Dural Arteriovenous Shunts

A New Classification of Craniospinal Epidural Venous

Anatomical Bases and Clinical Correlations
Sasikhan Geibprasert, MD; Vitor Pereira, MD; Timo Krings, MD, PhD; Pakorn Jiarakongmun, MD;
Frederique Toulgoat, MD; Sirintara Pongpech, MD; Pierre Lasjaunias, MD, PhD

Background and Purpose—The craniospinal epidural spaces can be categorized into 3 different compartments related to
their specific drainage role of the bone and central nervous system, the ventral epidural, dorsal epidural, and lateral
epidural groups. We propose this new classification system for dural arteriovenous shunts and compare demographic,
angiographic, and clinical characteristics of dural arteriovenous shunts that develop in these 3 different locations.
Methods—Three hundred consecutive cases (159 females, 141 males; mean age: 47 years; range, 0 to 87 years) were
reviewed for patient demographics, clinical presentation, multiplicity, presence of cortical and spinal venous reflux, and
outflow restrictions and classified into the 3 mentioned groups.
Results—The ventral epidural group (n⫽150) showed a female predominance, more benign clinical presentations, lower
rate of cortical and spinal venous reflux, and no cortical and spinal venous reflux without restriction of the venous
outflow. The dorsal epidural group (n⫽67) had a lower mean age and a higher rate of multiplicity. The lateral epidural
group (n⫽63) presented later in life with a male predominance, more aggressive clinical presentations, and cortical and
spinal venous reflux without evidence of venous outflow restriction. All differences were statistically significant
(P⬍0.001).
Conclusion—Dural arteriovenous shunts predictably drain either in pial veins or craniofugally depending on the
compartment involved by the dural arteriovenous shunt. Associated conditions (outflow restrictions, high-flow shunts)
may change that draining pattern. The significant differences between the groups of the new classification support the
hypothesis of biological and/or developmental differences in each epidural region and suggest that dural arteriovenous
shunts are a heterogeneous group of diseases. (Stroke. 2008;39:2783-2794.)
Downloaded from http://ahajournals.org by on October 7, 2019

Key Words: classification 䡲 cranial dural arteriovenous fistula 䡲 epidemiology 䡲 epidural veins, clinical aspects
䡲 spinal dural arteriovenous fistula

D ural arteriovenous shunts or fistulas (DAVSs or

DAVFs) are a group of diseases that share involvement
of the epidural space and adjacent dura mater and bony
structures.
Different classifications have been proposed for these
lesions,1–3 but the most widely used are those that focus on
the presence or absence of cortical venous reflux; it is
generally accepted that this angiographic finding constitutes
the major clinical consideration regarding the natural history
and, therefore, the therapeutic strategies for these lesions.4
Until now, the classification systems, being based on the
various topographies, arterial feeders, and the drainage pat-
tern, have been mainly descriptive. The different role played
by each epidural area involved, as revealed during develop-
ment and embryology of the venous system, were not
considered. Issues concerning spinal versus cranial homolo-
gies or classic sex dominance in some localizations were not
specifically addressed.
The basis for the new classification of DAVSs proposed in
this study relies on the characteristics of venous afferent
patterns of the various epidural spaces according to the
evolution and embryological development of the venous
drainage of the central nervous system and surrounding
structures (vertebra, skull base, calvarium) by which 3 dif-
ferent epidural spaces can be differentiated as outlined
subsequently.
Developmental Considerations of the Venous
System of the Brain, Spine, and Adjacent
Bony Structures
The notochord forms within the mesoderm in the third week
and is considered a “patterning” embryonic structure of the

Received February 5, 2008; accepted March 3, 2008.

From the Department of Radiology (S.G., P.J., S.P.), Ramathibodi Hospital, Mahidol University, Bangkok, Thailand; the Service de Neuroradiologie
Diagnostique et Thérapeutique (V.P., T.K., F.T., P.L.), Hôpital de Bicêtre, Le Kremlin-Bicêtre, Paris, France; and the Departments of Neuroradiology
and Neurosurgery (T.K.), University Hospital Aachen, Germany.
Correspondence to Timo Krings, MD, PhD, Department of Neuroradiology, University Hospital Aachen, Pauwelsstr 30, 52057 Aachen, Germany.
E-mail tkrings@ukaachen.de
© 2008 American Heart Association, Inc.
Stroke is available at http://stroke.ahajournals.org DOI: 10.1161/STROKEAHA.108.516757

