RESEARCHHUMANCLINICAL STUDIES

RESEARCHHUMANCLINICAL STUDIES

Effect of Prior Embolization on Cerebral

Arteriovenous Malformation Radiosurgery
Outcomes: A Case-Control Study
Eric K. Oermann, MD*
Dale Ding, MD
Chun-Po Yen, MD
Robert M. Starke, MD, MSc
Joshua B. Bederson, MD*
Douglas Kondziolka, MD,
MSc
Jason P. Sheehan, MD, PhD
*Mount Sinai Health System, Department
of Neurosurgery, New York City, New
York; University of Virginia, Department
of Neurosurgery, Charlottesville, Virginia;
New York University Langone Medical
Center, Department of Neurosurgery,
New York City, New York
Correspondence:
Jason P. Sheehan, MD, PhD,
University of Virginia,
Department of Neurosurgery,
P.O. Box 800212,
Charlottesville, VA 22908.
E-mail: jps2f@virginia.edu
Received, December 7, 2014.
Accepted, March 16, 2015.
Published Online, April 11, 2015.
Copyright 2015 by the
Congress of Neurological Surgeons.

BACKGROUND: Embolization before stereotactic radiosurgery (SRS) for cerebral arteriovenous malformations (AVM) has been shown to negatively affect obliteration rates,
but its impact on the risks of radiosurgery-induced complications and latency period
hemorrhage is poorly defined.
OBJECTIVE: To determine, in a case-control study, the effect of prior embolization on
AVM SRS outcomes.
METHODS: We evaluated a database of AVM patients who underwent SRS. Propensity
score analysis was used to match the case (embolized nidi) and control (nonembolized
nidi) cohorts. AVM angioarchitectural complexity was defined as the sum of the number
of major feeding arteries and draining veins to the nidus. Multivariate Cox proportional
hazards regression analyses were performed on the overall study population to
determine independent predictors of obliteration and radiation-induced changes.
RESULTS: The matching process yielded 242 patients in each cohort. The actuarial
obliteration rates were significantly lower in the embolized (31%, 49% at 5, 10 years,
respectively) compared with the nonembolized (48%, 64% at 5, 10 years, respectively)
cohort (P = .003). In the multivariate analysis for obliteration, lower angioarchitectural
complexity (P , .001) and radiologically evident radiation-induced changes (P = .016)
were independent predictors, but embolization was not significant (P = .744). In the
multivariate analysis for radiologic radiation-induced changes, lack of prior embolization
(P = .009) and fewer draining veins (P = .011) were independent predictors.
CONCLUSION: The effect of prior embolization on AVM obliteration after SRS may be
significantly confounded by nidus angioarchitectural complexity. Additionally, embolization could reduce the risk of radiation-induced changes. Thus, combined embolization and SRS may be warranted for appropriately selected nidi.
KEY WORDS: Embolization, Endovascular procedures, Gamma knife, Intracranial arteriovenous malformations,
Radiosurgery, Stroke, Vascular malformations
Neurosurgery 77:406417, 2015

DOI: 10.1227/NEU.0000000000000772

erebral arteriovenous malformations

(AVMs) are uncommon intracranial vascular lesions that possess an inherent propensity to hemorrhage, although the risk of
rupture varies by the occurrence of prior AVM
hemorrhage and by nidus size, location, and
angioarchitecture.1-5 Because of the neurological
devastation rendered by intracranial hemorrhage
ABBREVIATIONS: AVM, arteriovenous malformation; RBAS, radiosurgery-based AVM score; SRS,
stereotactic radiosurgery; VRAS, Virginia Radiosurgery AVM Scale

406 | VOLUME 77 | NUMBER 3 | SEPTEMBER 2015

www.neurosurgery-online.com

from a ruptured AVM, the primary goal of AVM

intervention is complete obliteration of the nidus
in order to eliminate subsequent hemorrhage
risk.6-12 Stereotactic radiosurgery (SRS) offers
a minimally invasive alternative to surgical resection and curative embolization for the obliteration
of AVMs, especially those situated in deep or
eloquent regions.13-18 However, the risk-tobenefit profile of SRS is unfavorable for large
AVMs, with relatively low rates of obliteration and
high rates of SRS-induced complications.19-24
Partial nidus embolization has become a commonly used adjunct in the multimodality management of AVMs to reduce the volume of large

www.neurosurgery-online.com

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

RADIOSURGERY OUTCOMES FOR EMBOLIZED AVMS

AVMs to facilitate definitive treatment with radiosurgery.25

Previous studies have shown that prior AVM embolization
reduces radiosurgical obliteration rates.26-31 However, the effect
of embolization on the incidence of SRS-induced complications
and the risk of latency period hemorrhage is inconsistently
reported and, thus, poorly understood. Additionally, the mechanisms by which embolization reduces obliteration rates following AVM SRS have not been thoroughly investigated.32 The aim
of this case-control study, which compares AVM patients with
and without prior embolization who were treated with SRS, is to
determine the effect of prior nidal embolization on radiosurgical
outcomes, with respect to obliteration, SRS-induced complications, and latency period hemorrhage.

METHODS
Participant Selection and Study Design
We performed a retrospective evaluation of a prospectively collected,
institutional review board approved database of approximately 1400
AVM patients who were treated with Gamma Knife SRS at a single
institution from 1989 to 2013. The inclusion criteria were (1) patients
with sufficient data regarding prior interventions (ie, surgical resection
and/or embolization), clinical presentation, AVM characteristics, and
post-SRS outcomes, and (2) minimum radiologic follow-up duration of 2
years or complete AVM obliteration on follow-up angiography or
magnetic resonance imaging (MRI). Patients who underwent dose- or
volume-staged SRS were excluded. The case cohort consisted of patients
who underwent initial embolization, followed by SRS. The control cohort
consisted of patients who underwent SRS alone.

Baseline Variables
The patient variables were: sex, age, clinical presentation, and radiologic
and clinical follow-up durations. The AVM variables were: modified
radiosurgery-based AVM score (RBAS), Virginia Radiosurgery AVM
Scale (VRAS), location (eloquent or noneloquent), associated aneurysms,
AVM size (maximum diameter and volume), and Spetzler-Martin
grade.33-35 The AVM size and AVM grading system classification
(Spetzler-Martin grade, RBAS, VRAS) for each patient were based on the
postembolization nidus volume and angioarchitecture. The maximum
dimension and volume of each nidus were determined based on review of
the patients MRI and angiography by a neurosurgeon and neuroradiologist at our institution. Because the primary goal of AVM embolization
was volume reduction of large AVMs (typically a maximum diameter
.3 cm or volume .12 cm3), the postembolization volume of each nidus
was typically less than its initial volume. Eloquent location was defined
based on the Spetzler-Martin grading system.33 Associated aneurysms
included intranidal or perinidal aneurysms. All patients were treated
using the Leksell Gamma Knife (Elekta AB, Stockholm, Sweden).
Details of the SRS procedures performed at the University of Virginia
have been previously reported.36 Patients treated before 1991 had planning
performed with biplane angiography, whereas patients treated after 1991
had volumetric computed tomography (CT) or MRI for supplemental
imaging as well. Planning was performed with the Kula software from May
1989 until June 1994, and with the Gamma Plan software thereafter. Only
the nonembolized, patent (ie, filled with contrast during angiography)
portion of each nidus was targeted with SRS. The SRS variables were:
margin dose, isodose line, and number of isocenters.

NEUROSURGERY

Patient Follow-up and Outcomes Assessment

Radiologic follow-up was obtained by MRI every 6 months for the first
2 years, and then annually thereafter. Obliteration was defined as the
absence of flow voids on MRI or the absence of anomalous arteriovenous
shunting on angiography. Angiography was performed, when possible,
after obliteration was noted on MRI to confirm nidus occlusion. Further
neuroimaging, either CT or MRI, was performed as necessary in patients
with new or worsening neurological symptoms.
Radiation-induced changes were defined as new perinidal T2 signal
change on MRI. Radiation-induced changes were monitored for associated
clinical symptoms and SRS-associated toxicity throughout the duration of
follow-up. Latency period hemorrhage was defined by neuroimaging as any
AVM hemorrhage following SRS treatment, regardless of clinical status.
The annual post-SRS hemorrhage risk was calculated by dividing the
total number of latency period hemorrhage events by the total number of
risk years (interval between SRS treatment and nidus obliteration or
last radiographic follow-up for incompletely obliterated nidi). Clinical
follow-up was obtained by a combination of clinic and hospital visits and
data collected by outside referring institutions and patients local physicians
from the University of Virginia.

