Spetzler-Martin Grade III arteriovenous malformations. Radiosurgery for Spetzler-Martin Grade III arteriovenous malformations.

Transcription

1 See the corresponding editorial in this issue, pp J Neurosurg 120: , 2014 AANS, 2014 Radiosurgery for Spetzler-Martin Grade III arteriovenous malformations Clinical article Dale Ding, M.D., Chun-Po Yen, M.D., Robert M. Starke, M.D., M.Sc., Zhiyuan Xu, M.D., Xingwen Sun, B.S., and Jason P. Sheehan, M.D., Ph.D. Department of Neurological Surgery, University of Virginia Health System, Charlottesville, Virginia Object. Intracranial arteriovenous malformations (AVMs) are most commonly classified based on their Spetzler- Martin grades. Due to the composition of the Spetzler-Martin grading scale, Grade III AVMs are the most heterogeneous, comprising 4 distinct lesion subtypes. The management of this class of AVMs and the optimal treatment approach when intervention is indicated remain controversial. The authors report their experience with radiosurgery for the treatment of Grade III AVMs in a large cohort of patients. Methods. All patients with Spetzler-Martin Grade III AVMs treated with radiosurgery at the University of Virginia over the 20-year span from 1989 to 2009 were identified. Patients who had less than 2 years of radiological follow-up and did not have evidence of complete obliteration during that period were excluded from the study, leaving 398 cases for analysis. The median patient age at treatment was 31 years. The most common presenting symptoms were hemorrhage (59%), seizure (20%), and headache (10%). The median AVM volume was 2.8 cm 3, and the median prescription dose was 20 Gy. The median radiological and clinical follow-up intervals were 54 and 68 months, respectively. Univariate and multivariate Cox proportional hazards and logistic regression analysis were used to identify factors associated with obliteration, postradiosurgery radiation-induced changes (RIC), and favorable outcome. Results. Complete AVM obliteration was observed in 69% of Grade III AVM cases at a median time of 46 months after radiosurgery. The actuarial obliteration rates at 3 and 5 years were 38% and 60%, respectively. The obliteration rate was higher in ruptured AVMs than in unruptured ones (p < 0.001). Additionally, the obliteration rate for Grade III AVMs with small size (< 3 cm diameter), deep venous drainage, and location in eloquent cortex was higher than for the other subtypes (p < 0.001). Preradiosurgery AVM rupture (p = 0.016), no preradiosurgery embolization (p = 0.003), increased prescription dose (p < 0.001), fewer isocenters (p = 0.006), and a single draining vein (p = 0.018) were independent predictors of obliteration. The annual risk of postradiosurgery hemorrhage during the latency period was 1.7%. Two patients (0.5%) died of hemorrhage during the radiosurgical latency period. The rates of symptomatic and permanent RIC were 12% and 4%, respectively. Absence of preradiosurgery AVM rupture (p < 0.001) and presence of a single draining vein (p < 0.001) were independent predictors of RIC. Favorable outcome was observed in 63% of patients. Independent predictors of favorable outcome were no preradiosurgery hemorrhage (p = 0.014), increased prescription dose (p < 0.001), fewer isocenters (p = 0.014), deep location (p = 0.014), single draining vein (p = 0.001), and lower Virginia radiosurgery AVM scale score (p = 0.016). Conclusions. Radiosurgery for Spetzler-Martin Grade III AVMs yields relatively high rates of obliteration with a low rate of adverse procedural events. Small and ruptured lesions are more likely to become obliterated after radiosurgery than large and unruptured ones. ( Key Words intracranial arteriovenous malformation stroke Spetzler-Martin Grade III stereotactic radiosurgery Gamma Knife vascular malformations Abbreviations used in this paper: AVM = arteriovenous malformation; DSA = digital subtraction angiography; RBAS = radiosurgery-based AVM scale; RIC = radiation-induced change; Virginia RAS = Virginia radiosurgery AVM scale. Spetzler-Martin Grade III arteriovenous malformations (AVMs) are the most heterogeneous class of AVMs by nature of the grading system, which accounts for AVM size, location eloquence, and venous drainage pattern. 22 The natural history of AVMs is not benign, with an annual hemorrhage rate of 2% 4% and combined annual morbidity and mortality risk of 2.7%. 2,7,17 Grade III AVMs straddle the line dividing low-grade AVMs (Grades I and II), the majority of which are treated aggressively, and high-grade AVMs (Grades IV and V), for which conservative management may be equivalent to if not superior to multimodality therapy in many cases. 8,10 While a growing consensus has evolved in the literature regarding the 959

2 D. Ding et al. management of low- and high-grade AVMs, the approach to Grade III lesions that yields the most favorable outcomes remains poorly defined. Among the currently accepted treatments for AVMs, including microsurgery, radiosurgery, and endovascular embolization, only radiosurgery is essentially devoid of immediate treatment-related complications. 25 Conversely, radiosurgery does not produce immediate angiographic results and is associated with a latency period, typically 2 to 3 years, prior to obliteration during which the AVM may rupture. 