2783
 2784 Stroke October 2008
neural tube that regulates the development of the surrounding
structures (such as nerves, blood vessels, and somites).
Among other roles, it also controls the arterial versus venous
identification of the major axial blood vessels and specifies a
variety of cell types in forming primitive segments. It extends
from the level of the sacrum to the basisphenoid and adjacent
portion of the sphenoid wings; it will contribute to the future
development of the vertebral bodies and bony structures
within these regions.5
During the third week, angioblasts initially form small cell
clusters (blood islands) within the embryonic and extraem-
bryonic mesoderm. Recent studies have shown that arteries
and veins are in part genetically determined by their endo-
thelial types before the blood starts to flow inside the lumen.6
The neural tube starts to develop into the ventricular spaces
at the fourth week.7 In the same period (corresponding to
Paget’s 5-mm stage embryo), the blood traversing the capil-
lary network at the surface of the brain is drained on either
side by 3 “meningeal” (leptomeningeal) venous plexuses
arranged craniocaudally (anterior, middle, and posterior);
cranially, these open into a corresponding “primary head
sinus” (epidural) and which will further develop by coales-
cence into “primitive sinuses.”8,9
The development of the choroid plexus starts at Paget’s 8-
to 11-mm stage embryo10 with venous drainage into the
median prosencephalic vein, which functions as the venous
drainage route for the early immature brain and choroid
plexuses. Secondarily, it may collect the deep cerebral venous
afferents and basal vein, thereafter becoming the great cere-
Downloaded from http://ahajournals.org by on October 7, 2019
bral vein or vein of Galen.
Development of the intrinsic venous drainage of the spinal
cord starts after closure of the neural tube. Two longitudinal
collector systems form in the subarachnoid space at the dorsal
and ventral surface of the cord, later joining the epidural
space laterally through numerous emissary-bridging veins.
Contrary to arteries, these veins are not embryologically
(metamerically) linked with the spinal nerves and thus do not
follow a nerve root nor do they exit the subarachnoid space
with them. Up to 40% of them exit through a separate dural
foramen between the spinal nerves8; thus, they should not be
confused with the minute radicular veins draining the roots.
They may instead be named emissary-bridging veins (trans-
dural) to differentiate them from the so-called emissary veins
(transosseous), which drain the intracranial venous sinuses
outside of the skull.
The spinal venous system, being the basic one, will exhibit
homologous features at its cranial counterpart for the 3
cephalic vesicles. However, the volume of developing brain
will progressively alter the simple pattern of the spinal and
spinal cord venous drainage. The emissary-bridging vein of
the rhombencephalon is the medullary vein opening laterally
into the condyloid vein at the level of the foramen magnum.
For the mesencephalon, it is the petrous vein that drains the
middle and upper brainstem. Longitudinal pial veins anasto-
moses will allow significant additional cranial contributions
from the basal vein system. At the prosencephalic level, the
base of the brain venous circle (interbasal venous anastomotic
circle)11 and its outlets will represent the most cranial
equivalent of the basic spinal and hindbrain pial venous
system. During development, the earliest venous outlet of the
anastomotic circle corresponds to the lateromesencephalic
vein opening into the petrous vein and in the superior petrosal
sinus as illustrated in the epsilon-shaped drainage in aneu-
rysmal malformations of the vein of Galen.12 Posteriorly, the
great cerebral vein (Galen vein) and anteriorly the deep
sylvian vein openings correspond to phylogenetically more
recent pathways as, respectively, shown by their late or
postnatal development in humans.
During development, the draining veins of the archicere-
bellum (flocculonodular and lateral recess veins) and archice-
rebrum (anterior cerebral, uncate, and olfactory veins) will
open into these older (previously existing) venous systems
(basal vein, lateromesencephalic vein, pontine vein, and
medullary vein) and thereafter through the petrosal vein or
medullary emissary-bridging vein to the superior petrosal
sinus or condyloid vein.
As the intrinsic venous system of the cortex further
develops with its increase in cerebral territory (volume), it
tends to be directed laterally to drain into the primitive
tentorial sinus.9 The embryonic “tentorial sinus” at that point
is likely to represent the most cranial portion of the lateral
epidural venous system of the spine. Subsequently, the
Breschet sinus, superior petrosal sinus, and straight sinuses
are also included in that group. In short, the basal vein
anastomotic circle; its afferent veins; and its mesencephalic,
cavernous, and galenic (telencephalic and diencephalic) out-
lets are the cranial homologs of the emissary-bridging veins
of the hindbrain and spinal cord.11
Considering the volume increase of the cerebrum and
cerebellum (paleo- and neopallium) observed throughout
phylogeny, the venous drainage will finally require additional
outlets, which will be illustrated by the appearance of new
dorsal pathways, like the superior sagittal (for the brain) and
transverse sinuses (for the cerebellum) that are not present at
the spinal level.
The development of the cranial vault sinuses into the
epidural space results from the coalescence of the leptomen-
ingeal vein contribution from the central nervous system with
the osseous venous drainage. The secondary modeling into
the adult-type sinuses is largely dependent on the growth
of the skull.8,9 At the spine and skull, 2 types of bone can be
distinguished, the endochondral type (cartilaginous bone–
vertebral bodies, basioccipital, basisphenoid, and the petro-
mastoid bone) and the intramembranous type (membranous
bone–spinous process, laminae, and cranial vault).13–15 The
creation of the dorsal midline-located sinuses and conflu-
ences is, nevertheless, linked to the falces (cerebri and
cerebelli) and the tentorium cerebelli16; both are associated
with an epidural confluence and coalescence of the venous
vascular spaces at their bony attachment later to become the
superior sagittal sinus, medial occipital sinus, and transverse
sinuses.17 Their role in the venous drainage of the paleo- and
neocerebrum and paleo- and neocerebellum is a recent
phylogenetic acquisition; they will further evolve postnatally
after regression of the medial occipital sinus and maturation
of the jugular bulbs.
The sigmoid sinuses and jugular bulbs are dependent on
the growth of the petrous bone and skull base as they collect
 Geibprasert et al Dural Arteriovenous Shunts 2785

