12 Spinal Dural Arteriovenous Fistulae

12.1
12.1.1
12.1.2
12.1.3
12.2
12.3
12.4
12.5
12.5.1
12.5.2
12.5.3
12.5.4
12.5.5

Pathology 850
Macroscopic Appearance 851
Microscopic Appearance 853
Pathological Changes in the Spinal Cord 854
Pathophysiology 854
Clinical Presentation of Spinal Dural Arteriovenous Fistulae 858
Imaging of Spinal Dural Arteriovenous Fistulae 859
Treatment of Spinal Dural Arteriovenous Fistulae 867
Indications 867
Embolization of Spinal Dural Arteriovenous Fistulae: Techniques 867
Results of Spinal Dural Arteriovenous Fistula Embolization 869
Surgery for Spinal Dural Arteriovenous Fistulae 871
PostoperativeFollow-up 872

As opposed to the so-called congenital arteriovenous malformations

(AVMs) involving cord and paraspinal structures, spinal dural arteriovenous fistulae (SDAVFs) are acquired shunts located within or adjacent to
dura along the spinal canal. They are by far the most frequent arteriovenous shunt (AVS) that occurs in older adults. They usually present after
the fourth o r fifth decade of life, with an 85% (5:l) male predominance
(Tables 12.1,12.2). The reason for this male predominance at the spinal
level is not known, and this dominance is reversed in dural and osteodura1 AVSs at the skull base and sphenoid level, where there is a female predominance. The location of the fistula has been reported throughout the
spinal canal, from the sacrum to the level of the foramen magnum. The
venous drainage may be very extensive and reach the intracranial dural
sinuses even if the shunt is at the sacral level (Fig. 12.1) or conversely reach
the thoracic perimedullary venous plexus level from an intracranial shunt.
Table 12.1. Age distribution in 352 SDAVF and SCAVM patients at time of diagnosis

Djindjian et al. 1977

Rosenblum et al. 1987a
Berenstein and Lasjaunias 1992
Symon et al. 1984

SDAVFs

Mean

SCAMVs Mean

130
81
81
60

34-73
25-72
35-87
29-75

52
49
56
57

3-57
4-58
2-42
14-40

26
27
22
31

SDAVF, spinal dural arteriovenous fistula; SCAVM, spinal cord arteriovenous malformation.

850

12 Spinal Dural Arteriovenous Fistulae

Fig. 12.1A-E. Progressive thoracic myelopathy. MRI investigation (A, B) shows evidence of increased signal changes in the cord extending up to the mid-thoracic level
(arrows) and prominent flowvoids,in particular along the dorsal aspect of the spinal
cord. Angiography shows this was caused by a spinal dural arteriovenous fistula fed
by the lateral sacral artery (C), with retrograde venous drainage via the sacral radicular vein (arrows) toward the dorsal perimedullary venous plexus at the conus level
(D, E)

Table 12.2. Sex and age distribution in 172 patients with SDAVF
(Patients) n Male
Djindjian et al. 1977
Merland et al. 1980a
Symon et al. 1984"
Rosenblum et al. 1987a
Berenstein and Lasjaunias 1992
Total

46
13
55
27
31
172

Female

38
6
10
3
49
6
23
4
25
5
145 (85%) 25 (15%)

Mean age
52
59
57
49
56
55

SDAVF, spinal dural arteriovenous fistula.

Includes Kendall's first ten patients.

12.1 Pathology
An SDAVF is an abnormal arteriovenous shunt in the dura, most commonly at the level of the intervertebral foramen (Kendall and Logue 1977;
Merland et al. 1980b; Symon et al. 1984; Rosenblum et al. 1987a). The
arterial supply mostly arises from a dural (radicular) branch of the dorsospinal artery (seeVol. 1) in the region of the intervertebral foramen. The
spinal cord veins (in contrast to the arteries) can pierce the dura quite far
from a nerve root. In this venous disposition, a potential bimetamere

Macroscopic Appearance

851

Fig. 12.2A-C. Spinal dural arteriovenous fistula supplied by intersegmental dural

anastomosis. Angiography of intercostal artery at T6 ( A ) demonstrates radicular
supply (short arrow) towards SDAVF (long arrow) with reflux down the intersegmental anastomosis (double arrows).Angiography of intercostal artery T7 (B) shows
radicular supply to the SDAVF via the same intersegmental anastomosis (double arrows).Injection of liquid glue via T6 ( C )shows deposition of embolic material down
the intersegmental anastomosis as well as across the arteriovenous fistula toward the
perimedullary venous plexus of the spinal cord, resulting in permanent obliteration
of the shunt and good clinical outcome

arterial supply to the dural arteriovenous fistula (DAVF) can be demonstrated (Figs. 12.2,12.3). However, in the great majority of patients, there
is a single extradural arterial pedicle that gives rise to a small (sometimes
microscopic) shunt that is within the dura itself. From this shunt, a highly
tortuous, single draining vein emerges. This vein pierces the dura several
millimeters from the accompanying nerve root (either above or below it)
to reach the perimedullary venous system (Benhaim et al. 1983) and then
produces venous hypertension of the medullary veins (see Sect. 12.2).

