A Proposed Grading Scale for Predicting Outcomes After Stereotactic Radiosurgery for Dural Arteriovenous Fistulas

Abstract

BACKGROUND

There are presently no grading scales that specifically address the outcomes of cranial dural arteriovenous fistula (dAVF) after stereotactic radiosurgery (SRS).

OBJECTIVE

To design a practical grading system that would predict outcomes after SRS for cranial dAVFs.

METHODS

From the International Radiosurgery Research Foundation (University of Pittsburgh [41 patients], University of Pennsylvania [6 patients], University of Sherbrooke [2 patients], University of Manitoba [1 patient], West Virginia University [2 patients], University of Puerto Rico [1 patient], Beaumont Health System 1 [patient], Na Homolce Hospital [13 patients], the University of Virginia [48 patients], and Yale University [6 patients]) centers, 120 patients with dAVF treated with SRS were included in the study. The factors predicting favorable outcome (obliteration without post-SRS hemorrhage) after SRS were assessed using logistic regression analysis. These factors were pooled with the factors that were found to be predictive of obliteration from 7 studies with 736 patients after a systematic review of literature. These were entered into stepwise multiple regression and the best-fit model was identified.

RESULTS

Based on the predictive model, 3 factors emerged to develop an SRS scoring system: cortical venous reflux (CVR), prior intracerebral hemorrhage (ICH), and noncavernous sinus location. Class I (score of 0-1 points) predicted the best favorable outcome of 80%. Class II patients (2 points score) had an intermediate favorable outcome of 57%, and class III (score 3 points) had the least favorable outcome at 37%. The ROC analysis showed better predictability to prevailing grading systems (AUC = 0.69; P = .04). Kaplan–Meier analysis showed statistically significant difference between the 3 subclasses of the proposed grading system for post-SRS dAVF obliteration (P = .001).

CONCLUSION

The proposed dAVF grading system incorporates angiographic, anatomic, and clinical parameters and improves the prediction of the outcomes following SRS for dAVF as compared to the existing scoring systems.

ABBREVIATIONS

AUC

area under the curve

AUROC

areas under the ROC

AVM

arteriovenous malformation

CCF

carotico-cavernous fistula

CI

confidence interval

CTA

computed tomography angiogram

CVR

cortical venous reflux

dAVF

dural arteriovenous fistula

DSA

digital subtraction angiography

GKRS

gamma knife radiosurgery

ICH

intracerebral hemorrhage

IRRF

international radiosurgery research foundation

MRA

magnetic resonance angiogram

MRI

magnetic resonance imaging

NHND

nonhemorrhagic neurological deficit

NS

not significant

OR

odds ratio

ROC

receiver operating curve

SRS

stereotactic radiosurgery

Current dural arteriovenous fistula (dAVF) grading scales have been useful in dAVF hemorrhage risk stratification that help providers make recommendations for best management.^1-3 These grading scales are based largely on dAVF angiographic features and presence of cortical venous reflux (CVR). Although endovascular embolization is often the first-line intervention for dAVFs, stereotactic radiosurgery (SRS) has become an important alternative, additive, or salvage intervention, especially for dAVFs that are surgically or endovascularly high risk, and in patients with significant medical comorbidities.^4,5 Despite the identification of several factors that predict dAVF obliteration, no grading scale that predict SRS outcomes for dAVF has been devised.⁶ The primary goal of the present study is to propose such a grading scale that is practical and can be easily applied in the clinical setting.

Objectives

The objectives of the current study were (1) to evaluate the factors that predict a favorable outcome of dAVF after SRS treatment and, based on these factors, design a grading system that would predict the outcomes after SRS treatment of dAVF; (2) to evaluate the predictability of existing scales for outcomes of dAVF treatment after SRS; and (3) to compare the proposed grading score with existing grading scores in predictability of outcomes after SRS.

METHODS

Study Design

This is an international multicenter retrospective study.

Setting and Participants

Patients who underwent SRS for dAVFs between 1988 and 2016, at 10 institutions that participate in the International Radiosurgery Research Foundation (University of Pittsburgh [41 patients], University of Pennsylvania [6 patients], University of Sherbrooke [2 patients], University of Manitoba [1 patient], West Virginia University [2 patients], University of Puerto Rico [1 patient], Beaumont Health System 1 [patient], Na Homolce Hospital [13 patients], the University of Virginia [48 patients], and Yale University [6 patients]), were identified and retrospectively reviewed under Institutional Review Board-approved protocols of each individual institution. The International Radiosurgery Research Foundation dAVF study methods have previously been described.⁷ Since the initial study, an additional institution contributed additional data to the database. The data from each individual institution were deidentified and pooled by an independent third party. Discrepancies and ambiguities in data were addressed by respective contributing institutions. The pooled data were subsequently sent to the institution of the first and senior authors for statistical analysis and drafting of the initial manuscript. Because this is a retrospective study, patient consent was not required.