6,13 Patient, physician, and institutional preferences continue to play large roles in the management of intermediate-grade AVMs for which the optimal treatment remains unclear. We present our radiosurgery experience with a large cohort of patients harboring Spetzler- Martin Grade III AVMs. Methods Patient Population We identified all patients harboring Spetzler-Martin Grade III AVMs from a prospective, institutional review board approved database of 1400 AVM cases involving patients treated with Gamma Knife radiosurgery at the University of Virginia from 1989 to Patients with less than 2 years of radiological follow-up were excluded except for those with radiological evidence of complete AVM obliteration prior to 2 years follow-up. This yielded 398 cases of Grade III AVMs for analysis. The patient population included 204 males (51.3%) and 194 females (48.7%) with a median age of 30.9 years. There were 75 pediatric patients (under the age of 18; 18.8% of the study group). Preradiosurgery therapies included embolization in 109 cases (27.4%) and microsurgical resection in 44 cases (11.1%). The majority of the AVMs, 252 (63.3%), were ruptured prior to radiosurgery. The most common presenting symptoms were hemorrhage in 234 patients (58.8%), seizure in 80 patients (20.1%), headache in 38 patients (9.5%), and focal neurological deficit in 25 patients (6.3%). Descriptive statistics of the patient population are summarized in Table 1. Radiosurgery Treatment Parameters and AVM Characteristics Our institution s Gamma Knife radiosurgery technique has been previously described and uses a combination of digital subtraction angiography (DSA) supplemented by MRI to define the AVM nidus. 25 Prior to 1991, MRI was not routinely used, and treatment plans were developed using only DSA. We have been using the Leksell GammaPlan software (Elekta AB) for dose planning since June 1994, prior to this point we used the Kula software. The initial radiosurgery median treatment parameters were AVM volume 2.8 cm 3, prescription dose 20 Gy, maximum dose 38 Gy, isodose line 50%, and 2 isocenters. The AVM location was deep, which we defined as basal ganglia, thalamus, and brainstem, in 184 cases (46.2%). The venous drainage pattern was deep in 337 cases (84.7%) and multiple in 169 (42.5%). Pandey et al. further classified Spetzler-Martin Grade III AVMs into 4 types based on the components TABLE 1: Patient characteristics* Characteristic Value sex male 204 (51.3) female 194 (48.7) age (yrs) mean 32.2 median 30.9 range pediatric patients (age <18 yrs) 75 (18.8) preradiosurgery hemorrhage 252 (63.3) preradiosurgery embolization 109 (27.4) preradiosurgery microsurgical resection 44 (11.1) presenting symptom hemorrhage 234 (58.8) seizure 80 (20.1) headache 38 (9.5) focal neurological deficit 25 (6.3) asymptomatic 11 (2.8) * Values indicate numbers of patients (%) unless otherwise indicated. of their grade (S = AVM size, E = eloquence of cortex, V = venous drainage pattern). 18 When this classification system was applied to our cases, 302 AVMs (75.9%) were Subtype 1, S1E1V1 (size < 3 cm, eloquent location, deep venous drainage); 61 (15.3%) were Subtype 2, S2E1V0 (size 3 6 cm, eloquent location, superficial venous drainage); 35 (8.8%) were Subtype 3, S2E0V1 (size 3 6 cm, noneloquent location, deep venous drainage); and 0 were Subtype 4, S3E0V0 (size > 6 cm, noneloquent cortex, superficial venous drainage). The radiosurgery-based AVM scale (RBAS, 2011 modification) 28 factors in patient age and AVM volume and location; the mean score on this scale was 1.23, the median was 1.18, and the range We recently described the Virginia radiosurgery AVM scale (RAS), which is composed of 3 components: AVM volume (< 2 cm 3 = 0 points, 2 4 cm 3 = 1 point, > 4 cm 3 = 2 points), preradiosurgery AVM hemorrhage (unruptured = 0 points, ruptured 1 = point), and eloquent location (noneloquent = 0 points, eloquent = 1 point). 24 Table 2 details the initial radiosurgery treatment parameters and AVM properties. Repeat Radiosurgery Repeat radiosurgery was performed for 64 AVMs (16.1%) that did not become obliterated after the initial procedure, including 60 AVMs treated with 1 repeat radiosurgery (15.1%) and 4 AVMs treated with 2 repeat radiosurgeries (1.0%). For the first repeat radiosurgery session, the mean AVM volume was 1.2 cm 3, the median was 0.9 cm 3, and the range cm 3 ; the mean prescription dose was 20.9 Gy, the median 22 Gy, and the range 4 28 Gy; the mean maximum dose was 37.4 Gy, the median 40 Gy, and the range 8 50 Gy; the mean isodose line was 57.9%, the median 50%, and the range 43% 91%; and the mean number of isocenters was 2.6, the median 2, and 960

3 Radiosurgery for Grade III arteriovenous malformations TABLE 2: AVM characteristics and treatment parameters* Characteristic or Parameter Value location superficial 214 (53.8) deep 184 (46.2) venous drainage pattern superficial 61 (15.3) deep 337 (84.7) no. of draining veins single 229 (57.5) multiple 169 (42.5) eloquence of location eloquent 363 (91.2) noneloquent 35 (8.8) maximum diameter (cm) mean 2.3 median 2.3 range volume (cm 3 ) mean 3.6 median 2.8 range maximum dose (Gy) mean 37.9 median 38 range prescription dose (Gy) mean 20.6 median 20 range 5 32 isodose (%) mean 55.8 median 50 range no. of isocenters mean 2.7 median 2 range 1 16 Spetzler-Martin Grade III subtype Subtype (75.9) Subtype 2 61 (15.3) Subtype 3 35 (8.8) Subtype 4 0 RBAS score mean 1.23 median 1.18 range < (28.9) (48.0) (18.6) > (4.5) (continued) TABLE 2: AVM characteristics and treatment parameters* (continued) Characteristic or Parameter Value Virginia RAS score 1 2 points 219 (55.0) 3 points 131 (32.9) 4 points 48 (12.1) * Values indicate numbers of patients (%) unless otherwise indicated. Deep location includes basal ganglia, thalamus, and brainstem. From Pandey et al. 18 Subtype 1 = S1E1V1, Subtype 2 = S2E1V0, Subtype 3 = S2E0V1, and Subtype 