12.1 .I MacroscopicAppearance

The nidus of the DAVF can be located anywhere along the dura but is most
commonly situated near the nerve root exit. The abnormal arteriovenous
shunt is usually not readily visible using an extradural surgical approach
(Benhaim et al. 1983).
The spinal radicular veins enter the dura close to the nerve root in 60%
of patients and away from the nerve root in 40%. Shunting of arterialized
blood from the DAVF into the radicular vein causes reversal of flow in this
vessel, resulting in enlargement of the radicular vein in the intradural
space. The ascending vein is usually a single, dilated, tortuous channel
that reaches the perimedullary venous network most frequently on the
dorsal surface of the spinal cord. Patients with both posterior and anterior medullary drainage tend to have more severe symptoms than those

852

12 Spinal Dural Arteriovenous Fistulae

Fig. 12.3A-D. A 48-year-old female patient presented with slowly progressive thoracic myelopathy and MRI compatible with venous congestion from spinal dural arteriovenous fistula (SDAVF). Spinal angiography demonstrates dual supply from T8
(A) and T7 (B) toward the intersegmental dural anastomosis (small arrows), which in
turn supplied the DAVF (arrowhead). Selectivecatheterization of the radicular artery
at the T8 level (C) and embolization with glue resulted in deposition of the embolic
material just proximal to the fistulous communication (arrowhead). Immediate
follow-up angiogram at the T7 level shows the continued opacification of the fistula
(arrowhead) from the radicular artery at that level (D), indicating the lack of venous
penetration of the embolic material. Surgical disconnection was achieved 2 days later
by dividing the radicular vein intradurally, resulting in good clinical recovery

Microscopic Appearance

853

with exclusively posterior drainage. The subsequent direction of the perimedullary venous drainage is most often upward toward the thoracic and
cervical levels.

12.1.2 Microscopic Appearance

The dural (feeding) arteries originating at the sacrolumbar or thoracic

levels often divide into small branches and then merge into a single artery
before entering the shunt. These small arteries show no structural abnormalities.
The large draining radicular vein is always single, intradural, and usually very dilated and tortuous, with irregular partial thickening and variable luminal narrowing. The normal dorsal medullary vein is rarely single
at the thoracolumbar level (see Vol. 1) but normally divides into three
channels of a smaller caliber at the lower mid-thoracic level. Due to this
normal disposition of venous architecture with limited venous outflow
capacity, recruitment of additional venous drainage often involves the
vasa corona and dorsolateral spinal venous channels. Wall thickening of
these venous channels, whether circumferential and/or limited to a cushion-like plaque, is nearly always present. Such thickening frequently involves the intima and/or the media. Atherosclerotic changes and calcifications can sometimes be found in the veins. Stenosis to near complete
occlusion is also found in some segments.
Various stages of thrombosis can be seen, which may range from occlusion to full recanalization. Similar changes have been described as a
frequent incidental finding at autopsy.
In general, the arteriovenous communication is microscopic and therefore may be difficult to identify at surgery. Benhaim et al. (1983) were able
to obtain two specimens of the fistula itself. Examination revealed the vascular lesion to be situated within the dura mater. The structure of these
fistulous vessels was not that of normal veins or arteries. Some of the vessel wall will appear similar to that of arteries, with a lamina elastica interna (LEI), prominent smooth muscle cells, poor connective tissue, and a
thick and rigid wall. The LEI frequently divides into two or more layers of
elastin; muscular atrophy and increased connective tissue were present
in the media. Other vessels looked more like veins without evidence of
anomalous changes in the media. Still others looked like veins with no
LEI, much more connective tissue, and a rather t h i n ~ n dflexible wall, with
modifications such as irregular thickening due to collagen and/or elastin
fibrosis and hypertrophy of smooth muscle cells. Interestingly, Benhaim
et al. (1983) found no calcifications or thromboses in the arteriovenous
fistula and in both specimens they were able to describe a free communication between vessels of different structure. The arteriovenous shunt
could be seen in serial sections, demonstrating two arterial structures
anastomosing and merging with a small-caliber vein through a short,narrow channel.
The histological description, and in particular the appearance of the
draining vein, would suggest evidence of a dystrophic aspect rather than
a dysplastic lesion, with changes secondary to local venous hypertension

854

12 Spinal Dural Arteriovenous Fistulae

resulting from the arteriovenous shunt. The histological changes seen in

these veins are very similar to those seen in saphenous veins used for
coronary bypass grafts. The atherosclerotic and calcific changes in the
veins are also quite compatible with an acquired lesion.

12.1.3 Pathological Changes in the Spinal Cord

Extensive pathological changes may be present within the cord itself, even
though the primary structural abnormality is entirely extramedullary. In
advanced lesions, histological abnormalities are evident throughout the
cord. The changes particularly involve the lateral corticospinal tract and
appear to spread gradually into adjacent portions of the white matter of
the lateral funiculus. More advanced changes progressively involve the anterior gray matter and the posterior columns. Gillilian (1970) pointed out
consistent sparing of the anterior median segment. A typical feature in
long-standing advanced lesions is the appearance of neocapillaries within the cord itself. This particular change has been confused with the presence of an arteriovenous nidus. Our current understanding of the venous
anatomy and the pathophysiology of the disease has taught us that these
neocapillaries are primarily the congested intrinsic venous network of the
spinal cord. This is different from secondary neovascularization, which
will result from chronic and extensive hypoxia secondary to long-standing venous congestion (sprouting angiogenesis from venules) (Folkman
1975). The end stage of this venous congestion, venous ischemia and its
impact on the spinal cord, will be similar to the syndrome described by
Foix and Alajouanine (1926).