The criteria for study inclusion were the following: (1) intracranial dAVF treated with SRS with sufficient baseline data to assess demographic information, fistula angioarchitecture, and radiosurgery parameters; (2) ≥6 mo of neuroimaging and clinical follow-up; and (3) SRS performed in a single stage.

Baseline Data and Variables

The baseline data comprised patient demographic, dAVF, and SRS variables, as described previously. Patient demographics included age, sex, prior dAVF treatment (radiotherapy, microsurgery, or endovascular embolization), and presenting symptoms. dAVF variables were presence of spinal venous drainage, associated aneurysms, presence of peri-dAVF edema on magnetic resonance imaging (MRI), multihole vs single hole (dAVF with single arteriovenous fistula are called single hole, and those with multiple arteriovenous fistulas seen on angiogram are called multihole dAVFs), presence of CVR, presence of venous ectasia, maximum diameter, and location. The Borden and Cognard grades were calculated for each dAVF.^1,2 dAVF obliteration was confirmed by angiogram study. Favorable outcome was defined as dAVF obliteration without the occurrence of post-SRS hemorrhage. If the primary treatment for dAVF given is SRS before endovascular or surgical management, it is referred to as “Upfront SRS.” The treatment volume was taken as a surrogate measure of dAVF size.

Radiosurgery Protocol

The SRS procedure has been previously described.⁸ SRS was performed using several models of the Leksell Gamma Knife unit (Elekta).⁴ The specific model use (including models U, B, C, 4C, Perfexion) varied by year and availability at each participating institution. Briefly, under local or monitored anesthesia, a Leksell Model G stereotactic frame (Elekta AB, Stockholm, Sweden) was affixed to the patient's calvarium. Digital subtraction angiography (DSA) and thin-slice MRI or computed tomography (CT) when MRI was contraindicated, with intravenous contrast, were performed for fistula definition and treatment planning. SRS treatment parameters included margin dose, maximum dose, isodose line, and number of isocenters.

Clinical and Neuroimaging Follow-up

Clinical and neuroimaging follow-ups were performed between 5 and 8 mo intervals for the first 2 yr after SRS, and yearly intervals thereafter. Patients were recommended to undergo DSA after MRI, magnetic resonance angiogram (MRA), or computed tomography angiogram (CTA) suggested dAVF obliteration to confirm the finding. Following dAVF obliteration, patients were recommended to undergo long-term follow-up neuroimaging with MRI at intervals of 1 to 5 yr for radiation-associated adverse effect surveillance. Favorable outcome was defined as dAVF obliteration without post-SRS hemorrhage.

Systematic Review of Literature

A systematic review of literature was performed in accordance to the PRISMA protocol from PubMed, Cochrane library, ScienceDirect, and Medline. A combination of keywords that included “Dural arteriovenous fistula,” “Gamma knife,” and “Radiosurgery” was used to build search strategy. An independent web search was also made. The references of the selected articles were scanned to include any skipped articles. Two investigators, Drs Mohammed and Hung, independently reviewed all the articles to reduce selection bias. Inclusion criteria were studies that had evaluated the outcomes of dAVF after SRS with 20 or more cases, and which had evaluated factors associated with post-SRS dAVF obliteration. Studies with less than 20 cases and those that had no description of factors associated with dAVF obliteration were excluded. The statistical methods used for the association in each study were noted (Figure 1).

FIGURE 1.

Study methodology and selection of cases.

Statistical Methods

Statistical analyses were performed using Stata/IC (StataCorp, version 15.1, College Station, Texas). Patients were dichotomized into those with and without favorable outcome. Continuous and categorical variables were presented as medians with ranges and percentages, respectively. Univariable predictors of favorable outcome were assessed using binary logistic regression, and those with P < .10 were entered into a backward stepwise multivariable regression model to identify independent predictors of favorable outcome after SRS. Pseudo R² values were compared between models, and the model with the highest value was selected as the final model.

