Patient population

Under an institutional review board approved protocol, a retrospective analysis was carried out on all patients with Spetzler-Martin (SM) grade I-II AVMs between 2002 and 2017. SM I-II AVMs were identified using an existing quality assurance database, tracking all patients and all neurovascular procedures. Patients were divided into those presenting with a ruptured or unruptured AVM. Baseline clinical and imaging characteristics were compared between the ruptured and unruptured AVM groups. AVM size, eloquence, and venous drainage pattern were assessed for each patient using CT scans, MR imaging, and cerebral angiograms. Patient demographics and clinical presentation (hemorrhage, headache, seizure, clinical deficit, asymptomatic/incidental) were noted.

Endovascular and surgical procedures

All neuroendovascular procedures were performed in a biplane flat-panel angiographic suite (Siemens Medical, Erlangen, Germany; Philips Medical, Best, The Netherlands) under monitored anesthesia care or general anesthesia. The purpose of staged presurgical embolization in the treatment plan was reduction of blood flow to the AVM, elimination of high-risk features, such as feeding artery aneurysms, elimination of deep arterial pedicles penetrating white matter tracts or deep nuclei, and progressive reduction in arteriovenous shunt vasculature potentially altering perfusion of normal brain.

Presurgical embolization was performed using either n-butyl cyanoacrylate (Trufill n-butyl cyanoacrylate (n-BCA); Cordis Neurovascular, Miami Lakes, Florida, USA), ethylene vinyl copolymer (Onyx 18 or 34; Medtronic, Minneapolis, Minnesota, USA), detachable coils, polyvinyl alcohol (PVA, Contour; Stryker Neurovascular, Fremont, California, USA; Trufill PVA; Cordis) or a combination thereof. In select cases, provocative anesthestic testing using superselective injection of 10–50 mg amobarbital followed by 10 cc 2% cardiac lidocaine (Xylocaine) was performed in eloquently located AVMs before embolization. Details of the endovascular procedures, including number of arteries supplying the AVM, number of branches embolized, high-risk features (feeding artery aneurysm, intranidal aneurysms, venous outflow stenosis), and embolic agent(s) used, were studied. Complications during and after the interventional procedure were recorded. Angiographic outcomes were reported in the health records. Clinical outcomes were independently assessed and recorded in the health record by vascular or critical care neurologists.

All neurosurgical operations were performed by four neurosurgeons (RAS, MBS, ESC, NAF) specializing in cerebrovascular disease and all radiosurgical procedures were performed by a radiation oncologist in conjunction with the attending neurosurgeon (RAS, MBS, ESC, NAF) and medical physicist. Radiosurgery was used primarily for patients with deep or eloquent AVMs with a compact nidus and volume <30 cm³.

Functional outcomes and statistical analysis

Functional outcome was determined using the modified Rankin Scale (mRS). For each patient, the mRS score was determined immediately before and after each endovascular embolization or microsurgical treatment. Long-term follow-up was obtained using a combination of routine postoperative clinic visits, other hospital admissions, or phone interviews in lieu of routine clinic assessment. The clinical condition was dichotomized into two groups: mRS score ≤2 was considered as a good outcome while mRS score >2 was considered as a poor outcome. Any change in mRS score between the initial evaluation and subsequent intervention was monitored through routine clinical follow-up, which showed whether a deficit was transient or permanent.

Statistical analysis was performed using SPSS (version 20, IBM, White Plains, New York, USA). Student’s t-test (2×2 contingency tables) was used to analyze continuous data and the χ2 test (larger contingency tables) for categorical variables. Statistical analysis was performed to assess the association between baseline clinical characteristics and functional outcome. These patient variables were entered into a logistic regression model to determine independent predictors of functional outcome. Two simple logistic regression models were performed with the outcome variable being immediate postintervention mRS score and long-term follow-up mRS score. Statistical significance was defined as p<0.05.