12.2 Pathophysiology
In 1972,Manelfe et al. described a glomerulus-like structure that he called
peloton vasculaire, or vascular balls (Fig. 12.4). These are normal vascular
structures, usually situated between two layers of the dura mater. In his
specimen, usually two or more afferent arterioles converged into this vascular ball and drained through a single vein located intradurally. Unfortunately, in Manelfe's specimen the dura mater was transected and the
vein could not be followed. These normally present structures, which we
will call the glomerulus of Manelfe, can routinely be found in the upper
lumbar or dorsal regions but are absent at the cervical level. The striking
resemblance between the glomerulus of Manelfe and the SDAVF was noted by Merland et al. (1980a). The former structures are believed to have
the function of preserving constant venous pressure in the spinal cord, regardless of changes in intra-abdominal or intrathoracic pressure. Therefore, it appears that SDAVFs may result from a loss of the normal physiological control of this system.
However, if this were the sole mechanism responsible for SDAVFs
one would expect the condition to occur far more frequently. Therefore,
additional factors must also contribute to the development of these fistulae.

Pathophysiology

855

Fig. 12.4. Glomerulus of

Manelfe specimen removed
from a normal dura of the
spinal cord, after injection with
a radiopaque material, demonstrates a glomerulus-like structure (arrowhead) situated between two layers of the dura
mater. Note the two afferent arterioles (arrows) reaching into
this vascular'ball, with a single
draining vein that has been cut.
(Courtesy of Prof. C. Manelfe)

Tadie et al. (1985),in their study of the morphological functional anatomy of the spinal cord veins, demonstrated that the lumbar and lower thoracic portions of the spinal cord normally drain cephalad into the radiculospinal veins ,which are small in caliber (see Vol. 1).This makes drainage
from the lumbar and thoracic cord somewhat tenuous and sensitive to
hemodynamic alterations. At the cervical level, symptomatic SDAVFs are
extremely rare, since the venous drainage is divergent and therefore favorable, in contrast to the convergent lumbothoracic venous drainage
(Moss et al. 1989).
During normal physiological changes of intra-abdominal and/or intrathoracic pressure (Valsalva, defecation, respiration, etc.), the pressure
in the spinal cord veins remains constant. Furthermore, Tadie claimed
that it is impossible to inject the spinal veins from the periphery. To explain this, various investigators have postulated the presence of valves in
the spinal venous system (Lazorthes 1978), although no anatomical proof
can be provided.
Tadie, in his histological sections, was also unable to find actual valves
in the spinal veins. He did, however, demonstrate an important narrowing
of the radiculospinal veins at the point where they cross the dura mater.
At this point, the vein loses its own wall, which is replaced by an arachnoid
cuff and by the dura itself. In addition, as the vein enters and exits the
dura, it does so in a zigzag fashion. Tadie postulated that, under normal
physiological conditions, the narrowing and zigzagging opposes any
blood flow from the periphery into the intradural space, so that blood flow
is only permitted in the physiological direction. Such a configuration is a
consistent disposition at the normal sinodural junction. If the venous
pressure were to increase abnormally, then the arachnoid cuff would become engorged and obstruct the vein. In addition, a glomerulus-like structure (similar to the glomerulus of Manelfe) was demonstrated at the point
where the normal draining vein crosses the dura mater. These observations have led to the term "the protective anti-back flow system."

856

12 Spinal Dural Arteriovenous Fistulae

Fig. 12.5A-D. Legend seep. 857

Pathophysiology

857

Fig. 12.5A-F. Spinal dural arteriovenous fistula (SDAVF): anterior spinal artery and
vein circulation. A Mid-arterial phase. B Late phase of the left T10 intercostal artery,
which gives rise to the spinomedullary artery; this phase failed to demonstrate the
radiculospinal vein. C Superselective injection of the right L2 lumbar artery demonstrates the SDAVF (arrowhead), the single ascending draining vein (curved arrows),
and the ascending venous drainage. D Plain film shows the radiopaque acrylic deposition with occlusion of the dural lesion and the first centimeter of the ascending
draining vein (white curved arrow). E, F Postembolization follow-up angiogram of
the anterior spinal axis shows normal arterial (E) and venous (F) circulation time after treatment, with opacification of the radiculomedullary vein (compare to A and B)

During selective spinal angiography, following the injection of the

anterior spinal artery, one can routinely demonstrate the drainage of the
spinal cord into the radiculospinal veins. In SDAVFs, such spinal cord
drainage is delayed (Fig. 12.5) and the venous return of the DAVS usually
takes an ascending course toward the cervical and/oraintracranialregions.
Furthermore, maneuvers that change intra-abdominal or intrathoracic
pressure can aggravate symptoms in this condition. Tadie has proven that
in SDAVFs one can reflux into the spinal veins from extraspinal injections,
testifying to an impaired venous protective system.
In one of our patients, an abnormal shunt at the epidural level was noted as causing root pain at that time. Some 4 years later, with venous
drainage towards the medullary veins, the patient developed a myelopathy.
An additional factor to consider is the requirement that a preexisting
restriction of the venous drainage of the spinal cord and/or meninges may
be present to make the acquired SDAVF symptomatic. The presence of
thrombosis and recanalization seen pathologically is consistent with the
pathophysiology of intracranial dural AVMs.