To develop the grading system that predicts favorable outcome, points were assigned to independent predictors in the final model based on their regression weights. The patients were then scored based on the grading system. To compare the ability of the developed grading system to those of the Borden grade and the Cognard classification in predicting favorable outcome after SRS, receiver operating curves (ROC) were generated, and areas under the ROC (AUROC) were calculated and compared. An AUROC of 0.5 indicates no discrimination; whereas, an AUROC of 1.0 indicates perfect discrimination. AUCs were compared using the Delong method⁹ with STATA statistical software. Time-dependent analysis of dAVF obliteration was performed using Kaplan–Meier method and cohorts were compared using the log-rank test. All tests were 2-tailed, and P values of <.05 were considered significant. Missing data were not imputed.

RESULTS

Demographics and Clinical Presentation

A total of 148 cases treated between 1988 and 2016 from 10 international medical centers were identified. Among these, 15 patients with less than 6 mo of follow-up were excluded. A further 13 patients were excluded because of insufficient outcome data. A total of 120 patients with minimum of 6 mo of follow-up who were treated with single-session SRS were selected for the present study. The goal of treatment was complete obliteration of dAVF.

Pooled data were screened for errors and was analyzed. Any inadequacies were queried back to the centers and clarified. The following centers participated in this study: University of Pittsburgh (41 patients), University of Pennsylvania (6 patients), University of Sherbrooke (2 patients), University of Manitoba (1 patient), West Virginia University (2 patients), University of Puerto Rico (1 patient), Beaumont Health System 1 (patient), Na Homolce Hospital (13 patients), the University of Virginia (48 patients), and Yale University (6 patients).

The median age of the patients was 55 yr (range: 12-80 yr). There were 50 (42%) females, and 70 (58%) were males. The median follow-up was 43 mo (6-210 mo). Neurodeficits was noted in 28 (23%) of the cases. CVR was present in 60 (50%) of the patients. Cavernous location was seen in 19 (16%) of the cases. The median margin target dose was 21 Gy (9-33 Gy). The median size was 12.5 mm (0.2-6 mm). Prior endovascular treatment was performed in 51 cases. Eventual long-term radiation induced changes were noted in 4 patients. Post-SRS persistent symptoms caused by the dAVF were noted in 23 (19.5%) of the cases.

Factors Predicting Favorable Outcome

Complete dAVF obliteration was noted in 76 (64%) of the cases. Favorable outcome was seen in 73 (62%) of the patients. Four patients developed post-SRS hemorrhage. Three of these achieved eventual obliteration. The factors that predicted favorable outcome in univariable analysis were absence of CVR, absence of venous ectasia, no prior history of ICH, or absence of neurological deficits to dAVF, cavernous sinus, and/or tentorial location. Of these, cavernous location (P = .03, OR: 9.57, 95% CI: 1.17-78.1), absence of CVR (P = .01, OR: 3.11, 95% CI: 1.27-7.61), and prior history of ICH or presence of neurological deficits because of dAVF (P = .03, OR: 2.55, 95% CI: 1.08-6.01) were found to be statistically significant in multivariable analysis (Table 1).

TABLE 1.

Univariate and Multivariate Logistic Regression Analysis for Factors Predicting a Favorable Outcome After SRS for dAVF

	Univariate			Multivariate
Variable	P value	Odds ratio	95% CI	P value	Odds ratio	95% CI
Demographic features
Female sex	.06	2.095	0.958-4.580		NS
Age more than 65 yr	.40	1.441	0.605-3.431		-
Presenting symptoms
Asymptomatic	.25	0.483	0.139-1.685		-
No seizure on presentation	.06	3.838	0.908-16.22		NS
No Prior ICH or absence of neurodeficit due to dAVF	.03	2.389	1.081-5.278	.03	2.555	1.086-6.015
Headache	.58	0.812	0.387-1.702		-
Tinnitus	.38	2.502	1.052-5.954		-
Angiographic characteristics
Absence of CVR	.02	2.330	1.089-4.983	.01	3.113	1.273-7.614
Absence of venous ectasia	.04	2.504	1.018-6.162		NS
Absence of Spinal drainage	.16	1.937	0.758-4.952		-
Single hole	.48	1.318	0.606-2.867		-
Size	.65	0.873	0.478-1.595		-
Location
Cavernous location	.03	3.954	1.083-14.43	.03	9.573	1.173-78.11
Transverse sinus location	.80	1.111	0.493-2498		-
Torcular location	.38	0.588	0.178-1.947		-
Tentorial location	.08	0.329	0.145-0.745		NS
Anterior fossa	.69	0.761	0.193-2.993		-
Middle fossa	.16	0.292	0.051-1.661		-
Convexity	.58	1.594	0.296-8.578		-
Treatment related
Margin dose	.70	1.023	1.018-6.162		-
Maximum dose	.44	1.021	0.968-1.077		-
Prior endovascular treatment	.88	0.945	0.446-2.002		-
Upfront SRS	.86	0.936	0.447-1.958		-

Bold font denotes statistically significant factors.