Patient demographics

A total of 258 patients with SM grade I-II AVMs were evaluated and treated at one university medical center between 2002 and 2017. SM grade I and grade II AVMs were found in 36.0% and 64.0% of our patients, respectively (table 1). The mean age of our patients was 38.3±16.7 years (mean ±SD). No statistically significant difference in demographics was found between patients with ruptured and unruptured AVMs. Unruptured AVMs were present in 133 (51.6%) of our patients. Among the patients with unruptured AVM, the most common presenting symptom was seizure (40.6) followed by headache (31.6%), asymptomatic/incidental (18.8%), and clinical deficit (9.0%). Patients with a hemorrhage attributable to their AVM (ruptured AVM) were more likely than unruptured patients to present with a mRS score >2 (p<0.05, table 1). Presurgical embolization was performed in 63.6% of patients. Microsurgery and radiosurgery was performed in 84.5% and 15.5% of patients, respectively.

Endovascular treatment and clinical outcome

In our cohort of patients, differences existed in the endovascular treatment of ruptured and unruptured AVMs (table 1). Presurgical embolization was used more frequently in unruptured AVMs than in ruptured AVMs (67.7% vs 59.2%; p<0.05). Despite a similar number of arterial feeding arteries between ruptured and unruptured AVMs, the mean number of arterial pedicles embolized was greater in the unruptured group than in the ruptured group (2.51 vs 1.82; p<0.05). No significant difference was found in the embolic agent used in the two groups, with the majority of AVMs embolized using n-BCA followed by Onyx, then PVA, and least often detachable platinum coils.

Immediate post-embolization clinical outcome was assessed in both unruptured and ruptured AVMs. Presurgical embolization in patients with unruptured AVMs did not change their clinical condition (table 2). In the group with ruptured AVMs, presurgical embolization worsened the clinical condition of one patient (1.4%) through a delayed post-embolization cerebellar hemorrhage requiring emergency hematoma evacuation with AVM resection (table 3).

Catheter cerebral angiography was performed as a component of surgical resection in each case. Immediate post-microsurgical resection angiography demonstrated angiographic cure in 100% of patients in both the ruptured and unruptured group (table 4). AVM resection was completed in a single stage in all but two patients. In these two patients, residual AVM was seen on immediate postoperative angiography under anesthesia, and the patients were immediately taken back to the operating room for complete resection. In the radiosurgery group of patients, 15 patients in the unruptured group and eight patients in the ruptured group had follow-up angiograms confirming complete obliteration of their AVMs (table 5). Among the unruptured radiosurgery patients, eight had no follow-up angiograms. Of these eight, four patients had brain MRI scans demonstrating residual AVM. Similarly, among the ruptured radiosurgery patients, four patients had no follow-up angiograms. Of these four, one patient received follow-up MRI demonstrating residual AVM while two other patients had MRI scans demonstrating no residual AVM.

Endovascular procedure-related complications

Complications from the 164 embolization procedures are detailed in table 6. They include two postprocedural hemorrhages, one resulting in a small asymptomatic subarachnoid hemorrhage and one in a large cerebellar hematoma requiring emergency evacuation. Additional complications included one guide catheter-related dissection of the extracranial internal carotid artery and one postprocedural infarct of the occipital lobe in a patient with an occipital pole AVM after embolization of the posterior cerebral artery branch to the AVM.

Long-term clinical outcomes

Two hundred and forty-two of 258 patients (93.8%) had long-term follow-up (>30 days) in our study (tables 4 and 5). In long-term follow-up, good outcome (mRS score ≤2) was seen in 123 of 133 patients with an unruptured AVM (92.5%) and 110 of 125 patients with a ruptured AVMs (88.0%). One patient in the ruptured group died 2 years after treatment from congestive heart failure. This patient had an immediate postsurgical resection mRS score of 2.

In the unruptured AVM group, clinical deficits were seen in 10 patients; eight were transient and resolved at long-term follow-up while two were permanent deficits. Of the two patients with permanent deficits, one patient had a hemorrhage in the area of the motor cortex 1 year after radiosurgery for AVM located in the motor cortex. In the other case, the patient requested surgical resection of a posterior frontal AVM in the motor cortex despite the original recommendations for radiosurgery. The patient developed permanent postoperative hemiparesis that had not resolved on last clinical follow-up.

