Xiaochuan Huo, PhD, Yuhua Jiang, PhD, Xianli Lv, PhD, Hongchao Yang, MD, Yang Zhao, MD, and

Transcription

1 clinical article J Neurosurg 124: , 2016 Gamma Knife surgical treatment for partially embolized cerebral arteriovenous malformations Xiaochuan Huo, PhD, Yuhua Jiang, PhD, Xianli Lv, PhD, Hongchao Yang, MD, Yang Zhao, MD, and Youxiang Li, PhD Department of Interventional Neuroradiology, Beijing Neurosurgical Institute and Beijing Tiantan Hospital, Capital Medical University, Beijing, China Objective A combination of embolization and radiosurgery is used as a common strategy for the treatment of large and complex cerebral arteriovenous malformations (AVMs). This study presents the experiences of partially embolized cerebral AVMs followed by Gamma Knife surgery (GKS) and assesses predictive factors for AVM obliteration and hemorrhage. Methods The interventional neuroradiology database that was reviewed included 404 patients who underwent AVM embolization. Using this database, the authors retrospectively analyzed all partially embolized AVM cases followed by GKS for a residual nidus. Except for cases of complete AVM obliteration, the authors excluded all patients with radiological follow-up of less than 2 years. Logistic regression analysis was used to analyze the predictive factors related to AVM obliteration and hemorrhage following GKS. Kaplan-Meier analysis was used to evaluate the obliteration with a cutoff AVM nidus volume of 3 cm 3 and 10 cm 3. Results One hundred sixty-two patients qualified for the study. The median patient age was 26 years and 48.8% were female. Hemorrhage presented as the most common symptom (48.1%). The median preembolization volume of an AVM was 14.3 cm 3. The median volume and margin dose for GKS were cm 3 and 16.0 Gy, respectively. The median radiological and clinical follow-up intervals were 47 and 79 months, respectively. The annual hemorrhage rate was 1.71% and total obliteration rate was 56.8%. Noneloquent area (p = 0.004), superficial location (p < 0.001), decreased volume (p < 0.001), lower Spetzler-Martin grade (p < 0.001), lower Virginia Radiosurgery AVM Scale (RAS; p < 0.001), lower Pollock-Flickinger score (p < 0.001), lower modified Pollock-Flickinger score (p < 0.001), increased maximum dose (p < 0.001), and increased margin dose (p < 0.001) were found to be statistically significant in predicting the probability of AVM obliteration in the univariate analysis. In the multivariate analysis, only volume (p = 0.016) was found to be an independent prognostic factor for AVM obliteration. The log-rank (Mantel-Cox) test of the Kaplan-Meier analysis (chisquare = , p < 0.001) showed a significantly decreased obliteration rate of different cutoff AVM volume groups of less than 3 cm 3, 3 10 cm 3, and more than 10 cm 3. No independent prognostic factor was found for AVM hemorrhage in multivariate analysis. Conclusions Partially embolized AVMs are amenable to successful treatment with GKS. The volume of the nidus significantly influences the outcome of radiosurgical treatment. The Virginia RAS and Pollock-Flickinger score were found to be reliable scoring systems for selection of patient candidates and prediction of partially embolized AVM closure and complications for GKS. Key Words cerebral; arteriovenous malformation; endovascular embolization; Gamma Knife surgery; stereotactic radiosurgery; vascular disorders Arteriovenous malformations (AVMs) are abnormalities of the intracranial vessels that constitute a connection between the arterial and venous systems and lack an intervening capillary bed. 7,19,22,25,30,34 According to reports, 0.1% of the population harbors an AVM, 3,8 and both sexes are equally affected. AVMs are the leading cause of nontraumatic intracerebral hemorrhage in people less than 35 years old. 37 Most lesions come to patients attention in their 40s and 75% of the hemorrhagic presentations occur before the age of 50 years. 7 In addition, AVMs can cause debilitating headaches, seizures, and progressive neurological deficits due to ischemia of the surrounding Abbreviations AVM = arteriovenous malformation; DSA = digital subtraction angiography; GKS = Gamma Knife surgery; HR = hazard ratio; NBCA = N-butyl cyanoacrylate; RAS = Radiosurgery AVM Scale; RIC = radiation-induced change; SRS = stereotactic radiosurgery. submitted November 27, accepted January 21, include when citing Published online August 7, 2015; DOI: / JNS AANS, 2016 J Neurosurg Volume 124 March

2 X. Huo et al. brain tissue, 22,28 with a significant effect on quality of life and working activity of individual patients. 4,29 Because AVMs are rare cerebral vascular diseases, prospective multicenter studies to determine the natural history of long-term treatment results are rare. Consequently, AVMs continue to be a significant challenge in the cerebrovascular community, and opinions differ concerning the optimal treatment of these lesions. 9,13,40,41 The treatment of cerebral AVMs requires a multidisciplinary approach that includes microsurgery, endovascular embolization, and stereotactic radiosurgery (SRS). The first case of AVM embolization was described by Luessenhop and Spence in 1960, which was achieved by indiscriminately injecting silicone microspheres directly into the carotid artery. 27 With the advancement of new materials, endovascular embolization has become an important method for treating AVMs. Symptoms of vascular steal phenomena, venous hypertension, and seizures may benefit from endovascular embolization. 