Article Text


Case series
Long term follow-up of endovascular management of spinal cord arteriovenous malformations with emphasis on particle embolization
  1. Adrien Collin1,
  2. Marc-Antoine Labeyrie1,2,
  3. Stephanie Lenck1,
  4. Akli Zetchi1,
  5. Armand Aymard1,
  6. Jean-Pierre Saint-Maurice1,
  7. Vittorio Civelli1,
  8. Emmanuel Houdart1,2
  1. 1 Department of Interventional Neuroradiology, Hopital Lariboisiere, Paris, France
  2. 2 EA 7334 REMES, Université Paris, Paris, France
  1. Correspondence to Dr Marc-Antoine Labeyrie, Service de neuroradiologie interventionnelle, Hôpital Lariboisière, Paris 75010, France; labeyriem{at}


Objective To determine long term safety and efficacy of endovascular treatment of spinal cord arteriovenous malformations (AVMs), with calibrated particle embolization as a firstline approach.

Methods We reviewed clinical and imaging data of consecutive patients who underwent endovascular treatment for both nidal and fistulous type spinal cord AVMs in our center, from 1990 to 2015. Outcome at the last follow-up was assessed by an independent observer.

Results Embolization of spinal cord AVMs was performed in 61 patients, including 46 (75%) with particles (exclusively in 29 patients), 30 (49%) with cyanoacrylate, and 6 (10%) with combined surgical treatments. Particle embolizations were iterative in 33 patients (median number of sessions 5 (range 3–6)). Neurological deterioration after treatment occurred in 5 patients (cyanoacrylate=4, surgery=1, particles=0; P<0.001). At a median follow-up of 6 years (range 3–13 years), angiographic cure was obtained in 11/61 (18%) patients (nidal type=6/53 (11%), fistulous type=5/8 (63%)). In progressive forms, neurological improvement occurred in 16/28 (57%) patients, stabilized in 9/28 (31%), and worsened in 3/28 (12%). In hemorrhagic forms, the rebleeding rate was 4/14 patient years without standard treatment, 0/322 patient years in partial iterative treatment, and 0/15 patient years in angiographically cured lesions (P=0.001).

Conclusion Our study suggests that particle embolization as a firstline therapy to treat spinal cord AVMs is safe and offers long term efficacy, especially for those with small, distal, and multiple shunts. Partial occlusion of the AVM may be sufficient to prevent rebleeding, without the potential risks of complete occlusion. Particle calibration and injection technique, ‘one by one’, are critical to safety. Cyanoacrylate embolization or surgery remains necessary if particle embolization fails to occlude large shunts.

  • arteriovenous malformation
  • spinal cord

Statistics from

Intradural spinal cord arteriovenous malformations (AVMs) are rare lesions. The largest series in the literature involved 108 patients recruited over a 20 year period.1 The best approach to treat spinal cord AVMs has not been validated to date, and there is no consensus.2 Particle embolization is a well known technique but is controversial as it is associated with a high rate of recanalization.3–5 Cyanoacrylate embolization or surgery may be associated with a high rate of severe neurological morbidity.1 3 6 7

For many years, our center has been using particle embolization as a firstline treatment for spinal cord AVMs. However, since 1990, our technique has evolved from polyvinyl alcohol (PVA) particle injection8 to non-PVA spherical calibrated particles.

The aim of our study was to report the long term efficacy and safety of endovascular treatment of spinal cord AVMs using spherical calibrated particle embolization as firstline treatment.



We analyzed all consecutive patients treated in our center for a spinal cord AVM from January 1990 (the date of the standardization of our spinal cord AVM embolization protocol using calibrated spherical particles) to December 2015. A spinal cord AVM was defined on the angiogram as an arteriovenous shunt fed by one or several radiculomedullary arteries. Fistulous (vs nidal) type was defined as an AVM with ≤3 shunts (vs >3 shunts) draining into the same vein.9 Patients with dural arteriovenous fistulas, epidural arteriovenous fistulas, cavernomas, hemangioblastomas, and those previously treated in another center or during the inclusion period were excluded, as well as those exclusively treated by surgery or those who received no treatment. Patient charts, spinal cord MRI, and angiographic data were retrospectively reviewed by two interventional neuroradiologists in consensus (M-AL, AC). The modified McCormick Scale before treatment and at the last follow-up was computed by an interventional neuroradiologist, independent of the procedure (AC).10 All patients were informed and provided their consent prior to each intervention. The study was approved by our institution’s patient protection committee.