Figure 1. Location of the 3 groups of DAVSs of the new classification in relation to the bony structures. The blue area depicts the ven-
tral epidural group that includes the basioccipital bone, the petrous pyramid, the basisphenoid with its adjacent sphenoid wings, and
Downloaded from http://ahajournals.org by on October 7, 2019

their related dural structures. The orange area relates to the dorsal epidural group, which includes the superior sagittal sinus (SSS), tor-
cular, transverse sinus, medial occipital sinus, and posterior marginal sinus. The green area relates to the lateral epidural group. Intra-
cranial locations are DAVS draining into emissary-bridging veins of the brainstem and their homologs draining deep cerebral structures
such as the condyloid vein at the foramen magnum, the superior petrosal vein, the basal vein, the vein of Galen, the veins of the ante-
rior cranial fossa, and the orbit. Modified with permission from Lasjaunias P, Berenstein A, ter Brugge KG. Surgical Neuroangiography.
1. Clinical Vascular Anatomy and Variations. Berlin: Springer; 2001.

petromastoid veins and do not receive any direct afferent
from the leptomeningeal space. Thus, they are rather behav-
ing like a conduit to bring the transverse and medial occipital
sinus blood to the jugular vein.
The cavernous plexus at birth and the basilar venous plexus
do not drain any cerebral vein either. After birth, the so-called
cavernous capture may bring the remnant of the tentorial
sinus or the deep sylvian vein to open into the cavernous
plexus.12 The sigmoid sinus and cavernous and basilar plex-
uses therefore correspond to the ventrally located venous
plexus of the spine. They share large midline communications
as well as their unique role in draining the spongiosus parts of
the related bones. Because the cranial sutures serve as
intramembranous growth sites,18 the evolution of the dorsal
dural sinuses closely follows development of the adjacent
sutures.
In summary, 3 main observations of the venous develop-
ment of the brain and spinal cord during embryology there-
fore contribute to the generation of the new classification:
1. The venous system of the notochord and corresponding
sclerotome extends from the basisphenoid (cavernous
plexus) to the sacrum and gives rise to the ventral
epidural drainage group. It collects the blood from
spongiosus bony structures and has no primary role in
the drainage of the central nervous system.
2. The dorsal epidural venous space is normally poorly
developed at the spinal level. Presence of dorsally located
dural sinuses intracranially is therefore the major differ-
ence between the venous systems of the brain and spine.
Their formation is linked to the appearance, during
evolution, of the dural falces and tentorium and associ-
ated with the development of the paleo- and neopallial
structures. These sinuses result from the confluence in
the epidural space of 2 different venous systems: the
osseous system draining the cranial vault and the
leptomeningeal system draining the brain.
3. The veins draining the central nervous system (both
spinal and cranial) are not related to the peripheral
nerves, as are the arteries; these “emissary-bridging
veins” join the lateral epidural venous spaces as con-
necting drainage system. This leptomeningeal venous
drainage has no direct confluent communication with
the ventral and dorsal epidural venous plexuses, which
drain the skull and spine.
Although the epidural space changes its configuration,
from the spinal to the intracranial regions at the level of
magnum foramen, it keeps the same analogous characteris-
 2786 Stroke October 2008
Downloaded from http://ahajournals.org by on October 7, 2019
Figure 2. Spinal VE DAVSs (A). Lateral sacral artery angiogram. Anteroposterior view reveals spinal VE DAVSs at the sacrum (B) for-
merly known as “epidural arteriovenous fistula.” Venous drainage is related to the bone and therefore typically drains through the epi-
dural plexus of the vertebral body without perimedullary reflux.
tics. All 3 epidural venous spaces remain compartmented and
converge secondarily to join the jugular, azygos, lumbar, or
sacral, paraspinal venous collectors. We suggest that DAVSs
in these 3 different areas will drain according to the specific
role of the venous system of each region, unless there are
associated conditions such as constraint placed on the venous
outlet or high-flow arteriovenous shunts. These venous outlet
restrictions (ie, occlusion or stenosis) can be either adjacent to
the shunting area, most commonly from thrombosis rather
than endothelial proliferation, and/or remote from the shunt
such as seen in jugular bulb dysmaturation in pediatric
patients.
New Classification
As pointed out previously, there are 3 venous subgroups
according to the relationship of the epidural venous spaces
with the afferent veins from the bone and central nervous
system (Figure 1): the ventral epidural (VE) or
osteo(cartilaginous)-epidural group, the dorsal epidural (DE)
or osteo(membranous)-epidural group, and the lateral epi-
dural (LE) or leptomeningeal– epidural group. Based on these
spaces, DAVSs are therefore classified into the following 3
groups.
Ventral Epidural Shunts
(osteo[cartilaginous]-epidural)
Dural arteriovenous shunts in these regions involve mainly
the epidural space and are in direct contact with the adjacent
osseous structures that they may invade or recruit the blood
supply from (Figures 2 and 3). The venous afferents of these
regions are closely related to the bony structures and they
drain outside the bony limits, thus resulting in no subarach-
noid (spinal or cortical) venous reflux (CVR). However,
some factors may precipitate development of CVR in these
 Geibprasert et al Dural Arteriovenous Shunts 2787
Downloaded from http://ahajournals.org by on October 7, 2019

Figure 3. A, Internal carotid artery and (B) external carotid artery angiograms in lateral views reveal cranial VE DAVSs at the cavernous
sinus in 2 different patients. Venous drainage is related to the bone at the skull base and therefore no CVR occurs as seen in the first
patient (A) unless there is associated outflow restriction (B), ie, thrombosis of the jugular bulb in the second patient (*) causing severe
CVR through the superficial middle cerebral vein.

