Article Text

Download PDFPDF

Original research
Pediatric perimedullary arteriovenous fistula: clinical features and endovascular treatments
  1. Jingwei Li1,
  2. Gao Zeng1,
  3. Xinglong Zhi1,
  4. Lisong Bian2,
  5. Fan Yang3,
  6. Jianxin Du1,
  7. Feng Ling1,
  8. Hongqi Zhang1
  1. 1 Department of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing, China
  2. 2 Department of Neurosurgery, Beijing Haidian Hospital, Beijing, China
  3. 3 Department of Neurosurgery, Beijing United Family Hospital, Beijing, China
  1. Correspondence to Dr Hongqi Zhang, Department of Neurosurgery, Xuanwu Hospital, Capital Medical University, Beijing 100053, China; xwzhanghq{at}


Objective Pediatric spinal perimedullary arteriovenous fistula (PMAVF) is rare but may cause permanent disability. We aim to summarize the clinical features of pediatric PMAVFs and our clinical experience in their treatment and to evaluate the effect of endovascular treatment in a large cohort.

Methods From 2008 to 2017, 51 PMAVFs in pediatric patients (<14 years' old) treated with endovascular techniques were retrospectively reviewed, including 24 type IVb (47.1%) and 27 type IVc (52.9%) lesions. Clinical features, radiological findings, treatment, and outcomes were evaluated.

Results Thirty-eight boys and thirteen girls were included, and the mean age at presentation was 5.6±4.1 years. Acute neurological deterioration was identified in 33 patients, and 21 of those patients (63.6%) suffered from bleeding. The annual bleeding rate before treatment was 2.55%. After transarterial embolization with coils and glue, 46 PMAVFs (90.2%) were completely occluded, and five (9.8%) were obliterated by supplementary microsurgery. During a follow-up period of 6 to 119 months (mean 58.4±16.7 months), the patients’ clinical states were improved in 42 cases (82.4%), stationary in nine cases (17.6%), and aggravated in none. Type IVc patients had a longer preoperative period, more chronic symptoms, a lower cure rate by embolization, and less improvement of symptoms than type IVb patients had (P<0.05).

Conclusions Pediatric PMAVF is a special subgroup of intradural arteriovenous shunt that should be treated early. Endovascular embolization is safe and effective in the treatment of pediatric PMAVFs.

  • perimedullary arteriovenous fistula
  • arteriovenous malformation
  • pediatric
  • treatment
  • embolization
  • prognosis

Statistics from

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.


Perimedullary arteriovenous fistulas (PMAVFs), defined as pathological connections between spinal arteries and medullary veins without an intervening nidus, account for 17% to 39% of spinal arteriovenous malformations (SAVMs).1 2 According to the angioarchitecture and effect on the spinal cord, these lesions are classified as type IVa to type IVc SAVMs.3 4 Type IVa, consisting of a small feeder and a small fistula, mostly occurs in adult patients. In contrast, types IVb and IVc occur more often in children than in adults: these PMAVFs are composed of multiple large feeders, a giant fistula, and a markedly distended venous network.4 Additionally, the presentation and natural history of PMAVF differ between the pediatric and adult populations in many significant ways. However, many larger series in the literature group all patients together, and only a few pediatric patients have been reported in larger series dealing with PMAVFs.1 2 5–8

In the past few years, as new diagnostic tools and interventional materials have been rapidly introduced, knowledge and clinical experience of this rare disease have also changed greatly.3 9 10 Nonetheless, pediatric PMAVFs are extremely rare, thus, there have been very few large pediatric series focusing on their current treatments and long-term outcomes. Therefore, we reviewed the pediatric cases diagnosed and treated by our SAVM Study Group. The purposes of this study were as follows: to summarize the clinical features and the radiological findings, revealing the specific clinical parameters necessary for the treatment of PMAVFs; to evaluate the safety and efficacy of current endovascular embolization methods; and to analyze the long-term outcomes and identify risk factors that affect the prognosis.

Patients and methods


We conducted a retrospective study of 51 consecutive pediatric patients with PMAVFs (≤14 years' old) who were diagnosed and treated by our SAVM Study Group between 2008 and 2017. Five other patients were excluded due to lack of follow-up data.