Endovascular treatment

An angiogram of spinal artery feeders was performed under general anesthesia. Treatment was only considered for symptomatic patients (hemorrhagic forms or progressive worsening). The treatment aims to occlude ruptured aneurysms and the maximum number of arteriovenous shunts, but does not necessarily aim to achieve complete occlusion as this may be associated with an unacceptable risk of neurological morbidity. An endovascular approach was considered whenever possible. Surgery was considered for large and superficial fistulous types, especially at the filum terminalis, where endovascular access is challenging or dangerous.

Embopheres were initially used for embolization with non-PVA spherical calibrated particles, which were then replaced by colored particles (Embogold, Merit Medical, Salt Lake City, Utah, USA). Because of their coloration, we were able to count them easily and inject them one by one (see online supplementary figure). They were injected through a 0.013 inch internal diameter microcatheter (Marathon; MTI-ev3 Neurovascular, Irvine, California, USA). We used particles of 300–500 µm in the first instance. Particle diameter was increased to 500–700 µm if no significant reduction of the shunts was noted after about 15 particles of 300–500 µm. Particles of 500–700 µm were injected through the 0.013 inch microcatheter because they can distort, or in rare cases through a 0.017 inch microcatheter (Echelon 10; MTI-ev3 Neurovascular, Irvine, California, USA). Injection of the particles was stopped when all of the shunts were occluded or when the spinal artery presented with significant slowdown. The tip of the catheter was placed as distally as possible, making sure to preserve antegrade flow. Depending on the size of the feeders, it could be positioned either in the trunk of the radiculomedullary artery (figure 1) or in the anterior or posterior spinal arteries. A follow-up angiogram was performed under general anesthesia after a period ranging from 1 to 3 years after each session. New embolization was considered depending on the recanalization of each feeder.

Supplementary file 1

Figure 1

Partial and iterative embolization of a cervical spinal cord arteriovenous malformation (AVM), exclusively with particles. Sagittal T2 weighted MRI of the cervical spinal cord (A, F). Front view angiogram of the right deep cervical artery (B–E). Initial MRI (A) shows an intramedullary hematoma with edema of the cervical spinal cord. Initial angiogram (B) shows the cervical spinal cord AVM fed by a posterior radiculomedullary artery (white arrow) giving rise to several small secondary feeders and arteriovenous shunts which drain into the perimedullary veins (black arrow). There is a nidal aneurysm (white arrowhead) corresponding to the rupture area. The angiographic control immediately after the first embolization with particles (C) shows subtotal occlusion of the shunts and exclusion of the nidal aneurysm. The MRI control at 1 year (D) shows hallmarks of hematomyelia and regression of the edema. The angiographic control at 1 year (E) shows partial recanalization of the AVM with early enhancement of the perimedullary veins (black arrow), without recurrence of the nidal aneurysm. The angiographic control immediately after a second particle embolization (F) shows complete occlusion of the shunts.

In large shunts, especially when particle embolization did not achieve successful occlusion, cyanoacrylate injection or surgery (in superficial shunts) was considered. Cyanoacrylate embolization was performed using Glubran (Aspide Medical, La Talaudière, France) or Histoacryl (B Braun, Melsungen AB, Melsungen, Germany) through a Magic 1.2 F or 1.5 F microcatheter (Balt, Montmorency, France), after dilution with ethiodol (Lipiodol Ultra-fluid, Guerbet, France) using a 1:1 to 1:3 ratio depending on the residual flow in the pedicle. Injection of cyanoacrylate was only used in our center under specific conditions. For an arteriovenous shunt fed by a commissural artery, the tip of the microcatheter was placed distally from the trunk of the anterior spinal artery. For an arteriovenous shunt fed by a circumflex artery, injection was only done if the tip of the microcatheter was located within 10 mm of the shunt and was not occlusive. A wedged flow situation was a contraindication to cyanoacrylate injection due to the risk of opening arterio-arterial anastomoses.