Case Matching and Statistical Analysis

Propensity scores were calculated for case and control patients by using
a logistic regression model consisting of the following input variables: sex,
age, follow-up time, RBAS, margin dose, prior AVM hemorrhage,
Spetzler-Martin grade, nidus volume, and maximum nidus diameter.
We matched the case cohort to the control cohort using a 1:1, fuzzy
(tolerance of 0.6 standard deviation), greedy matching algorithm run over
100 iterations, with selection of the maximally matched set for final use.37
The final matched cohort was validated for appropriate matching by
using means testing or x2 testing as appropriate on a per variable basis to
confirm equivalence of individual variables between each cohort. To
account for potential variability due to propensity score matching over
a small sample size, the analysis was confirmed empirically by repetition
with a second matched cohort.
The number of feeding arteries and number of draining veins were
analyzed both as ordinal variables, and as binarized categorical variables
looking at individual or multiple vessels. We quantified AVM angioarchitectural complexity for inclusion in our vascular analysis. Angioarchitectural complexity was defined as the sum of the number of major draining
veins and major feeding arteries, under the assumption that higher flow
may be associated with a greater amount of feeding and draining vessels.
The calculation of angioarchitectural complexity was based solely on
angiography, which was assessed by the treating neurosurgeon and
a neuroradiologist at the time of the Gamma Knife procedure as well as
by one of the coauthors, who is also a neurosurgeon, at the time of data
collection. No formal vessel caliber was used to qualify an artery or vein as
major. In general, each feeding artery proximal and each draining vein
distal to the nidus which were readily discernible on angiography were
considered a major vessel, although the final decision to classify an artery
or vein as major was left to a consensus of the evaluating physicians (ie,
2 neurosurgeons and a neuroradiologist).
Between-group differences were assessed using the log-rank test or x2 test
within each of the study cohorts for categorical data. Ordinal variables were
assessed by using gamma (a type of rank correlation). Time to events was
assessed by using the Wilcoxon signed rank test or Friedman test as
appropriate. Cumulative event rates were calculated according to the
product limit method of Kaplan and Meier with differences in survival

VOLUME 77 | NUMBER 3 | SEPTEMBER 2015 | 407

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

OERMANN ET AL

curves assessed by using the log-rank method. Cox proportional hazards

regression analysis was performed on the overall study population to
construct a multivariable model of obliteration rates by using variables found
to be significant (a = .05) on univariate analysis not including the matching
variables, which were matched on embolization status. Angioarchitectural
complexity was analyzed as part of a separate set of multivariable models, not
inclusive of the number of feeding arteries and number of draining veins,
owing to their linear dependence. A computerized binning algorithm was
used to identify the cutoff for angioarchitectural complexity, below which
AVM obliteration was significantly more likely. To do so, we used the
Minimum Description Length Principle Cut strategy, a supervised binning
algorithm that works by minimizing the interbin informational entropy.38
All reported ranges are interquartile ranges defined as the range between the
25th percentile and the 75th percentile. All reported P values are 2-sided,
with an a of .05. All data management and analyses were conducted using
SPSS 21.0 (IBM, Inc, Armonk, New York) as well as the open source
SCikit-learn library in Python.39

TABLE 1. Comparison of Baseline Patient Characteristics, AVM

Features, and Radiosurgery Parameters Between the Case and
Control Cohortsa
Patient Characteristic
Sex, n (%)
Male
Female
Age, y (IQR)
Radiologic follow-up time,
mo (IQR)
RBAS, median score (IQR)
VRAS, n (%)
0
1
2
3
4
Location, n (%)
Noneloquent
Eloquent
Associated aneurysms
Margin dose, Gy (IQR)
Prior hemorrhage, n (%)
Maximum diameter, cm
(IQR)
Volume, cm3 (IQR)
Spetzler-Martin
classification, n (%)
I
II
III
IV
V

RESULTS
Case and Control Cohorts
Of the 1010 AVM patients in the eligible data set, 242
underwent treatment with a combination of embolization and
SRS (24.0%), whereas 768 underwent treatment with SRS alone
(76.0%) and were eligible to serve as matched controls. The
matching process yielded 242 patients in each of the case and
control cohorts (Figure 1). Patients were predominately embolized with N-butyl cyanoacrylate (79%) and coils (13%). Other
techniques, including Onyx (ethylene vinyl alcohol copolymer,
ev3 Endovascular, Irvine, California), constituted the remaining
9% of the cases. The number of feeding arteries was discernible in
226 nidi in the case cohort (93.4%) and in 110 nidi in the control
cohort (45.5%). The case and control cohorts were equivalent
across all matched variables (Table 1). However, there were
statistically significant differences between the 2 cohorts with

SRS 1 Embo
(n = 242)

SRS Only
(n = 242)

P
Value
.856

122
120
32
56

(50)
(50)
(23-42)
(36-106)

125
117
30
53

(52)
(48)
(21-42)
(31-98)

1.19 (0.97-1.45) 1.17 (0.88-1.52)

8
32
79
84
39

(3)
(13)
(33)
(35)
(16)

6
32
87
85
32

.418
.142
.485
.850

(2)
(13)
(36)
(35)
(13)
.213

90
152
29
20
121
2.7

(37)
(63)
(12)
(18-22)
(50)
(2.0-3.4)

4.6 (2.5-6)

34
67
109
32
0

(14)
(28)
(45)
(13)
(0)

76
166
25
20
123
2.5

(31)
(69)
(10)
(18-23)
(51)
(2.0-3.0)

4 (2.3-5.5)

31
81
87
42
1

.224
.671
.928
.095
.091
.179

(13)
(33)
(36)
(17)
(,1)

AVM, arteriovenous malformation; Embo, embolization; IQR, interquartile range;

SRS, stereotactic radiosurgery; RBAS, radiosurgery-based AVM score; VRAS, Virginia
Radiosurgery AVM Scale.

regard to unmatched AVM and SRS variables (Table 2). Nidi in

the case cohort were more frequently fed by multiple arteries
(63% vs 39%, P = .001) and treated using more isocenters
(median 3 vs 2, P = .008).

FIGURE 1. Flow chart of patient selection process for the case and control
cohorts. AVM, arteriovenous malformation; RBAS, radiosurgery-based AVM
score; SRS, stereotactic radiosurgery.

408 | VOLUME 77 | NUMBER 3 | SEPTEMBER 2015

AVM Obliteration
Of the 484 patients included for analysis in both cohorts, AVM
obliteration was determined by MRI alone in 71 patients (14.7%)
and confirmed by angiography in 207 patients (42.8%), yielding
a cumulative obliteration rate of 57.4% (278/484 patients). The
median time to obliteration for all patients was 94 months (7.8
years). The actuarial obliteration rates for all AVM patients were
22.8% at 3 years, 39.8% at 5 years, and 56.4% at 10 years. For the
AVMs treated with embolization and SRS (case cohort), the
actuarial obliteration rates were 19.4% at 3 years, 30.9% at 5 years,
and 49.0% at 10 years. The cumulative obliteration rate of the case

www.neurosurgery-online.com

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

RADIOSURGERY OUTCOMES FOR EMBOLIZED AVMS

TABLE 2. Angiographic Differences Between Groupsa

Characteristic
Number of major feeding
arteries, n (%)b
1
2
3
Number of draining veins, n (%)
1
2
31
% .1 major feeding arteries, %
% .1 draining veins, %
Median number of isocenters,
n (IQR)

SRS 1 Embo
(n = 242)

SRS Only
(n = 242)

TABLE 3. Analysis of Matched Factors Influencing Obliteration for

All Patientsa,b

P
Value

Patient
Characteristic

,.001
84 (37)
108 (48)
34 (15)

67 (61)
35 (32)
8 (7)

117 (48)
68 (28)
57 (24)
63
52
3 (2-3)

126 (52)
72 (30)
44 (18)
39
48
2 (2-3)

Age
Sex
Prior AVM
hemorrhage
Maximum diameter
Volume
Marginal dose
Follow-up duration
RBAS
VRAS
Spetzler-Martin
Grade

.656

.001
.467
.008

Univariate P
Value
.626
.559
.007
,.001
,.001
,.001
,.001
,.001
,.001
,.001

Univariate Hazard Ratio

(95% CI)
1.00 (0.99-1.02)
0.90 (0.63-1.29)
1.65 (1.15-2.37)
0.93
0.87
1.24
0.99
0.51
0.69
0.66

(0.91-0.95)
(0.83-0.91)
(1.16-1.32)
(0.98-0.99)
(0.34-0.74)
(0.57-0.83)
(0.54-0.81)

Embo, embolization; IQR, interquartile range; SRS, stereotactic radiosurgery.