14,23,26,31 In the last several decades, Gamma Knife surgery (GKS) has been demonstrated to be a safe and highly effective treatment for AVMs; the reported obliteration rate of the AVM nidus varies from 70% to 94% using this method. 11,16,18,24,35 For large and complex AVMs, when microsurgery is considered too risky, a combination of embolization and SRS is used as a routine strategy and may render an originally untreatable AVM in a lesion potentially curable. 1,6,45 In this study, we present our combined treatment experience with a large cohort of patients harboring partially embolized AVMs followed by GKS and assess the predictive factors for AVM obliteration and hemorrhage after GKS. Grading systems for radiosurgery were also evaluated by our series. Methods Patient Population Between May 2002 and March 2012, 404 consecutive patients with cerebral AVMs underwent coil embolization at Beijing Neurosurgical Institute and Tiantan Hospital. Of these 404 patients, 180 cases of partial embolization were further treated with GKS at the Beijing Gamma Knife Radiosurgery Center. A retrospective review of these collected databases was performed. Except for cases of completely obliterated AVMs, we excluded all patients with radiological follow-up of less than 2 years. One hundred sixty-two (90%) of the 180 patients underwent follow-up. The study protocol was approved by the Institutional Review Board at Beijing Tiantan Hospital. After hospitalization, all patients in our department sign an informed consent form with the understanding that the patient s clinical and radiological data may be used for scientific research or for publication; the privacy of patients was strictly protected. Treatment Strategy for AVMs For ruptured AVMs, the choice of treatment was based on the patient s anatomy and condition. Surgery and embolization were the first options because they are associated with an immediate result. When embolization was chosen as the first-line treatment, the first embolization session was usually performed at least 1 month after the bleeding event. For ruptured AVMs, GKS was used in selected patients as a complementary treatment in cases of incomplete treatment with surgery or embolization, or as firstline treatment if the AVM was untreatable by surgery or embolization. For unruptured AVMs, indications for treatment were based on the balance between natural history of hemorrhage and the risks associated with the treatment. Embolization was often used as the first-line treatment followed by GKS in cases of incomplete embolization. Endovascular Embolization Endovascular treatment was performed under general anesthesia. The femoral artery was punctured and a 6-Fr femoral artery sheath was placed. According to the AVM s location, a 6-Fr guiding catheter was inserted into the responsible artery. A flow-directed microcatheter (Marathon; ev3) was navigated through a feeding artery into the AVM nidus under roadmap guidance. Superselective angiography was performed through a microcatheter to analyze the structure of the AVM nidus and to ensure the suitable position of the microcatheter. When intranidal aneurysms or hemodynamic aneurysms were found by superselective angiography, they were embolized during the procedure. We usually use Onyx (Covidien) as the embolic agent. However, in cases with high-flow fistulas or short-feeding pedicles, coils or N-butyl cyanoacrylate (NBCA) was used for embolization. When Onyx was used, we first used normal saline to flush the microcatheter, and then filled the empty space with 0.25 ml of dimethyl sulfoxide. Then we slowly inject Onyx under a blank roadmap. When we observed the Onyx reflux or passage in the draining vein, the reflux-hold-reinjection technique was used. At the end of the treatment, the microcatheter was gently removed. Embolization was stopped when it was impossible to get the microcatheter into the AVM or inject Onyx into the nidus. Gamma Knife Surgery GKS treatment was performed using the Leksell Gamma Knife Perfexion machine (Elekta). The radiosurgical procedure began with the positioning of the Leksell stereotactic helmet (G model, Elekta Instrument AB). The patient was then transferred to neuroradiology for brain MRI (1.5-T Magneton Vision, Siemens). The acquired images were transferred to the workstation (Leksell Gamma- Plan) for the definition of the target. The treatment target (residual nidus) was defined by 3D stereotactic MRI and digital subtraction angiography (DSA). Using a combination of DSA and MRI, the nidus was delineated and treated with radiosurgery. The embolized nidus was usually not included in the radiation targeting; however, the origin of the draining veins was usually included in the radiation targeting. Each patient s neuroimaging studies were evaluated by an attending neurosurgeon and neuroradiologist at Beijing Tiantan hospital (X.H. and Y.L.). If a patent residual nidus was demonstrated 3 years after GKS, a second GKS treatment was considered. Follow-up Protocol Patients underwent MRI every 6 months after GKS. Control DSA was performed 2 years after GKS but could 768 J Neurosurg Volume 124 March 2016