Results of Literature Review

From a total of 1865 articles, 7 studies with 736 patients were finally selected. The factors that were found to positively effect post-SRS obliteration in these studies were the absence of CVR, target volume <1.5 cc, Cognard's type III and IV, ICH at presentation, cavernous location, longer imaging follow-up, older age, and duration of symptoms before SRS. These are summarized in Table 2. These actors were added to the predictive model along with the factors obtained by logistic regression on the multicenter database.

TABLE 2.

The Factors Associated With dAVF Obliteration After SRS in Various Studies

Study	No. of patients	Factors associated with obliteration	Factors not associated with obliteration	Statistical method of association
Hanakita et al,25 2012	22	No CVR Target vol <1.5 cc Cognard's type III or IV ICH at presentation	Age, sex, location, prior treatment	Multivariate cox proportional hazard model
Yang et al,13 2013	40	Cavernous location	NS	Fisher exact test
Söderman et al,32 2006	41	NS	Radiation dose	Mann–Whitney U test
Cifarelli et al,4 2010	55	No CVR	Sex, prior endovascular or surgical treatment, location, multihole	Chi-square test
Park et al,20 2016	30	Longer image follow-up No CVR	NS	Cox proportional hazard model
Wu et al,17 2006	227	Older age, duration of symptoms before SRS	Sex, diabetes mellitus, hypertension, CVR	Cox proportional hazard model
Pan et al,18 2013	321	Cavernous sinus	NS#	NS

^#NS: Not specified

Formulation of a Predictive Model

Logistic regression analysis was carried out on the multicenter database in univariable setting. The factors that were found to be significant in univariable analysis were then pooled with the factors shortlisted from literature review. Stepwise hierarchical multivariable regression analysis was then carried out on the multicenter database. The factors obtained from the systematic review of literature were pooled in the final predictive model. The complete model incorporating the 3 factors, namely the absence of CVR, Prior ICH or nonhemorrhagic neurological deficit (NHND), and cavernous location had the best pseudo R² value (Pseudo R² = 0.17). Other factors like venous ectasia, spinal drainage, age, and sex were removed from the predictive model to get the best pseudo R² value.

Assignment of Scores and Classification into Subgroups

The presence of CVR, history of prior ICH or NHND, and noncavernous location were assigned 1 point each. The patients were then scored based on the point system developed. The score ranged from a lowest of zero to a highest of 3 points. A lower score was reflective of a more favorable outcome. Patients were then classified into 3 grades. Those scoring 0 to 1 point were included in grade I, those with 2 points in grade II, and those scoring 3 points were included in grade III. (Table 3 and Figure 2).

TABLE 3.

The Scoring System With Assignment of Points

Characteristic	Presence or absence	Points given
Cortical venous reflux	Absent	0
	Present	1
Prior ICH or neurological deficit due to dAVF	Absent	0
	Present	1
Location	Cavernous	0
	Noncavernous	1
Grade I	0-1	80%	63%	84%
Grade II	2	57%	47%	79%
Grade III	3	37%	29%	35%

FIGURE 2.

Proposed grading system with details of score assignment.

ROC Analysis and Area under the Curve

Receiver operating characteristic analysis showed area under the curve of 0.69 for the proposed scoring system. The ROC analysis was then performed for Borden's, Cognard's, and Zipfel's grading systems and compared with the AUC of the proposed grading system. The AUC for Borden's, Cognard's, and Zipfel's grades were 0.58, 0.55, and 0.57, respectively. Each of the Borden's, Cognard's, and Zipfel's grading showed a smaller AUC than the proposed grading system (P = .04, Figure 3).

FIGURE 3.

Receiver operating curve analysis predicting favorable outcome in the present grading score, Cognard's and Borden's grade.