Statistical analysis

Continuous variables are described as median (Q1–Q3). Categorical variables were compared using the χ2test or Fisher’s exact test, as appropriate (SPSS 19). A P value ≤0.05 was considered significant.



Sixty-eight patients were managed in our center for spinal cord AVMs. We included 61 patients and excluded 7 patients (5 were exclusively treated by surgery and 2 received no treatment). The characteristics of the patients are reported in table 1. Among patients treated only with particle embolization, 41/46 (89%) had a nidal type and 24/46 (52%) had a cervical topography. In patients with iterative particle embolization (figures 1 and 2), the median number of sessions was 5 (3–6).

Table 1

Characteristics of patients with endovascular management of spinal cord arteriovenous malformations (n=61)

Figure 2

Partial and iterative embolization of a cervical spinal cord arteriovenous malformation (AVM), exclusively with particles. Sagittal T2* weighted MRI of the cervical spinal cord (A). Front view angiogram of the right vertebral artery (B–D). Initial MRI (A) shows an intramedullary hematoma of the cervical spinal cord. Initial angiogram (B) shows a cervical spinal cord AVM fed by a tortuous and narrow posterior radiculomedullary artery (white arrow) arising from the V4 segment of the right vertebral artery, giving two arteriovenous shunts (white asterisk). Distal catheterization of this artery for cyanoacrylate injection was not possible. Then, the ostium of the artery was catheterized with a 0.013 inch internal diameter microcatheter, and a balloon was inflated (black asterisk) in the downstream segment of the vertebral artery to protect it from particle embolism (C). The angiographic control immediately after first embolization with ten 300–500 µm particles (D) shows occlusion of the shunts. The control angiogram at 3 years (not shown) shows persistent occlusion of the shunts.

Treatment morbidity

Neurological worsening following treatment was observed in five patients (cyanoacrylate=4, surgery=1, particle=0), including two with persistent disability at the last follow-up. Considering each session of embolization, it was significantly less frequent with particle embolization (P<0.001).

Long term outcome

Outcomes at the last follow-up (median 6 years (3–13)) are reported in table 2. Long term follow-up was missing for 10/61 (16%) patients (median follow-up 2 years (1–3)), including six patients living abroad. Among the patients with the progressive form who experienced worsening (n=3), two had persistent shunt without endovascular or surgical access. The rebleeding rate was 4/14 patient years among patients already diagnosed but waiting for our standardized protocol of treatment (including one patient who had a first session embolization but did not attend the follow-up appointment and presented with rebleeding 13 years later), 0/322 patient years for partial iterative treatment, and 0/15 patient years in angiographically cured lesions (P=0.001). It must be emphasized that intranidal ruptured aneurysms were permanently occluded after the first particle embolization session in five patients (figure 1).

Table 2

Outcome at the last follow-up of patients with endovascular management of spinal cord arteriovenous   malformations (n=61)


Our study is one of the largest series in the recent literature of spinal cord AVMs treated with particle embolization as a firstline approach and as first treatment. Comparison with previous series published after 2000 with more than 10 patients is reported in the online supplementary table.1 3 6 7 11–15