3 Combined treatment for cerebral AVMs sometimes be conducted earlier or delayed depending on the results of MRI. The MRI and DSA studies were performed at a combination of facilities, including those at the interventional neuroradiology department of Tiantan Hospital, Gamma Knife Center of the Beijing Neurosurgical Institute, and outside hospitals. All imaging studies were reviewed by 3 attending neurosurgeons and an attending neuroradiologist at Beijing Neurosurgical Institute (X.H., Y.J., X.L., and Y.L.). Computed tomography or MRI was performed for neurological deterioration. Hemorrhage was defined by imaging studies. For the annual hemorrhage rate, the number of risk years was defined as the total time across all patients from radiosurgery to either AVM obliteration or to the last clinical follow-up evaluation in patients without hemorrhage symptoms. Patients were evaluated by DSA to confirm AVM obliteration only after MRI demonstrated the absence of a residual lesion. AVM obliteration was defined by the absence of abnormal arteriovenous shunting on DSA. Radiation-induced changes (RICs) were identified as T2-weighted perinidal hyperintensities on postradiosurgical MRI. These changes were considered to be symptomatic if correlated with clinical presentations such as headache and seizure. Clinical status before and after treatment, and at the time of follow-up, was evaluated using the modified Rankin Scale scoring system. AVM closure was censored at the date of DSA demonstrating no residual nidus; AVM obliteration time was calculated from the date of GKS to the date of AVM closure. For annual hemorrhage rate, the number of risk years was defined as the total time across all patients from radiosurgery to either AVM obliteration or to last clinical follow-up evaluation in patients without hemorrhagic symptoms. Statistical Analysis We performed the statistical analysis using SPSS (version 17.0, SPSS Inc.). The descriptive data analysis was conducted using a t-test for continuous variables and the chi-square test for categorical variables. A Kaplan-Meier univariate analysis was performed to analyze AVM closure at follow-up and to find a cutoff volume for obliteration. Predictive factors that were analyzed included the following: age, sex, hemorrhage history, AVM volume, AVM location, draining vein location, number of veins, maximum and margin GKS dose, presence of RICs, Spetzler- Martin grade, Pollock-Flickinger score, modified Pollock- Flickinger score, and Virginia Radiosurgery AVM Scale (RAS) score. The predictive factors listed above were first analyzed by univariate analysis and then by logistic regression analysis. The hazard ratio (HR), 95% confidence interval (CI), and p value were determined for each predictive factor. All statistical analyses were 2-sided. An HR with a 95% CI not including 1.0, and a p value < 0.05, was defined as statistically significant. Results Patient Population One hundred sixty-two patients treated with a combination of embolization and GKS qualified for the study. We analyzed 79 females (48.8%) and 83 males (51.2%), and the median age was 26 years (Table 1). Thirty-two patients (19.8%) were less than 18 years old. Hemorrhage presented as the most common clinical symptom in 78 cases (48.1%); other symptoms included headache in 49 cases (30.2%), seizure in 32 cases (19.8%), and neurological deficit in 15 cases (9.3%). Endovascular Embolization and GKS Treatment Parameters The parameters for endovascular procedures before GKS were as follows: median nidus volume 14.3 cm 3 ; median Onyx volume 1.8 ml; embolization < 50% for 76 cases (46.9%), 50% 90% for 80 cases (49.4%), and > 90% for 6 cases (3.7%); embolization-related complications in 24 cases (14.8%); and median duration before GKS of 3 months. Patients underwent embolization once in 124 cases, twice in 40 cases, 3 times in 12 cases, and 4 times in 6 cases. The time interval between embolization and GKS ranged from 1 day to 1236 days, with a mean of 151 days, and a median of 99 days. The median GKS parameters were as follows: residual nidus volume 9.01 cm 3, margin dose 16 Gy, and maximum dose 33.3 Gy. There was a deep AVM location (basal ganglia, thalamus, or brainstem) in 86 cases (53.1%) and a superficial location in 76 cases (46.9%); and a noneloquent area in 53 cases (32.7%) and an eloquent area in 109 cases (67.3%). Superficial venous drainage occurred in 85 cases (52.5%) and deep venous drainage in 77 cases (47.5%). There were multiple draining veins for 95 cases (58.6%) and a single draining vein for 67 cases (41.4%). The Spetzler-Martin grade was I in 20 cases (12.3%), II in 47 cases (29%), III in 50 cases (30.9%), and IV in 45 cases (27.8%). The Virginia RAS score which is calculated by preradiosurgical hemorrhage history, AVM nidus volume, and AVM eloquence location was 0 or 1 in 18 cases (11.1%), 2 in 47 cases (29%), 3 in 63 cases (38.9%), and 4 in 34 cases (21%). The Pollock-Flickinger score which is based on age, AVM nidus volume, and location was a mean of 1.84, median of 1.55, and range of The modified Pollock-Flickinger score was a mean of 1.61, median of 1.27, and range of The details of AVM and radiosurgical management are shown in Table 1. Radiological Outcomes Following GKS The mean duration of radiological follow-up was 48 months (4 years) with a median of 47 months (3.9 years), and a range of months ( years). Every AVM nidus obliteration was confirmed by DSA, and the total AVM obliteration rate was 56.8% (92 cases) at follow-up (Fig. 1 upper). A cutoff of 3 cm 3 and 10 cm 3 in AVM nidus volume before GKS was identified using the most informative cutoff methodology from our series. According to the literature and the findings of our series, there were 3 volume ranges of the AVM nidus relevant for radiosurgery: < 3 cm 3, 3 10 cm 3, and > 10 cm 3. Patients with AVM volumes in different groups had a statistically significant different rate of AVM obliteration according to the logrank (Mantel-Cox) test of the Kaplan-Meier analysis (chisquare = , p < 0.001; Fig. 1 lower). The oblitera- J Neurosurg Volume 124 March