858

12 Spinal Dural Arteriovenous Fistulae

Thiebot et al. (1986) reported the case of a 24-year-old female patient

with multiple SDAVFs draining into spinal veins without venous restrictions or flow impairment. The patient had exhibited minimal nonprogressive symptoms since childhood and was followed for an additional
5 years. The angiographic studies in this patient showed multiple SDAVFs
with vascular structures similar to the glomerulus of Manelfe. Clear arteriovenous shunting was noted in the angiograms but no associated outflow restrictions were seen. The lesions drained through spinoradicular
veins. This finding is highly suggestive of a congenital anomaly of the normal anti-back flow mechanism described by Tadie et al. (1985).In contrast
to the acquired (symptomatic) type of SDAVF, this lesion was in a very
young patient, the arteriovenous fistulae were multiple (and only in the
thoracolumbar area), the paraspinal venous drainage had a normal appearance, and an unimpaired, normal-to-increased medullary transit
time was present without stagnation.
The usual clinical findings of the slowly progressive mixed motor and
sensory myelopathy are reversible if properly treated in its early stage, although its effects may become irreversible in later stages when it is associated with a necrotizing myelopathy. This myelopathy can be explained
by chronic venous hypertension (Aminoff et al. 1974b; Merland et al.
1980a; Symon et d. 1984).
The arteriovenous shunting into the coronary venous plexus of the
spinal cord may extend to the dorsal and/or ventral surface of the cord.
Extension may also occur upward to the cervical or intracranial dural sinuses. One will not see the radicular spinal veins at the lumbar or thoracic
area draining these lesions or the spinal cord itself. As the pressure in the
venous system increases, it is progressively transmitted toward the intrinsic veins of the spinal cord, mainly those of the posterolateral white matter and lateral corticospinal tract, and subsequently affects the lateral funiculus, posterior column, and anterior gray. The relative sparing of the
anterior median segment of the spinal cord is probably related to the
anatomical organization, which preserves the central veins and of the anterior median spinal vein (Symon et al. 1984).
Venous hypertension reduces the arteriovenous pressure gradient and
decreases tissue perfusion, resulting in progressive hypoxia to the spinal
cord. The raised venous pressure further decreases blood flow by producing progressive intramedullary vasodilatation, with possible progressive exhaustion of autoregulation in the affected areas. Progressive vascular dilatation in this uncontrolled fashion reaching the capillaries results in the transmission of undamped pulsations to the cord, decreased
tissue perfusion with edema formation, and progressive loss of cord tissue function.

12.3 Clinical Presentation

of Spinal Dural Arteriovenous Fistulae
The clinical presentation of DAVFs has been documented in various series
in the literature (Kendall and Logue 1977; Merland et al. 1980a; Djindjian
1978; Symon et al. 1984; Rosenblum et al. 1987a) and was documented

Imaging of Spinal Dural Arteriovenous Fistulae

859

Table 12.3. Symptoms of spinal dural arteriovenous fistulaea

Symptoms

Initial symptoms (%)

Symptoms at diagnosis (%)

Back andlor root pain

Paresis
Impotence
Bowel disturbance
Bladder disturbance
Hemorrhage

28
40

37
88
43
75
85
0

4
5
0

Total number of patients, 172. Based on Djindjian et al. 1977; Merland et al. 1980a;
Symon et al. 1984; Rosenblum et al. 1987a; Berenstein and Lasjaunias 1992.

in the first edition of this book (Vol. 5,Chap. 1) (Tables 12.1-12.3). Subsequently, additional series and case reports have been published (Brunerau et al. 1996;Hurst et al. 1995;van Dijk et al. 2002).
There is a 5:l male-to-female ratio in SDAVFs and age ranges from
25 to 78 years. The vast majority of the patients, however, were older than
50 years of age at the time of presentation, and the average age at presentation was 68 in the van Dijk series (2002).
In van Dijk's series the clinical history showed that the first signs of
DAVFs were spastic gait in 55%, paresthesias in 47%, and pain in 33%. The
time interval between initial symptoms and diagnosis was on average
10.5 months.
At the time of presentation for medical consultation, leg weakness or
paraparesis was documented in 96%, sensory numbness or paresthesia
occurred in 9096,urinary incontinence or retention in 82% bowel problems in 65%, and pain in 55% of patients. Sexual dysfunction is a frequent
presenting symptom but often not discussed and poorly recorded. Intradural hemorrhage, as a rule, does not occur with SDAVFs but has been
reported in exceptional circumstances (Do et al. 1999).

