Cerebral arteriovenous malformations in the era of embolization for angiographic cure: a single-center experience in Egypt

Over the past decades, the target of embolization has evolved, starting as preoperative embolization to facilitate AVM resection (Crowley et al. 2014). Although it is not approved yet by the US Food and Drug Administration (FDA), pre-radiosurgical embolization is performed to eliminate high-risk features or to reduce nidal size (Pierot et al. 2013a). With evolution in both materials [e.g., ethylene-vinyl alcohol co-polymer (EVOH)] and techniques, the indications for endovascular embolization of cerebral arteriovenous malformations (AVMs) have expanded to include curative embolization (Mounayer et al. 2007; Panagiotopoulos et al. 2009).

Despite the wide variety of the embolic agents that can be used, only few agents are commonly used [n-butyl cyanoacrylate (NBCA), ethylene-vinyl alcohol copolymer (EVOH) (Onyx) (Covidien/ev3, Plymoth, Massachusetts, USA), and, to a lesser extent, polyvinyl alcohol (PVA) particles] (Vaidya et al. 2008). Glue (NBCA) has several privileges, including the immediate complete permanent vascular occlusion. Unfortunately, it requires high degree of experience for its proper use. Several complications can occur, including catheter entrapment (Debrun et al. 1997), distal spread, and proximal reflux beyond the intended location of the injection. Onyx (EVOH) in comparison to NBCA is less thrombogenic with minimal inflammatory reaction, since there is no protein denaturation (Duffner et al. 2002; Vinters et al. 1985). The major advantage of Onyx is the non-adhesive nature, which allows longer injection time, with the ability to perform control angiography when needed. Despite the non-adhesive nature of Onyx, still there is a risk of catheter entrapment due to reflux around the catheter tip. This article reported our experience in the treatment of cerebral AVMs using Onyx mainly. Our policy was to use embolization as a curative option whenever possible.

A retrospective review of all consecutive patients who had underwent endovascular embolization for cerebral AVM between January 2011 and June 2017 in Suez Insurance Hospital was performed. Twenty-one AVMs in 21 patients were included in the current study. All patients were fulfilling the following inclusion criteria: ruptured or unruptured AVM (either symptomatic or discovered accidently) with high bleeding risk due to small size (< 3 cm), deep venous drainage, and/or associated aneurysm (Fig. 1). Patients with recent intracranial hemorrhage were postponed for at least 1 month (hematomas may interfere with good angiographic assessment) (Pierot et al. 2013b). They were considered for neurosurgical opinion, in case of intraventricular hemorrhage or hemorrhage that cause mass effect; otherwise, a conservative therapy was implemented. None of those patients developed rebleeding with these conservative measures (Reitz et al. 2016).

All patients underwent clinical assessment and modified Rankin Scale (mRS) before and after all embolization sessions and during follow-ups, and before treatment, all AVMs were assessed by six-vessel digital subtraction cerebral angiography (DSCA) according to their eloquent site, nidus type, presence of fistulas, feeding arteries and draining veins, number of the main feeders, presence of aneurysms, and presence of venous stenosis. Spetzler-Martin grading (Spetzler and Martin 1986) was identified for all AVMs. Eloquent areas were identified according to Pollock and Flickinger’s (2002) definition: frontal and temporal (not invading the speech centers and motor strip) are non-eloquent; fronto-parietal, speech, parietal, occipital, intraventricular, corpus callosum, and cerebellar are intermediate; and basal ganglia, internal capsule, thalamic, and brain stem are considered eloquent. After the last embolization session, a follow-up angiography was performed 3 months later, except when other treatment modality is needed, follow-up DSA was performed after 1 month. A “poor” outcome was defined as a mRS score of 3–6 at discharge. Cure was defined as angiographic complete obliteration of the AVM nidus with disappearance of early veins.

The target of brain AVM treatment is to eliminate the risk of bleeding. This target could be achieved by complete occlusion of AVM. Various options are available (embolization, surgery, and radiosurgery) either alone or in combination. Embolization usually used as the first step in treatment plan, but in two different strategies. In the first strategy, the target is to achieve occlusion as much as possible before thinking in other options. The second strategy is to use embolization to prepare the patient to other modality of treatment. In the current study, we used the first strategy whenever possible.

