours.

Figure 1.

A 32-year-old male patient. Computed tomography (CT) of the brain showing a mass lesion located in the right mesencephalen (a, b, and c).

Figure 2.

CT angiography showing an arteriovenous fistula with a large venous pouch draining into straight sinus (a). Angiogram of the right external carotid artery (ECA) showing a tentorial DAVF with dilated venous aneurysm (larger than 2.5 cm) fed by right middle meningeal artery and ascending pharyngeal artery (b, c and d). A marathon microcatheter was navigated into the fistula point via the middle meningeal artery and a volume of 2.0 ml Onyx was injected (e). The right ECA angiogram (f) showing the complete occlusion of the fistula.

Figure 3.

Six hours later, the patient suffered severe headache and a sudden onset of unconsciousness and acute CT scan (a, b and c) showed a hyperdensity within the venous aneurysm consistent with acute thrombosis and hematoma. The following day, the patient’s neurological condition continued to decline and CT scan showing the hyperdense region enlarged (d, e and f).

Figure 4.

A 69-year-old male presented with intermittent onset of headache. The right external carotid artery (ECA) (a and b) and left vertebral artery (VA) angiograms (c and d) showing a tentorial DAVF with dilated venous aneurysm (larger than 2.5 cm) fed by right middle meningeal artery, meningohypophyseal trunk of the internal carotid artery (ICA), and small dural branch of the posterior cerebral artery. A volume of 7.0 ml Onyx embolization was injected via the middle meningeal artery (e). The right common carotid artery (CCA) (f and g) and the left VA (H and I) angiograms showing the complete occlusion of the fistula.

Figure 5.

Two hours later, the patient suffered a sudden onset of unconsciousness and acute CT scan (a, b and c) showing a hyperdensity within the venous aneurysm consistent with hematoma. Emergency surgery was performed and control CT scan showing the hematoma as well as the venous pouch was completely removed (d, e and f).

Univariate and multivariate logistic regression analysis

In univariate analysis of the differences between the hemorrhage group and non-hemorrhage group (Table 2), gender (χ2 = 8.2, P = 0.004), non-sinus location (χ2 = 24.3, P < 0.001), giant venous aneurysm (χ2 = 15.2, P < 0.001), and the Cognard types (Z = −6.543 P < 0.001) were significantly associated with intracranial hemorrhage prior to the endovascular treatment. Multivariate logistic regression analysis confirmed that giant venous aneurysm (OR 21.069; 95% CI, 3.961–112.078; P < 0.001) was the only significant risk factor for intracranial hemorrhage prior to the endovascular treatment.

Table 2.

Demographic, clinical and angiographic characteristics of patients with dural arteriovenous fistula in hemorrhagic and nonhemorrhagic groups prior to endovascular treatment.

Variables		Total(n = 267)	Hemorrhagic group	Non-hemorrhagic group	P value	Multivariate analysisOR (95% CI)	P value
Gender	Male	166	42 (79.2%)	124 (79.2%)	0.004	0.948 (0.201–4.481)	0.946
Age	Years	47.4 ± 13.2	49.2 ± 12.4	47.0 ± 13.4	0.274
Location	Non-sinus	107	37 (69.8%)	70 (32.7%)	<0.001	0.703 (0.038–13.137)	0.813
	Sinus	160	16 (30.2%)	144 (67.3%)
Cognard classification	I	16	0	16 (7.5%)	<0.001	0.888 (0.087–9.144)	0.920
	II	123	6 (11.3%)	117 (54.7%)
	III and IV	128	47 (88.7%)	81 (37.9%)
Pial arterial supplier	With	97	21 (39.6%)	76 (35.5%)	0.578
	Without	170	32 (60.4%)	138 (64.5%)
Giant venous aneurysm (≥ 2.5cm)	Yes	21	11 (20.8%)	10 (4.7%)	<0.001	21.069 (3.961–112.078)	<0.001
	No	246	42 (79.2%)	204 (95.3%)

In univariate analysis of the differences between the hemorrhagic complication group and non-hemorrhagic complication group (Table 3), non-sinus location (χ2 = 6.4 P = 0.012), the presence of pial arterial supplier (χ2 = 16.6 P < 0.001), giant venous aneurysm (χ2 = 11.2 P = 0.001), Onyx volume ≥ 6 ml in one session (χ2 = 40.6 P < 0.001), and the Cognard types (Z=−1.967 P = 0.049) were significantly associated with post-procedural hemorrhagic complications. Multivariate logistic regression analysis confirmed that the pial arterial supplier (OR 13.630; 95% CI, 1.556–119.368; P = 0.018), giant venous aneurysm (OR 15.196; 95% CI, 2.505–92.183; P = 0.003) and Onyx volume ≥ 6 ml (OR 1.138; 95% CI, 1.006–1.288; P = 0.040) were significant risk factors of post-procedural hemorrhagic complications.