For arterial analysis, n = 110 for the SRS only patients and n = 226 for SRS 1 Embo
patients. Statistically significant P values (,.05) are in bold.

cohort was 47.5%, with a median time to obliteration of 122

months (10.2 years). For the AVMs treated with SRS alone
(control cohort), the actuarial obliteration rates were 25.4% at 3
years, 48.2% at 5 years, and 62.8% at 10 years. The cumulative
obliteration rate of the control cohort was 67.4%, with a median
time to obliteration of 63 months (5.3 years). Embolization and

FIGURE 2. Kaplan-Meier plots of obliteration rate over time for AVM patients
who underwent treatment with embolization and SRS vs SRS alone. For patients
who underwent treatment with embolization and SRS, the 3-, 5-, and 10-year
obliteration rates were 19.4%, 30.9%, and 49.0%, respectively. For patients
who underwent treatment with SRS alone, the 3-, 5-, and 10-year obliteration
rates were 26.1%, 48.4%, and 63.5%, respectively. AVMs treated with
embolization and SRS had significantly lower obliteration rates (P = .003, logrank test). AVM, arteriovenous malformation; SRS, stereotactic radiosurgery.

NEUROSURGERY

AVM, arteriovenous malformation; CI, confidence interval; RBAS, radiosurgerybased AVM score; VRAS, Virginia Radiosurgery AVM scale.
b
Statistically significant P values (,.05) are in bold.

SRS resulted in decreased rate of AVM obliteration over time

(P = .003; Figure 2).
Univariate Cox proportional hazards regression analysis of the
overall study population for the matched variables found prior AVM
hemorrhage (P = .007), lower nidus maximum diameter
(P , .001), lower nidus volume (P , .001), higher margin
dose (P , .001), shorter follow-up duration (P , .001), lower
RBAS (P , .001), lower VRAS (P , .001), and lower SpetzlerMartin grade (P , .001) to be significantly associated with
obliteration (Table 3). Univariate Cox regression analysis of the
overall study population for the unmatched variables found the lack
of prior embolization (P , .001), fewer isocenters (P , .001),
fewer feeding arteries (P , .001), fewer draining veins (P , .001),
and radiologic evidence of radiation-induced changes (P , .001) to
be significantly associated with obliteration (Table 4).
In the multivariate Cox proportional hazards regression analysis
of the overall study population, which included only unmatched
variables found to be statistically significant in the univariate
analysis, fewer feeding arteries (P = .006), fewer draining veins
(P , .001), lower angioarchitectural complexity (P , .001), and
radiologic evidence of radiation-induced changes (P = .016) were
identified as independent predictors of obliteration (Table 4).
Notably, prior embolization was not predictive of obliteration in
the multivariate model (P = .744). The actuarial obliteration rates
were significantly higher for nidi with a single feeding artery (P =
.006; Figure 3A) and those with a single draining vein (P , .001;
Figure 3B). The actuarial obliteration rates for AVMs with an
angioarchitectural complexity less than 3 were 36.4% at 3 years,
52.3% at 5 years, and 68.0% at 10 years, and for AVMs with an
angioarchitectural complexity of 3 or higher were 13.8% at 3
years, 24.1% at 5 years, and 39.5% at 10 years. Nidi with an

VOLUME 77 | NUMBER 3 | SEPTEMBER 2015 | 409

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

OERMANN ET AL

TABLE 4. Analysis of Nonmatched Factors Influencing Obliteration for All Patientsa,b

Patient Characteristic

Univariate P Value

Prior embolization
Isodose
Number of isocenters
Number of feeding arteries
Number of draining veins
Angioarchitectural complexityc
Radiologically evident RIC

,.001
.057
,.001
,.001
,.001
,.001
,.001

Univariate Hazard
Ratio (95% CI)
0.44
1.02
0.78
0.55
0.55
0.68
2.13

(0.30-0.63)
(0.99-1.05)
(0.69-0.87)
(0.38-0.76)
(0.45-0.68)
(0.58-0.79)
(1.45-3.14)

Multivariate P Value

Multivariate Hazard
Ratio (95% CI)

.744

.549
.006
,.001
,.001
.016

1.06 (0.75-1.5)

0.96 (0.85-1.09)
0.70 (0.55-0.90)
0.68 (0.55-0.84)
0.69 (0.59-0.80)
1.50 (1.08-2.09)

CI, confidence interval; RIC, radiation-induced changes.

Statistically significant P values (,.05) are in bold. All hazard ratios are from a model not inclusive of angioarchitectural complexity, but inclusive of its constituent variables
(number of feeding arteries, number of draining veins).
c
In a separate model without the number of feeding arteries and number of draining veins due to its colinearity dependency on those variables.
b

angioarchitectural complexity less than 3 had significantly higher

obliteration rates (P , .001; Figure 3C).
Radiosurgery-Induced Complications and Latency
Period Hemorrhage
In the AVM patients treated with embolization and SRS,
cumulative (transient and permanent) radiation-induced changes
were radiologically evident in 88 patients (36.4%), symptomatic
in 15 patients (6.2%), and permanent in 6 patients (2.5%;
Table 5). In the AVM patients treated with SRS alone,
cumulative radiation-induced changes were radiologically evident
in 109 patients (45.0%), symptomatic in 35 patients (14.5%),
and permanent in 6 patients (2.5%). There was a trend toward
a lower incidence of radiologically evident radiation-induced
changes in the case cohort (P = .052). Patients in the case cohort
were significantly less likely to develop symptomatic radiationinduced changes (P = .003), although the incidence of permanent
(P = 1.000) radiation-induced changes was the same in both
cohorts. In the case cohort, the actuarial rates of cumulative
radiologically evident radiation-induced changes were 20% at 1
year, 35% at 2 years, and 42% at 5 years. In the control cohort,
the actuarial rates of cumulative radiologically evident radiationinduced changes were 36% at 1 year, 47% at 2 years, and 52% at
5 years. The development of radiologic radiation-induced
changes over time was significantly lower in the case cohort
(P = .003; Figure 4A).
Univariate Cox proportional hazards regression analysis of the
overall study population for the matched variables found the lack
of prior AVM hemorrhage (P = .001) and larger maximum nidus
diameter (P = .021) to be significantly associated with radiologically evident radiation-induced changes (Table 6). Univariate
Cox regression analysis of the overall study population for the
unmatched variables found the lack of prior embolization (P =
.013) and fewer draining veins (P = .011) to be significantly
associated with radiologic radiation-induced changes (Table 7).
In the multivariate Cox regression analysis of the overall study
population, which included only unmatched variables found to

410 | VOLUME 77 | NUMBER 3 | SEPTEMBER 2015

be statistically significant in the univariate analysis, the lack of

prior embolization (P = .009) and fewer draining veins (P = .011)
were identified as independent predictors of radiologic radiationinduced changes (Table 7). Notably, the number of major
feeding arteries (P = .380) and angioarchitectural complexity (P =
.118) were not significantly associated with radiologic radiationinduced changes. The actuarial rates of cumulative radiologically
evident radiation-induced changes for AVMs with an angioarchitectural complexity less than 3 were 45% at 1 year, 47% at 2
years, and 47% at 5 years, and for AVMs with an angioarchitectural complexity of 3 or higher were 39% at 1 year, 42% at 2
years, and 42% at 5 years (P = .363; Figure 4B).
In the embolized cohort, a total of 30 hemorrhages occurred in
27 patients over a combined latency period of 1498 risk-years,
yielding an annual post-SRS hemorrhage risk of 2.0%. In the
control cohort, a total of 29 hemorrhages occurred in 25 patients
over a combined latency period of 1459 risk-years, yielding an
annual post-SRS hemorrhage risk of 2.0%. Post-SRS cyst
formation occurred in 5 patients in the case cohort (2.1%) and
in 6 patients in the control cohort (2.5%). The incidences of both
post-SRS hemorrhage (P = .769) and post-SRS cyst formation
(P = .760) were similar between the 2 cohorts (Table 5).

