A proposed grading system for endovascular treatment of cerebral arteriovenous malformations: Buffalo score.

BACKGROUND: The Spetzler-Martin arteriovenous malformation (AVM) grading system has proven to be useful in guiding treatment of cerebral AVMs with craniotomy. It is based on anatomical characteristics each of which makes surgical resection of an AVM more difficult, namely, deep venous drainage, eloquence of surrounding tissue, and large nidus size. A higher score correlates with more complications after treatment. Although this grading system has proven reliable over time, it does not reflect the major determinants of risk associated with endovascular treatment. The authors developed a grading system unique to endovascular treatment of cerebral AVMs.
METHODS: The proposed grading system accounts for the principal AVM anatomical and physiological features that make endovascular embolization more difficult and, thus, the likelihood of complications greater. These include number of arterial pedicles, diameter of arterial pedicles, and eloquent location of AVM nidus. The proposed grading system was retrospectively applied to 50 patients undergoing endovascular AVM embolization, and its ability to predict complications was compared to the Spetzler-Martin grading system.
RESULTS: Perioperative complications among the 50 patients included 4 major and 9 minor complications. The proposed grading system was predictive of complication risk, with an increasing rate of perioperative complications associated with an increasing AVM grade. An improved correlation of perioperative complication incidence was noted with the proposed system (P = 0.002), when compared with the Spetzler-Martin grading system (P = 0.33).
CONCLUSION: This grading system for the endovascular treatment of AVMs is simple, easily reproduced, and clinically valuable.

INTRODUCTION

The arteriovenous malformation (AVM) grading system introduced by Spetzler and Martin in 1986[9] has proven to be useful in guiding treatment of cerebral AVMs with craniotomy.[2] It is based on anatomical characteristics, each of which makes surgical resection of an AVM more likely to result in complications, including deep venous drainage, eloquence of surrounding tissue, and nidus size. A higher Spetzler–Martin score correlates with more complications after treatment.

Although this grading system has proven reliable over time and even in prediction of deficits after endovascular treatment,[5] it does not reflect the major anatomical issues at hand in conjunction with endovascular treatment.[13] This is relevant, given advancements in endovascular and radiosurgical treatments that have made multimodal treatment of AVMs favorable over surgical resection alone.[10] To account for this, a separate grading system for radiosurgical treatment has been devised,[78] and another grading system for endovascular treatment has been suggested, although this system has not been applied or validated.[1]

A grading system to estimate the risks of endovascular treatment of AVMs should take into account the anatomical difficulties unique to that therapeutic approach. We propose a grading system, the Buffalo score, that accounts for the principal anatomical and functional AVM features that we have observed to make endovascular embolization procedures more difficult and prone to complication. Such features include the number and diameter of arterial pedicles and the anatomical (functional) location of the AVM nidus.

METHODS

Description of the grading system

Important features of AVM embolization

Endovascular treatment of an AVM requires microcatheter selection of an artery feeding directly into the AVM. A liquid embolic agent is then delivered directly into the AVM directed at embolization of the abnormal vasculature. Hazards of treatment come from selective catheterization of the targeted artery or embolization of arterial supply to normal brain tissue. The proposed grading system is based on a combination of factors that make the likelihood of complication greater: Number of arterial pedicles, diameter of arterial pedicles, and eloquent location of the AVM nidus. These factors were chosen based on the operative experience of the authors and others and are discussed in detail.

Graded variables

Number of arterial pedicles

The number of arterial pedicles is estimated based on angiography. The number of arterial pedicles is determined to be 1-2, 3-4, or 5 or more and is based on an estimation of the number of arterial pedicles contributing flow into the AVM nidus. Embolization of each arterial pedicle essentially comprises a separate surgical approach as, with each embolization, the microcatheter must be removed and replaced for another arterial pedicle embolization. Whether performed during one prolonged procedure or in staged fashion, each microcatheterization and embolization carries technical risks of arterial dissection or wire perforation, as well as potential neurological damage. Thus, with an increasing number of arterial pedicles, the risk of procedural complications increases.

Arterial pedicle diameter

The diameter of arterial pedicles is estimated on the basis of angiography. Arterial pedicles are determined to be mostly large (diameter >1 mm) or small (diameter ≤1 mm). This measurement is assessed at a distal segment of the arterial pedicle, within 1 cm of the AVM nidus. A smaller vessel caliber will make catheterization-related complication more likely, as a more delicate vessel is more prone to injury during wire manipulation. In addition, reflux of glue polymer may be more likely, with less penetration into the AVM nidus owing to blood flow demand.