regions such as thrombosis within the epidural space sur-
rounding, remote, or distal to the shunt. This will produce
reflux in the lateral epidural space as seen in rare cases of
ventral epidural spinal shunts with perimedullary venous
reflux.19,20 CVR can also be encountered in high-flow shunts
forcing the emissary-bridging vein opening after reflux in the
lateral epidural space.21 In this group, the following locations
are encountered: “vertebral body,” basioccipital, sigmoid
sinus, petrous pyramid, basisphenoid and adjacent sphenoid
wings, and related dural structures.
Dorsal Epidural Shunts
(osteo[membranous]-epidural)
At the spinal level, they are exceptional and clinically present
with epidural hematoma (Figures 4 and 5). The venous
pressure within the dural sinuses is usually low, thus ante-
grade, craniofugal flow of the shunts is normally observed.
CVR may occur after associated venous outflow restriction,
or with high-flow shunts; in case of variations of the venous
opening into this system such as presence of an accessory
epidural sinus, distal restriction of the sinus opening will
rapidly produce CVR. In this group, the following DAVS
locations are encountered: “dorsal spinal epidural DAVS,”
marginal sinus (dorsal portion), medial occipital sinus, torcu-
lar, transverse and accessory epidural sinuses, and superior
sagittal sinus.
Lateral Epidural Shunts
At the spinal level, these shunts correspond to the typical
spinal DAVFs and are located lateral where the pial emissary
bridging vein pierces the dura (Figures 6 and 7). Intracranial
locations include DAVS draining into emissary-bridging
veins of the brainstem and their homologs draining deep
cerebral structures. The drainage is always directed to the
cortical or perimedullary veins; therefore, the DAVS devel-
oping there are always considered to be aggressive.22–24 In
this group, the following locations are encountered: spinal
dural arteriovenous shunts, marginal sinus (lateral portion-
foramen magnum) with the emissary-bridging vein to the
condyloid vein, falcotentorial (vein of Galen), petrosal and
 2788 Stroke October 2008
Downloaded from http://ahajournals.org by on October 7, 2019

Figure 4. Spinal DE DAVSs (A). Reconstructed CT angiography (B) and right cervical artery angiogram in anteroposterior (C) and lateral
(D) views reveal a spinal DE DAVS at the cervical level. Note the relationship of the posterior epidural space with the lamina and spi-
nous process (*). Spinal DE shunts are uncommon and lesions within this space often present with spontaneous epidural hematomas.

basitentorial, Breschet sinus, paracavernous region (embryonic
tentorial sinus remnants), intraorbital shunts, and lamina
cribriformis.
The goal of our study was to compare in 300 consecutive
patients the demographic, angiographic, and clinical aspects of
DAVSs that develop in these 3 different epidural spaces and to
assess whether they truly present according to the predicted role
of each epidural space in the venous drainage of the central
nervous system and adjacent bony structures. Although the latter
would confirm the validity of the classification system, the
former may demonstrate that the 3 different DAVSs also
constitute separate clinical entities with possible different age
and sex distributions and pathomechanisms.
Materials and Methods
This is a retrospective review of 300 consecutive patients, 150
patients each from 2 different institutions (Ramathibodi Hospital,
Bangkok, Thailand, during July 1998 to December 2006 and Bicêtre
Hospital, Paris, France, during September 1992 to December 2006)
for patient demographics, clinical symptoms, shunt localization,
multiplicity, presence of CVR, and outflow modifications (stenosis/
occlusion). The study was approved by the local ethics committees
of both hospitals.
The clinical symptoms were separated into 2 groups, benign and
aggressive.25 The benign group consists of incidental diagnosis,
nonspecific headaches, cranial nerve deficits, chemosis/proptosis,
midline bruit, mass lesions, and cardiac insufficiency. Aggressive
symptoms included seizures, intracranial hemorrhage, motor or
sensory deficits, visual field defects, aphasia, global neurological
deficits (dementia, delayed psychomotor development, macrocra-
nia), and other nonhemorrhagic neurological deficits such as
incontinence.
On angiographic terms, benign and aggressive lesions were
defined by absence or presence of CVR, respectively,26 and were
grouped also using the classification systems of Borden2 and
Cognard.3 Borden 1 (sinus drainage only), Cognard I (antegrade
sinus drainage) and IIa (antegrade and retrograde sinus drainage)
 Geibprasert et al Dural Arteriovenous Shunts 2789
Downloaded from http://ahajournals.org by on October 7, 2019

Figure 5. External carotid artery angiograms in lateral views in 2 patients reveal cranial DE DAVSs at the superior sagittal sinus without
(A) and with (B) thrombosis. Drainage is mainly outside the cranial cavity due to relatively low pressure of the dural sinuses and jugular
bulb unless there is a high flow shunt (A) or there is restriction of the venous system (B). Under these circumstances, CVR into the
frontal cortical veins (*) and frontoparietal cortical veins (arrows) are demonstrated.