DISCUSSION
AVMs continue to pose as one of the most significant challenges
a cerebrovascular surgeon must face because of their complex
physiology, variable angioarchitecture across patients, and propensity to cause neurological injury over the course of their natural
history. Combined therapy with embolization and SRS is
a minimally invasive approach for large or morphologically
complex AVMs that are associated with high surgical risk and
are not amenable to successful treatment with either embolization
or SRS alone. However, the unfavorable effect of embolization on
obliteration rates following SRS reported in prior studies,
including those from the treating institution, has dampened the
enthusiasm for combining the 2 modalities.26-28,30 By quantifying

www.neurosurgery-online.com

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

RADIOSURGERY OUTCOMES FOR EMBOLIZED AVMS

the angioarchitectural complexity of an AVM nidus (ie, the sum

of the number of major feeding arteries and draining veins), we
present new evidence that supports the use of combined
embolization and SRS for appropriately selected AVMs.
Risks and Benefits of AVM Embolization
The role of AVM embolization varies widely across institutions.
For compact nidi with favorable angioarchitecture, embolization
alone can be curative. Saatci et al40 reported a 51% complete
occlusion rate in 350 AVMs embolized with the permanent
embolic agent Onyx. The permanent morbidity and mortality
rates were 7.1% and 1.4%, respectively, and the recanalization rate
of completely occluded nidi at a mean follow-up duration of 47
months was 1.1%. However, typical curative embolization rates are
substantially lower, at 10% to 20%.41-46 Given the relatively
modest complete obliteration rate with embolization compared
with surgical resection or SRS, and that the combined morbidity
and mortality rates for embolization have been reported in excess
of 10%, the AVM embolization is used as an adjunctive therapy
at the majority of centers.46-48 As endovascular technology and
liquid embolic agents continue to improve, the role of
embolization in the treatment of AVMs will require reevaluation in future studies.49 In the current series, we tended to use
pre-SRS embolization as a means of reducing the nidus volume
to a more manageable level for single-session SRS (typically
a residual nidus volume of 12 cm3 or less) and to obliterate highrisk features (eg, perinidal aneurysms) before SRS.

FIGURE 3. Kaplan-Meier plots of obliteration rate over time,

stratified by nidus angioarchitectural features. A, the actuarial
obliteration rates at 3, 5, and 10 years for AVMs with a single
feeding artery were 46.1%, 52.2%, and 60.8%, respectively, and
for AVMs with multiple feeding arteries were 33.8%, 38.1%,
and 51.5%, respectively. Nidi with a single feeding artery had
significantly higher obliteration rates (P = .006, log-rank test). B,
the actuarial obliteration rates at 3, 5, and 10 years for AVMs
with a single draining vein were 30.7%, 49.4%, and 67.1%,
respectively, and for AVMs with multiple draining veins were
14.3%, 29.6%, and 44.5%, respectively. Nidi with a single
draining vein had significantly higher obliteration rates (P ,
.001, log-rank test). C, the actuarial obliteration rates at 3, 5, and
10 years for AVMs with an angioarchitectural complexity less than
3 were 36.4%, 52.3%, and 68.0%, respectively, and for AVMs
with an angioarchitectural complexity of 3 or higher were 13.8%,
24.1%, and 39.5%, respectively. Nidi with an angioarchitectural
complexity less than 3 had significantly higher obliteration rates
(P , .001, log-rank test). AVM, arteriovenous malformation.

NEUROSURGERY

Mechanisms of Radiosurgical Failure in

Embolized AVMs
A number of different causes for the detrimental effect of
embolization on AVM obliteration rates following treatment with
SRS have been proposed, but no single theory has prevailed.32
Despite the capability to selectively catheterize individual AVM
feeding vessels, the ability of a neurointerventionalist to control
the distribution of an embolic agent is limited. Additionally, the
radiodensity of certain embolic agents, such as Onyx, may hinder
angiographic definition of a nidus. Thus, AVM embolization can
increase the difficulty of SRS treatment planning by creating an
irregular target with obscured angioarchitecture.50
Embolization has been shown to promote neoangiogenesis by
increasing the expression of hypoxia-inducible factor-1a and
vascular endothelial growth factor, although whether this occurs
to an appreciable degree to achieve clinical relevance is
unknown.51,52 The alteration of radiosurgical beams, either
absorption or scattering, by embolic agents has not been
substantiated by in vitro studies, which have shown negligible
dose reduction following beam penetration of embolic agents.53
It is likely that inherent distinctions between the vascular biology
of embolized vs nonembolized AVMs contribute to the differences in SRS outcomes. However, in totality, the mechanisms
underlying the lower obliteration rates of embolized AVMs
remain largely hypothetical.

VOLUME 77 | NUMBER 3 | SEPTEMBER 2015 | 411

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

OERMANN ET AL

TABLE 5. Summary of Postprocedural Complicationsa,b

Patient Characteristic
Post-SRS hemorrhages, n (%)
Total
1 Hemorrhage
2 Hemorrhages
Radiologically evident RIC
Symptomatic RIC
Total
Headaches
Neurological deficit
Seizures
Permanent RIC, n (%)
Postradiosurgery cyst
formation, n (%)

SRS 1 Embo
(n = 242)

SRS Only
(n = 242)

P
Value
.769

30
24
3
88

(12)
(10)
(1)
(36)

29
21
4
109

(12)
(9)
(2)
(45)

15
3
10
2
6
5

(6)
(1)
(4)
(1)
(2)
(2)

35
8
23
4
6
6

(14)
(3)
(10)
(2)
(2)
(2)

.052
.003

1.000
.760

Embo, embolization; SRS, stereotactic radiosurgery; RIC, radiation-induced

changes.
b
Statistically significant P values (,.05) are in bold.