In the ruptured AVM group, clinical deficits were seen in six patients; five were transient and one permanent. The permanent deficit occurred in a patient who had a small occipital lobe hemorrhage from an occipital AVM that ruptured near the visual cortex. Postoperatively, the patient’s visual fields did not improve on long-term clinical follow-up.

In the radiosurgery group, one patient with an unruptured AVM had a hemorrhage in the motor cortex 1 year after radiosurgery for an unruptured motor cortex located AVM. This patient continued radiosurgery treatments after recovering from the hemorrhage.

When controlling for the covariates of SM grade, age, gender, rupture status, preoperative mRS score was the only significant predictor of both immediate and long-term functional outcome (p<0.05).

Microsurgery for low-grade AVMs

The SM grading scale was designed to predict the morbidity of surgical resection of brain AVMs.14 Microsurgical resection is considered the ‘gold standard’ for curing low-grade AVMs (SM I-II). With respect to the grading system, grades I and II can have only two architectures: (1) If the AVM is <3 cm maximum diameter, they can have either deep venous drainage or be located in the eloquent cortex, but never both; (2) If neither eloquent cortex nor deep venous drainage is present, then the AVM must be <6 cm in maximum diameter. Because of these uncomplicated architectural features, low-grade AVMs are often deemed safe for microsurgical resection with minimal morbidity and mortality. Based on review of the medical literature, the morbidity and mortality rates of curative surgical resection of these lesions is 2.2% and 0.3%, respectively.5–7 13–16

ARUBA compared any form of intervention for unruptured brain AVMs with their natural history in 223 patients for an average of 33 months over a 5-year enrollment period. The study demonstrated a 10.1% and 30.7% risks of stroke or death in the conservative and interventional management arms, respectively.4 Consequently, ARUBA concluded that conservative management was superior to intervention in unruptured brain AVMs of all grades for preventing death and stroke.4

Much criticism has been made of the ARUBA trial. First, only a small number of patients in the study underwent microsurgery compared with other forms of intervention, such as embolization and radiosurgery (81% of patients). The results of ARUBA need to be considered because embolization or radiosurgery alone lead to low complete occlusion rates of 13% and 38%, respectively, compared with 96% obliteration rate with surgery.9 17 Second, the mean follow-up length was only 33 months, which is too short a period in which to assess both long-term hemorrhage risks of AVMs and the efficacy of radiosurgery. With radiosurgery, there is a latency period of 2–4 years, on average, from the start of treatment to radiographic obliteration of an AVM.2 Third, the ARUBA trial was conducted at 65 certified sites, including medical centers that were not high-volume AVM centers. Additionally, the study lacked rigorous physician experience with treating brain AVMs, with a minimum requirement of having treated only 10 brain AVMs for a physician to participate in the study. Fourth, the study drew conclusions by analyzing only unruptured AVMs while, in real-world practice, many patients also present with ruptured AVMs. When studied in this context, it is possible to misapply the study’s results and conclude that any interventional modality, including microsurgical resection, of AVMs has a higher morbidity than conservative management. Fifth, ARUBA failed to specifically investigate the microsurgical outcomes of patients with SM I-II AVMs, who because of the particular anatomical features of the AVM, are better treated with microsurgery.2 Subgroup analysis of ARUBA and other published case series on the treatment of lower SM grades have shown that microsurgery is safe and effective.

Overall in our study, 233 patients (90.3%) in total, with 123 patients (92.5%) in the unruptured group and 110 patients (88%) in the ruptured group, experienced good outcomes (mRS 0–2). Sixteen patients (6.2%) were lost to long-term clinical follow-up. For patients lost to follow-up, the social security death index was searched, and no deaths were reported for these patients. Fifteen patients (5.8%) had a neurological deficit immediately after surgical resection. Of these 15 patients, 13 (86.7%) experienced resolution of their deficit at long-term clinical follow-up. Of the two surgical patients who did not improve on clinical follow-up, one patient elected for surgical resection of a motor cortex located AVM despite having the option of radiosurgery while the other experienced a visual field deficit after resection of an occipital lobe AVM. In the radiosurgery group of patients, one patient had a large intraparenchymal hemorrhage in the motor cortex 1 year after radiosurgery for a motor cortex located AVM. Our data suggest that many immediate deficits encountered after surgery are transient and can improve over time. In another large series of 288 surgically resected AVMs, the authors also found that early postoperative clinical worsening was seen in nearly 40% of their patients but when followed up long-term, only 12% of the patients had a permanent deficit.7 This supports our finding that many clinical deficits improve over time.