Table 3.

The demographic, angiographic and treatment information of the 267 DAVFs and logistic regression analysis results.

Basic characteristics			Total(n = 267)	Without-hemorrhagic complication(n = 255)	Hemorrhagic complication(n = 12)	P value	Multivariate analysisOR (95% CI)	P value
Gender		Male	166	157 (61.6%)	9 (75.0%)	0.349
Age		Years	47.4 ± 13.2	47.53 ± 13.1	45.3 ± 15.6	0.574
Prior hemorrhage		Yes	53	51 (20.0%)	2 (16.7%)	0.777
		No	214	204 (80.0%)	10 (83.3%)
Embolization outcome		Complete	187	181 (71.0%)	6 (50.0%)	0.121
		Partial	80	74 (29.0%)	6 (50.0%)
Location		Non-sinus	107	98 (38.4%)	9 (75%)	0.012	0.435 (0.015–12.347)	0.626
		Sinus	160	157 (61.6%)	3 (25%)
Cognard classification		I	16	16 (6.3%)	0 (0%)	0.049	0.659 (0.042–10.328)	0.767
		II	123	120 (47.1%)	3 (25.0%)
		III and IV	128	119 (46.7%)	9 (75.0%)
Pial arterial supplier		With	97	86 (33.7%)	11 (91.7%)	<0.001	13.630 (1.556–119.368)	0.018
		Without	170	169 (66.3%)	1 (8.3%)
Giant venous aneurysm (≥ 2.5cm)		Yes	21	14 (5.5%)	7 (58.3%)	<0.001	15.196 (2.505–92.183)	0.003
		No	246	241 (94.5%)	5 (41.7%)
Onyx volume in one session		≥ 6ml	36	27 (10.6%)	9 (75%)	<0.001	1.138 (1.006–1.288)	0.040
		<6ml	231	228 (89.4%)	3 (25%)

Follow-up outcomes

As to the 12 patients with hemorrhagic complication, two patients died of severe intracranial hemorrhage, two patients suffered severe disability, and the other eight patients all recovered uneventfully with no neurological deficits.

During mean 12.3 ± 13.7 months (range, 2–84 months) angiographic follow-up, 105 (70%) complete DAVF obliteration were found in 150 follow-up angiograms. Over a mean clinical follow-up of 66.6 ± 14.0 months (range, 48–96 months), 3 (1.1%) deaths were found in this series, 258 (96.6%) had no disability (mRS, 0 and 1), 4 (1.5%) had mild or moderate disability (mRS, 2 or 3), and only 2 (0.8%) had unfavorable outcome because of the initial hemorrhage (mRS, 4).

Discussion

Hemorrhagic complication after the endovascular treatment of the DAVFs has been reported in recent literatures with an incidence rate of 0-3.8%.²,³ In this study, the hemorrhagic complication rate after the endovascular embolization was 4.5% (12/267), which was not eligible in DAVF treatment. In 2017, Sadeh-Gonike U et al.⁵ conducted a systematic review of all previous transarterial embolization studies using Onyx published between January 2005 and December 2015 and found the rate of hemorrhagic complications rate was 0%. In 2018, Mantilla D et al.⁶ reported the outcome of transarterial embolization of 54 DAVFs and hemorrhagic complications occurred in two patients (3.8%).

The classification of sinus (Cognard I and II) and non-sinus (Cognard III and IV) type could reflect the clinical course of DAVFs.¹ Although non-sinus DAVFs carried higher risk of post-procedural hemorrhage in the univariate analysis, multivariate analysis did not show this association. This contradiction may be caused by other confounding factors: pial arterial supplier, giant venous aneurysm and Onyx volume.