12.4 Imaging of Spinal Dural Arteriovenous Fistulae

Advances in noninvasive imaging have greatly contributed to our ability
to establish the timely diagnosis of spinal DAVFs. The role of plain films,
myelography, and computed tomography has been taken over by MRI,
which reliably demonstrates the presence of signal changes within the
cord, signifying the imaging equivalent of the clinical symptomatology.
The absence of significant mass effect and the nonspecific slight enhancement following intravenous contrast administration all favor the
nonneoplastic or nondemyelinating cause of these MRI findings (Masaryk
et al. 1987;Gilbertson et al. 1995).In addition, abnormal flow voids can be
seen, representing the prominent perimedullary venous system associated with the retrograde venous reflux from the DAVE These findings on
MRI are nonspecific for the actual location of the dural fistula (Figs. 12.1,
12.6) but are highly specific for their inipact. Epidural, dural, and perimedullary AVSs may all produce similar MRI findings, and therefore
their clinical symptoms may also be similar (Figs. 12.7-12.9). Spinal

860

12 Svinal Dural Arteriovenous Fistulae

Fig. 12.6A, B. A 52-year-old

man presented with a 5-month
history of progressive thoracic
myelopathy. MRI ( A )shows
increased signal changes within the thoracic spinal cord
(arrows)and possible flow
voids along posterior aspect of
the cord. Spinal angiogram
with verification of the anterior
spinal axis at the thoracic level
showed delayed venous phase,
requiring a search for a dural
arteriovenous fistula, which, at
the time of a second angiogram
(B) was shown to be located at
the foramen magnum and fed
by the ascending pharyngeal
artery and to be draining
downward toward the cervical
and thoracic levels (arrow)

Fig. 12.7. Epidural arteriovenous shunt at the low thoracic

level (long arrow) with drainage
outward toward the hemiazygos system (short arrow) as
well as toward the epidural
venous plexus within the spinal
canal (double arrows) and associated with retrograde pial
spinal cord venous drainage
(arrowheads)

Imaging of Spinal Dural Arteriovenous Fistulae

861

Fig. 12.8A-C. A 72-year-old

man presented with a 2-month
history of progressive myelopathy. MRI ( A ) demonstrates evidence of prominent flow voids
(arrows) along posterior aspect
of the cord and increased signal at the conus level. Selective
angiography into the lateral
sacral artery (double arrows, B)
shows arterial supply (short
arrows) converging toward
an epidural sacral arteriovenous shunt (arrowhead) draining toward the epidural venous
plexus (open arrowhead).
C Drainage toward the ascending lumbar venous system was
noted as well as reflux toward
the sacral radicular vein, which
drained cranially toward the
conus along the R S1 nerve root
(long arrow)

angiography will be needed to separate out these different etiologies and

establish the exact nidus location of the AVS.Improvements in MRI technique (Farb et al. 2001,2002) have resulted in our ability to noninvasively
demonstrate the actual location of the AVF along the dura, which, when
present, will help focus the spinal angiogram on a specific segmental level
(Figs. 12.10,12.11).
Spinal angiography is done when MRI has demonstrated evidence
compatible with a venous congestive myelopathy as the explanation for
the clinical symptomatology. While in the past the angiogram was done as
part of the investigative process, nowadays it is done with the diagnosis of

862

12 S ~ i n aDural
l
Arteriovenous Fistulae

Fig. 12.9. Paravertebral, epidural arteriovenous malformation presenting with myelopathy, which on MRI (A) shows evidence of increased signal changes within
the spinal cord (arrow).At selective angiography (B, C), this proved to be related to
retrograde radicular venous drainage from the epidural venous plexus (open straight
arrow) to the perimedullary veins of the cord (arrows).The draining vein was surgically clipped (arrows, D) to disconnect the epidural venous system from that of the
spinal cord, leaving the epidural arteriovenous shunt to continue to drain toward the
paravertebral hemiazygos venous system (open curved arrow)

DAVF already strongly suspected or even confirmed by MRI. It can therefore be more efficient and focused on the potential for endovascular treatment. It remains important to assess the blood supply to the spinal cord
and in particular the origin of the main supply to the anterior spinal artery (radicular-medullary arterial supply). This will allow for assessment
of the circulation time within the spinal cord, which should be delayed if
the myelopathy is caused by venous congestion (Fig. 12.5) (Launay 1979;
Merland et al. 1980; Merland and Reizine 1987; Willinsky 1990b). It has
been our routine to iniect nonionic contrast material at 1 ccls for 10 s into
the intercostal or lumbar artery that supplies the anterior spinal axis
(radicular-medullary arterial system). Delayed opacification of the venous
phase (perimedullary and radicular-venous systems) of the region (supplied by the anterior spinal artery) beyond 20 s confirms the probable venous cause of the myelopathy. If the myelopathy is at the cervical level the
cervical anterior spinal axis should be examined rather then the radicular-medullary arterial system at the thoracolumbar level.
Another important reason to examine and to establish the origin of the
anterior spinal artery is that it can be located at the same level as the radicular supply to the dural fistula. This information may preclude safe endovascular treatment of the DAVF (Figs. 12.12,12.13) (Agarwal et al. 1992).
While the majority of the DAVFs are located in the dura adjacent to the
nerve root, occasionally the location is along the dura between two adjacent nerve roots (intersegmental). This will invariably result in dual arterial supply to these types of DAVSs and make it imperative at the time of
treatment to reach the venous outlet of the shunt, as otherwise the shunt
will continue to be fed by the untreated arterial pedicle.