were considered as “benign” DAVSs, whereas all the higher grades
(which have cortical and spinal drainage with or without sinus
drainage) were grouped as “aggressive” DAVSs.27
Venous outflow restrictions (stenosis/occlusion) were defined as
adjacent to the shunt, remote from it, or both (mixed). In addition, we
evaluated whether the restriction occurred proximal or distal to the
shunt or both. The cavernous sinus lesions were excluded from this
item, because nonvisualization of the outflow pathways could mean
either thrombosis or compartmentalization of the cavernous sinus.
However, if there was stenosis or occlusion of the venous system
elsewhere, they were added and classified as a remote type. The
spinal lesions were also excluded because nonvisualization of the
draining spinal cord veins is part of the diagnosis of this location and
reflects the spinal cord venous congestion leading to neurological
symptoms.28 In addition, thrombosis or fibrosis of pial emissary-
bridging veins of the cord occur with normal aging.8
The DAVSs were classified into the 3 groups of the new
classification. Crosstabulations with the sex, age groups, clinical
symptoms, presence of CVR, and types of venous outflow modifi-
cations were performed.
Statistical significance was calculated for each group using ␹2 and
one-way analysis of variance tests.
Results
Of the 300 patients included in this article, the general patient
data are summarized in Table 1. There were 150 patients
(50%) who had lesions classified to the VE group, 67 patients
(22,3%) in the DE group, and 63 patients (21%) in the LE
group. There were 20 (6.7%) patients with multiplicity of
lesions: 16 had both a VE and a distant DE DAVF, one had
VE ⫹ LE types, and 3 patients had DE ⫹ LE types. When
calculating the statistics to evaluate the differences for each
of the 3 groups (VE, DE, LE), the patients having multiplicity
of lesions were excluded. Refer to supplemental Table I for
details of the Borden and Cognard classification and presence
of CVR.
Male versus Female
There was a female predominance in patients having VE
lesions (117 of 167 [70%]; P⬍0.001) and a strong male
predominance in patients with LE-type lesions (54 of 67
[81%]; P⬍0.001). There was no definite sex predominance
(female:male⫽41:46; P⫽0.225) for the DE lesions.
Age
The mean consultation ages of patients with only VE, DE,
and LE lesions were 51, 28, and 56 years, respectively. The
mean age in the DE group was significantly lower than the
 2790 Stroke October 2008
Downloaded from http://ahajournals.org by on October 7, 2019

Figure 6. Spinal LE DAVSs (A). T8 intercostal artery angiogram in anteroposterior view reveals a spinal LE DAVSs at the thoracic level,
also known as a “typical spinal DAVF” (B) supplied by a radiculomeningeal artery draining directly into a pial emissary bridging vein and
subsequently retrogradely into perimedullary veins (asterisk).

VE and LE lesions (P⬍0.001) with 31 of 67 (46.3%) being in
the pediatric population. Patients with LE-type lesions also
presented later in life with no patients presenting before the
age of 30.
Benign Versus Aggressive Clinical Symptoms
In 150 patients of the VE group, 139 (92%) had benign
clinical symptoms, whereas in the LE group, 54 of 63 (86%)
patients had aggressive clinical symptoms (P⬍0.001). There
was a slight trend to more aggressive symptoms in the DE
group (42 of 67 [63%]) without statistical significance at 99%
CI (P⫽0.038). Concerning the patients with multiplicity of
DAVS, whenever a LE lesion was present in these patients,
the symptoms were aggressive, demonstrating that these
lesions are dominant in terms of risks for the patient, whereas
in only 7 of 16 patients who had both VE and DE shunts,
aggressive clinical symptoms were present.
Multiplicity
The DE shunts had a higher rate of multiple lesions, 31 of 86
(36%), whereas only 20 of 167 (12%) shunts of the VE type
and 6 of 67 (9%) shunts of the LE type had multiple lesions
(P⬍0.001).
Comparison With Cognard/Borden
Because the LE lesions are located within the laterally located
epidural space, which drains primarily into the cortical or
spinal veins, all of them were classified as aggressive and,
 Geibprasert et al Dural Arteriovenous Shunts 2791
Downloaded from http://ahajournals.org by on October 7, 2019

Figure 7. Internal carotid artery angiogram (A) and occipital artery angiogram (B) reveal cranial LE DAVSs in 2 patients at the anterior
cranial fossa (A) and tentorium (B), respectively. In these locations, CVR is always present. Branches of the ophthalmic artery (A) supply
the ethmoidal LE DAVSs with drainage into ectatic venous pouches of the frontal cortical vein before entering the superior sagittal
sinus (arrows). Stylomastoid branches of the occipital artery (B) supply the tentorial LE DAVSs and drain into the basal vein group,
which joins the vein of Galen (arrowheads).

subsequently, had classifications of Types III, IV, and V
according to Cognard and Type 3 according to Borden. If
multiplicity of shunts was present, the lateral epidural loca-
tion always determined the aggressiveness and could there-
fore be determined as the primary risk factor. Drainage
pattern of the cranial VE and DE type of DAVSs was
primarily through the sinuses, resulting in Types I, IIa, IIb,
and IIab according to the Cognard classification and Type 1
and Type 2 according to the Borden classification depending
on outflow restrictions. Refer to supplemental Tables II and
III, available online at http://stroke.ahajournals.org, for a
detailed comparison of the Cognard and Borden classifica-
tions with the new classification scheme.
Cortical/Spinal Venous Reflux
Cortical/spinal venous reflux was less common in patients
with only VE shunts, being present in only 45 of 150 (30%)
of the patients, all of the noncavernous shunts having steno-
sis/thrombosis of the venous outflow. The percentage of CVR
increased in patients with DE and LE shunts, being 43 of 67
(64%) of the only DE shunts and all (100%) of the LE shunts
(P⬍0.001). In the multiple type of VE⫹DE lesions, CVR
was present in 13 of 16 (81%) cases.
Venous Outflow Restriction
In the noncavernous VE lesions, 39 of 52 cases (75%) had
evidence of venous outflow restriction, which occurred sur-
rounding the shunt in 30 of 39 cases (77%; P⬍0.001). Sixty
of 67 cases (90%) with only DE lesions had stenosis/
occlusion of the venous outflow and the majority (35 of 60
[58%]) were of the multiple type that had thrombosis sur-
rounding and remote to the shunt location (P⬍0.001). In the
35 patients with nonspinal LE type of DAVSs, only 6 patients
(17%) had restrictions of venous outflow, being remote from
the shunt location in 4 of those (P⬍0.001).
In the VE group, no cortical or spinal reflux was present
when there was no modification of the venous outflow pattern
(13 of 13 [100%]). In the remaining patients in the VE group,
the venous outflow restriction usually occurred surrounding
the shunt and most of the patients with this type did not have
 2792 Stroke October 2008