4 X. Huo et al. TABLE 1. Pre-GKS patient and AVM characteristics in 162 partially embolized cases Characteristic Value Sex Male 83 Female 79 Age (yrs) Mean Median Range 8 51 More than 1 embolization 38 Location Superficial 76 Deep* 86 Venous drainage pattern Superficial 85 Deep 77 No. of draining veins Single 67 Multiple 95 Max diameter (cm) Mean 3.19 Median 3.15 Range Residual nidus volume (cm 3 ) Mean Median 9.01 Range Maximum dose (Gy) Mean Median Range Margin dose (Gy) Mean 16.3 Median 16 Range Spetzler-Martin grade I 20 II 47 III 50 IV 45 Pollock-Flickinger score < > Mean 1.84 Median 1.55 Range (continued) TABLE 1. Pre-GKS patient and AVM characteristics in 162 partially embolized cases (continued) Characteristic Value Modified Pollock-Flickinger score <1 (%) (%) (%) 28 >2.00 (%) 39 Mean 1.61 Median 1.27 Range Virginia RAS score * Deep location includes thalamus, basal ganglia, and brainstem. tion rates in the < 3 cm 3, 3 10 cm 3, and > 10 cm 3 volume groups were 94.12%, 61.29%, and 33.33%, respectively. Predictors of AVM Obliteration By univariate regression analysis, we found that noneloquent area, superficial location, decreased volume, lower Spetzler-Martin grade, lower Virginia RAS score, lower Pollock-Flickinger score, lower modified Pollock- Flickinger score, increased maximum dose, and increased margin dose were statistically significant in predicting the probability of AVM obliteration. According to multivariate analysis, only decreased AVM volume (p = 0.005) was revealed to be the factor independently associated with AVM obliteration. The univariate and multivariate regression analyses of predictors of obliteration are detailed in Table 2. Predictors for Post-GKS Hemorrhage Over a total of risk years, 18 hemorrhages occurred in 18 patients after GKS, and the annual postradiosurgical hemorrhage rate was 1.81%. Five patients died of hemorrhage, and the annual mortality rate was 0.5%. Seven patients (4.3%) experienced permanent clinical deterioration (5 deaths, 2 with limb numbness) at follow-up because of post-gks hemorrhage. No hemorrhage happened after complete obliteration. Deep location, aneurysm, increased volume, decreased maximum dose, decreased margin dose, higher Virginia RAS score, higher Pollock- Flickinger score, and higher modified Pollock-Flickinger score were associated with postradiosurgical hemorrhage based on univariate logistic regression analysis. However, no predictor was independently associated with post-gks hemorrhage after multivariate analysis. The details of the logistic regression analyses are shown in Table 3. Clinical Outcomes Following GKS The mean clinical follow-up duration was 74.8 months 770 J Neurosurg Volume 124 March 2016

5 Combined treatment for cerebral AVMs (6.2 years), with a median of 79 months (6.6 years), and a range of months (1 9.1 years). For cases with seizure symptoms, the frequency decreased in 10 cases, and 19 cases were seizure-free after GKS. In 4 cases (2.5%), the frequency of seizures increased. There were no newonset seizures after GKS. In 55 cases (34%), clinical improvement occurred following GKS. There were 5 cases (3.1%) with permanent clinical deterioration combined with 25 cases (15.4%) with transient clinical deterioration. The rate of both temporary and permanent clinical morbidity was 18.5% following GKS. Radiosurgical Grading Systems for Obliteration Rate and Complication Events The Pollock-Flickinger, modified Pollock-Flickinger, and Virginia RAS scores were divided into subgroups according to different score values. The obliteration rates according to Pollock-Flickinger scores of < 1, , , and > 2.00 were 75.0%, 68.2%, 64.9%, and 28.6%, respectively. The obliteration rates according to these same score ranges for the modified Pollock-Flickinger score were 80.8%, 51.2%, 64.3%, and 25.6%, respectively. The obliteration rates according to Virginia RAS scores of 1, 2, 3, and 4 were 88.9%, 68.1%, 50.8%, and 35.3%, respectively. These grading systems could be reliable scoring systems for the prediction of partially embolized AVM closure after GKS. Table 4 details the associations between radiosurgical grading systems and AVM obliteration. Post-GKS complications including hemorrhage, seizure, and headache were assessed to evaluate the effectiveness of radiosurgical grading systems in predicting complications after radiosurgery. The percentage of complications according to Pollock-Flickinger scores of < 1, , , and > 2.00 were 12.5%, 4.5%, 13.5%, and 38.8%, respectively. The percentage of complications according to modified Pollock-Flickinger scores of < 1, , , and > 2.00 were 11.5%, 7.0%, 14.3%, and 43.6%, respectively. The complication rates of Virginia RAS scores 0 2, 3, and 4 were 7.7%, 23.8%, and 29.4%, respectively. Table 5 details the associations between radiosurgical grading systems and post-gks complications. Increasing Virginia RAS score resulted in a gradually increased complication rate. Discussion The goal of treating AVMs with SRS is to reduce and ultimately, with obliteration, eliminate the future hemorrhage risk. After radiosurgery, the mechanisms of AVM obliteration include progressive intimal thickening, thrombosis of irradiated vessels, and eventual occlusion of the vascular lumen. 38 This result can usually be obtained after a period of 2 3 years after radiosurgery, although some AVMs may require a latency of 3 5 years after treatment to obtain a complete closure. 