Imaeine of Spinal Dural Arteriovenous Fistulae

863

Fig. 12.10. Example of contrastenhanced MRI axial view,

at the level of the intercostal arteries, demonstrating normal
vascular appearance (A) and
findings in a case of spinal dura1 arteriovenous fistula (SDAVF,
B). A prominent vascular channel enters the spinal canal at
the foramina1 level (long arrow)
and runs toward the posterior
aspect of the spinal cord, showing the draining radicular vein
from an SDAVF at the foramen
level refluxing toward the
dorsal perimedullary venous
plexus of the spinal cord (short
arrow)

Fig. 12.11A, B. Example of contrast-enhanced MRI (coronal

view, A) in a patient investigated for possible spinal dural
arteriovenous fistula (SDAVF),
demonstrating prominent vascular channel (arrow) entering
intervertebral foramen as well
as tortuous prominent channels within the spinal canal.
Selective angiography of the
radicular artery at that same
level demonstrates excellent
correlation with the MRI findings, confirming the presence
of SDAVF (arrow) and the refluxing radicular vein towards
the perimedullary venous
plexus of the spinal cord

While the location of the DAVF can be anywhere along the dura, a cervical localization is extremely uncommon. In the Toronto series (van Dijk
et al. 2002), the most frequent location was mid-thoracic and 70% of
DAVFs were located on the left side. It is also accepted that the location of
the dural shunt may be at the level of the foramen magnum or even intracranial (Figs. 12.6,12.14) and yet present with spinal cord myelopathy
due to the rerouting of the venous drainage of the shunt toward the perimedullary venous system of the spinal cord (Willinsky et al. 1990~;
Mahagne et al. 1992; Bret et al. 1994). Part of the angiographic protocol

864

12 Svinal Dural Arteriovenous Fistulae

Fig. 12.12A-D. A 48-year-old

woman presented with progressive myelopathy of several
months duration. MRI (A)
shows evidence of increased
signal changes within the spinal
cord and prominent flow voids
along the dorsal and ventral
aspect of the cord. Angiogram
at L2 (B)shows evidence of
opacification of spinal dural arteriovenous fistula (SDAVF) on
the left side (arrow). Selective
catheterization of the radicular
branch of the L2 lumbar artery
(C, D) shows that the same
radicular artery supplies the
AVF and the spinal c6rd (radiculospinal artery, arrows), a
contraindication for endovascular treatment

Fig. 12.14A, B. Intracranial dural arteriovenous malformation at the level of the left
vetrous bone and drainage into the anterior and posterior spinal medullary veins.
The patient presented with progressive myelopathy involving the lower extremities
and sphincter dysfunction. A Cervical myelographic
examination demonstrates pro.
- .
minent vascular structures (arrows). B Lateral subtraction angiogram of the superselective injection of the stylomastoid artery (small arrow). The site of fistulization is
at the basal tentorial edge of the cerebellopontine angle (arrowhead), draining into
the petrosal vein (curved arrows) and reaching the anterior and posterior medullary
veins (long arrows). There is also reflux into cerebellar vermian veins and drainage
into the straight sinus (curved arrows). Note filling of the middle meningeal artery
via its tentorial branches

.,

Imaging of Spinal Dural Arteriovenous Fistulae

865

Fig. 12.13A, B. Selective angiography into the radicular branch of the L2 lumbar artery demonstrates supply to both a spinal dural arteriovenous fistula (long arrow, A)
and the anterior spinal axis (radiculomedullary spinal artery, short arrows, B), a contraindication to endovascular treatment

Fig. 12.14A, B. Legend see p. 864

866

12 Spinal Dural Arteriovenous Fistulae

Fig. 12.15A-C. Multifocal spinal dural arteriovenous fistula (SDAVF). A Selective injection of the left T10 fills the radiculomedullary artery and an SDAVF on the level
above the arrowhead, draining into the anterior spinal vein (arrow).B Late phase of
the same injection. Note the descending and ascending drainage (arrows) and the
dilution at the highest level (curved arrow). C Injection of T7 shows a second dural
lesion (double arrowhead) draining cephalad. Note downward drainage to the same
vein that drains the TI0 lesion (long arrow) and the exact site of venous anastomosis
(curved arrow)

Embolization of Spinal Dural Arteriovenous Fistulae: Techniques

867

while investigating patients suspected of harboring a DAVF is to assess

the intra- and extracranial circulation, in particular when no shunt can be
identified at the thoracic, lumbar, or sacral levels.
Multiple DAVFs in the same patient at the same time or occurring over
time are exceptional (Fig. 12.15) (Pierot et al. 1993; Chaloupka et al. 1995)
and they occur much less often than in intracranial DAVFs.
Spontaneous closure of an SDAVF without treatment is extremely rare
(Meder et al. 1995).

12.5 Treatment of Spinal Dural Arteriovenous Fistulae

12.5.1 Indications

The presence of an SDAVF is an excellent indication for treatment in all

patients, as the risk of endovascular or surgical treatment is minimal and
the possible benefits are significant. This applies even to patients presenting with what appear to be fmed neurological deficits.