A recent meta-analysis showed that the average cure rate for all NBCA-used studies was 13.7%. The cure rate increased to 24% in a subgroup analysis (387 patients) for studies conducted after the year 2000 (Elsenousi et al. 2016). Valavanis and Yaşargil (1998) embolized 7.7 feeders per AVM using NBCA with a cure rate of 40%. On the other hand, ethylene-vinyl alcohol copolymer (EVOH) (Onyx; eV3-Covedien, Irvine, California, USA) average cure rate was 30.5% in all studies, and 30.1% in the studies that did not use detachable microcatheters (Elsenousi et al. 2016a). The BRAVO trial was a European prospective multicenter trial to assess the safety and efficacy of Onyx. It showed 23.5% complete occlusion rate using onyx (Pierot et al. 2013b). Meanwhile, the range of complete occlusion rates in published case series was 8.3–53.9% (Pierot et al. 2013b). In the current study, complete occlusion was achieved in 9 patients (42.9%), this cure rate is similar to the previously quieted figures, and this high cure rate could be explained by the prior intention to completely occlude the lesion whenever possible, the use of Oynx with detachable tip catheters which permit longer injection time, and the use of staged approach. With case selection, van Rooij et al. (2012) reported achievement of cure for all the selected patient (24 patients) using Onyx without morbidity or mortality; also, de Castro-Afonso et al. (2016) reported (91.3%) cure rate in their selected AVM pediatric patients (23 patients) with (13%, 3 patients) procedure-related complications, yet without significant morbidity or mortality.

A new trend of embolization of brain AVMs with intent to angiographic cure was noticed. In some centers, curative embolization is offered to selected patients (van Rooij et al. 2012; de Castro-Afonso et al. 2016). This trend was encouraged by the expansion of the use of Onyx. This was reflected by higher cure rates of onyx compared with glue, but this was claimed to be associated with worse neurological outcomes. On the other hand, NBCA showed higher cure rates with time also, but without increase in poor neurological outcome. The assumption that the introduction of detachable microcatheter was associated with better neurological outcome was not supported by the results of meta-analysis. This may be explained by the fact that the use of the detachable microcatheters was associated with more aggressive Onyx embolization that may lead to venous occlusion and hemorrhagic complications (Elsenousi et al. 2016). It seems that the results of meta-analysis in this area should not be generalized, as it depends on operator experience which is variable, and cautious slow controlled injections that avoid the draining veins can minimize the procedural risk even in eloquent locations with longer injection time especially with the use of detachable tip microcatheters.

A meta-analysis showed that poor neurological outcomes for glue and Onyx embolization were 5.2 and 6.8%, respectively. When the results for patients embolized with Onyx using detachable catheters were excluded, poor outcome associated with Onyx embolization increased to 7.1% (Elsenousi et al. 2016). The BRAVO trial showed that Onyx embolization is associated with 4.3% mortality and 5.1% morbidity (Pierot et al. 2013b). The range of hemorrhagic complication in published case series was 4.0–12.2% while the range of mortality and morbidity was 0.0–3.2 and 3.5–15.5%, respectively (Pierot et al. 2013b). In the current study, early hemorrhage occurred in 3 sessions out of 43 sessions (7.0%) for 2 patients (9.5%). Only 2 patients (9.5%) were at grade 3 mRS at discharge (one of them was already on grade 3 on admission) which goes in line with the previously published data.

In our series, 17 AVMs (80.9%) were classified as high bleeding risk lesions. Thirteen AVMs (61.9%) were < 3 cm in size, 5 AVMs (23.8%) had deep venous drainage (not-exclusive), and 4 AVMs (19.0%) were associated with venous aneurysms. Crawford et al. (1986) described that 7% of the AVMs were associated with aneurysms and 75% of these aneurysms located on the major feeder vessels. AVM-associated aneurysms tend to present 10 years later than AVMs (Ondra et al. 1990). D’Aliberti et al. in their series found that the different types of AVM-associated aneurysms had different bleeding risk. The bleeding risk of the unrelated type and the remote flow-related type should be considered almost the same as any else unruptured aneurysm, where the INSUIA parameters were applicable. On the other hand, the adjacent flow-related type and the venous type carried higher bleeding risk (D’Aliberti et al. 2015). Choi et al. (2005) had suggested that AVMs with deep or infratentorial location, single or few draining veins, venous stenosis, and intranidal aneurysms may had increased risk of bleeding. Guo et al. (1995) found that small AVMs (< 3 cm) carried higher risk of bleeding. Meanwhile, deep venous drainage was found to be independent risk factor for AVM bleeding (Stapf et al. 2006). Although infratentorial AVM in adults account for only 7 to 15% of all brain AVMs, it was found to be an independent risk factor for bleeding; the bleeding rates were high up to 92% (Arnaout et al. 2009; Khaw et al. 2004; Westphal and Grzyska 2000).