4 = S3E0V0. S = size (1: < 3 cm; 2: 3 6 cm; 3: > 6 cm), E = eloquent location (0: noneloquent; 1: eloquent), and V = venous drainage pattern (0: superficial venous drainage; 1: deep venous drainage). the range 1 9. For the second repeat radiosurgery session, the mean AVM volume was 0.6 cm 3, the median 0.6 cm 3, and the range cm 3 ; the mean prescription dose was 22.3 Gy, the median 23 Gy, and the range Gy; the mean maximum dose was 44.5 Gy, the median 46 Gy, and the range Gy; the isodose line was 50% for all 4 AVMs; and the mean number of isocenters was 5.5, the median 4.5, and the range Radiological and Clinical Follow-up Standard radiological follow-up in our series included an MRI every 6 months for the first 2 years and an annual MRI thereafter. Obliteration was defined by lack of flow voids on MRI or absence of pathological arteriovenous shunting on DSA. Digital subtraction angiography was only performed to confirm total AVM obliteration after follow-up MRI showed a lack of flow voids. The overall mean duration of radiological follow-up was 71.7 months (6.0 years), the median was 54.3 months (4.5 years), and the range months ( years). Additional imaging, including CT and MRI, was obtained if patients showed neurological decline. Radiological complications following radiosurgery were hemorrhage and radiationinduced changes (RICs), defined on MRI as perinidal areas of signal hyperintensity on T2-weighted images. Symptomatic RICs typically manifested clinically as headache, focal neurological deficits, and seizures. Favorable outcome was defined as complete AVM obliteration, no permanently symptomatic RICs, and no postradiosurgery hemorrhage. 24 Since the University of Virginia is a tertiary cerebrovascular and radiosurgery referral center, clinical follow-up was obtained both directly and indirectly. Direct clinical follow-up included return clinical appointments and hospital admissions to the University of Virginia Health System. Indirect clinical follow-up included correspondence with outside referring hospitals and patients local primary care physicians. The overall mean duration of clinical follow-up was 81.5 months (6.8 years), the median was 67.5 months (5.6 years), and the range months ( years). Statistical Analysis The time to obliteration and actuarial obliteration 961

4 D. Ding et al. rates were determined with Kaplan-Meier analysis. Since DSA was only performed subsequent to AVM obliteration on MRI, time to obliteration was based on the date of the MRI identifying obliteration. The univariate log-rank test was used to compare the actuarial obliteration rates of unruptured versus ruptured AVMs and the obliteration rates among the different Grade III AVM subtypes. Cox proportional hazards regression analysis was used to identify factors associated with AVM obliteration. Logistic regression analysis was used to identify factors associated with RIC and favorable outcome following radiosurgery. The variables analyzed were sex, age, preradiosurgery hemorrhage, postradiosurgery hemorrhage, preradiosurgery embolization, AVM volume, eloquent location, prescription dose, number of isocenters, AVM location (superficial vs deep), venous drainage pattern (superficial vs deep), number of draining veins (single vs multiple), radiological evidence of RIC, Spetzler-Martin Grade III subtype, RBAS score, and Virginia RAS score. For each variable, a hazard or odds ratio was reported for Cox or logistic analysis, respectively, as well as the 95% confidence interval (CI) and p value. Statistically significant predictors had a 95% confidence interval not including 1.0 and a p value less than Univariate analysis was performed for all variables. Multivariate analysis was only performed if more than 1 variable was statistically significant in the univariate analysis. All statistical analyses were performed with the SPSS version 20 statistical software program. The annual postradiosurgery hemorrhage risk was calculated by dividing the total number of individual hemorrhages following radiosurgery by the sum of the postradiosurgery risk years. Risk years were defined by the amount of time from the treatment date to the obliteration date or the date of the most recent radiological follow-up in patients with AVMs that were not obliterated. Results Radiological Outcomes Following Radiosurgery Complete radiological AVM obliteration was evident on MRI only in 54 patients (13.6%) and confirmed by DSA in 222 patients (55.8%) for a cumulative obliteration rate of 69.4%. The obliteration rates stratified by preradiosurgery treatment were as follows: microsurgical resection only, 82.9% (29 of 35 cases); embolization only, 43.0% (43 of 100 cases); both microsurgical resection and embolization, 77.8% (7 of 9 cases); and no preradiosurgery treatment, 77.6% (197 of 254 cases). Excluding all patients with less than 2 years of radiological follow-up, even those with completely obliterated nidi, AVM obliteration was demonstrated on MRI only in 13.5% of cases (45 of 334) and confirmed by DSA in 50.0% of cases (167 of 334) for a cumulative obliteration rate of 63.5% (212 of 334 cases). The mean time to obliteration was 79.2 months (6.6 years) and the median 45.5 months (3.8 years). The actuarial obliteration rates at time intervals of 3 and 5 years following radiosurgery were 38% and 60%, respectively. Figure 1 demonstrates the obliteration rate over time for all Grade III AVMs. The mean and median times to obliteration were 96.4 months (8.0 years) and 63.9 months (5.3 years), respectively, for unruptured AVMs and 69.9 months (5.8 years) and 39.3 months (3.3 years), respectively, for ruptured AVMs. The rate of obliteration over time was significantly higher for ruptured lesions, 44% at 3 years and 68% at 5 years, compared with unruptured ones, 28% at 3 years and 48% at 5 years (p < 0.001), based on univariate logrank test. Figure 2 demonstrates the obliteration rate over time for unruptured versus ruptured Grade III AVMs. The mean and median times to obliteration were 66.3 months (5.5 years) and 39.9 months (3.3 years), respectively, for Subtype 1 (S1E1V1) AVMs; months (9.9 years) and months (10.1 years), respectively, for Subtype 2 (S2E1V0) AVMs; and months (9.9 years) and months (9.7 years), respectively, for Subtype 3 (S2E0V1) AVMs. The actuarial obliteration rates were significantly higher for Subtype 1 lesions, 44% at 3 years and 69% at 5 years, than for Subtype 2 lesions, 18% at 3 years and 35% at 5 years, (p < 0.001) or Subtype 3 lesions, 24% at 3 years and 27% at 5 years (p < 0.001). The actuarial obliteration rates were similar in Subtype 2 compared with Subtype 3 lesions (p = 0.829). Pairwise comparisons among the 3 different Grade III AVM subtypes were made utilizing the univariate log-rank test. Figure 3 demonstrates the obliteration rate over time for subtypes 1, 2, and 3 Grade III AVMs. Predictors of AVM Obliteration From the univariate Cox proportional hazards regression analysis, we determined preradiosurgery AVM rupture (p < 0.001), no postradiosurgery latency period hemorrhage (p = 0.006), no preradiosurgery embolization (p < 0.001), decreased volume (p < 0.001), eloquent location (p = 0.002), increased prescription dose (p < 0.001), fewer isocenters (p < 0.001), deep location (p < 0.001), deep venous drainage (p < 0.001), single draining vein (p < 0.001), lower Spetzler-Martin Grade III subtype, lower RBAS score (p < 0.001), and lower Virginia RAS score (p < 0.001) to be significantly associated with AVM obliteration following radiosurgical treatment. In the multivariate analysis, preradiosurgery AVM rupture (p = 0.016), no preradiosurgery embolization (p = 0.003), increased prescription dose (p < 0.001), fewer isocenters (p = 0.006), and a single draining vein (p = 0.018) were identified as independent predictors of obliteration. Table 3 details the results of the univariate and multivariate Cox proportional hazards regression analyses for variables associated with obliteration. Postradiosurgery Hemorrhage There were a total of 33 postradiosurgery hemorrhages in 28 patients 1 hemorrhage in each of 23 patients and 2 hemorrhages in each of 5 patients. Dividing by a total of 1897 risk years, the annual postradiosurgery hemorrhage rate was 1.7%. Two patients died of postradiosurgery hemorrhage (0.5%). Due to the relatively small fraction of patients with postradiosurgery latency period hemorrhage (7.0%), it was not statistically appropriate to perform logistic regression analysis for factors associated with postradiosurgery hemorrhage. 962

5 Radiosurgery for Grade III arteriovenous malformations Fig. 1. Kaplan-Meier analysis demonstrating the obliteration rate of Spetzler-Martin Grade III AVMs over time. The number of Grade III AVM patients with available radiological follow-up remaining at each time interval is shown under the x axis. Fig. 2. Kaplan-Meier analysis demonstrating the obliteration rate over time of ruptured versus unruptured Spetzler-Martin Grade III AVMs. The actuarial obliteration rate was significantly higher for ruptured Grade III AVMs than for unruptured ones (p < 0.001). The number of ruptured and unruptured Grade III AVM patients with available radiological follow-up remaining at each time interval is shown under the x axis. 963

6 D. Ding et al. Fig. 3. Kaplan-Meier analysis demonstrating the obliteration rate over time of Subtypes 1 (S1E1V1), 2 (S2E1V0) and 3 (S2E0V1) Grade III AVMs. The actuarial obliteration rate was significant higher for Subtype 1 AVMs compared with both Subtype 2 (p < 0.001) and Subtype 3 (p < 0.001) AVMs. The obliteration rates for Subtype 2 and 3 AVMs were similar (p = 0.829). The number of Subtype 1, 2, and 3 Grade III AVM patients with available radiological follow-up remaining at each time interval is shown under the x axis. Radiation-Induced Changes Following Radiosurgery Radiologically evident RIC occurred in 138 patients following radiosurgery (34.7%). The time interval following radiosurgery to RIC development was a mean of 12.9 months, median 10.0 months, and range months, and the duration of RIC after onset was a mean of 20.5 months, median 15.3 months, and range months. Radiation-induced change was symptomatic in 47 patients (11.8%), including 30 patients with focal neurological deficits (7.5%), 15 patients with headache (3.8%), and 2 patients with seizures (0.5%). The clinical manifestations of RIC were temporary in 31 patients (7.8%) and permanent in 16 patients (4.0%). The permanent symptoms of RIC were focal neurological deficits in 15 patients and new onset seizures in 1 patient. Excluding all patients with less than 2 years radiological follow-up, including those with complete AVM obliteration, the rates of cumulative, symptomatic, and permanent RIC were 34.3% (115/334 patients), 10.2% (34/334 patients), and 3.9% (13/334 patients), respectively. No preradiosurgery AVM rupture (p < 0.001), superficial venous drainage (p = 0.012), a single draining vein (p < 0.015), and higher RBAS score (p = 0.048) were significantly associated with RICs based on univariate logistic regression analysis. Multivariate analysis determined the lack of preradiosurgery AVM rupture (p < 0.001) and presence of a single