With long-term clinical follow-up, our overall surgical morbidity was 1.2% and one mortality. Our results are consistent with other large case series on the surgical resection of low-grade AVMs. Potts et al, in their series of 232 patients with low-grade AVM managed surgically, reported a surgical morbidity of 3%.6 Similarly, Morgan et al reported a morbidity rate of 1.4%5 and Schramm et al a morbidity of 2.6% for their surgically resected low-grade AVMs.7 In their operative series of 67 AVMs treated with surgery, Sisti et al reported a surgical morbidity of 1.5%.13

A few details of our low surgical morbidity should be noted. First, when deemed feasible, we readily use presurgical embolization at our institution before microsurgical resection. We believe embolization helps to eliminate deep arterial pedicles that would otherwise require greater dissection. This is especially useful in AVMs where the arterial supply may be deep in the AVM nidus, or in AVMs located in eloquent cortex where less surgical dissection would reduce postoperative morbidity. We used embolization in 64% of our patients compared with Potts et al who used it in 43%, Morgan et al who used it in 8%, and Schramm et al who used it in 13.5% of their patients.5–7 Second, we have long-term clinical follow-up on our patients, which shows that many of the immediate clinical deficits are transient, with many patients recovering at later follow-up. Lastly, we believe that not all grade I and II AVMs have an equal surgical safety profile. Various combinations of AVM characteristics (size, venous drainage, eloquence) exist within a given SM grade that result in subtypes with different surgical risks. Similar to the heterogeneity within SM grade III AVMs, there is heterogeneity within SM grade II AVMs. In their study of functional outcomes across different subtypes of 208 SM grade II AVMs, Hung et al found that deep venous draining SM grade II AVMs had a worse outcome than medium sized grade II AVMs and eloquently located grade II AVMs.10 Similarly, Schaller et al showed that 29.5% of patients with deep drainage had permanent post-treatment deficits compared with 7.5% patients with superficial drainage.18 Although deep venous drainage has been a predictor of risk of hemorrhage, it can also make it difficult and dangerous to gain access to the AVM.18 19 In our patients, only 29% of those with grade II AVMs had deep drainage. This might be responsible for the reduced morbidity seen in our study.

Presurgical embolization of low-grade AVM

Endovascular embolization of AVMs requires selective catheterization of arterial pedicles to the AVM in order to occlude the AVM nidus and dedicated perinidal feeding arteries, while preserving venous outflow.2 The ultimate goal is to reduce blood flow to the AVM, eliminate high-risk features (ie, arterial aneurysms), and occlude deep arterial pedicles that would otherwise require more brain dissection to visualize and control, reduce the nidal size or eliminate high-risk features such as ruptured aneurysms, to allow for radiosurgery. The role of presurgical embolization has evolved as endovascular techniques continue to be refined, and newer embolic agents (eg, Onyx) are introduced. Many case series have been published demonstrating the low morbidity associated with endovascular embolization, with some rivaling the morbidity of surgical resection alone, but also showing complete angiographic cure with embolization alone.20–23

Although effective, use of presurgical embolization is still restricted because of the concern that embolization may add risk.