All cases were diagnosed by more than three experienced spinal vascular specialists on the basis of MRI and selective spinal digital subtraction angiography (DSA). The cases were classified into three subtypes: type IVa (no cases): a small feeder and small fistula; type IVb (24 cases): dilated feeder, intermediate-sized fistula, and dilated draining veins (figure 1); and type IVc (27 cases): multipediculated feeders, giant fistula, and dilated tortuous draining veins (figure 2).4 On the basis of location, the fistulas were divided into three regions (cervical region: C1-C7 vertebral level; thoracic region: T1-T10 vertebral level; and medullary cone region: below T11 vertebral level). Bleeding events were defined as: patients presented with acute symptoms, including sudden onset of pain or spinal dysfunction; and CT/MRI images at onset showed subarachnoid hemorrhage or hematomyelia. The patients’ motor function and sphincter function were quantified by using the modified Aminoff–Logue Scale (mALS).8

Figure 1

A 6-year-old boy suffered from severe headache, nausea, and vomiting for 7 days, accompanied with neck pain and slight weakness of four extremities. The mALS on admission was 2 (G2U0F0). He was diagnosed as a type IVb PMAVF with acute symptoms. Two weeks after the transarterial embolization, his symptoms relieved completely and the mALS decreased to 0. (A) Axial brain CT scan imaging, shows hematocele in the fourth ventricle (white arrowhead) and SAH; (B) Sagittal T2-weighted MR imaging, shows flow-voids (white arrowhead) behind the cervical spinal cord in C1–C5 level, the spinal cord was compressed by a ectasia and mild edema can be seen around it; (C–D:) Posteroanterior view of left vertebral artery angiogram (B, early arterial phase; C, late arterial phase), shows type IVc PMAVF with two venous ectasias (star) was fed by the posterior spinal artery (red arrow) and drained via the perimedullary vein (blue arrow), the deep cervical artery (white arrow) developed retrogradely; (E) Posteroanterior view of cast (black arrow) post-embolization, shows the terminal of feeding artery as well as the first venous ectasia were occluded by coils and 25% Glubran-II; (F) Posteroanterior view of left vertebral artery post-embolization angiogram, shows the fistula was completely occluded; (G) Posteroanterior view of left vertebral artery angiogram 1 year post-embolization, shows the fistula was completely obliterated.

Figure 2

A 22-month-old girl suffered from progressive bilateral leg weakness, frequent urination, and constipation for 12 months, which were considered as a delayed development and no examination was performed. The mALS on admission was 7 (G4U1F2). She was diagnosed as a type IVc PMAVF with chronic symptoms. After the transarterial embolization, her symptoms relieved a lot and the mALS decreased to 3 (G2U0F1) at follow-up period (5-year-old). (A) Sagittal T2-weighted MR imaging, shows giant flow-voids (white arrowhead) compressing the thoracic spinal cord in T9–T11 level; (B–C) Lateral view of left T10 intercostal artery angiogram (B, early arterial phase; C, venous phase), shows the type IVc PMAVF with a giant venous ectasia (star) was feed by the posterior spinal artery (red arrow) and drained via the perimedullary vein (blue arrow); (D) Lateral view of cast (black arrow) post-embolization, shows the terminal of feeding artery was occluded by coils and 33% Glubran-II; (E) Lateral view of left T10 intercostal artery post-embolization angiogram, shows the fistula was completely occluded, the posterior radicular artery did not develop due to the blood retention; (F) Follow-up MR-T2 imaging 1 year post-embolization, shows the flow-voids disappeared and the giant venous ectasia shrank obviously after its thrombosis (white arrowhead); (G) Posteroanterior view of left T10 intercostal artery angiogram 2 years' post-embolization, shows the fistula was completely obliterated.

Treatment and follow-up

PMAVFs can be treated by endovascular embolization, microsurgery, or hybrid operation in our group. Considering the particular context of pediatric PMAVFs, especially the immature skeletal system and low blood volume, endovascular therapy was preferred as the primary treatment modality in most cases. Systemic heparinization was performed only during the embolization process (heparin, 2/3 mg/kg). When the tip of the microcatheter drew near the fistula, superselective angiography was conducted, from which the angioarchitecture as well as the blood flow velocity could be evaluated. In patients with low-blood-flow fistulas, coils, or glue (n-butyl cyanoacrylate (NBCA) or Glubran 2) alone can be considered for complete occlusion. In patients with high-blood-flow fistulas, coiling combined with glue may be needed for permanent occlusion. Sparse coiling inside the proximal end of the dilated draining vein and the distal end of the feeding artery is suggested to prevent displacement of the embolic glue. Usually, a safe reflux distance should be reserved before glue embolization to avoid occlusion of the anterior/posterior spinal artery: otherwise, dense packing with coils inside the distal end of the feeding artery is recommended. For patients with multiple high-blood-flow fistulas, staged occlusion is suggested to avoid the undesired thrombogenesis of draining veins and the displacement of embolic materials. Generally, the Echelon-10 microcatheter (EV3, Irvine, CA, USA) can be used to deliver coils or glue, while the Marathon microcatheter (Medtronic, Irvine, CA, USA) is used only for the glue injection. If endovascular embolization cannot occlude the fistula completely, microsurgery or hybrid surgery is considered. Usually, electrophysiological monitoring (motor and sensory evoked potentials) and indocyanine green angiography would be used to guide the disconnection. After the occlusion of the fistula, volume expansion therapy (40–60 mL/kg, crystalloid solution:colloidal solution=1:2) was used to prevent undesired thrombogenesis in the draining vein.