11 Some studies show a radiosurgical obliteration rate of 70% 94%. 11,17,43 In our series, the obliteration rate was 56.8%, lower than most reported radiosurgical series. 11,17,43 Through a matched case-control approach, Kano et al. 20 found that patients with pre-srs embolization had a lower obliteration rate than patients who had not undergone embolization. All the AVMs in Fig. 1. Graphs of Kaplan-Meier estimates demonstrating the time to AVM obliteration defined on angiographic study. Upper: Kaplan-Meier estimate of time to AVM obliteration, showing the median obliteration time in this series was 65.3 months. The table row beneath the x-axis shows the number of patients with obliteration at each time point on the axis. Lower: Kaplan-Meier estimate for those with pre-gks AVM volume less than 3 cm 3, 3 10 cm 3, and greater than 10 cm 3. The median obliteration times for 3 groups were 25.8 months, 65.3 months, and 73.7 months, respectively. The log-rank test revealed that there was a significant difference among 3 curves, and the obliteration rate of the group with an AVM volume less than 3 cm 3 was higher than the other 2 groups (chi-square = , p < 0.001). The table beneath the x-axis shows the number of patients with obliteration at each time point on the axis for the 3 groups. Figure is available in color online only. our group are partially embolized before GKS. Our results showed that pre-gks partial embolization decreased the rate of obliteration of AVMs after radiosurgery. However, patients in our group included more AVMs with maximum diameters larger than 3 cm, which are those with a lower probability of closure. J Neurosurg Volume 124 March

6 X. Huo et al. TABLE 2. Factors predicting AVM obliteration after GKS Factor Univariate Analysis Multivariate Analysis HR 95% CI p Value HR 95% CI p Value Female sex Increased age Previous hemorrhage Single embolization pre-gks No RICs Noneloquent area * Superficial location < 0.001* Single draining vein No deep drainage Decreased volume < 0.001* * Increased maximum dose < 0.001* Increased margin dose < 0.001* Lower Spetzler-Martin grade < 0.001* Lower Pollock-Flickinger score < 0.001* Lower modified Pollock-Flickinger score < 0.001* Lower Virginia RAS score < 0.001* * Statistically significant (p <0.05). In our study, multivariate analysis indicated that only decreased AVM volume (p = 0.005) was independently associated with obliteration. This is understandable because radiation dose depends mainly on nidus volume, provided that radiation dose follows international standards. We further identified a cutoff nidus volume separating obliterated from nonobliterated cases by our experience and previously published studies. An AVM nidus < 3 cm 3 responds with a high obliteration rate to single-stage radiosurgery, whereas an AVM nidus > 10 cm 3 is significantly less likely to respond with obliteration, and the obliteration rate of an AVM nidus of 3 10 cm 3 is somewhere in between. This should be kept in mind when selecting patients with AVMs for radiosurgery. A smaller nidus volume allows for higher radiation dosage and could increase the obliteration rate with fewer radiation complications. However, although we should decrease the AVM volume as much as we can when it does TABLE 3. Factors predicting hemorrhage after GKS Factor Univariate Analysis Multivariate Analysis HR 95% CI p Value HR 95% CI p Value Female sex Decreased age Previous hemorrhage Multi-embolization pre-gks Eloquent area Deep location * Single draining vein Deep drainage Aneurysm * Increased volume < 0.001* Decreased maximum dose * Decreased margin dose * Higher Spetzler-Martin grade Higher Pollock-Flickinger score * Higher modified Pollock-Flickinger score * Higher Virginia RAS score * * Statistically significant (p < 0.05). 772 J Neurosurg Volume 124 March 2016

7 Combined treatment for cerebral AVMs TABLE 4. Association between radiosurgical grading systems and AVM obliteration Grading System Obliteration (%) OR 95% CI p Value Pollock-Flickinger score <1 24/32 (75.0) /44 (68.2) /37 (64.9) > /49 (28.6) 1 Modified Pollock- Flickinger score <1 42/52 (80.8) /43 (51.2) /28 (64.3) > /39 (25.6) 1 Virginia RAS score /18 (88.9) /47 (68.1) /63 (50.8) /34 (35.3) 1 not add to the complication rate, there are some important principles we should follow during embolization. We follow the principle of targeted embolization during our endovascular procedure. The goal of targeted embolization is not to eliminate the nidus in 1 procedure, but to eliminate the aneurysms and fistulae with a high hemorrhage risk when left untreated. If it is impossible to decrease the AVM nidus volume to a suitable size for GKS by embolization, we use targeted embolization to decrease the hemorrhage risk for AVMs, and this provides a safe latency for GKS to take effect. In addition, the obliteration rate was lower for deeply located AVMs when compared with cortical AVMs. Therefore, during embolization of large-volume AVMs, we should try to eliminate the deeper part of the nidus and attempt to make the residual nidus compact after embolization. Some of the AVMs that were treated with GKS were very small, and most of these were located in the basal ganglia, thalamus, brainstem, or cerebellum. During embolization, reflux of the liquid embolic agent may lead to fatal complications, and our goal for these AVMs was to eliminate the fistula or aneurysms for reducing hemorrhage risk. Our results contribute to the understanding of GKS treatment of AVMs and