Eloquent location

Eloquent location of the AVM nidus is determined based on nidus location on angiography and correlated to noninvasive magnetic resonance imaging. Eloquent location of the AVM nidus is defined according to the grading system of Spetzler–Martin.[9] Any portion of the AVM nidus located within the motor or sensory cortex areas, including language and vision, as well as deep eloquent areas comprising the hypothalamus, thalamus, brainstem, and cerebellar peduncles, is considered eloquent in location. With arterial embolization of an AVM, embolysate may be inadvertently injected to a nearby or en passage vessel supplying normal brain. Such inadvertent embolization is more likely to cause a neurological deficit if the AVM is situated in eloquent cortex.

Determination of grade

The grade of the AVM is calculated by determining the individual scores of arterial pedicle number, arterial pedicle diameter, and eloquent brain location. A numerical value is assigned to each of the categories [Table 1], and the grade is determined by summing the points assigned to each category. The lowest possible score is Grade I; such a lesion would have a single, large arterial pedicle in noneloquent cortex. The highest possible score is Grade V; such a lesion would have five or more arterial pedicles, mostly ≤1 mm in diameter and be located within eloquent cortex. Other gradations comprise variations of arterial pedicle number (N), pedicle diameter (D), and eloquence of adjacent cortex (E). Angiograms displaying examples of the grades of the proposed system are shown in Figure 1. Schematic representations of the features and grades of the proposed system and the Spetzler–Martin scheme are provided in Figure 2a and b, respectively, for the purpose of comparison.

Table 1

Determination of cerebral arteriovenous malformation grade according to the proposed (Buffalo) grading system

Figure 1

Lateral views of early arterial phase angiograms display examples of possible grades from the series of 50 patients retrospectively tested with the proposed grading system. No patients treated had an arteriovenous malformation (AVM) with 1 or 2 small vessels and noneloquence of adjacent brain, and thus an example of this is not shown. N: Number (of arterial pedicles); D: Diameter (of arterial pedicles); E0: Noneloquence; E1: Eloquence

Figure 2

(a) The Buffalo system takes into account the following features: Number of arterial pedicles (N), diameter of those pedicles (D), and eloquent location (E). This schematic representation of supratentorial AVMs provides examples of different AVM types and the grade (1-5) determined by summing the points for each graded feature. In this figure, arterial pedicles and nidus are represented by black lines and shading, respectively. A higher complication incidence would be expected for patients with a higher score. See also Table 1. (b)The Spetzler–Martin system takes into account the features of venous drainage (V), size (S), and eloquence (E). This schematic representation of supratentorial AVMs provides examples of different AVM types and the grade (1-5) determined by summing the points for each graded feature. In this schematic, the AVM nidus is represented with black shading and draining veins are represented with lines (dashed lines represent deep drainage whereas solid lines represent superficial drainage). A higher complication incidence would be expected for patients with a higher score. See also table 1

Application of the grading system

To test the predictive value of this grading scale, we conducted a retrospective analysis of 50 consecutive cases of AVMs in which patients were treated at our hospital between January 2006 and December 2011. Embolization was performed with the intention of cure in all cases, although in most cases this was not possible. We attempted catheterization and obliteration of all AVM pedicles until cure was achieved or when the treating physician felt that the risk of endovascular treatment was outweighed by the potential benefit, in which case adjunctive treatment with stereotactic radiosurgery or surgical resection was performed. At our hospital, endovascular embolization of one or two pedicles is generally performed at 6-week intervals for patients requiring multiple embolization procedures. Patients were generally treated under conscious sedation unless patient safety mandated intubation for the procedure. A superselective Wada test is performed to assess the eloquence of each pedicle prior to embolization, as previously described.[11] Liquid embolic agents are used for nearly all arterial pedicles embolized. At our hospital, we favor Onyx (Covidien, Mansfield MA) as the liquid embolic agent of choice; however, we frequently employ n-butyl cyanoacrylate (NBCA) in cases where the microcatheter is not delivered into an ideally distal position proximal to the AVM nidus, as we have had superior penetration of the AVM nidus with NBCA in this setting. Patients presenting with AVM-related hemorrhage were treated during the presenting hospitalization if high-flow characteristics, such as a prenidal aneurysm, associated arteriovenous fistula, or venous outlet stenosis, were encountered on angiography (defined below).