Role of Combined Embolization and Radiosurgery in

the Management of AVMs
The present study confirms the previously described phenomenon that pre-SRS embolization decreases obliteration rates (P =
.003).27,28 This study also highlights the difficulties in ascertaining
the impact of embolization upon AVMs treated with radiosurgery
because of the significant differences in the angioarchitecture of the
AVMs that undergo combined embolization and SRS vs SRS
alone. Nidi that underwent combined therapy were more
commonly fed by multiple arteries (P , .001) and were treated
with more isocenters (P = .038), suggesting that AVMs in the case
cohort had higher vascular and anatomic complexity than in the
control cohort. This is consistent with the primary goals of
embolization at the treating institution, which are to (1) reduce the
volume of large AVMs (.12 cm3), (2) occlude high-flow feeding
arteries harboring intranidal perinidal aneurysms, and (3) occlude
intranidal arteriovenous shunts, which may be relatively resistant to
radiosurgically induced obliteration. However, it appears that the
postembolization nidi retain a high degree of angioarchitectural
complexity relative to nonembolized nidi.
Similar prior studies that analyzed the effect of AVM embolization on obliteration following SRS did not account for the nidus
angioarchitecture.26-28 We show, for the first time, that the
inclusion of nidus angioarchitecture (ie, angioarchitectural
complexity) in a multivariate model results in a nonsignificant
interaction between embolization and AVM obliteration (P =
.744). Instead, fewer feeding (P = .006) and draining (P , .001)
vessels, lower angioarchitectural complexity (P , .001), and
radiologic radiation-induced changes (P = .016) were independent predictors of obliteration. Although further studies
are necessary to determine the extent to which nidus

412 | VOLUME 77 | NUMBER 3 | SEPTEMBER 2015

angioarchitecture affects obliteration, our study at least suggests

that nidus angioarchitecture may significantly confound the
effect of embolization on SRS-induced AVM obliteration.
In addition to challenging prior associations between embolization and decreased radiosurgical obliteration, our findings
indicate that embolization may abrogate the risk of radiationinduced changes. The development of radiologically evident
radiation-induced changes over time was significantly lower in
the case cohort (P = .003), and there was a trend toward a lower
cumulative incidence of radiologic radiation-induced changes
(P = .052). Furthermore, symptomatic radiation-induced changes
were also significantly less common in patients with embolized
AVMs (P = .003). In the multivariate model, lack of prior
embolization (P = .009) and fewer draining veins (P = .011) were
independent predictors of radiation-induced changes. Yen et al54
previously found that large, unruptured AVMs with a single
draining vein are particularly prone to the development of
radiation-induced changes. It is possible that the presence of an
embolic cast provides a physical separation between normal
cortex and the radiosurgical dose, thus acting as a shield from
radiation-induced changes. Embolization also decreases arterial
flow into the nidus, thereby relieving venous congestion, which
has also been suggested as an etiology of radiation-induced
changes.
Although the radiobiology of radiation-induced changes is
incompletely understood, they are believed to result from gliosis,
arteriosclerosis, and other reactive vascular and parenchymal
changes in response to radiation.55-57 This is a radiobiological
model similar to that of AVM obliteration, wherein endothelial
disruption and radiation-induced arteriopathy is thought to be the
primary mechanism of SRS-induced obliteration.56-60 Clinical
evidence of this potentially shared radiobiology can be found in the
present results, which show that radiation-induced changes are
predictive of AVM obliteration. However, this appears to be an
embolization-dependent effect. Given that embolization resulted
in both a decrease in the overall incidence of radiation-induced
changes and the symptomatic severity of radiation-induced
changes and the correlation between radiation-induced changes
and obliteration, a common mechanism may underlie the 2 SRSinduced phenomena.
Although the precise mechanisms underlying radiation-induced
changes after AVM SRS are unknown, venous congestion has been
proposed as a potential etiology, secondary to SRS-induced
endothelial damage and resultant occlusion of an AVMs draining
veins.61 A similar mechanism, specifically intimal proliferation
after SRS-induced endothelial injury, results in progressive
occlusion of an AVMs feeding arteries and eventual obliteration
of the nidus.57,62 Thus, embolization-mediated devascularization
of an AVM may abrogate the effect of SRS on the venous
component of the nidus to a greater extent than the arterial
component, hence resulting in a significant reduction of radiationinduced changes but similar rates of obliteration, after adjusting for
confounding factors, such as angioarchitectural complexity. One
should also consider the temporal differences in the development

www.neurosurgery-online.com

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

RADIOSURGERY OUTCOMES FOR EMBOLIZED AVMS

NEUROSURGERY

VOLUME 77 | NUMBER 3 | SEPTEMBER 2015 | 413

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

OERMANN ET AL

FIGURE 4. Kaplan-Meier plots of cumulative radiologically evident radiation-induced changes (RIC) rate over
time. A, AVM patients who underwent treatment with combined embolization and SRS had significantly lower
incidences of radiologic radiation-induced changes (P = .003, log-rank test). For patients who underwent treatment
with embolization and SRS, the actuarial rates of radiologic radiation-induced changes at 1, 2, and 5 years were
20%, 35%, and 42%, respectively. For patients who underwent treatment with SRS alone, the actuarial rates of
radiologic radiation-induced changes at 1, 2, and 5 years were 36%, 47%, and 52%, respectively. B, AVM
patients with a nidus angioarchitectural complexity less than 3 had similar incidences of radiologic radiationinduced changes to those with a nidus angioarchitectural complexity of 3 or higher (P = .363, log-rank test). For
AVMs with an angioarchitectural complexity less than 3, the actuarial rates of radiologic radiation-induced changes
at 1, 2, and 5 years were 45%, 47%, and 47%, respectively. For AVMs with an angioarchitectural complexity of 3
or higher, the actuarial rates of radiologic radiation-induced changes at 1, 2, and 5 years were 39%, 42%, and
42%, respectively. AVM, arteriovenous malformation; SRS, stereotactic radiosurgery.

of radiation-induced changes (typical interval, 6-18 months after

SRS) compared with the attainment of nidal obliteration (typical
interval, 2-3 years after SRS). It is conceivable that embolization
preferentially attenuates SRS-induced effects that occur earlier in
the postradiosurgery interval (ie, radiation-induced changes), thus
supporting the findings of our statistical analyses.
Despite the goal of targeting rupture-prone components of the
nidus with embolization, there has yet to be any evidence to
support a reduction in the natural history hemorrhage risk
following embolization. In fact, data extrapolated from 2 recent
prospective, multicenter studies suggests that partial embolization
may, in fact, increase the hemorrhage risk of AVMs.63,64 In our
series, the incidence of post-SRS hemorrhage was similar between
embolized and nonembolized nidi (P = .769). Additionally, the
annual post-SRS hemorrhage risk was the same in both cohorts
(2.0%) and not appreciably different from the natural history of
untreated AVMs.65-68
Based on the present study, it may be time to revisit the issue of
preradiosurgery embolization of AVMs. Multimodal therapy with
embolization and SRS subjects the patient to the cumulative risks

TABLE 6. Analysis of Matched Factors Influencing Radiologic RIC

for All Patientsa,b
Patient Characteristic

Univariate P Value

Age
Sex
Prior AVM hemorrhage
Maximum diameter
Volume
Margin dose
Follow-up duration
RBAS
VRAS
Spetzler-Martin grade

.207
.704
.001
.021
.139
.253
.218
.703
.609
.531

Univariate Hazard
Ratio (95% CI)
1.01
0.92
0.53
1.03
0.99
0.98
0.99
0.93
0.96
0.94

(0.99-1.02)
(0.63-1.34)
(0.36-0.77)
(1.00-1.05)
(0.94-1.04)
(0.92-1.03)
(0.99-1.00)
(0.63-1.36)
(0.83-1.11)
(0.76-1.15)

AVM, arteriovenous malformation; CI, confidence interval; RIC, radiation-induced

changes; RBAS, radiosurgery-based AVM score; VRAS, Virginia Radiosurgery AVM
Scale.
b
Statistically significant P values (,.05) are in bold.