Table 1. General Data of the 300 Patients Table 3. Dorsal Epidural Shunts (osteo关membranous兴-epidural)
Demographics/Characteristics No. of Patients No sex predominance
Sex Lower mean age (pediatric group)
Female 159 (53%) Multiplicity
Male 141 (47%) Usually no CVR unless
Age, mean years (range) 47.2 (0–7) years Extensive thrombosis (of the mixed type, both proximal and distal)
Clinical symptoms High-flow shunts
Benign 183 (61%) Special anatomic variations; compartmentalization of the sinusal space if
full coalescence of the venous lumen does not occur
Aggressive 117 (39%)
History of procoagulatory states, sinus thrombosis 21 (7%)
Multiple shunts 57 (19%) same regions such as cavernous aneurysms.29 The characteristics
are summarized in Table 2. Cranial lesions usually present with
Adjacent 19 (6.3%)
benign clinical symptoms attributed to the low rate of CVR. In
Separate 18 (6%)
the spine, these shunts are usually asymptomatic; there have
Bilateral cavernous lesions 20 (6.7%) been only few case reports describing associated perimedullary
CVR 168 (56%) reflux causing congestive myelopathy and stimulated consider-
Venous outflow restriction (VOR) (excluded 93 129/179 (72%) ations about a possible valve-like mechanisms normally imped-
cavernous and 28 spinal lesions) ing retrograde flow from the epidural plexus to perimedullary
Type of VOR veins; however, in our opinion, it is more likely an extensive
Adjacent (surrounding) to the shunt 47 thrombosis of the normal epidural outlets that leads to retrograde
drainage to perimedullary veins as a seldom variant of spinal VE
Remote to the shunt 27
arteriovenous shunts.
Mixed (both) 49
Location of VOR Dorsal Epidural Shunts
Upstream to the shunt 11 There was no significant sex predominance. The characteris-
Distal to the shunt 68 tics are summarized in Table 3. The infantile type and dural
Both 48 sinus malformations of the pediatric DAVSs are within this
group,30,31 therefore significantly reducing the average age of
Downloaded from http://ahajournals.org by on October 7, 2019