to the goal of minimizing individual patients risks and optimizing treatment results. Most of our patients were treated with Onyx as the only embolization agent (86 cases) or at least once (12 cases) during more than 1 endovascular treatment procedure. Some patients had NBCA as the only embolization agent (60 cases). A few patients (4 cases) were treated using silk or coils combined with a liquid embolization agent. When comparing the outcome of patients treated at least once with Onyx for (98 cases) with non-onyx patients (using NBCA, silk, coils, or in combination; 64 cases), we did not observe a statistically significant difference between the 2 groups (p > 0.05). However, the obliteration rate with Onyx (53.1%, 52/98) was lower than in the non-onyx group (62.5%, 40/64). A larger study population is needed to further elucidate the difference between these 2 subgroups. The annual incidence of hemorrhage of unruptured and untreated brain AVMs is 2% 4%, with approximately 38% to 71% of patients with brain AVMs presenting with intracranial hemorrhage. 12 Hemorrhage risk during the obliteration time is the major drawback of radiosurgical treatment. In the first 5 years after radiosurgery, AVMs had an annual bleeding rate of 2% 3% with a 3.7% hemorrhage risk and a 2.5% 4.5% bleeding frequency. 5,15 In 1604 patients suffering from AVMs and treated with GKS, Karlsson et al. TABLE 5. Risk of complications according to radiosurgical grading systems after GKS during follow-up Complications Events (%) Hemorrhage Seizure Headache OR 95% CI p Value Pollock-Flickinger score <1 4/32 (12.5) /44 (4.5) /37 (13.5) > /49 (38.8) * Modified Pollock-Flickinger score <1 6/52 (11.5) /43 (7.0) /28 (14.3) > /39 (43.6) * Virginia RAS score 0 2 5/65 (7.7) /63 (23.8) * 4 10/34 (29.4) * * Statistically significant (p < 0.05). J Neurosurg Volume 124 March

8 X. Huo et al. Fig. 2. Graphs of different scoring systems before GKS in predicting the obliteration rate and complication events of AVMs. Predictions of obliteration rate (A C) and complications (D F) in patients with partial embolization undergoing GKS for AVM, as determined using the Pollock-Flickinger score (A and D), modified Pollock-Flickinger score (B and E), and the Virginia RAS (VRAS) score (C and F). observed 49 cases (3%) of bleeding within 2 years after radiosurgery. 21 In our series, 17 hemorrhages (10.5%) occurred during risk-years at follow-up. Among these 17 patients, 12 had already experienced bleeding at the onset, while 5 had never bled. The annual bleeding rate was 1.71%. Five cases died due to hemorrhage and the mortality rate was 3.1%. Most patients who experienced bleeding (15 cases, 88.2%) harbored AVMs with deep locations or AVM-related aneurysms (9 cases, 52.9%). During univariate analysis, deep location (p = 0.01) and aneurysms (p = 0.017) were significantly correlated with hemorrhage. However, in multivariate analysis, this association did not reach a statistically significant value, probably because of the low number of events. In our cases, patients presenting with seizures experienced decreased seizure frequency or were seizure-free following GKS in 17.9% (29/162) of all the cases. In 4 patients the seizure frequency increased, and half of these patient underwent further operations for intractable epilepsy with antiepileptic drugs. Headache was observed in 7 cases and 4 of these cases were caused by radioactive cerebral edema. All of the headache symptoms disappeared at follow-up. Five patients experienced permanent clinical deterioration because of hemorrhage, and 25 patients showed transient clinical decline. In 1986, Speztler and Martin proposed a classification system aimed to predict the clinical outcome after resection of AVMs. 39 They also argued that the proposed classification did not have the same accuracy in predicting clinical outcome in patient candidates for radiosurgical or endovascular treatment. The present study confirms that the Spetzler-Martin classification system is not suitable for 774 J Neurosurg Volume 124 March 2016

9 Combined treatment for cerebral AVMs radiosurgical cases in predicting hemorrhage complication (p > 0.05). In univariate analysis of factors predicting obliteration after GKS, lower Spetzler-Martin grade was related to higher obliteration rate. However, this significance was not noted in multivariate analysis, mainly because only the volume (p = 0.016) was found to be an independent prognostic factor for AVM obliteration in the multivariate analysis, and volume was 1 component of the Spetzler-Martin classification system. In 2002, Pollock and Flickinger proposed a classification system for AVM candidates for radiosurgery and showed that this system was able to predict the outcome in patients undergoing GKS for cerebral AVMs treated at the Mayo Clinic between 1990 and This classification system was very simple and consisted of the use of 3 main parameters (volume, age, and location). Over the last 5 years, many authors confirmed the importance of the system for deeply located AVMs, 2,36 in pediatric patients, 10,44 and for AVMs undergoing radiosurgery with a linear accelerator. 2,44 In a later study, published in 2008, the authors simplified the previous classification drafted in 2002, using parameters such as the age of patients, and the volume and location of AVMs. The latter parameter, instead of being divided into 3 subcategories, was divided into only 2 subcategories: deeply located AVMs versus all other locations. 