The average age at presentation in this study was 34 years, which was in keeping with the previously quoted figure of 33 years (Ondra et al. 1990). In a recent meta-analysis for the outcome of embolization of cerebral AVMs, the mean for patients’ age was 35.8 and 35 years in the glue and Onyx groups, respectively (Elsenousi et al. 2016). AVMs were known to show male predominance, which went in the same direction with the current study findings [13 patients (61.9%) were males] (Ondra et al. 1990).

In the current series, 15 patients (71.4%) had a history of intracranial hemorrhage, 10 patients (47.6%) had a history of seizures, and 4 patients (19.0%) were complaining of chronic headache. In a meta-analysis, hemorrhage was the presenting symptom in 47% of patients in the glue and 41.5% of patients in the Onyx group. Epileptic seizures were the presenting symptom in 28% of the glue group and in 33% of the Onyx group (Elsenousi et al. 2016). Hemorrhagic presentation accounted for 50–61% of cases (Drake 1979; Perret and Nishioka 1966). AVM bleeding was associated with neurological deficit in 30–50% of cases only (Hetts et al. 2014). The high rate of bleeding as a presenting event in our series could be explained by the fact that 17 AVMs (80.9%) were classified as high bleeding risk lesions or due to small size of the sample. An association between age and the incidence of seizures was described by Crawford et al. (1986), with a seizure risk 44% in patients aged 10–19 years, 31% in patients aged 20–29 years, and 6% risk in patients aged 30–60 years.

Interventional treatment of AVMs is generally accepted; the annual risk of bleeding is 2–4%, which occurs over time, with high lifetime risk of bleeding that can result in morbidity or mortality (Young et al. 2015). The ARUBA (a randomized trial for unruptured brain arteriovenous malformation) has challenged this concept. The study concluded that medical treatment alone is better than interventional options, for prevention of stroke or death in patients with unruptured brain AVMs (Mohr et al. 2014). Despite the importance of the study, criticism to its design and follow-up duration undermined the value of its conclusion (Russin and Spetzler 2014). The poor outcome in the intervention arm of ARUBA trial is 30%, which is much higher than what was reported in a meta-analysis (5.2% for NBCA and 6.8% for EVOH) (Elsenousi et al. 2016); on the other hand, this poor outcome is also higher than the published rates of poor outcome for surgical resection of AVMs (0.34–2.2% in grades I–II) (Spetzler and Martin 1986; 2013). It was not clear how much of the poor outcome was related to surgical resection, embolization, or delayed hemorrhage from partially embolized AVM. On the other hand, a curative embolization with Onyx with high cure rates as recently published (van Rooij et al. 2012; de Castro-Afonso et al. 2016) that was associated with low morbidity and mortality (< 10%) may serve as a good treatment option. A low complication rate (< 10%) is lower than the complication rate of the medical arm of the ARUBA trial (Mohr et al. 2014). Six patients (28.57%) in our case series had unruptured AVM. Embolization treatment was considered for these cases due to the presence of high bleeding risk criteria of these AVMs.

In the current study, we used the staged approach in AVMs with multiple feeders. Despite that the value of this approach was debated, we thought it was associated with less hemorrhagic complications (Mounayer et al. 2007). This approach allowed smooth normalization of hemodynamics and prevented extensive venous thrombosis. The use of NBCA (glue) was limited to high flow fistulas in 4 patients (19.0%), where we thought it was more simple and effective method.

We compared between patients who achieved angiographic cure and patients who had partial occlusion. The comparison included the site, the total number of feeders, the size, the presence in eloquent site, and the presence of deep venous drainage. Only the presence of AVM in eloquent site was significantly associated with partial occlusion. This finding was logic, but we thought that larger sample size of study was needed to identify more factors that were associated with angiographic cure.

We think that curative embolization should be offered for selected cases based on certain criteria that need further research. For the cases that are not fulfilling these criteria, embolization should be offered as preparing step for other modality of treatment. For curative embolization, the AVM should be small sized (< 3 cm), supplied by one vascular territory, with feeders that can tolerate reflux up to 2–3 cm, with clear proximal parts of the draining veins, and not located in deep structure or brain stem (van Rooij et al. 2012).

Onyx embolization could serve as a curative option with accepted morbidity and mortality. Understanding the angio-architectural characteristics of the treated AVM could help in better differentiation between lesions suitable for curative Onyx embolization, and those non-suitable for curative embolization, with fewer treated feeders and high occlusion rate. Good selection of intranidal catheter tip position with slow, controlled injections that avoid the draining veins can minimize the procedural risk even in eloquent locations.

Several limitations should be considered in the current study: the retrospective nature, the potential patient selection for treatment bias, the small size of the sample, the absence of other modalities of interventional treatment, and the lack of cases that were treated on emergency basis.