Clinical presentations were recorded via the inpatient, outpatient, or telephone follow-up. By comparing the preoperative and follow-up mALS, we classified the clinical results into four groups: normal (normal functional status); improved (improved but slight neurological deficits remained); unchanged; and deteriorated. Usually, MRI was used to monitor the changes in spinal cord and perimedullary flow void signals. DSA at more than 6 months’ follow-up was used to confirm the complete obliteration of the lesion.

Statistical analysis

Statistical analysis was performed via the SPSS for Windows (Version 20.0; IBM, Armonk, New York). Variables are expressed as the mean ±SD or number of patients (percentage) as appropriate. Descriptive data were compared by using the χ2 test for proportions and a Mann–Whitney test for continuous measurements. Differences were considered significant at P<0.05 for a 95% CI.


Twenty-four type IVb and 27 type IVc were included in this study (tables 1–2).

Table 1

Population and angioarchitecture features of pediatric patients with PMAVF

Table 2

Treatment and results of pediatric patients with PMAVF

Patient population

Thirty-eight boys and 13 girls were included (M/F ratio was 2.92:1). The mean age at presentation was 5.6±4.1 years (ranging from 6 months to 14 years' old). The period from onset to treatment ranged from 5 days to 13 years (mean 15.6±9.8 months).

Acute neurological deterioration was identified in 33 patients (64.7%): 17 patients suffered from hematomyelia; four patients presented with subarachnoid hemorrhage (SAH); and 12 patients had acute paraplegia without evidence of bleeding. Assuming that the PMAVFs had been present since birth, the annual bleeding rate before treatment reached 2.55%. Recurrent hemorrhage before treatment was observed in three patients (14.3%, SAH in two cases and hematomyelia in one case). Chronic symptoms were observed in 18 patients; progressive limb weakness in 11 patients; and dysfunction of urination and defecation in seven patients.

No difference was found in mean age at presentation or in sex ratio (P>0.05). No correlation was found between age and clinical presentation, including the first symptom, bleeding history, and onset type (P>0.05). Nearly half of the type IVc patients presented with chronic symptoms (48.1%), which was a much higher rate than we found in type IVb (20.8%). This difference may be responsible for the delay of diagnosis and treatment in type IVc, as the preoperative period of type IVc (18.2±11.1 months) was longer than that of type IVb (12.8±7.7 months). There was no significant difference in the bleeding history between two type PMAVFs (P<0.05). In addition, the chronic symptoms, including limb weakness and sphincter dysfunction, occurred with an equal proportion in the two groups (P>0.05) (table 1).


Among the 51 patients, 24 cases (47.1%) were diagnosed as type IVb, and 27 cases (52.9%) were type IVc. No patients with type IVa were found in this study. The fistulas were located throughout the spinal axis: cervical region in six cases (11.8%); thoracic region in 15 cases (29.4%); and medullary cone region in 30 cases (58.8%). The number of feeding arteries ranged from one to 10. A fistula located in the ventral side of the spinal cord was identified in 15 cases, and these fistulas were fed solely by the anterior spinal artery (ASA). In addition, 16 malformations were supplied by the posterior spinal artery (PSA), and another 20 patients were supplied by both the ASA and the PSA. Among all patients, 12 lesions were supplied by a single feeder (ASA in eight cases, PSA in four cases). There was no difference in the fistula locations or the number of feeding arteries between the two subtypes (P>0.05; table 1).

Concomitant vascular syndromes were found in 13 patients (25.4%), including hereditary hemorrhagic telangiectasia (HHT) in seven patients, Cobb syndrome in four patients, and Kippel–Trenaunay–Weber (KTW) syndrome in two patients. No family history of PMAVF was found in this group, however, a family history of HHT was found in two cases. No difference was found among the patients with HHT, Cobb, and KTW syndromes. However, 10 of those patients had multiple feeding arteries (especially PSAs) and fistulas.