12.5.2 Embolization of Spinal Dural Arteriovenous Fistulae: Techniques

The main objective of SDAVF treatment is to occlude the draining radicular vein as it exits from the arteriovenous shunt, thereby insuring disconnection of the microfistula from the spinal cord venous system. It is
occluded (at the time of embolization) within the first millimeters of the
exiting vein (Figs. 12.16, 12.17) or (at the time of surgery) prior to its
opening into the pial network (Fig. 12.9). In sacral lesions, the long draining vein can be occluded over several centimeters, as it only reaches the
medullary veins at the conus level. The only significant contraindication
to endovascular treatment of SDAVF occurs in those patients in whom the
anterior or posterior spinal artery originates from the same pedicle as the
SDAVF (Figs. 12.12,12.13) (Merland et al. 1980a,b; Agarwal et al. 1992).In
the great majority of patients, embolization can close the lesion in a very
safe manner.
In view of the small size of the afferent arteries and the fistula itself (in
the 40-60 pm range), only a low-viscosity liquid agent will reach the nidus
and the proximal vein. Particulate agents, such as polyvinyl alcohol (PVA)
or dura mater, will not be effective or may only give temporary good results. Almost invariably their use results in recanalization and recurrence
(Hall et al. 1989; Morgan and Marsh 1989).Particles will not penetrate effectivelyto obtain complete cure and we consider them contraindicated in
the endovascular management of SDAVFs.
Closure of SDAVFs using a liquid embolic agent is best accomplished
with a mixture that will have a long polymerization time, such as '1,
N-butyl-cyanoacrylate (NBCA) with 2/3 iophendylate (lipiodol). The injection can be done with a simple superselective catheter or with a coaxial assembly system. The acrylic-iophendylate mixture can be injected in a
continuous column or using a so-called sandwich or push technique. The
technical goal of therapy is occlusion of both the nidus and the proximal
portion of the afferent vein.

868

12 Spinal Dural Arteriovenous Fistulae

Fig. 12.16A-D. Spinal dural arteriovenous fistula (SDAVF) of the sacrum. A Angiography using a coaxial system with a variable stiffness microcatheter (arrow)demonstrates the arterial component (small arrows) and the site of shunting (arrowhead)
with its ascending venous drainage (curved arrow). Note the retrograde filling of the
contralateral dural supply to the sacrum (arrow).B Later phase of the same injection
with better filling of the right lateral sacral artery (large arrows). C Digital subtraction angiography image of the acrylic deposition at the time of embolization shows
the cast of the SDAVF as well as the proximal aspect of the radicular vein (arrowhead)
and the distal segment of the right lateral sacral contribution. D Follow-up left internal iliac injection. There is no filling of the SDAVF and good opacification of both the
right lateral sacral artery (arrow) and the medial sacral artery (large arrow). This
confirms that there is no fdling of the lesion

Results of Spinal Dural Arteriovenous Fistula Embolization

869

Fig. 12.17. Spinal dural arteriovenous fistula (SDAVF) shown at 3-D digital subtraction angiogram on frontal view (A) and axial view (B) to be located at T12 (left side)
and supplied by the radicular artery (longarrows) and to be draining first toward the
ventral medullary venous plexus (short arrows). C-G see p. 870

As in other anatomical locations, the collateral circulation of the dura

of the involved region must be examined for the presence of natural arterial collaterals: ipsilateral above and below the level of the fistula as well
contralateral at the same level of the DAVF and repeated on the immediate postembolization studies (Fig. 12.3).
The absence of radicular vein opacification at the thoracic level, during
the ascending venous drainage phase of a thoracolumbar or lumbosacral
SDAVF might favor the use of heparin following embolization. Similarly,
patients with descending venous drainage in a thoracic SDAVF may benefit from postembolization anticoagulation therapy for at least several days.
As opposed to the treatment of spinal epidural AVFs,;here is no role for
the venous endovascular approach to SDAVFs (Willinsky et al. 1993).

12.5.3 Results of Spinal Dural Arteriovenous Fistula Embolization

Nimii et al. (1997) (49 cases), Song et al. (2001a) (27 cases), Westphal and
Koch (1999) (47 cases) and van Dijk et al. (2002) (49 cases) reported the
initial success rate of endovascular treatment with liquid adhesive embolic materials to vary between 25% and 90% of all cases. Nimii et al. (1997)
considered penetration of the liquid adhesive into the fistula without penetration into the proximal vein as adequate, which is reflected in their
high recurrence rate of 23%. Song et al. (2001a) reported a failure rate of
25%, but follow-up angiography was done in only 65% of the patients.
Up to 80% of patients will show clinical benefit from this form of treatment, ranging from clinical improvement of various degrees to stabiliza-