patients compared with the other groups in which children are

CVR (27 of 30 [90%]; P⫽0.001). There was also a higher
chance of CVR when the restriction occurred surrounding the
shunting zone (5 of 8 [63%]) or in mixed type patterns (31 of
35 [89%]) in this group than when the restriction occurred
remote to the shunt (6 of 17 [35%]; P⬍0.001). All of the LE
lesions drained directly into the cerebral or spinal cord veins
with CVR and in the majority (29 of 35 [83%]) of the nonspinal
patients, there was no venous outflow restriction. Refer to
supplemental Table IV for tabulation of the data concerning the
type of thrombosis in relation to CVR in each group.
In the DE shunts, the venous stenosis was usually distal to
the shunt or of the mixed type with the rate of CVR being
higher in the mixed type group (20 of 35 [57%] versus 22 of
25 [88%]; P⫽0.001). No statistical significance was found in
the other groups. Refer to supplemental Table V for tabula-
tion of the data concerning the location of thrombosis in
relation to CVR in each group.
Discussion
Ventral Epidural Shunts
There was an approximately 2.3:1 female predominance in
this group, which is also seen in other lesions involving the
Table 2. Ventral Epidural Shunts (osteo关cartilaginous兴-epidural)
Female predominance
No reflux unless
Extensive thrombosis of epidural drainage
High-flow shunts
poorly represented. CVR is more frequently observed in
patients who have venous outflow restriction; the more
extensive the thrombosis or venous occlusion, the higher the
percentage of CVR.
Multiplicity of the shunts was more commonly found in the
DE group (36% increasing up to 46% with the pediatric group
excluded), frequently being within adjacent regions of the
same epidural type. Multifocality of dural arteriovenous
lesions in children has already been reported previously.32
Although the etiology of these shunts is not known, there
seems to be clinical and radiological evidence that they are
evolutionary lesions and it has been suggested that they may
have been induced by venous sump from the sinus draining
the DAVSs.32 In the adult population, multifocality may be
due to venous hypertension, typically from sinus thrombo-
sis,33 although the exact pathomechanism is as yet unclear. In
the literature, it has been reported that the subgroup with
multiple DAVFs had a 3 times higher rate of hemorrhagic
presentation compared with a single DAVFs due to a higher
rate of CVR.34
Lateral Epidural Shunts
There was a strong male predominance, approximately 4:1,
within this group. The characteristics are summarized in
Table 4. Lateral Epidural Shunts
Male predominance
Present at later age
Always aggressive lesion; CVR always present
 Geibprasert et al
Table 4. The patients also present at a later age compared
with the other groups. As previously described in patients
with spinal DAVFs, these shunts most likely manifest them-
selves after spinal cord draining veins have been obliterated
by fibrosis from aging.35,36 The drainage into the cortical or
perimedullary spinal cord veins results in increased pial
venous pressure with a high rate of neurological symptoms
ranging from venous infarctions, global neurological deficits
to hemorrhage in the brain or congestive venous myelopathy
in the cord.37,38 Due to the role played by veins in cranial
locations, the shunts will involve the subarachnoid portion of
the draining cortical vein causing venous ectasias that may be
associated with hemorrhage from rupture of the venous
pouches.22–24
Treatment Considerations
The decision to treat or not to treat still depends on the
presence of cortical or spinal venous reflux regardless of the
symptoms present.26 As multiple other previous reports have
already emphasized, the presence of cerebral or spinal venous
reflux should prompt immediate treatment, because their
natural history is always aggressive with a high rate of
morbidity and mortality.4,27,39
Conclusions
The roles of venous afferents of the 3 epidural spaces
established during embryological development of the central
nervous system and craniospinal structures are different. This
reference to embryology allows analyzing both cranial and
Downloaded from http://ahajournals.org by on October 7, 2019
spinal homologs and, therefore, is the basis for a classifica-
tion system that can be applied to all types of lesions
developing in these spaces.
This new classification has the ability to predict the
drainage of shunts located in different locations based on the
craniospinal venous anatomy and the presence of associated
venous outflow restriction (stenosis/occlusion), usually
thrombosis, which could be either an etiology or comorbidity
of the disease. Because venous thrombosis is uncommon in
the general population, it may reflect an underlying biological
disorder in these patients with possible different geographic
(population) impact.
The significant difference of the patient characteristics (sex
and age) between the groups of the new classification
suggests biological/hormonal differences within the epidural
venous system in each region and may point toward a
different etiology of the 3 different groups of arteriovenous
shunts. Grouping DAVFs into a single entity may hide the
fundamental differences existing in each location. We pre-
sume that the new classification proposed in this study will
contribute to a better understanding of these lesions and to
their respective disease mechanisms.
Source of Funding
SG received a Research Grant of the World Federation of Interven-
tional Neuroradiology (WFITN).
Disclosures
None.
Dural Arteriovenous Shunts 2793
References
1. Barrow DL, Spector RH, Braun IF, Landman JA, Tindall SC, Tindall GT.
Classification and treatment of spontaneous carotid– cavernous sinus
fistulas. J Neurosurg. 1985;62:248 –256.
2. Borden JA, Wu JK, Shucart WA. A proposed classification for spinal and
cranial dural arteriovenous fistulous malformations and implications for
treatment. J Neurosurg. 1995;82:166 –179.
3. Cognard C, Gobin YP, Pierot L, Bailly AL, Houdart E, Casasco A, Chiras
J, Merland JJ. Cerebral dural arteriovenous fistulas: clinical and angio-
graphic correlation with a revised classification of venous drainage.