Treatment and results

Transarterial embolization was attempted in all cases, and complete occlusion was achieved in 43 cases (90.2%). Among these patients, 34 patients received only one embolization each, six patients received two embolizations each, and three patients received three embolizations each. The average number of embolizations was 1.8. Eight other cases (9.8%) were cured by complementary microsurgery. One patient suffered from a treatment-related complication, which was caused by thrombosis of the draining vein after the complementary microsurgery. The patient’s symptoms were markedly aggravated after surgery (mALS increased from 2 to 9) but were resolved almost completely in the 7-year follow-up (mALS decreased to 1).

During 6 to 119 months of follow-up (mean 58.4±16.7 months), no recurrence was observed. The average mALS decreased significantly at 6 months and during the follow-up period (3.2±2.8 and 1.7±1.1 respectively). Finally, the patients’ clinical states were improved in 42 cases (82.4%), stationary in nine cases (17.6%), and aggravated in none. At the last follow-up, 19 patients (37.3%) recovered to normal (mALS=0), 24 patients (47%) had slight disabilities (1≤mALS≤3), and eight patients (15.7%) had moderate to severe disability (mALS=4 in three cases, mALS=5 in one case, mALS=7 in three cases, and mALS=8 in one case). No significant difference was found in the follow-up mALS or clinical results between the bleeding group and the nonbleeding group (P>0.05).

More patients with type IVb PMAVFs (95.8%) than with type IVc (85.2%) were cured by embolization. Although the follow-up mALS in the two types of patients was similar (P>0.05), the proportion of improvement was higher in type IVb patients than in type IVc (89.8% vs 75.8%, P<0.05). However, no difference in spinal function between the two types of fistulas was found at follow-up (P>0.05) (table 2).



PMAVF in children usually presents as type IVb or IVc (also known as giant or macro-PMAVF), which is usually thought as a congenital spinal vascular disease.11–14 The exact incidence in all children is still unknown. According to our previous study, these types account for 25% of pediatric SAVMs and 14.73% of PMAVFs at all ages.8 Rodesch et al also report that the incidence of PMAVFs in children accounts for one-third of the population and includes almost all giant PMAVFs.15 In particular, there is a significant association between high-flow PMAVFs and other systemic vascular abnormalities, such as HHT, Cobb syndrome, and KTW syndrome. In addition, the incidence of such abnormalities in high-flow PMAVFs may reach 29% to 50%.15 16 In this study, the incidence of concomitant systemic vascular abnormalities was 25.4%: we suspect that the true value is higher because symptoms for diagnosis (especially HHT) may not be manifested at a very young age.

In children below 14 years' old, the age of onset may be as early as from birth and the mean age was 5.4 to 5.9 years' old.7 8 14 Due to the rarity, most of such cases had their treatment delayed for treatment, which may range from 0.5 to 164.5 months.8 17 With the development of imaging examinations and the increased knowledge of PMAVFs, the time to treatment was shortened markedly, reaching 15.6 months in our study.

The male:female ratio among pediatric PMAVF patients has been reported as 2:1 to 3.75:1.7–9 15 Similarly, male predominance was also observed in our study: the ratio was 2.92:1. These ratios are different from what has been reported in adult patients, among whom an equal gender ratio or a slight predominance of females has been reported.2 17 18 In obvious contrast to the PMAVFs of adult patients, no type IVa PMAVFs were found in pediatric patients. The proportions of type IVb and IVc among children are not known precisely. Type IVc patients accounted for 68.4% in a previous study and 52.9% in the current study.8 There is some degree of uniformity in the site of giant PMAVFs, for the thoracolumbar region is the preferred site (78.9%–83.3%) and the cervical region is less common (16%–27%).8 9 12 14 Our finding was similar to this result. Furthermore, we found that the lumbar enlargement (58.8%) was the most commonly involved region, indicating the possibility that active angiogenesis and a rich blood supply may play a role in the formation of AVF during the embryonic development period.12 19


High-flow fistulas and the subsequent changes in feeding arteries and draining vein will act on the spinal cord in four aspects: compression; the arterial steal phenomenon; hematomyelia; and venous hypertension. Acute onset of PMAVFs, caused by bleeding, mass effect from the fast-changing venous ectasia, or acute changes of draining vein, generally account for a higher proportion of cases in children (50%–57.9%) compared with adult patients.8 9 12 16 20 Additionally, chronic symptoms in pediatric PMAVFs, caused by progressive venous hypertension, the arterial steal phenomenon, or compression of venectasia, are not obvious in the early stage and can sometimes delay the treatment.7 8 21

Pediatric PMAVFs are associated with a strong tendency to bleed (50%–70%), and patients with hematomyelia have a more severe deficit than the patients with SAH.15 21 22 In the present study: 64.7% of cases presented with acute symptoms, and in 63.6% of those patients (21/33), the symptoms were caused by bleeding. The annual bleeding rates before treatment reached 2.55%, which were higher than the overall rates for all-aged PMAVFs (2.5% and 5.6%).23 This result may suggest that the complex and high-flow types of PMAVFs are unstable. Although most of the patients fell into paraplegia immediately after the bleeding event, patients’ bleeding history appeared not to be associated with the long-term prognosis, as no difference was found in the follow-up results between the bleeding and nonbleeding groups in this study.