The role of pial arterial supplier for predicting post-procedural hemorrhage in previous reported studies remained contradictory and controversial.²,³,⁷ In 2016, Wu Q et al.³ reported that the incidence of post-procedural hemorrhage rate in the DAVFs with pial arterial supplier was significantly higher than that in the cases without pial artery supply (2/6, 33.3% vs 1/47, 2.1%). On the contrary, in 2019, Osada T et al⁷ found patients with a DAVF with a pial arterial supplier in their series showed no intracranial hemorrhage during or after interventional therapy. Although hemodynamic changes after the procedure caused by occlusion of the normal draining veins had been reported as the main causes for post-procedure hemorrhage, the rupture site may also fed by a pial artery, which had been neglected in previous studies. In 2017, Sato K et al.² had reported a patient with tentorial DAVF who developed a severe intracranial hemorrhage after curative Onyx embolization. They found a continuous arterial bleeding from a glomus-like vascular structure around the proximal part of the embolized draining vein fed by a pial arterial branch from the posterior cerebral artery during emergency craniotomy.

In our study, the presence of giant venous aneurysm was a risk factor for hemorrhagic complication. The presence of large or giant venous aneurysm may prone to develop venous thrombosis after the decreasing the shunt flow. In a recent study, Brzozowski K et al.⁹ reviewed angiographic features of 25 DAVF patients and speculated mean outflow vein diameter might be associated with risk of hemorrhage before endovascular treatment. However, statistical analysis failed to recognize the association confined to the limited sample size. Our study demonstrated that the large or giant venous aneurysm instead of the pure venous ectasia contributed to the intracranial hemorrhage. In 2013, Gonzalez et al.¹² confirmed antegrade progressive thrombosis of the venous system by MRI after DAVF embolization. Venous thrombosis was caused by blood flow stagnation after occlusion of significant feeders. Venous thrombus initially developed from the venous pouch and continued to propagate in the venous system, which caused venous hypertension and venous rupture. Blood stagnation is a very powerful trigger of the coagulation cascade despite therapeutic anticoagulation.¹² Once the coagulation cascade was unleashed, it could not be halted even with local administration of fibrinolytic agents. In 2008, Cognard et al.¹³ reported a series of 30 patients with DAVF treated with Onyx embolization. Despite strict anticoagulation, the author identified 1 case of extensive thrombosis of the draining vein with cerebellar bleeding following successful obliteration of the fistula. In another large series that included 121 DAVF treated with transarterial glue embolization, Kim et al.¹⁴ noted some degree of venous thrombosis in the postoperative period in 5 patients. One of these patients recovered spontaneously and the other patients were treated with anticoagulant or steroid therapy. All but 1 patient made a full recovery. In our center, anticoagulation was not routinely utilized after the uneventful embolization. To our knowledge, the use of anticoagulation may be one way to reduce the risk of this complication, however, given this complication is rare, it remains unclear which patients should be anticoagulated.

The higher volume of Onyx has been investigated as a significant risk factor in the brain arteriovenous malformation (AVM) patients with hemorrhagic complications. In 2012, Ovalle et al.¹⁰ worked on 13 potential factors to find predictive factors for delayed hemorrhagic complication after uneventful embolization of AVM. The higher volume of Onyx was the only one predictive factor for the delayed hemorrhagic complication. In 2014, Baharvahdat H et al.¹⁵ also found that higher injected volume of Onyx and deposition on the venous outflow before complete occlusion of the AVM accounted for severe hemorrhagic complications. In this study, the higher Onyx volume seems also a predictive factor of hemorrhagic complication associated with DAVF embolization. These results highlighted the awareness of the higher Onyx volume injected into vessels associated with the higher volume of dimethyl sulfoxide (DMSO). Onyx, mixed in a DMSO solvent, was approved by FDA for endovascular embolization of cerebral AVMs or DAVFs. However, the possible toxicity of DMSO on blood vessels was still unclear.¹⁶ DMSO may evoke an inflammatory response and endothelial necrosis following intra-arterial injection and dissipation of the DMSO in the blood stream, and then led to potential toxicity to blood vessels both at the site of injection and may result in the post-procedural hemorrhage.¹⁶ Hence, moderate Onyx volume and slow Onyx injection rate were recommended.

Conclusions

Intracranial hemorrhage is not a negligible devastating complication associated with endovascular DAVF treatment. The presence of pial arterial supplier, giant venous aneurysm and higher Onyx volume in one session are the risk factors for hemorrhagic complication. Recognizing this potentially fatal complication is very important in endovascular DAVF embolization.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Beijing Municipal Administration of Hospitals Incubating Program (PX2020039) and the National Natural Science Foundation of China, grant number (81901197).

ORCID iDs

Peng Liu https://orcid.org/0000-0002-1849-9634

Xianli Lv https://orcid.org/0000-0001-8270-8464