Supplementary file 2

Our rate of complications with permanent neurological deficit was low compared with other recent series using surgery or embolization with cyanoacrylate as the main approach. This may be related to the zero rate of complications obtained using particle embolization in our series, despite a higher proportion of procedures. The higher safety of particle embolization may be explained by the fact that cyanoacrylate is more prone to progress into small functional arteries because of its liquid property, whereas progression of calibrated particles is limited by their size. Furthermore, cyanoacrylate injection carries the risk of refluxing into and occluding the spinal arteries, whereas particles are aspirated through the AVM shunts. A zero rate of complications with particle embolization was not confirmed in previous series reporting this technique.6 8 16 We hypothesize that this finding in our current series is related to the use of better calibrated particles. Also, non-PVA particles are less prone to aggregate, as suggested in vitro.17 The quality of calibration and the risk of aggregation are important in spinal cord AVM embolization, as particles <80 µm may occlude functional sulco-commissural arteries (diameter varies from 60 to 72 µm), and particle aggregates >1000 µm may occlude the spinal arteries. Therefore, we recommend using non-PVA and strictly calibrated particles with a diameter of 300–500 µm or 500–700 µm. Similarly, we recommend injecting the particles one by one in order to decrease the risk of particle aggregation. The use of colored particles for proper visualization is necessary in that setting (see online supplementary figure).

The rate of angiographically cured AVMs was low in our series, especially for nidal types. This could be explained by the well known transient efficacy of particles to occlude arteries.5 8 18–20 This is also explained by the fact that all of the shunts of the spinal cord AVMs were not necessarily accessible for embolization. However, comparison with others series reveals that even when using cyanoacrylate or surgery as the first treatment, the rate of complete occlusion of nidal types of spinal cord AVMs is <50% in most of the series (ranging from 16% to 88%). Thus we consider that the efficacy of particle embolization compared with cyanoacrylate or surgery should not be judged on the sole criterion of long term rate of complete occlusion.

There was no long term rebleeding in our series when considering only patients who benefited from multiple embolization sessions following our protocol described above, even in cases of partial occlusion. Rebleeding was significantly higher in patients without standard treatment, suggesting that our approach compares favorably with the natural history of ruptured spinal cord AVMs. A recent meta-analysis reported a rebleeding rate of 10% in untreated hemorrhagic forms of spinal cord AVMs.19 Our results are comparable with that of others series in the literature concerning patients who were mainly treated with cyanoacrylates or surgery. We suggest that our protocol of particle embolization has a similar efficacy as cyanoacrylate or surgery in preventing rebleeding. The mechanism by which partial embolization is effective in preventing rebleeding of spinal cord AVMs remains unclear and is outside the scope of this study. Aneurysms on the arterial feeders of the AVMs are known to be associated with an increased risk of rebleeding.21 They are usually considered as a target of embolization with cyanoacrylate or surgery.22 We reported in our series five patients with such aneurysms which were definitely occluded using particle embolization (figure 1).

Despite favoring particle embolization as firstline therapy, 25% of patients were not eligible for this treatment modality and 31% required combined treatment. This could be explained by the inefficacy of particles (even of 500–700 µm) to occlude the largest shunts. Also, in rare selected cases it was felt that a single session of cyanoacrylate injection met our safety conditions described above and led to permanent occlusion of the shunts.

The rate of long term favorable outcome in progressive forms in our study was high. Thus 88% of patients with progressive forms had a stable or improved modified McCormick score at the last follow-up. Worsening in two patients occurred despite complementary cyanoacrylate embolization or surgery. These results suggest that iterative embolizations with particles are also effective in treating congestive symptoms even without complete occlusion.

Our study has several limitations. This was a monocentric, retrospective, non-controlled analysis with a relatively small number of patients. However, given the rarity of this pathology and the high level of specialization of its treatment, we consider that a multicenter randomized study to compare the different therapeutic approaches could not succeed. Our study, based on consecutive cases, may suggest that particle embolization is not an obsolete approach in further meta-analyses or systematic reviews. The long study period (25 years) could have introduced biases in the modality of treatment, although our technique remained unchanged during that time. Considering our rate of patients lost to follow-up, the safety and efficacy results may have been underestimated although the total follow-up of 349 patient years is considerably high and most of the lost to follow-up patients lived abroad.


We thank Caroline Plougoulen from the Interventional Neuroradiology Department.


View Abstract


  • AC and M-AL contributed equally.

  • Contributors All listed authors contributed to the design, data collection, data analysis, drafting of the manuscript, and final approval. They agree to be accountable for all aspects of the work.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests None declared.

  • Patient consent Not required.

  • Ethics approval The study was approved by CPP Ile de France VI.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement No data sharing.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.