draining vein (p < 0.001) to be independent predictors of RIC following radiosurgery. Table 4 details the results of the univariate and multivariate logistic regression analyses for variables associated with RIC. Clinical Outcomes Following Radiosurgery In patients presenting with seizures, 56 patients (14.1% of the total group) had decreased seizure frequency (38 patients, 9.6%) or had no seizures (18 patients, 4.5%) following radiosurgery. In contrast, 3 patients presenting with seizures had increased seizure frequency (0.8% of the total group) and 4 patients without preradiosurgery seizures had new-onset seizures (1.0%), including 1 patient with seizures resulting from RIC, after radiosurgery. Clinical improvement was observed in 79 (19.8%) of 398 patients. Two patients (0.5%) died during the radiosurgical latency period from AVM rupture. Favorable outcome, or successful AVM obliteration without permanent RICs or postradiosurgery hemorrhage, was observed in 251 patients (63.1%). Univariate logistic regression analysis identified no preradiosurgery hemorrhage (p = 0.010); no preradiosurgery embolization (p < 0.001); decreased volume (p < 0.001); eloquent loca- 964

7 Radiosurgery for Grade III arteriovenous malformations TABLE 3: Factors predicting obliteration after radiosurgery* Factor Univariate Analysis Multivariate Analysis HR 95% CI p Value HR 95% CI p Value male sex decreased age preradiosurgery hemorrhage < no postradiosurgery hemorrhage no preradiosurgery embolization < decreased volume < eloquent location increased prescription dose < <0.001 fewer isocenters < deep location < deep venous drainage < single draining vein < RICs lower Spetzler-Martin Grade III subtype < lower RBAS score < lower Virginia RAS score < * HR = hazard ratio. Statistically significant (p < 0.05) on multivariate analysis. tion (p = 0.029); increased prescription dose (p < 0.001); fewer isocenters (p < 0.001); deep location (p = 0.023); deep venous drainage (p < 0.001); single draining vein (p < 0.001); Spetzler-Martin Grade III Subtypes 1, 3, then 2 in decreasing order (p < 0.001); and lower Virginia RAS score (p < 0.001) to be predictors of favorable outcome. In the multivariate analysis, no preradiosurgery embolization (p = 0.014), increased prescription dose (p = 0.001), fewer isocenters (p = 0.014), deep location (p = 0.014), single draining vein (p = 0.001), and lower Virginia RAS score (p = 0.016) were determined to be independent predictors of favorable outcome. Table 5 details the results of the univariate and multivariate logistic regression analyses for variables associated with favorable outcome. Discussion The traditional management of intracranial AVMs has been to aggressively pursue complete microsurgical resection due to the relatively young age of the patient population and initial natural history studies. 2,9,17,27 More recently, the importance of patient selection, conservative management, and multimodality therapy when treatment is indicated have emerged as crucial aspects of modern AVM management. 5,15,18 Despite the same rating on the Spetzler- Martin scale, Grade III AVMs are a highly diverse cohort of lesions encompassing those small and large in size, those with superficial and deep venous drainage, and those located in eloquent and noneloquent cortex. Microsurgical Resection of Grade III AVMs Although microsurgical resection offers excellent obliteration rates for Grade III AVMs, many surgical series are subject to significant selection biases. 3,23 Lesions deemed unsuitable for resection were referred for radiosurgical treatment. For a cohort of 141 patients with Spetzler-Martin Grade III AVMs, de Oliveira et al. reported better outcomes following microsurgical resection of large Grade III lesions, subclassified by the authors as IIIA (96% good outcome), than of small lesions, subclassified as IIIB (70% good outcome). 3 In a modern series of 76 patients harboring Grade III AVMs treated with microsurgical resection, Dr. Lawton reported an obliteration rate of 97% with an accompanying combined permanent morbidity and mortality rate of 8%. 12 In stark contrast to the surgical outcomes described by de Oliveira et al., the risk of surgical complications in Dr. Lawton s series was mildest for Subtype 1, moderate for Subtype 2, and most severe for Subtype 3 AVMs. Just as in our radiosurgery series, there were no Subtype 4 AVMs in his surgical series. All except 1 of the patients had presurgical embolization; there were 2 hemorrhagic complications (3%) necessitating emergent surgical evacuation. The author eloquently describes the combination of technical skill and careful patient selection required to obtain optimal microsurgical outcomes for Grade III AVMs. Endovascular Embolization of Grade III AVMs The development of ethylene vinyl alcohol copolymer, or Onyx (ev3 Neurovascular), has expanded the ability of neurointerventionalists to treat a wider range of AVMs. 16 For smaller AVMs, specifically Subtype 1, embolization may cure up to 50% of highly preselected lesions. 20 More practical cure rates are significantly lower, approximately 10% 20%. 26 Aggressive endovascular treatment should be avoided to limit embolization-related complications associated with early draining vein occlusion and inadvertent embolization of critical feeders of normal cortex. 11 The 965