414 | VOLUME 77 | NUMBER 3 | SEPTEMBER 2015

of both procedures, which can grow significantly over time if

multiple embolization sessions are required. However, embolization also appears to confer protection against radiation-induced
changes, which may offset some of the additional risk associated
with embolization. It also remains an open question as to what
extent the difference in outcomes is due to the effects of
embolization vs pretreatment differences in AVM vasculature,
flow rate, and volume. A detailed study including preembolization
AVM volume and angioarchitecture will be better able to isolate
the effect of embolization on radiosurgical obliteration. Ultimately, a multimodality cerebrovascular approach must balance
the added risk of one or more embolization procedures with its
potential effects on SRS outcomes. However, the judicious use of
embolization remains a crucial component in the therapeutic
armamentarium for the management of AVMs.
Limitations
This study has a number of limitations. The single-center,
retrospective design of this study subjects it to the selection and
management biases of the institution and its physicians. The
nature of being a tertiary referral center for SRS resulted in a lack of
detailed clinical outcomes, including limited documentation of
the complications related to the embolization procedures. Furthermore, we were unable to account for changes in endovascular
devices and embolic agents throughout the study period. Given
the rapid evolution of these embolization technologies in recent
years, this may restrict the generalizability of our findings. Thus,
our ability to extrapolate the results of this study to the overall
management of AVM patients is incomplete.
Perhaps the most significant limitation of our study, which has
also plagued prior analyses of the same topic, is the inherent bias
toward poorer SRS outcomes for embolized AVMs, owing to their
larger original size, and, in most instances, higher Spetzler-Martin
grade, RBAS, and VRAS. Because the longitudinal nature of this
study and the nature of being a tertiary SRS referral center, we were
unable to determine the angioarchitectural characteristics of the
initial AVMs in the case cohort before embolization. Additionally,
a proportion of the embolizations were performed at other
institutions, which further complicates our ability to determine
the original AVM characteristics.

www.neurosurgery-online.com

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

RADIOSURGERY OUTCOMES FOR EMBOLIZED AVMS

TABLE 7. Analysis of Unmatched Factors Influencing Radiologic RIC for All Patientsa,b
Patient Characteristicb
Prior embolization
Isodose
Number of isocenters
Number of feeding arteries
Number of draining veins
Angioarchitectural complexityc

Univariate P Value
.013
.098
.908
.380
.011
.118

Univariate Hazard
Ratio (95% CI)
0.62
0.98
1.01
1.16
0.77
0.89

Multivariate P Value

Multivariate Hazard
Ratio (95% CI)

.009

.011

0.69 (0.52-0.91)

0.81 (0.68-0.95)

(0.43-0.90)
(0.96-1.00)
(0.91-1.12)
(0.83-1.63)
(0.63-0.94)
(0.77-1.03)

CI, confidence interval; RIC, radiation-induced changes.

Statistically significant P values (,.05) are in bold. All hazard ratios are from a model not inclusive of angioarchitectural complexity, but inclusive of its constituent variables
(number of feeding arteries, number of draining veins).
c
In a separate model without the number of feeding arteries and number of draining veins due to its colinearity dependency on those variables.
b

Another limitation was that our modeling of angioarchitectural

complexity did not account for the presence of intranidal or
perinidal aneurysms, despite the fact that their presence could be
seen as an indicator of a higher flow state. We acknowledge that
quantifying angioarchitectural complexity, by summing the number of major feeding arteries and draining veins, may be overly
simplistic. However, angioarchitectural complexity was not
intended to be a sophisticated feature for modeling AVM
hemodynamics, but rather a simple feature composed of a linear
combination of 2 readily ascertained components of an AVMs
vascular anatomy (ie, number of major feeding arteries and number
of draining veins). We included angioarchitectural complexity as
a variable in the present analysis to potentially guide future studies
that may incorporate quantitation of blood through a nidus into
predictive models of AVM radiosurgery outcomes.
We also note that prior studies have found smaller nidi, with
likely fewer feeding arteries and draining veins than their larger
counterparts, to be associated with higher rupture risks because of
higher intranidal pressures.69 Additionally, only 45% of patients
in the control cohort had information regarding the number of
arterial feeders because of the long time period over which the
data were collected and the variable quality of angiograms,
especially those from patients treated earlier in the study period.
However, even when the number of arterial feeders was excluded
from the multivariable regression model, embolization was still
not a significant independent predictor of obliteration. We
attribute this to the finding that in the analysis for predictors of
obliteration, the more significant factor was radiation-induced
changes, of which embolization was a major statistical determinant. Next, the inclusion of patients with less than 2 years of
radiologic follow-up with AVM obliteration may have biased our
outcomes, although we believe this was mitigated by our
matching process, which accounted for follow-up duration.
Finally, angiographic documentation of obliteration was unavailable for 15% of patients. However, prior studies have shown that
the accuracy of MRI closely parallels that of angiography.70,71

NEUROSURGERY

CONCLUSION
Our analysis suggests that the adverse impact of embolization on
AVM SRS outcomes may be overestimated by prior studies.
Whereas embolized AVMs had lower obliteration rates than nonembolized ones, we found that nidus angioarchitecture may be
a significant confounding factor in the statistical evaluation of the
effect of prior embolization on AVM obliteration after SRS because
of the inverse correlation between nidus angioarchitectural complexity and post-SRS obliteration rates. Thus, higher angioarchitectural complexity may partially account for the lower obliteration rates
observed in embolized AVMs. Furthermore, embolization appears
to confer protection against both radiologic and symptomatic
radiation-induced changes, thereby potentially offsetting some of
the risk inherent to the embolization procedure. Thus, although the
risk-to-benefit profile of combined embolization and SRS remains
controversial, we suggest that its role in the in the multimodality
management of AVMs is warranted, especially for nidi that are not
amenable to safe surgical resection or single-session SRS alone.
Disclosure
The authors have no personal, financial, or institutional interest in any of the
drugs, materials, or devices described in this article.

REFERENCES
1. Al-Shahi R, Warlow C. A systematic review of the frequency and prognosis of
arteriovenous malformations of the brain in adults. Brain. 2001;124(pt 10):1900-1926.
2. Morgenstern LB, Hemphill JC III, Anderson C, et al. Guidelines for the
management of spontaneous intracerebral hemorrhage: a guideline for healthcare
professionals from the American Heart Association/American Stroke Association.
Stroke. 2010;41(9):2108-2129.
3. Pollock BE, Flickinger JC, Lunsford LD, Bissonette DJ, Kondziolka D. Factors
that predict the bleeding risk of cerebral arteriovenous malformations. Stroke.
1996;27(1):1-6.
4. Gross BA, Du R. Natural history of cerebral arteriovenous malformations: a metaanalysis. J Neurosurg. 2013;118(2):437-443.
5. Stapf C, Mast H, Sciacca RR, et al. Predictors of hemorrhage in patients
with untreated brain arteriovenous malformation. Neurology. 2006;66(9):1350-1355.