870

12 Spinal Dural Arteriovenous Fistulae

Fig. 12.17C-G.
Legend see p. 871

Surgery for Spinal Dural Arteriovenous Fistulae

871

tion. Merland and Reizine (1987) reported on 63 patients treated at Lariboisiere Hospital (27 operated on and 36 embolized). In 50% of patients,
there was a significant improvement of symptoms, in 20% only minor improvement occurred, in 16% the progressive myelopathy stabilized, and in
4% there was an aggravation or continued deterioration. Clinical improvement was noted in days, weeks, or within 3 months. The best results
were seen in patients with the shortest time between onset of symptoms
and treatment. Clinical improvement occurred primarily in motor function, deep sensory function, and strength. Superficial sensation and
sphincter improvement was noted later and was less satisfactory. These
results are similar to those reported by Merland and Reizine (1987),Barth
et al. (1984), and more recently by Song et al. (2001b). Significant improvement in motor and bladder function was also demonstrated by van
Dijk et al. (2002), with a mean follow-up of 32.3 months. Song et al.
(2001b) demonstrated that if the patient was treated within 13 months of
symptom onset, improvement in gait and to a lesser extend bladder function could be anticipated.
Complications associated with endovascular treatment could be related
to the failure to recognize that the same radicular artery supplying the
AVF is also contributing to the supply of the anterior or posterior spinal
arterial systems with inadvertent embolization of this supply.Another potential cause for complication is the inadvertent deposition of the embolic material beyond the nidal-venous junction and into the perimedullary
venous system of the spinal cord, causing thrombosis of the spinal cord
venous system. Both these potentially devastating complications are
avoidable with proper analysis of the anatomical situation and choice and
method of injection of the embolic material. Recent experience by welltrained teams has shown the complication rate associated with endovascular treatment to be extremely low or even absent (van Dijk et al. 2002).

12.5.4 Surgery for Spinal Dural Arteriovenous Fistulae

If endovascular therapy is unsuccessful, surgery should be performed

during the same hospital stay. As indicated above, the goal of surgical
treatment will be the same: disconnection (transection, division) of the
radicular vein as it runs intradurally from the dura toward the spinal cord
perimedullary venous plexus (Fig. 12.13).
4

Fig. 12.17 (continued). Selective angiography in to the intercostal artery (C-E) showed
radicular branches to be converging toward a DAVF, with retrograde drainage into the
radicular vein and subsequently to the ventral perimedullary venous plexus (long arrow) and then downward around the conus and upward to the dorsal venous plexus
(short arrow). Selective injection into the radicular artery (F) showed converging
branches towards the DAVF (arrowhead) draining into the ascending radicular vein
(long arrow). Embolization with a liquid mixture of glue ( G ) demonstrates that the
embolic material has reached the proximal portion of the radicular vein (long arrow),
resulting in permanent obliteration and cure

872

12 Spinal Dural Arteriovenous Fistulae

In the past, excision of the dura at the site of fistula was proposed
(Hurth et al. 1978; David 1982; Oldfield 1989; Symon et al. 1984; Morgan
and Marsh 1989);however, this approach has for the most part been abandoned (Afshar et al. 1995).The results of surgery are very much the same
as glue embolization. Of 55 patients reported on by Symon et al. (1984) (50
with SDAVFs),7 (13%) deteriorated after surgery. Symon ascribed this deterioration to excision of some of the coronary venous plexus of the spinal
cord in association with division of the radicular vein emptying into the
medullary venous system. Stripping the vein on the dorsal surface of the
spinal cord must therefore be discouraged at all costs, as it is the wrong
type of operation and may lead to clinical worsening. In Symon's series, in
65% of the 31 severely disabled patients and in 80% of 15 moderately disabled patients, there was appreciable improvement of gait and sphincter
control. Similarly good results were noted in five out of six patients reported on by Oldfield (1989) and in seven of eight operated on by Morgan
and Marsh (1989) (one patient was not operated on by the author). A good
surgical response was also more recently noted by Westphal et al. (1999)
and van Dijk et al. (2002). In patients with intracranial shunts draining
downward toward the spinal cord veins, interruption of the vein draining
the fistula will give similar good results.

12.5.5 Postoperative Follow-up

In the great majority of patients with SDAVFs who have undergone endovascular or open surgery, some improvement or arrest in progression
follows. In those with no improvement within 4-6 weeks after treatment,
repeat MRI and angiographic investigation should be considered. MRI
may be able to show evidence of progressive venous thrombosis as an extremely rare unfavorable progression of previously successful treatment.
MRI at 4-8 weeks after treatment rarely shows improvement in signal
changes within the spinal cord, even in patients who respond favorably to
treatment (Willinsky et al. 1995).Therefore, the absence of improvement
as shown on MRI does not necessarily reflect residual or recurrent DAVF,
as the MRI findings tend to lag behind the clinical findings. On the other
hand, if MRI clearly shows the persistence of flow voids along the spinal
cord in a patient who did not respond to therapy, then repeat angiography
is indicated (Mascalchi et al. 2001). The repeat angiographic work-up
should start with the injection of the anterior spinal artery to verify the
circulation time again. If prolonged, the ipsilateral and contralateral arterial pedicles at the level of the DAVF as well as those above and below that
level should be reevaluated. If no residual fistula is demonstrated a search
should be made to exclude the occurrence of a second fistula. Similarly,in
those patients showing initial improvement but subsequent cessation of
improvement or deterioration, reevaluation is mandatory, as recanalization or development of a second fistula, although extremely rare, may be
found.