Radiology. 1995;194:671– 680.
4. van Dijk JM, ter Brugge KG, Willinsky RA, Wallace MC. Clinical course
of cranial dural arteriovenous fistulas with long-term persistent cortical
venous reflux. Stroke. 2002;33:1233–1236.
5. Fleming A, Keynes RJ, Tannahill D. The role of the notochord in
vertebral column formation. J Anat. 2001;199:177–180.
6. Eichmann A, Yuan L, Moyon D, Lenoble F, Pardanaud L, Breant C.
Vascular development: from precursor cells to branched arterial and
venous networks. Int J Dev Biol. 2005;49:259 –267.
7. O’Rahilly R, Muller F. Neurulation in the normal human embryo. Ciba
Found Symp. 1994;181:70 – 82; discussion 82– 89.
8. Lasjaunias P, Berenstein A, ter Brugge KG. Surgical Neuroangiography.
1. Clinical Vascular Anatomy and Variations. Berlin: Springer; 2001.
9. Padget DH. The cranial venous system in man in reference to devel-
opment, adult configuration, and relation to the arteries. Am J Anat.
1956;98:307–355.
10. O’Rahilly R, Muller F. Ventricular system and choroid plexuses of the
human brain during the embryonic period proper. Am J Anat. 1990;189:
285–302.
11. Cullen S, Ozanne A, Alvarez H, Mercier PH, Brassier G, Lasjaunias P.
The anastomotic venous circle of the base of the brain. Interv Neuro-
radiol. 2005;11:325–332.
12. Alvarez H, Garcia Monaco R, Rodesch G, Sachet M, Krings T,
Lasjaunias P. Vein of Galen aneurysmal malformations. Neuroimaging
Clin N Am. 2007;17:189 –206.
13. Friede H. Normal development and growth of the human neurocranium
and cranial base. Scand J Plast Reconstr Surg. 1981;15:163–169.
14. Labrom RD. Growth and maturation of the spine from birth to ado-
lescence. J Bone Joint Surg Am. 2007;89(suppl 1):3–7.
15. Wilson DB. Embryonic development of the head and neck: part 5, the
brain and cranium. Head Neck Surg. 1980;2:312–320.
16. Lasjaunias P, Kwok R, Goh P, Yeong KY, Lim W, Chng SM. A devel-
opmental theory of the superior sagittal sinus(es) in craniopagus. Childs
Nerv Syst. 2004;20:526 –537.
17. Piske RL, Lasjaunias P. Extrasinusal dural arteriovenous malformations.
Report of three cases. Neuroradiology. 1988;30:426 – 432.
18. Opperman LA. Cranial sutures as intramembranous bone growth sites.
Dev Dyn. 2000;219:472– 485.
19. Krings T, Mull M, Bostroem A, Otto J, Hans FJ, Thron A. Spinal epidural
arteriovenous fistula with perimedullary drainage. Case report and patho-
mechanical considerations. J Neurosurg Spine. 2006;5:353–358.
20. Pirouzmand F, Wallace MC, Willinsky R. Spinal epidural arteriovenous
fistula with intramedullary reflux. Case report. J Neurosurg. 1997;87:
633– 635.
21. Goyal M, Willinsky R, Montanera W, terBrugge K. Paravertebral arte-
riovenous malformations with epidural drainage: clinical spectrum,
imaging features, and results of treatment. AJNR Am J Neuroradiol.
1999;20:749 –755.
22. Kai Y, Hamada J, Morioka M, Yano S, Mizuno T, Kuratsu J. Arterio-
venous fistulas at the cervicomedullary junction presenting with sub-
arachnoid hemorrhage: six case reports with special reference to the
angiographic pattern of venous drainage. AJNR Am J Neuroradiol. 2005;
26:1949 –1954.
23. Komotar RJ, Connolly ES Jr, Lignelli AA, Mack WJ, Mocco J, Harbaugh
RE. Clinicoradiological review: bilateral ethmoidal artery dural arterio-
venous fistulas. Neurosurgery. 2007;60:131–135; discussion 135–136.
24. Kothbauer K, Huber P. Dural arteriovenous fistula in the falx cerebri.
Neuroradiology. 1994;36:616 – 618.
25. Lasjaunias P, Chiu M, ter Brugge K, Tolia A, Hurth M, Bernstein M.
Neurological manifestations of intracranial dural arteriovenous malfor-
mations. J Neurosurg. 1986;64:724 –730.
26. Lasjaunias P, ter Brugge K, Chiu M. Dural AVM. Dural AVM.
Neurosurgery. 1985;16:435– 436.
 2794 Stroke October 2008
27. Davies MA, ter Brugge K, Willinsky R, Coyne T, Saleh J, Wallace MC.
The validity of classification for the clinical presentation of intracranial
dural arteriovenous fistulas. J Neurosurg. 1996;85:830 – 837.
28. Krings T, Mull M, Gilsbach JM, Thron A. Spinal vascular malformations.
Eur Radiol. 2005;15:267–278.
29. Stiebel-Kalish H, Kalish Y, Bar-On RH, Setton A, Niimi Y, Berenstein A,
Kupersmith MJ. Presentation, natural history, and management of carotid
cavernous aneurysms. Neurosurgery. 2005;57:850 – 857; discussion
850 – 857.
30. Barbosa M, Mahadevan J, Weon YC, Yoshida Y, Rodesch G, Alvarez H,
Lasjaunias P. Dural sinus malformations (DSM) with giant lakes, in
neonates and infants: review of 30 consecutive cases. Interv Neuroradiol.
2003;9:407– 417.
31. Lasjaunias P, ter Brugge K, Berenstein A. Surgical Neuroangiography. 3:
Clinical and Interventional Aspects in Children, II Edition. Berlin:
Springer; 2006.
32. Garcia-Monaco R, Rodesch G, Terbrugge K, Burrows P, Lasjaunias P.
Multifocal dural arteriovenous shunts in children. Childs Nerv Syst. 1991;
7:425– 431.
33. Kusaka N, Sugiu K, Katsumata A, Nakashima H, Tamiya T, Ohmoto T.
The importance of venous hypertension in the formation of dural arte-
Downloaded from http://ahajournals.org by on October 7, 2019
riovenous fistulas: a case report of multiple fistulas remote from sinus
thrombosis. Neuroradiology. 2001;43:980 –984.
34. van Dijk JM, ter Brugge KG, Willinsky RA, Wallace MC. Multiplicity of
dural arteriovenous fistulas. J Neurosurg. 2002;96:76 –78.
35. Andersson T, van Dijk JM, Willinsky RA. Venous manifestations of
spinal arteriovenous fistulas. Neuroimaging Clin N Am. 2003;13:73–93.
36. Van Dijk JM, ter Brugge KG, Willinsky RA, Farb RI, Wallace MC.
Multidisciplinary management of spinal dural arteriovenous fistulas:
clinical presentation and long-term follow-up in 49 patients. Stroke.
2002;33:1578 –1583.
37. Awad IA, Little JR, Akarawi WP, Ahl J. Intracranial dural arteriovenous
malformations: factors predisposing to an aggressive neurological course.
J Neurosurg. 1990;72:839 – 850.
38. Willinsky R, ter brugge K, Montanera W, Mikulis D, Wallace MC.
Venous congestion: an MR finding in dural arteriovenous malformations
with cortical venous drainage. AJNR Am J Neuroradiol. 1994;15:
1501–1507.
39. van Dijk JM, ter Brugge KG, Willinsky RA, Wallace MC. Selective
disconnection of cortical venous reflux as treatment for cranial dural
arteriovenous fistulas. J Neurosurg. 2004;101:31–35.