32 The new classification system proved to be more accurate in predicting the clinical outcome in the authors series. In the present series, the Pollock-Flickinger score and modified Pollock-Flickinger score were both demonstrated to be reliable tools in predicting obliteration and hemorrhage in univariate analysis. In this study, by grouping these 2 scoring systems we found that the obliteration rate gradually decreased from the lower score group to the high score group (Fig. 2). In 2013, Starke et al. 42 performed a study to analyze the outcomes following GKS for AVMs and to make a practical scale for predicting long-term outcomes. The Virginia RAS score was created such that patients were assigned 1 point each for having an AVM volume of 2 4 cm 3, eloquent AVM location, or a history of hemorrhage, and 2 points for having an AVM volume greater than 4 cm 3. Lower Virginia RAS score was significantly related to higher obliteration rate by univariate analysis in our series. We further analyzed the efficacy of grouped Virginia RAS scores for obliteration and post-gks complications. Pre-GKS Virginia RAS score also showed good predictive abilities in both obliteration rate and complications during follow-up (Fig. 2). In summary, previous studies have identified nidus volume, patient age, location, venous drainage pattern, and previous embolization as predictors of outcome after GKS. Various AVM grading systems relevant to radiosurgery have been based on combinations of those parameters. We presented a retrospective review of our own data searching for experiences of GKS treatment for partially embolized AVMs and predictors of obliteration and hemorrhage. Our study had many limitations. The study is monocentric and retrospective, with a relative small population. The goal was to analyze a population of patients treated in a homogeneous way, with a standardized technique of endovascular embolization and a homogeneous technique for radiosurgery. However, combined treatments for AVMs were complicated and most of the combined treatment strategies were individualized. Post-GKS hemorrhage was only diagnosed by CT when patients experienced headache, and asymptomatic hemorrhage was rarely detected by MRI during follow-up. Conclusions GKS treatment is safe and efficient for partially embolized AVMs. However, for partially embolized AVMs, the obliteration rate following GKS is still low. Using multivariate analysis, the only independent factor found to be predictive for AVM obliteration was nidus volume. Reported radiosurgical grading systems (Pollock-Flickinger, modified Pollock-Flickinger, and Virginia RAS) were reliable scoring systems for selection of candidates with partially embolized AVMs for GKS. When considering a patient harboring a partially embolized AVM as a candidate for GKS, these parameters could be used to predict AVM closure and complications. Acknowledgments We would like to thank Dr. Ali Liu and Jian Pan of the Gamma Knife Center in Tiantan Hospital (China) for providing GKS treatment information, and Dr. Zhenghui Qian and Xiaoyun Sun in Tiantan Hospital affiliated with Capital Medical University (China) for follow-up and summary. References 1. Andrade-Souza YM, Ramani M, Scora D, Tsao MN, ter- Brugge K, Schwartz ML: Embolization before radiosurgery reduces the obliteration rate of arteriovenous malformations. Neurosurgery 60: , Andrade-Souza YM, Zadeh G, Scora D, Tsao MN, Schwartz ML: Radiosurgery for basal ganglia, internal capsule, and thalamus arteriovenous malformation: clinical outcome. Neurosurgery 56:56 64, Arteriovenous Malformation Study Group: Arteriovenous malformations of the brain in adults. N Engl J Med 340: , Bamford JM, Sandercock PA, Warlow CP, Slattery J: Interobserver agreement for the assessment of handicap in stroke patients. Stroke 20:828, 1989 (Letter) 5. Betti OO: Treatment of arteriovenous malformations with the linear accelerator. Appl Neurophysiol 50:262, Blackburn SL, Ashley WW Jr, Rich KM, Simpson JR, Drzymala RE, Ray WZ, et al: Combined endovascular embolization and stereotactic radiosurgery in the treatment of large arteriovenous malformations. J Neurosurg 114: , Brown RD Jr, Wiebers DO, Torner JC, O Fallon WM: Frequency of intracranial hemorrhage as a presenting symptom and subtype analysis: a population-based study of intracranial vascular malformations in Olmsted County, Minnesota. J Neurosurg 85:29 32, Brown RD Jr, Wiebers DO, Torner JC, O Fallon WM: Incidence and prevalence of intracranial vascular malformations in Olmsted County, Minnesota, 1965 to Neurology 46: , Cockroft KM: Unruptured brain arteriovenous malformations should be treated conservatively: no. Stroke 38: , Cohen-Gadol AA, Pollock BE: Radiosurgery for arteriovenous malformations in children. J Neurosurg 104 (6 Suppl): , 2006 J Neurosurg Volume 124 March