Each AVM was graded according to the Spetzler–Martin scheme[9] as well as the proposed grading scale [Table 1]. All patients underwent biplane cerebral digital subtraction angiography, and graded variables were derived from preoperative imaging. AVM nidus size, venous drainage pattern, and eloquence were calculated based on the criteria of the Spetzler–Martin grading system.[9] Number of arterial pedicles and pedicle diameter were estimated based on angiography, and classified according to the proposed grading system. Other relevant or potentially relevant clinical variables analyzed include age,[3] presentation (presence of hemorrhage or seizure at time of initial presentation), presence of neurological deficits at the time of the initial embolization procedure,[3] and abnormal angioarchitectural structures.[4] Abnormal angioarchitectural structures included prenidal aneurysm, fistulous or “high-flow” component[4] (defined according to the criteria of Yuki et al.),[13] or venous outlet stenosis (defined as >50% narrowing of a draining vein on cerebral angiography). Other demographic data and anatomical characteristics previously noted to be correlated with perioperative complications or morbidity were noted and included age,[3] perioperative hemorrhage,[6] and lack of neurological deficit at baseline.[3] All complications, including intraoperative and postoperative events, were noted and tabulated for each grade for both grading systems and reported as a percentage. Complications were arbitrarily deemed “major” complications if the result was a permanent neurological deficit or death. Other complications were considered “minor” complications and were temporary.

Complete obliteration of the AVM was defined as no evident contrast filling of preexistent AVM nidus and absence of early venous draining on a catheter angiogram that was performed at least 3 months after the last embolization. Cases without complete obliteration were assessed for percentage of embolization of the original nidus by two of the authors (TMD and PK). As this is a somewhat subjective measurement, no statistical assessment was made with respect to this outcome, although this data is provided as an approximate measure of completeness of obliteration in the series.

Independent variables and the grading systems were analyzed for predictive value of complications and tested for statistical significance with Fisher's Exact test or analysis of variance. Independent variables found to be significant to P < 0.20 were included in a multivariate analysis. Data were analyzed with SAS 9.2 (SAS Institute Inc., Cary, NC). The local institutional review board approved this study.

The test group consisted of 50 consecutive patients treated with endovascular AVM embolization [Table 2]. Most patients were men (54%), with varied presentation overall (including hemorrhage in 44% and seizure in 27%). Endovascular embolization with a goal of AVM obliteration was performed in all patients. A total of 194 arterial pedicles were embolized in 177 procedures, with an additional 21 planned pedicle embolizations aborted for the following reasons: Inadequate catheter position (14 cases), patient inability to cooperate (3 cases), positive Wada test (3 cases), or iatrogenic vessel injury (1 case). Embolic material used in the completed procedures was Onyx (78% of cases), NBCA (56% of cases), and detachable coils (4% of cases). The median number of pedicles embolized was 3.5, with a minimum of 1 pedicle (in 8 patients) and a maximum of 11 pedicles (in 1 patient). In the course of treatment of these patients, a total of 13 complications were encountered, of which 4 were major and 9 were minor [Table 2].

Table 2

Details of the test cases

RESULTS

In the test group, independent variables found to be correlated with complications included number of pedicles (P = 0.01), eloquent location (P = 0.04), and small pedicle diameter (P = 0.05) [Table 3]. In univariate analysis, these factors were found to be robust predictors for complication incidence. Features of the Spetzler–Martin grading system other than eloquent location,[9] namely, deep venous drainage (P = 0.74) and AVM diameter (P = 0.43), were not found to be independently correlated with perioperative complications. Several other independent variables previously noted to be correlated with perioperative complications by other authors were not found to be correlated with perioperative complications in the test group and included patient age in years (P = 1.0), lack of neurologic deficits at baseline (P = 0.33), and preprocedural hemorrhage (P = 0.75). Of additional note, no significant correlation to perioperative complications and abnormal angioarchitectural structures was found (P = 0.66). This is summarized in Table 3. In multivariate analysis, number of pedicles and eloquent location were found to be significant prediction variables for complications in the test group, whereas vessel diameter was not [Table 4].

Table 3

Univariate analysis of case features comparing cases with and without complications

Table 4

Multivariate analysis of case features as correlates of procedural complications

The incidence of complications was tabulated for each AVM score [Table 5]. Complication rates were seen to increase with increasing grade in the proposed grading system (P<0.0001) but not the Spetzler–Martin system (P = 0.28). In the test cohort, the complication rate for each grade in the proposed scoring system is as follows: Grades 1 and 2, 0%; Grade 3, 14%; Grade 4, 50%; and Grade 5, 75%.