Microsurgical disconnection, endovascular embolization, and a combination of the two have been used for the treatment of PMAVFs. Surgical treatment is always challenging, as a fistula embedded between congested veins is difficult to deal with in high-flow PMAVF.2 With the development of interventional therapy, embolization is currently considered the first choice to occlude the fistula.2 3 8 14 21 Although some special cases require the transvenous approach or the direct percutaneous puncture approach for catheterization, transarterial approach embolization can be achieved in most cases.8 14 16 In this study, we attempted the transarterial approach for all cases, and 90.2% of them were cured completely. For cases with some small feeders that are inaccessible from an endovascular approach, complementary microsurgery may be a safe choice.9

For PMAVFs, the embolization of the vein at the fistula site is critical for successful endovascular treatment.24 25 To achieve this goal, surgeons have used many embolic materials (including balloons, cellulose acetate polymer solution, and platinum coils with or without silk): the main materials in current use are coils, Onyx, NBCA, and Glubran.1 8 9 11 13 14 16 21 26 27 For pediatric high-flow PMAVFs, coils may be the most appropriate because they are more stable than glue and cannot be washed away easily. However, the mass effect on the spinal cord should be considered during the coiling. In our study group, we usually fill the venous ectasia and the terminal of the feeding artery loosely with coils and then inject some thrombotic glue inside (NBCA or Glubran). We rarely use Onyx in our cases for the following reasons: Onyx embolization is a relatively slow and gradual process, during which the hemodynamic stress may cause a hemorrhage; the solidified mass will compress the spinal cord permanently; and the angiotoxicity of dimethyl sulfoxide carries a potential risk of perimedullary phlebitis, and severe vasospasms in children. As an alternative, a sparse filling of thrombotic glue (alone or combined with loose coiling) will cause thrombosis of the venous ectasia and occlude the fistula as well. The mass effect will be relieved after the shrinkage of the thrombosis and venous ectasia.

Results and follow-up

Satisfactory outcomes can be achieved with most giant PMAVFs after appropriate treatments. According to previous studies, the cure rates for type IVb and type IVc range from 50% to 90.9%7 12 17 28 29 In our series, anatomical cure was achieved in all patients, of which 90.2% underwent endovascular embolization alone and the other 9.8% underwent embolization combined with microsurgery. This result is much better than our previous result (for cases in years 1986–2007) for three potential reasons: advanced embolic materials improved the chances of successful catheterization and occlusion; multiple intraoperative assistive techniques (indocyanine green angiography, vascular ultrasound, intraoperative DSA, and neurophysiological monitoring) can ensure the safe and complete obliteration of the lesion; and unsatisfactory occlusion rates and functional recovery in the first series prompted us to treat the lesions more aggressively in the second series.

Even after the complete obliteration of a PMAVF, only a subset of patients recover normal spinal function (26.3%–50%).8 14 28 29 In our series, 82.4% of patients were clinically improved, but only 45.5% (19/44) achieved fully normal spinal function. Several studies indicate that difficult recoveries are typically caused by secondary spinal cord injuries, such as long-term compression, venous hypertension, and the arterial steal phenomenon.9 16 17 29 The comparison between subtypes also indicates that the period from onset to treatment is an important determinant of recovery.


Pediatric PMAVF is a special subgroup of intradural arteriovenous shunt that has a high disability rate and should be diagnosed and treated early. Endovascular embolization is safe and effective in the treatment of pediatric PMAVFs.



  • Contributors Conception and design: JW-L, LS-B, F-Y. Acquisition of the data, analysis, and interpretation of the data, and critically revising the article: all authors. Drafting the article: JW-L and G-Z. Reviewed submitted version of manuscript: XL-Zh, JX-D, F-L. Approved the final version of the manuscript on behalf of all authors: HQ-Zh. Administrative/technical/material support: all authors. Study supervision: HQ-Zh.

  • Funding This work was supported by Beijing Municipal Science and Technology Commission, grant number D161100003816001.

  • Competing interests None declared.

  • Patient consent Parental/guardian consent obtained.

  • Ethics approval This study was approved by our institutional ethics committee (Xuanwu Hospital, No.2016032).

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