Obliteration Rates Between the Subgroups

Kaplan–Meier analysis was done for obliteration of dAVF with the 3 subclasses of the proposed grading score. There was a statistically significant difference between the obliteration rates (P = .001) in the different subclasses of the proposed grading score (Figure 4). The 3-yr and 5-yr obliteration rates in grade I were 63% and 84%, respectively. In grade II, were 47% and 79%, respectively, and in grade III, they were 29% and 35%, respectively. The difference in outcomes in the subgroups were linear in the proposed grading system (Figure 5A). In Borden's grade, the outcomes between grade II (50%) and III (55%) were very similar (Figure 5B), with grade III having a better outcome than grade II. There was no linear relationship for outcomes and subgroups in Cognard's and Zipfel's grading systems. (Figure 5C and 5D).

FIGURE 4.

Kaplan–Meier analysis showing the difference in the dAVF obliteration rates in the radiosurgery-based scoring for grades I, II, and III.

FIGURE 5.

Graphical representation of percentage of patients with favorable outcome in, A, the proposed grading system, B, the Borden's grading, C, the Cognard's grading, D, the Zipfel's grading systems.

DISCUSSION

Borden and Shucart¹ in their paper in 1995 proposed a classification based on endovascular drainage patterns of the dAVF on a series of 14 patients treated primarily by surgical excision. Although not originally intended by the authors, Borden's grade has also proven its value in both risk stratification and prediction of outcomes after SRS treatment of dAVFs. The Borden's grade does not take into consideration any of the radiosurgical and clinical parameters that may influence the outcome after SRS. Its predictability and applicability to radiosurgery series is, perhaps, surprising considering that it is based exclusively on the angiographic features of venous drainage and not on other factors frequently impacting radiosurgical delivery and outcomes.

Cognard et al² analyzed 205 patients with dAVFs and proposed a classification that introduced venous ectasia and spinal perimedullary drainage into the grading system. The Cognard's grade is more cumbersome and does not have a smooth gradation from low risk or most favorable outcome category to the high risk or worst outcome category. Zipfel et al³ proposed a clinically relevant modified classification based on the Borden's grading that subdivided those with CVR into symptomatic and asymptomatic categories. Symptomatic CVR was defined by them as presentation with ICH or NHND. The grading by Zipfel et al³ improves the risk stratification and its application to radiosurgery series may only enhance risk stratification, but not outcome predictability. Our proposed grading system incorporates clinical and anatomic characteristics, thereby increasing the predictability of the grading score as shown by the ROC analysis.

Several studies have shown that low-grade dAVFs of cavernous sinus location respond promptly and favorably to SRS than in noncavernous location.^4,10-13 A literature review suggests a dAVF obliteration rate in cavernous sinus to range between 75% and 93%, which is higher than that of noncavernous dAVFs.^10,12,14-18 In our multicenter database, cavernous location of dAVF was present in 19 cases, and 16 patients (84%) achieved obliteration. In a large study of 321 patients, Pan et al¹⁸ found the obliteration rates of cavernous dAVF was better (70%) as compared to noncavernous dAVF (59%). The studies^19,20 that have reported less favorable outcomes of cavernous dAVF have included small number of patients. The lower number of cases may reduce the statistical significance and may give rise to nonsignificant results in these studies. Larger series^13,17,18,21 have all shown better obliteration rates of cavernous dAVFs.

The absence of CVR has been shown to have positive effect on dAVF obliteration post-SRS by several studies.^4,5,20,22,23 The exact reason for this is not clear. The role of vessel diameter in the outcome of SRS for arteriovenous malformations (AVMs) has been studied. In AVM, radiosurgery causes endothelial damage causing progressive stenosis of the vessels by smooth muscle and collagen proliferation of the intimal layer leading to obliteration of the vessels.²⁴ The fistulous vessels in dAVF are considered to be normal anastomotic channels, but the processes, like thrombosis, endothelial damage, and intimal reaction, may be common to both diseases.⁵ It can be hypothesized that the dAVF, which have CVR, may have higher flow rates and larger pressures with larger diameter of the fistulous vessels. Larger vessel sizes may not achieve sufficient obliteration by intimal changes after radiosurgery. This is corroborated by studies on AVM. In dAVF, CVR may represent a surrogate marker of fistulous vessel diameter and hence be predictive of outcomes after SRS.

Presence of NHND or prior ICH has been described earlier as an important consideration for treating dAVF with SRS.^3,25 CVR is linked to the risk of hemorrhage and NHND because of dAVF.^26,27 Soderman et al²⁷ reported an annual risk of hemorrhage of 1.5% in unruptured dAVF and a 7.5% annual risk of hemorrhage in those with prior history of bleed from dAVF. Thus, history of prior ICH because of dAVF is reflective of hemodynamic changes that carry a future risk of bleeding.