Table 5

Incidence of complications and endovascular cure tabulated according to arteriovenous malformation grading systems

Complete obliteration was accomplished in 5 patients (10%), with >90% obliteration in 17 patients (34%) and >75% obliteration in 21 patients (42%) [Table 5]. Fifty percent or more obliteration was obtained in all but eight cases. No obvious correlation between complete endovascular obliteration and Buffalo grade (P = 0.51) or Spetzler–Martin grade (P = 0.99) was evident. Definitive treatment included stereotactic radiosurgery in 27 patients and surgical resection in 14 patients. Four patients were not offered adjunctive therapies due to perceived futility.

DISCUSSION

The contemporary multimodality AVM treatment paradigm has resulted in new issues in operative management. In conjunction with surgical resection of an AVM, deep venous drainage, eloquence of surrounding tissue, and nidus size are of greatest import, and thus comprise the Spetzler–Martin grading system. In endovascular treatment, we have found the diameter of arterial pedicles, number of arterial pedicles, and eloquent nidus location to be of greatest significance for complications, and thus these variables comprise the proposed grading system. The proposed grading system is simple in design, based on angiographic and magnetic resonance features, and valuable in prediction of complications.

In this series, the value of the Spetzler–Martin grading system is reduplicated for AVM treatment, as the incidence of complications was greater in grades IV and V AVMs compared with lower-grade AVMs [Table 5]. However, a better correlation between grade and complications was noted in the proposed grading system. The utility of a new AVM grading system for endovascular treatment is in cases where the Spetzler–Martin grade is very high but the grade in the proposed grading system is low, or vice-versa. Two cases illustrate the utility of the proposed grading system. Case 29, with superficial drainage, intermediate nidus size (5 cm), and noneloquent location, is a Spetzler–Martin grade II AVM. However, with more than 5 mostly small arterial pedicles, it is a proposed grade IV AVM. A minor complication was encountered with incomplete embolization of this AVM. Thus, one could argue that with low risk for surgical resection (expected rate of minor deficit, 5%)[9] and relatively high risk with embolization (among the proposed grade IV patients in the present series: Incidence of major complication, 6%; incidence of any complication, 50%), surgical resection without embolization should have been performed preferentially. Conversely, Case 40 with noneloquent location, deep venous drainage, and diameter greater than 6 cm, is a Spetzler–Martin grade IV AVM. However, with only two large arterial pedicles, it is a proposed grade II AVM. This AVM was treated with two embolization procedures with complete obliteration without complication. Given its grade in the proposed system, the risk of neurological deficit or any complication is low (among the proposed grade II patients in the present series: Incidence of any complication, 0%) compared with the risk of surgery given its Spetzler–Martin grade (27% risk of any neurological deficit).[9]

The proposed grading system is based on the observed risk with endovascular treatment of AVMs. Simply stated, smaller vessels are more prone to injury with catheterization, a greater number of arterial pedicles produces more potential risk with each embolization, and eloquent location increases risk of neurological deficit. The actual size of the AVM nidus and venous drainage pattern of the AVM are of less importance during endovascular embolization but of chief importance during surgical resection. The proposed grading scheme provides a simple confluence of these concepts that will enable neurointerventionists to estimate the risk of endovascular AVM embolization.

Limitations

Limitations of the grading scheme include no prospective assessment to date and no evident correlation with complete endovascular obliteration. Of additional note, in multivariate analysis, the most relevant factor was number of arterial pedicles, whereas small pedicle diameter was not a significant factor in multivariate analysis (although it was in univariate analysis). We suspect this may be an artifact of small sample size (type II error), and we consider small vessel caliber as an important consideration when planning an embolization procedure as this feature when present typically results in more difficult microcatheterization. Furthermore, this experience is limited to a high-volume center and may not be accurate when applied universally. However, it should be noted that with the technology presently available for endovascular AVM embolization, the anticipated rate of complete obliteration is on the order of 10%.[12] Thus, a grading scheme to predict complete obliteration is of less importance. Prospective assessment of this grading scheme is planned.

CONCLUSION

The proposed grading system for endovascular treatment of cerebral AVMs is simple, easily reproduced, and clinically relevant.