8 D. Ding et al. TABLE 4: Factors predicting RICs following radiosurgery* Factor Univariate Analysis Multivariate Analysis OR 95% CI p Value OR 95% CI p Value male sex increased age no preradiosurgery hemorrhage < <0.001 no postradiosurgery hemorrhage preradiosurgery embolization increased volume eloquent location decreased prescription dose more isocenters superficial location superficial venous drainage single draining vein <0.001 higher Spetzler-Martin Grade III subtype higher RBAS score higher Virginia RAS score * OR = odds ratio. Statistically significant (p < 0.05) on multivariate analysis. deleterious effect of embolization on radiosurgical obliteration is well described, including in the current series, although the underlying mechanism of decreased obliteration is not completely understood. 1,4,21 Nevertheless, 27% of the AVMs in this series were treated with embolization prior to radiosurgery. Nearly all cases of preradiosurgery embolization in our series were by intention to reduce the volume of a relatively large or diffuse AVM prior to radiosurgery, occlude high-flow feeding arteries harboring perinidal or intranidal aneurysms, or eliminate intranidal arteriovenous shunts. Additionally, we believed that reducing the size of these AVMs with embolization would allow us to deliver a more optimal dose to the residual nidi. Because dose and volume are intimately related to the benefits and risks of AVM radiosurgery, multimodality treatment for a select subset of Grade III lesions may be justified. Finally, the role of embolization after radiosurgery has yet to be clearly delineated. TABLE 5: Factors predicting favorable outcome following radiosurgery* Factor Univariate Analysis Multivariate Analysis OR 95% CI p Value OR 95% CI p Value male sex decreased age no preradiosurgery hemorrhage no preradiosurgery embolization < decreased volume < eloquent location increased prescription dose < fewer isocenters < deep location deep venous drainage < single draining vein < Spetzler-Martin Grade III subtype (1, 3, 2) < higher RAS score lower Virginia RAS score < * Favorable outcome was defined as complete AVM obliteration, no permanent RIC, and no postradiosurgery hemorrhage. Statistically significant (p < 0.05) on multivariate analysis. Subtype was placed in order from most to least likely to have favorable outcome. 966

9 Radiosurgery for Grade III arteriovenous malformations Multimodality Treatment of Grade III AVMs Pandey et al. reported on a series of 100 patients with Spetzler-Martin Grade III AVMs treated with microsurgery (64%), radiosurgery (49%), or embolization either alone or in combination (78%). 18 Single modality treatment was undertaken in 20%, including only 1 patient treated with embolization alone, compared with multimodality treatment in 80%, consisting of 2 modalities in 69 patients and 3 modalities in 11 patients. The rates of morbidity and mortality were 14% and 1%, respectively. In the 89 patients with adequate follow-up, there was an overall 88% obliteration rate. While the concept of multimodality treatment is appealing, the effects of each prior therapy upon the subsequent ones must be carefully considered to minimize adverse effects and maximize cure rates. For example, while endovascular embolization is frequently used prior to microsurgical resection to reduce intraoperative blood loss, it may decrease radiosurgical obliteration rates. 14,21 For patients in whom microsurgical resection is undertaken after failed radiosurgery, these radiated AVMs are, in general, easier to resect than nonradiated ones due to the radiosurgery-induced gliotic margin surrounding the nidus. However, Lawton describes 2 cases of Grade III AVMs in which the difficulty of microsurgical resection was increased by previous radiosurgical treatment. 12 Hence one should note that while multimodality treatment combines the advantages of each individual treatment approach, the treatment-specific complications may be additive or even multiplicative. Therefore close collaboration across the disciplines of neurosurgery, neurointerventional radiology, and radiation therapy is crucial. The Role of Radiosurgery as a Primary Treatment for Grade III AVMs Radiosurgery offers a very good chance of obliteration for Grade III AVMs, with an overall cure rate of 69% in our series. The AVMs with preradiosurgery hemorrhage had higher obliteration rates than those without (p < 0.001), and preradiosurgery hemorrhage was an independent predictor of successful radiosurgery-induced obliteration (p = 0.016). The reason for increased obliteration for ruptured versus unruptured AVMs is not known. We hypothesize that ruptured AVMs may develop a better-defined nidus on subsequent imaging, thereby increasing the precision of radiosurgical targeting. At our institution we typically wait 12 weeks following initial AVM hemorrhage to perform radiosurgery so that acute or subacute blood products do not obscure the true borders of the nidus. Based on the classification proposed by Pandey et al., the Grade III AVMs in our patient cohort were primarily comprised of Subtype 1 AVMs (76%) with fewer Subtype 2 (15%) or 3 (9%) AVMs and no Subtype 4 AVMs. 18 Subtype 1 Grade III AVMs, which are small (< 3 cm in diameter) with deep venous drainage and located in eloquent cortex, represent the ideal radiosurgical target and had a significantly higher obliteration rate than Subtype 2 (p < 0.001) or 3 (p < 0.001) lesions. In addition to preradiosurgery hemorrhage, no preradiosurgery embolization (p = 0.003), increased prescription dose (p < 0.001), fewer isocenters (p = 0.006), and the presence of a single draining vein (p = 0.016) were independent predictors of obliteration, which is consistent with previously reported findings. 