VOLUME 77 | NUMBER 3 | SEPTEMBER 2015 | 415

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

OERMANN ET AL

6. Choi JH, Mast H, Sciacca RR, et al. Clinical outcome after first and recurrent
hemorrhage in patients with untreated brain arteriovenous malformation. Stroke.
2006;37(5):1243-1247.
7. Lawton MT, Du R, Tran MN, et al. Effect of presenting hemorrhage on outcome
after microsurgical resection of brain arteriovenous malformations. Neurosurgery.
2005;56(3):485-493; discussion 485-493.
8. Ding D, Liu KC. Predictive capability of the Spetzler-Martin versus supplementary
grading scale for microsurgical outcomes of cerebellar arteriovenous malformations. J Cerebrovasc Endovasc Neurosurg. 2013;15(4):307-310.
9. Chen CJ, Chivukula S, Ding D, et al. Seizure outcomes following radiosurgery for
cerebral arteriovenous malformations. Neurosurg Focus. 2014;37(3):E17.
10. Ding D, Yen C, Starke RM, Xu Z, Sheehan JP. Effect of prior hemorrhage on
intracranial arteriovenous malformation radiosurgery outcomes. Cerebrovasc Dis.
2014;39(1):53-62.
11. Ding D, Xu Z, Yen CP, Starke RM, Sheehan JP. Radiosurgery for unruptured
cerebral arteriovenous malformations in pediatric patients. Acta Neurochir (Wien).
2015;157(2):281-291.
12. Przybylowski CJ, Ding D, Starke RM, et al. Seizure and anticonvulsant outcomes
following stereotactic radiosurgery for intracranial arteriovenous malformations.
J Neurosurg. 2015:1-7. Epub 2015 Jan 23.
13. Ding D, Yen CP, Xu Z, Starke RM, Sheehan JP. Radiosurgery for patients with
unruptured intracranial arteriovenous malformations. J Neurosurg. 2013;118(5):
958-966.
14. Ding D, Starke RM, Yen CP, Sheehan JP. Radiosurgery for cerebellar
arteriovenous malformations: does infratentorial location affect outcome? World
Neurosurg. 2014;82(1-2):e209-e217.
15. Ding D, Yen CP, Xu Z, Starke RM, Sheehan JP. Radiosurgery for primary motor
and sensory cortex arteriovenous malformations: outcomes and the effect of
eloquent location. Neurosurgery. 2013;73(5):816-824.
16. Kano H, Kondziolka D, Flickinger JC, et al. Stereotactic radiosurgery for
arteriovenous malformations, part 4: management of basal ganglia and thalamus
arteriovenous malformations. J Neurosurg. 2012;116(1):33-43.
17. Kano H, Kondziolka D, Flickinger JC, et al. Stereotactic radiosurgery for
arteriovenous malformations, Part 5: management of brainstem arteriovenous
malformations. J Neurosurg. 2012;116(1):44-53.
18. Wang YC, Huang YC, Chen HC, et al. Linear Accelerator stereotactic
radiosurgery in the management of intracranial arteriovenous malformations:
long-term outcome. Cerebrovasc Dis. 2014;37(5):342-349.
19. Kano H, Kondziolka D, Flickinger JC, et al. Multistaged volumetric management
of large arteriovenous malformations. Prog Neurol Surg. 2013;27:73-80.
20. Ding D, Yen CP, Starke RM, Xu Z, Sun X, Sheehan JP. Outcomes following
single-session radiosurgery for high-grade intracranial arteriovenous malformations. Br J Neurosurg. 2014;28(5):666-674.
21. Moosa S, Chen CJ, Ding D, et al. Volume-staged versus dose-staged radiosurgery
outcomes for large intracranial arteriovenous malformations. Neurosurg Focus.
2014;37(3):E18.
22. Pan DH, Guo WY, Chung WY, Shiau CY, Chang YC, Wang LW. Gamma knife
radiosurgery as a single treatment modality for large cerebral arteriovenous
malformations. J Neurosurg. 2000;93(suppl 3):113-119.
23. Xiao F, Gorgulho AA, Lin CS, et al. Treatment of giant cerebral arteriovenous
malformation: hypofractionated stereotactic radiation as the first stage. Neurosurgery. 2010;67(5):1253-1259; discussion 1259.
24. Yen CP, Ding D, Cheng CH, Starke RM, Shaffrey M, Sheehan J. Gamma Knife
surgery for incidental cerebral arteriovenous malformations. J Neurosurg. 2014;121
(5):1015-1021.
25. Miller RA, Jankowitz B. Endovascular embolization in combination with radiosurgery
for treatment of arteriovenous malformations. Prog Neurol Surg. 2013;27:81-88.
26. Pollock BE, Flickinger JC, Lunsford LD, Maitz A, Kondziolka D. Factors
associated with successful arteriovenous malformation radiosurgery. Neurosurgery.
1998;42(6):1239-1244; discussion 1244-1237.
27. Andrade-Souza YM, Ramani M, Scora D, Tsao MN, terBrugge K, Schwartz ML.
Embolization before radiosurgery reduces the obliteration rate of arteriovenous
malformations. Neurosurgery. 2007;60(3):443-451; discussion 451-442.
28. Kano H, Kondziolka D, Flickinger JC, et al. Stereotactic radiosurgery for
arteriovenous malformations after embolization: a case-control study. J Neurosurg.
2012;117(2):265-275.
29. Ding D, Yen CP, Starke RM, Xu Z, Sheehan JP. Radiosurgery for ruptured
intracranial arteriovenous malformations. J Neurosurg. 2014;121(2):470-481.

416 | VOLUME 77 | NUMBER 3 | SEPTEMBER 2015

30. Ding D, Yen CP, Starke RM, Xu Z, Sun X, Sheehan JP. Radiosurgery for SpetzlerMartin Grade III arteriovenous malformations. J Neurosurg. 2014;120(4):959-969.
31. Ding D, Yen CP, Xu Z, Starke RM, Sheehan JP. Radiosurgery for low-grade
intracranial arteriovenous malformations. J Neurosurg. 2014;121(2):457-467.
32. Buell TJ, Ding D, Starke RM, Webster Crowley R, Liu KC. Embolizationinduced angiogenesis in cerebral arteriovenous malformations. J Clin Neurosci.
2014;21(11):1866-1871.
33. Spetzler RF, Martin NA. A proposed grading system for arteriovenous
malformations. J Neurosurg. 1986;65(4):476-483.
34. Wegner RE, Oysul K, Pollock BE, et al. A modified radiosurgery-based
arteriovenous malformation grading scale and its correlation with outcomes. Int
J Radiat Oncol Biol Phys. 2011;79(4):1147-1150.
35. 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. 2013;119(4):981-987.
36. Steiner L, Lindquist C, Adler JR, Torner JC, Alves W, Steiner M. Clinical outcome of
radiosurgery for cerebral arteriovenous malformations. J Neurosurg. 1992;77(1):1-8.
37. Dyer M, Frieze A, Pittel B. The Average Performance of the greedy matching
algorithm. Ann Appl Probab. 1993;3(2):526-552.
38. Fayyad UM, Irani KB. Multi-interval discretization of continuous-value attributes for
classification learning. Paper presented at: Proc. Thirteenth International Joint Conference
on Artificial Intelligence; 1993; San Mateo, California. Available at: http://ijcai.org/Past%
20Proceedings/IJCAI-93-VOL2/PDF/022.pdf. Accessed April 9, 2015.
39. Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: machine learning in
Python. J Mach Learn Res. 2011;12:2825-2830.
40. 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. J Neurosurg. 2011;115(1):78-88.
41. Katsaridis V, Papagiannaki C, Aimar E. Curative embolization of cerebral
arteriovenous malformations (AVMs) with Onyx in 101 patients. Neuroradiology.
2008;50(7):589-597.
42. van Beijnum J, van der Worp HB, Buis DR, et al. Treatment of brain arteriovenous
malformations: a systematic review and meta-analysis. JAMA. 2011;306(18):2011-2019.
43. Panagiotopoulos V, Gizewski E, Asgari S, Regel J, Forsting M, Wanke I.
Embolization of intracranial arteriovenous malformations with ethylene-vinyl
alcohol copolymer (Onyx). AJNR Am J Neuroradiol. 2009;30(1):99-106.
44. van Rooij WJ, Sluzewski M, Beute GN. Brain AVM embolization with Onyx.
AJNR Am J Neuroradiol. 2007;28(1):172-177; discussion 178.
45. Gobin YP, Laurent A, Merienne L, et al. Treatment of brain arteriovenous
malformations by embolization and radiosurgery. J Neurosurg. 1996;85(1):19-28.
46. Jayaraman MV, Marcellus ML, Hamilton S, et al. Neurologic complications of
arteriovenous malformation embolization using liquid embolic agents. AJNR Am J
Neuroradiol. 2008;29(2):242-246.
47. Ledezma CJ, Hoh BL, Carter BS, Pryor JC, Putman CM, Ogilvy CS. Complications
of cerebral arteriovenous malformation embolization: multivariate analysis of
predictive factors. Neurosurgery. 2006;58(4):602-611; discussion 602-611.
48. Taylor CL, Dutton K, Rappard G, et al. Complications of preoperative embolization
of cerebral arteriovenous malformations. J Neurosurg. 2004;100(5):810-812.
49. Spiotta AM, James RF, Lowe SR, et al. Balloon-augmented Onyx embolization of
cerebral arteriovenous malformations using a dual-lumen balloon: a multicenter
experience. J Neurointerv Surg. Epub Aug 12 2014.
50. Valle RD, Zenteno M, Jaramillo J, Lee A, De Anda S. Definition of the key target
volume in radiosurgical management of arteriovenous malformations: a new dynamic
concept based on angiographic circulation time. J Neurosurg. 2008;109(suppl):41-50.
51. Sure U, Battenberg E, Dempfle A, Tirakotai W, Bien S, Bertalanffy H. Hypoxiainducible factor and vascular endothelial growth factor are expressed more
frequently in embolized than in nonembolized cerebral arteriovenous malformations. Neurosurgery. 2004;55(3):663-669; discussion 669-670.
52. Sure U, Butz N, Siegel AM, Mennel HD, Bien S, Bertalanffy H. Treatmentinduced neoangiogenesis in cerebral arteriovenous malformations. Clin Neurol
Neurosurg. 2001;103(1):29-32.
53. Mamalui-Hunter M, Jiang T, Rich KM, Derdeyn CP, Drzymala RE. Effect of
liquid embolic agents on Gamma Knife surgery dosimetry for arteriovenous
malformations. Clinical article. J Neurosurg. 2011;115(2):364-370.
54. Yen CP, Matsumoto JA, Wintermark M, et al. Radiation-induced imaging
changes following Gamma Knife surgery for cerebral arteriovenous malformations.
J Neurosurg. 2013;118(1):63-73.