We used presurgical embolization in 64% of our patients (figures 1 and 2). Despite the large percentage of patients who were embolized, a significant percentage of patients did not receive presurgical embolization. A few reasons may explain this finding. First, a larger percentage of non-embolized patients were patients with ruptured AVM who presented with symptomatic hemorrhages and did not have time for presurgical embolization. They went immediately for surgical decompression, intracranial hemorrhage evacuation, and AVM resection. This is supported by our data, which shows that embolization was performed in 67.7% of unruptured patients compared with 59.2% of ruptured patients. Second, we did not perform presurgical embolization on a large number of patients scheduled to receive radiosurgery (16% of our patients). While still controversial, embolization may reduce the efficacy of radiosurgery because the embolic material used obscures the borders of the AVM, thus limiting the target volume for radiotherapy, and reducing the effects of radiation on the AVM.2 Third, because safe embolization is achieved through superselective microcatheter access to feeding arteries, certain angiographic features of AVMs pose greater challenges for embolization than others. Many of the AVMs we treated either harbored small pedicles that were difficult to catheterize or were supplied by deep pedicles such as lenticulostriate or thalamoperforating arteries that are more dangerous to embolize without collateral tissue injury. Fourth, if an AVM is small and superficial, our experience has been to proceed directly to surgical resection and not incur any potential risk of embolization for marginal benefits. Lastly, the presence of en passant vessels going to eloquent cortex or normal brain was a reason to be more judicious with our embolizations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Images obtained in a patient with sudden-onset headache. (a) Non-contrast head CT demonstrated a cerebellar hemorrhage, extending into the fourth ventricle with obstructive hydrocephalus. (B, C) Anteroposterior and lateral projection angiograms show a (insert size) midline cerebellar arteriovenous malformation supplied by the left superior cerebellar artery (SCA) with an arterial aneurysm and drainage into the straight sinus. (D, E) Superselective arteriography with the microcatheter in the SCA proximal to the aneurysm and (F) subsequent injection of n-butyl cyanoacrylate with obliteration of the aneurysm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

This is a patient who presented with seizures. MRI brain demonstrated an arteriovenous malformation (AVM) in the right frontal lobe. (A, B, C) Anteroposterior and lateral projection angiograms show a 2.5 cm right frontal AVM with supply from an enlarged callosomarginal branch of the anterior cerebral artery and superficial drainage into the superior sagittal sinus. (D, E, F) Superselective arteriography with the microcatheter tip placed proximal to the nidus and subsequent injection of n-butyl cyanoacrylate into the nidus of the AVM.

At our institution, n-BCA was the primary liquid embolic used for AVM embolization for several reasons. First, there has been a longer history of use with n-BCA, first off label and then after FDA approval in 2000 for the treatment of cerebral AVM. Onyx was FDA-approved several years later in 2005. The senior endovascular authors, SDL and PMM, trained during the era when n-BCA was the only available agent and became very familiar and comfortable with its use in AVM embolization. Second, many of our patients underwent provocative neurological testing at the time of embolization and were not under general anesthesia. Use of an Onyx-compatible, over-the-wire microcatheter, such as the Echelon, Marathon or Apollo, can often cause vessel distortion and discomfort in an awake patient. The flow-directed catheters used in n-BCA embolizations often did not require guidewire direction. Additionally, studies have shown that Onyx embolization generally requires more fluoroscopy time and therefore higher radiation doses, and the patient must remain still for longer than with n-BCA embolization.24 Third, Onyx embolization can be associated with complications, such as extraction of an embedded microcatheter due to Onyx reflux, which we have not encountered with n-BCA. Catheter entrapment has been substantially addressed by the introduction of the Medtronic Apollo catheter in 2014.

In our study, we saw an overall presurgical embolization morbidity of 2.4%. These numbers are lower than in many other large case series studying the safety of embolization. Iosif et al reported a procedure-related morbidity of 2.7% in their series of 74 patients undergoing curative embolization of low-grade AVMs.3 Crowley et al reported a 11.2% morbidity rate in their series of presurgical embolization of AVMs of all grades, with 9.6% permanent deficits and 1.8% transient deficits.20 Moon et al reported a permanent morbidity of 2.7%.11 Potts et al used embolization in only 43% of their patients and saw no morbidity in their series.6 Starke et al reported a permanent morbidity of 2.5% after embolization of AVMs across all grades.8 They found that predictors of morbidity after embolization were AVMs requiring more than one embolization session, diameter <3 cm or >6 cm, deep venous drainage, and AVMs located in eloquent locations.8