Current Role of SRS in Management of dAVF

Endovascular management is the preferred treatment in most cases of cranial dAVFs.^28,29 Endovascular and surgical management achieve an immediate reduction in risk of hemorrhage or NHND with successful obliteration. Whereas, SRS requires a latent period during which the risk of hemorrhage remains. SRS is preferred in instances where endovascular approach may be limited because of vessel tortuosity, presence of peripheral vascular disease or renal impairment, or when surgical management may be difficult in deep seated locations, or because of medical comorbidities.³⁰ It is also considered in residual dAVF or recurrent cases.^16,31 Multimodal management of dAVF is now a recognized treatment strategy.^6,32 Gamma knife radiosurgery (GKRS) remains as a valuable adjunct and a second-line treatment in the management of dAVF.⁴

Comparison Between the Proposed Grading System and Other Prevalent Scoring Systems

The proposed grading was developed on a database of patients treated with SRS unlike the other grading systems. The factors peculiar to SRS were included in the proposed grade. Including the clinical, angiographic, and anatomic characteristics improves the predictability of the proposed grade unlike the other grading systems that incompletely incorporate these parameters. The ROC analysis yielded better predictability of the proposed grade in caparison with the other grades. The proposed grading could help clinicians to predict post-SRS outcomes on a case-by-case basis and aid the clinical decision-making process.

Limitations of the Study

The low number of cases in the derivation cohort is an important limitation of this study. Increasing the number of cases in the derivation cohort could have yielded more factors. We tried to circumvent this by including factors found in previous other SRS series after a systematic review of literature into the predictive model. Because of the rarity of SRS cases on dAVF, the validation of proposed grading system could not be performed. Splitting the multicenter database into derivation and validation cohorts would have further reduced the number of cases in the derivation cohort reducing the statistical significance. Short duration of follow-up is another important limitation.

CONCLUSION

The proposed dAVF grading system incorporates angiographic, anatomic, and clinical parameters and improves the prediction of the outcomes following SRS for dAVF as compared to the existing scoring systems. Further study, refinement, and validation of the currently proposed system will be needed.

Disclosures

Dr Kano reports he has an Elekta Research Grant. Dr Grills has stock ownership and serves on the Board of Directors in a company called Greater Michigan Gamma Knife. In addition, Dr Lunsford is a shareholder in Elekta AB. The other authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article. Dr Vargo reports he is a speaking honorarium for Brainlab.

COMMENTS

The authors devised a new grading scale to assess obliteration rates after SRS for patients with dAVF. This is a more comprehensive scale than others that exist. The likelihood of dAVF obliteration varied depending on the combined score used, with higher scores correlating with a lower chance of obliteration and worse clinical outcomes. Key features that made up the score included prior ICH, worse pre-treatment neurological status, cortical venous reflux, and location outside of the cavernous sinus.

Neurosurgeons who treat patients with dAVF should use this score as their experience grows so that over time the results of the authors can be confirmed (or not). If they probably will, given the data presented in this paper, the question will remain: how will the dAVF score influence treatment? Will patients with higher scores not be offered SRS, even if other treatments seem prohibitively risky?

Michael Schulder

Manhasset, New York

In the present manuscript, the authors propose a novel classification for dAVFs for prediction of improved outcomes after stereotactic radiosurgery (SRS). In reviewing a multi-institutional cohort, as well as the published literature, they proposed the following features to include in their grading scheme: presence of cortical venous reflux, prior ICH or neurological symptoms, and non-cavernous sinus location. Higher scores correlated with lower rates of obliteration on long-term follow up after SRS and significantly worse outcomes on Kaplan-Meier analysis. Further, the overall accuracy of the grading scale was found to be superior to existing scales with regard to predicting outcome after SRS.

With the emphasis today on value-based care, it is imperative to better understand how specific patients will respond to various treatment modalities. The goal of personalized medicine is to match a specific patient's pathology with a customized treatment regimen, both to improve patient outcomes and limit unnecessary costs of healthcare. To improve prediction of outcomes after SRS for dAVF, the authors have taken a comprehensive approach to risk factor identification. As a result, we as a neurosurgical community may better make evidence-based decisions to help guide treatment planning for this complex disease.

Kurt Yaeger

J. Mocco

New York, New York