4 Singular venous drainage and fewer isocenters are likely to be surrogate indicators of a more compact nidus, which facilitates radiosurgical targeting. It is not surprising that higher radiosurgical margin doses increase obliteration. However, one must balance the benefit of the increased likelihood of obliteration with the risk of increased radiosurgery-related complications that accompany treatment with higher doses. 6 The annual postradiosurgery hemorrhage rate was 1.7%, which is slightly better than the generally accepted natural history rupture risk of 2% 4% per year. 2,7,17 While published reports regarding the effect of radiosurgery on the natural history of AVMs, specifically on hemorrhage risk, remain conflicting, we have demonstrated in large cohorts of AVM patients that, at the very least, treatment with radiosurgery does not appear to increase the risk of AVM rupture. 4,30 The only postradiosurgery deaths were due to AVM rupture during the latency period (2 patients [0.5%]). Radiation-induced changes were symptomatic in 12% of patients and permanent in 4%. No preradiosurgery hemorrhage (p < 0.001) and the presence of a single draining vein (p < 0.001) were independently associated with postradiosurgery RIC. Yen et al. described similar findings in a larger cohort encompassing all AVMs treated at our institution. 29 Favorable outcome, which represents the combined ideal radiological and clinical outcome following radiosurgery, occurred in 63% of patients in this series. In addition to the independent predictors of obliteration, which, with the exception of preradiosurgery hemorrhage, were also found to be independently associated with favorable outcome, deep AVM location (p = 0.014) and lower Virginia RAS (p = 0.016) were independent predictors of favorable outcome. Initially described by Starke et al., the Virginia RAS is the radiosurgery analog to the Spetzler-Martin scale, which was originally intended to predict outcomes of microsurgical treatment of AVMs. 22,24 In this current series of Spetzler-Martin Grade III AVMs, the Virginia RAS was superior to the RBAS in predicting favorable outcome, although neither the Virginia RAS nor RBAS were independently associated with obliteration or RIC in multivariate analyses. Ultimately, external validation of the Virginia RAS in large series of AVMs treated with radiosurgery at other institutions is necessary. The overall complication rate following radiosurgery, including symptomatic temporary and permanent RICs, increased seizure frequency or new-onset seizures, and morbidity and mortality secondary to latency period AVM rupture, is comparable to the complication rates reported in the aforementioned studies of Grade III AVMs treated with microsurgical and multimodality approaches. 12,18 Study Limitations Despite the relatively large size of our patient cohort, the data are derived from a single-center, retrospective study. Therefore the treatment biases of the physicians and the institution will be pervasive throughout this study. Although our selection criteria were not strict or formu- 967

10 D. Ding et al. laic, we tended to favor treatment of small, deep-seated Grade III AVMs, or Subtype 1, with radiosurgery. While this study was not designed to compare our radiosurgical results to those of similar microsurgical or multimodality AVM series, it certainly beckons the juxtaposition of different treatment approaches for Grade III AVMs. In further emphasizing the selection bias inherent to our study, we note that the AVMs were relatively small, with a median volume of 2.8 cm 3 and the majority with a maximum diameter less than 3 cm (76%). This critical AVM characteristic should be kept in mind when comparing our results to AVM treatment outcomes in the published literature. Patients with larger lesions, Subtypes 2, 3, and 4, with favorable angioarchitecture, often underwent partial embolization prior to radiosurgery. Patients with ruptured AVMs presenting to our institution with significant hemorrhage-related mass effect or rapid clinical deterioration were treated surgically and received subsequent radiosurgery only if complete microsurgical extirpation was not achieved. Unruptured, minimally symptomatic, or asymptomatic large lesions that we believed were unlikely to be obliterated with radiosurgical treatment were managed conservatively. Additionally the nature of being a tertiary radiosurgery referral center subjects our patient cohort to the selection biases associated with local, regional, and international referral patterns. The biases of our study, as well as those of other studies describing treatment outcomes for Grade III AVMs, complicate comparisons of our data to those from other centers. Furthermore, we acknowledge that MRI, which was the only neuroimaging evidence of AVM cure in 14% of AVMs in this study, remains suboptimal in comparison to the gold standard of DSA for evaluating AVM obliteration. However, Pollock et al. demonstrated that MRI had 100% specificity and 91% negative predictive value compared with DSA for determining the patency of an AVM nidus. 19 Therefore, we believe that in instances where DSA follow-up is unavailable, MRI represents a generally acceptable alternative for radiological follow-up. Conclusions Radiosurgery offers an excellent risk-to-benefit profile for Spetzler-Martin Grade III AVMs. In appropriately selected cases, combination treatment with microsurgical resection may further increase the cure rate for these lesions. Small Grade III AVMs located in eloquent cortex with deep venous drainage are likely conferred the greatest advantage by radiosurgery compared with other Grade III AVM subtypes and therefore represent optimal targets for this treatment modality. Disclosure The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper. Author contributions to the study and manuscript preparation include the following. Conception and design: Sheehan, Ding. Acquisition of data: Ding, Yen. Analysis and interpretation of data: all authors. Drafting the article: Ding. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Sheehan. Statistical analysis: Ding, Starke, Xu, Sun. 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: , Brown RD Jr, Wiebers DO, Forbes G, O Fallon WM, Piepgras DG, Marsh WR, et al: The natural history of unruptured intracranial arteriovenous malformations. J Neurosurg 68: , de Oliveira E, Tedeschi H, Raso J: Comprehensive management of arteriovenous malformations. Neurol Res 20: , Ding D, Yen CP, Xu Z, Starke RM, Sheehan JP: Radiosurgery for patients with unruptured intracranial arteriovenous malformations. Clinical article. J Neurosurg 118: , Fiehler J, Stapf C: ARUBA beating natural history in unruptured brain AVMs by intervention. Neuroradiology 50: , Flickinger JC, Kondziolka D, Pollock BE, Maitz AH, Lunsford LD: Complications from arteriovenous malformation radiosurgery: multivariate analysis and risk modeling. Int J Radiat Oncol Biol Phys 38: , Graf CJ, Perret GE, Torner JC: Bleeding from cerebral arteriovenous malformations as part of their natural history. J Neurosurg 58: , Han PP, Ponce FA, Spetzler RF: Intention-to-treat analysis of Spetzler-Martin grades IV and V arteriovenous malformations: natural history and treatment paradigm. J Neurosurg 98:3 7, Heros RC, Korosue K, Diebold PM: Surgical excision of cerebral arteriovenous malformations: late results. Neurosurgery 26: , Kano H, Kondziolka D, Flickinger JC, Park KJ, Parry PV, Yang HC, et al: Stereotactic radiosurgery for arteriovenous malformations, Part 6: multistaged volumetric management of large arteriovenous malformations. Clinical article. J Neurosurg 116:54 65, Katsaridis V, Papagiannaki C, Aimar E: Curative embolization of cerebral arteriovenous malformations (AVMs) with Onyx in 101 patients. Neuroradiology 50: , Lawton MT: Spetzler-Martin Grade III arteriovenous malformations: surgical results and a modification of the grading scale. Neurosurgery 52: , Maruyama K, Kawahara N, Shin M, Tago M, Kishimoto J, Kurita H, et al: The risk of hemorrhage after radiosurgery for cerebral arteriovenous malformations. N Engl J Med 352: , Miller RA, Jankowitz B: Endovascular embolization in combination with radiosurgery for treatment of arteriovenous malformations. Prog Neurol Surg 27:81 88, Mohr JP, Moskowitz AJ, Stapf C, Hartmann A, Lord K, Marshall SM, et al: The ARUBA trial: current status, future hopes. Stroke 41:e537 e540, Mounayer C, Hammami N, Piotin M, Spelle L, Benndorf G, Kessler I, et al: Nidal embolization of brain arteriovenous malformations using Onyx in 94 patients. AJNR Am J Neuroradiol 28: , Ondra SL, Troupp H, George ED, Schwab K: The natural history of symptomatic arteriovenous malformations of the brain: a 24-year follow-up assessment. J Neurosurg 73: , Pandey P, Marks MP, Harraher CD, Westbroek EM, Chang SD, Do HM, et al: Multimodality management of Spetzler- Martin Grade III arteriovenous malformations. Clinical article. J Neurosurg 116: , Pollock BE, Kondziolka D, Flickinger JC, Patel AK, Bisson- 968

11 Radiosurgery for Grade III arteriovenous malformations ette DJ, Lunsford LD: Magnetic resonance imaging: an accurate method to evaluate arteriovenous malformations after stereotactic radiosurgery. J Neurosurg 85: , Saatci I, Geyik S, Yavuz K, Cekirge HS: Endovascular treatment of brain arteriovenous malformations with prolonged intranidal Onyx injection technique: long-term results in 350 consecutive patients with completed endovascular treatment course. Clinical article. J Neurosurg 115:78 88, Schwyzer L, Yen CP, Evans A, Zavoian S, Steiner L: Long-term results of gamma knife surgery for partially embolized arteriovenous malformations. Neurosurgery 71: , Spetzler RF, Martin NA: A proposed grading system for arteriovenous malformations. J Neurosurg 65: , Spetzler RF, Ponce FA: A 3-tier classification of cerebral arteriovenous malformations. Clinical article. J Neurosurg 114: , 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. Clinical article. 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, van Rooij WJ, Sluzewski M, Beute GN: Brain AVM embolization with Onyx. AJNR Am J Neuroradiol 28: , Viñuela F, Dion JE, Duckwiler G, Martin NA, Lylyk P, Fox A, et al: Combined endovascular embolization and surgery in the management of cerebral arteriovenous malformations: experience with 101 cases. J Neurosurg 75: , Wegner RE, Oysul K, Pollock BE, Sirin S, Kondziolka D, Niranjan A, et al: A modified radiosurgery-based arteriovenous malformation grading scale and its correlation with outcomes. Int J Radiat Oncol Biol Phys 79: , Yen CP, Matsumoto JA, Wintermark M, Schwyzer L, Evans AJ, Jensen ME, et al: Radiation-induced imaging changes following Gamma Knife surgery for cerebral arteriovenous malformations. Clinical article. J Neurosurg 118:63 73, Yen CP, Sheehan JP, Schwyzer L, Schlesinger D: Hemorrhage risk of cerebral arteriovenous malformations before and during the latency period after GAMMA knife radiosurgery. Stroke 42: , 2011 Manuscript submitted May 22, Accepted December 9, Please include this information when citing this paper: published online January 24, 2014; DOI: / JNS Address correspondence to: Jason Sheehan, M.D., Ph.D., Department of Neurosurgery, University of Virginia Health System, P.O. Box , Charlottesville, VA jsheehan@virginia. edu. 969