www.neurosurgery-online.com

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited

RADIOSURGERY OUTCOMES FOR EMBOLIZED AVMS

55. Kondziolka D, Lunsford LD, Claassen D, Maitz AH, Flickinger JC. Radiobiology of
radiosurgery: part I. The normal rat brain model. Neurosurgery. 1992;31(2):271-279.
56. Achrol AS, Guzman R, Varga M, Adler JR, Steinberg GK, Chang SD.
Pathogenesis and radiobiology of brain arteriovenous malformations: implications
for risk stratification in natural history and posttreatment course. Neurosurg Focus.
2009;26(5):E9.
57. Schneider BF, Eberhard DA, Steiner LE. Histopathology of arteriovenous
malformations after gamma knife radiosurgery. J Neurosurg. 1997;87(3):352-357.
58. Kamiryo T, Lopes M, Berr S, Lee K, Kassell N, Steiner L. Occlusion of the anterior
cerebral artery after Gamma Knife irradiation in a rat. Acta Neurochir (Wien).
1996;138(8):983-991.
59. Lawton MT, Arnold CM, Kim YJ, et al. Radiation arteriopathy in the transgenic
arteriovenous fistula model. Neurosurgery. 2008;62(5):1129-1139.
60. Mouchtouris N, Jabbour PM, Starke RM, et al. Biology of cerebral arteriovenous
malformations with a focus on inflammation. J Cereb Blood Flow Metab. 2015;35
(2):167-175.
61. Yen CP, Khaled MA, Schwyzer L, Vorsic M, Dumont AS, Steiner L. Early
draining vein occlusion after gamma knife surgery for arteriovenous malformations. Neurosurgery. 2010;67(5):1293-1302; discussion 1302.
62. Chang SD, Shuster DL, Steinberg GK, Levy RP, Frankel K. Stereotactic
radiosurgery of arteriovenous malformations: pathologic changes in resected
tissue. Clin Neuropathol. 1997;16(2):111-116.
63. Mohr JP, Parides MK, Stapf C, et al. Medical management with or without
interventional therapy for unruptured brain arteriovenous malformations (ARUBA):
a multicentre, non-blinded, randomised trial. Lancet. 2014;383(9917):614-621.
64. Al-Shahi Salman R, White PM, Counsell CE, et al. Outcome after conservative
management or intervention for unruptured brain arteriovenous malformations.
JAMA. 2014;311(16):1661-1669.
65. Graf CJ, Perret GE, Torner JC. Bleeding from cerebral arteriovenous malformations as part of their natural history. J Neurosurg. 1983;58(3):331-337.
66. Crawford PM, West CR, Chadwick DW, Shaw MD. Arteriovenous malformations of the brain: natural history in unoperated patients. J Neurol Neurosurg
Psychiatry 1986;49(1):1-10.
67. Brown RD Jr, Wiebers DO, Forbes G, et al. The natural history of unruptured
intracranial arteriovenous malformations. J Neurosurg. 1988;68(3):352-357.
68. 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. 1990;73(3):387-391.
69. Spetzler RF, Hargraves RW, McCormick PW, Zabramski JM, Flom RA,
Zimmerman RS. Relationship of perfusion pressure and size to risk of hemorrhage
from arteriovenous malformations. J Neurosurg. 1992;76(6):918-923.
70. Pollock BE, Kondziolka D, Flickinger JC, Patel AK, Bissonette DJ, Lunsford LD.
Magnetic resonance imaging: an accurate method to evaluate arteriovenous
malformations after stereotactic radiosurgery. J Neurosurg. 1996;85(6):1044-1049.
71. OConnor TE, Friedman WA. Magnetic resonance imaging assessment of cerebral
arteriovenous malformation obliteration after stereotactic radiosurgery. Neurosurgery. 2013;73(5):761-766.

Acknowledgments
The authors would like to acknowledge the late Professor Steiner and Professors Kassell, Jensen, and Evans for their roles in the care and treatment of some
of the patients in this study.

COMMENTS

he authors report the results of a matched case control study examining

outcomes (obliteration rate and radiation-induced complications) for
AVMs treated with SRS with and without prior embolization. The perennial
difficulty in performing this comparison is the fundamentally different
nature of AVMs that are selected for prior embolization as opposed to those
treated with SRS alone. The authors make an elegant effort to overcome, or
at least mitigate, this issue by matching cases based on propensity scoring
incorporating a number of key differentiating features including dose, nidus
volume, RBAS, etc. Based on the direct comparison of cases and controls,
prior embolization was found to be associated with both lower obliteration

NEUROSURGERY

rates and lower incidence of radiation-induced complications. However

subsequent multivariate analysis introducing a surrogate marker of angioarchitecture, a variable the authors term angioarchitectural complexity,
appears to negate the importance of prior embolization, emerging, instead
as a key predictor of obliteration itself.
This series brings to light the complexity of assessing the outcomes of
SRS, and the authors should be congratulated for their in depth and
detailed approach. However, the limitations of the analysis must not be
forgotten. The surrogate marker for AVM complexity is not validated.
The retrospective nature of the data and the differential ascertainment of
features such as angioarchitecural complexity are also major drawbacks.
Thus, conclusions as to the impact of prior embolization on SRS outcome
need to remain circumspect. Prospective data collection, as may become
more widely feasible with adjudicated multicenter registries in the future,
will undoubtedly be important in helping to settle the debate regarding the
effects of prior embolization on SRS efficacy.
Sepideh Amin-Hanjani
Chicago, Illinois

ith this publication, the authors make important contributions

toward our comprehension of the effect of prior embolization on
obliteration rates after stereotactic radiosurgery. Using a newly proposed
measure of angioarchitectural complexity, defined as the sum of the number
of major draining veins and major feeding arteries, the authors demonstrate
that the effect of prior embolization on obliteration after radiosurgery may
be significantly confounded by nidus angioarchitectural complexity.
Relative to prior publications on this subject, the advantages of this
article are:
1. a sound statistical method of case-control design with propensity
score matching
2. a robust volume of more than 200 patients in each cohort
3. multivariate analysis of numerous factors that could confound the
association between prior embolization and radiosurgery outcomes
However, in addition to the limitations offered by the authors, their
analysis suffers from a few pitfalls:
1. Some of their assumptions have not been validated. As defined, the
measurement of angioarchitectural complexity might be too simplistic,
or it might have paradoxical implications. For instance, smaller nidi
have been reported to have higher rupture rates due to higher intranidal
pressures, yet they would likely have fewer feeding arteries and draining
veins.
2. Studies like this inherently lack enough granularity to look at specific
treatment parameters.
For instance, it is not certain how the AVM volume and maximal
dimension were calculated in each case relative to the prior embolized
portion. If the embolization resulted in reduced total volume and only
the filling portion of the nidus were targeted, outcomes would be different than if the embolic material simply occupied portions of the
central volume within the target.
3. The approach leads to some paradoxical results. Prior embolization is
associated with both decreased radiation-induced complications (RIC)
and decreased obliteration rates, yet RIC is an independent predictor of
obliteration success (presumably this effect is not confounded by dose).
Despite these limitations, the article is well written and offers additional
insight to a subject of great importance and uncertainty.
Arun Paul Amar
Los Angeles, California

VOLUME 77 | NUMBER 3 | SEPTEMBER 2015 | 417

Copyright Congress of Neurological Surgeons. Unauthorized reproduction of this article is prohibited