10 X. Huo et al. 11. Colombo F, Pozza F, Chierego G, Casentini L, De Luca G, Francescon P: Linear accelerator radiosurgery of cerebral arteriovenous malformations: an update. Neurosurgery 34:14 21, da Costa L, Wallace MC, terbrugge KG, O Kelly C, Willinsky RA, Tymianski M: The natural history and predictive features of hemorrhage from brain arteriovenous malformations. Stroke 40: , Fiehler J, Stapf C: ARUBA beating natural history in unruptured brain AVMs by intervention. Neuroradiology 50: , Fox AJ, Girvin JP, Viñuela F, Drake CG: Rolandic arteriovenous malformations: improvement in limb function by IBC embolization. AJNR Am J Neuroradiol 6: , Friedman WA, Blatt DL, Bova FJ, Buatti JM, Mendenhall WM, Kubilis PS: The risk of hemorrhage after radiosurgery for arteriovenous malformations. J Neurosurg 84: , Friedman WA, Bova FJ, Bollampally S, Bradshaw P: Analysis of factors predictive of success or complications in arteriovenous malformation radiosurgery. Neurosurgery 52: , Friedman WA, Bova FJ, Mendenhall WM: Linear accelerator radiosurgery for arteriovenous malformations: the relationship of size to outcome. J Neurosurg 82: , Hadjipanayis CG, Levy EI, Niranjan A, Firlik AD, Kondziolka D, Flickinger JC, et al: Stereotactic radiosurgery for motor cortex region arteriovenous malformations. Neurosurgery 48:70 77, Hladky JP, Lejeune JP, Blond S, Pruvo JP, Dhellemmes P: Cerebral arteriovenous malformations in children: report on 62 cases. Childs Nerv Syst 10: , Kano H, Kondziolka D, Flickinger JC, Park KJ, Iyer A, Yang HC, et al: Stereotactic radiosurgery for arteriovenous malformations after embolization: a case-control study. J Neurosurg 117: , Karlsson B, Lindquist C, Steiner L: Effect of Gamma Knife surgery on the risk of rupture prior to AVM obliteration. Minim Invasive Neurosurg 39:21 27, Kjellberg RN, Hanamura T, Davis KR, Lyons SL, Adams RD: Bragg-peak proton-beam therapy for arteriovenous malformations of the brain. N Engl J Med 309: , Kusske JA, Kelly WA: Embolization and reduction of the steal syndrome in cerebral arteriovenous malformations. J Neurosurg 40: , Levy EI, Niranjan A, Thompson TP, Scarrow AM, Kondziolka D, Flickinger JC, et al: Radiosurgery for childhood intracranial arteriovenous malformations. Neurosurgery 47: , Loeffler JS, Alexander E III, Siddon RL, Saunders WM, Coleman CN, Winston KR: Stereotactic radiosurgery for intracranial arteriovenous malformations using a standard linear accelerator. Int J Radiat Oncol Biol Phys 17: , Luessenhop AJ, Mujica PH: Embolization of segments of the circle of Willis and adjacent branches for management of certain inoperable cerebral arteriovenous malformations. J Neurosurg 54: , Luessenhop AJ, Spence WT: Artificial embolization of cerebral arteries. Report of use in a case of arteriovenous malformation. JAMA 172: , Lunsford LD, Kondziolka D, Flickinger JC, Bissonette DJ, Jungreis CA, Maitz AH, et al: Stereotactic radiosurgery for arteriovenous malformations of the brain. J Neurosurg 75: , Morgan MK, Johnston IH, Hallinan JM, Weber NC: Complications of surgery for arteriovenous malformations of the brain. J Neurosurg 78: , Mori K, Murata T, Hashimoto N, Handa H: Clinical analysis of arteriovenous malformations in children. Childs Brain 6:13 25, Ogilvy CS, Stieg PE, Awad I, Brown RD Jr, Kondziolka D, Rosenwasser R, et al: AHA Scientific Statement: Recommendations for the management of intracranial arteriovenous malformations. Stroke 32: , Pollock BE, Flickinger JC: Modification of the radiosurgerybased arteriovenous malformation grading system. Neurosurgery 63: , Pollock BE, Flickinger JC: A proposed radiosurgery-based grading system for arteriovenous malformations. J Neurosurg 96:79 85, Pollock BE, Flickinger JC, Lunsford LD, Bissonette DJ, Kondziolka D: Factors that predict the bleeding risk of cerebral arteriovenous malformations. Stroke 27:1 6, Pollock BE, Flickinger JC, Lunsford LD, Bissonette DJ, Kondziolka D: Hemorrhage risk after stereotactic radiosurgery of cerebral arteriovenous malformations. Neurosurgery 38: , Pollock BE, Gorman DA, Brown PD: Radiosurgery for arteriovenous malformations of the basal ganglia, thalamus, and brainstem. J Neurosurg 100: , Ruíz-Sandoval JL, Cantú C, Barinagarrementeria F: Intracerebral hemorrhage in young people: analysis of risk factors, location, causes, and prognosis. Stroke 30: , Schneider BF, Eberhard DA, Steiner LE: Histopathology of arteriovenous malformations after gamma knife radiosurgery. J Neurosurg 87: , Spetzler RF, Martin NA: A proposed grading system for arteriovenous malformations. J Neurosurg 65: , Stapf C, Mohr JP: Unruptured brain arteriovenous malformations should be treated conservatively: yes. Stroke 38: , Stapf C, Mohr JP, Choi JH, Hartmann A, Mast H: Invasive treatment of unruptured brain arteriovenous malformations is experimental therapy. Curr Opin Neurol 19:63 68, Starke RM, Yen CP, Ding D, Sheehan JP: A practical grading scale for predicting outcome after radiosurgery for arteriovenous malformations: analysis of 1012 treated patients. J Neurosurg 119: , Steiner L, Lindquist C, Adler JR, Torner JC, Alves W, Steiner M: Clinical outcome of radiosurgery for cerebral arteriovenous malformations. J Neurosurg 77:1 8, Zabel-du Bois A, Milker-Zabel S, Huber P, Schlegel W, Debus J: Pediatric cerebral arteriovenous malformations: the role of stereotactic linac-based radiosurgery. Int J Radiat Oncol Biol Phys 65: , Zabel-du Bois A, Milker-Zabel S, Huber P, Schlegel W, Debus J: Risk of hemorrhage and obliteration rates of LINACbased radiosurgery for cerebral arteriovenous malformations treated after prior partial embolization. Int J Radiat Oncol Biol Phys 68: , 2007 Disclosure This project was supported by the Capital Medical Development Research Foundation. Author Contributions Conception and design: Li. Acquisition of data: Huo, Zhao. Analysis and interpretation of data: Huo. Drafting the article: Huo. Critically revising the article: Huo, Lv. Reviewed submitted version of manuscript: Huo, Jiang, Lv. Statistical analysis: Huo, Yang. Administrative/technical/material support: Huo, Jiang, Zhao. Study supervision: Li, Jiang. Correspondence Youxiang Li, Beijing Neurosurgical Institute, No. 6, Tiantan Xili, Chongwen, Beijing , China. doctorliyouxiang@163.com. 776 J Neurosurg Volume 124 March 2016