Many of the procedure-related complications in our study are within the established published risks of AVM embolization and cerebral angiography in general. One patient had an asymptomatic extracranial internal carotid artery dissection. Another had a small occipital lobe infarct and two others postintervention hemorrhages. Reported hemorrhage rates due to embolization of AVMs range from 1% to 2%.3 8 In addition, it is significant to note that many of the immediate deficits due to embolization are transient rather than permanent. Starke et al showed that 14% of their post-embolization patients developed a new clinical deficit but only 2.5% of patients had persistent deficits.8 Similar to their data, the one post-embolization patient in our study who had a visual field deficit after embolization of a deep posterior cerebral artery feeder for an occipital lobe AVM had improvement of vision postoperatively.

We attribute our low morbidity to a number of reasons. First, we judiciously used embolization to facilitate microsurgical resection rather than to attempt curative embolization. Thus, our intentions were to reduce blood loss, occlude deep feeders that would make surgery more difficult, and eliminate arterial aneurysms that would risk rerupture. These techniques differ from those needed when embolization is used as primary cure, where overembolization of an AVM runs the risk of occluding important venous outflow. Second, when treating AVMs located in eloquent cortex, we performed our embolizations under monitored anesthetic care and used provocative anesthetic testing through selective injection of amobarbital and lignocaine into arterial feeders that supplied the AVM. This allowed us to perform neurological assessments before permanent embolization.

In this series, we exclusively used a transarterial approach to AVM embolization. Although the transarterial route is routinely used for AVM embolization, not all AVMs have accessible arterial feeders. Newer techniques and liquid embolic agents have allowed transvenous embolization to become a feasible option in select patients. Indications for transvenous embolization are based largely on anatomical features of the AVM: deep location, small nidus, single draining vein, and absence of arterial access.25 The introduction of Onyx, with improved injection control and longer duration of injection, expanded the transvenous route for AVM embolization. The transvenous route is often used together with transarterial embolization techniques. New techniques, such as temporary balloon occlusion of the arterial pedicle to reduce arterial pressure in the AVM, allow improved Onyx penetration of the nidus transvenously.25 Furthermore, the pressure cooker technique, in which two microcatheters for embolization are placed into the AVM nidus and draining vein in conjunction with control of arterial pressure through temporal balloon occlusion of the internal carotid artery, coil embolization of arterial pedicle, and reduction of systolic blood pressure, improves the efficacy of the transvenous approach.25 Viana et al described their success with transvenous embolization of 12 patients. They reported a 91.6% immediate angiographic occlusion of the AVM without any procedural or clinical complications.25 Similarly, Mendes et al showed a 92.6% angiographic cure of 41 AVMs that were treated transvenously.26

The transvenous approach when combined with transarterial embolization has led to more complete obliteration of AVMs. Recently, endovascular embolization with the intention of cure has been described in the medical literature. Historically, the rates of complete endovascular obliteration of an AVM are <40%.23 26 In highly select patients with specific anatomical AVM features, complete obliteration approaches 90%. Wu et al, in their systematic review of the literature, discuss the angiographic characteristics of an AVM that can be associated with complete endovascular obliteration. These were AVM nidus size <30 mm, located in non-eloquent cortex, deep location, superficial or large arterial feeders with a single arterial pedicle, anatomy allowing for 2–3 cm of embolic material reflux, and unobstructed views of proximal draining veins.23

Our study has a number of limitations. First, this is a single institution retrospective analysis that may have sampling biases, which limit generalizability of our results. Second, we used mostly n-BCA embolic agent, and so we cannot confirm how safe embolization would be if Onyx were used instead. Third, functional status was based on mRS, which is inherently subjective, depending on a rater’s judgment, and also would not account for higher cortical function, which is better assessed by neuropsychiatric testing. Finally, a number of patients were lost to follow-up in the radiosurgery group limiting our ability to generalize our results to patients with low-grade AVM treated in that way.