Article Text

Download PDFPDF

Management and outcome of vein of Galen malformation
  1. N McSweeney1,
  2. S Brew2,
  3. S Bhate1,
  4. T Cox2,
  5. D J Roebuck2,
  6. V Ganesan1,3
  1. 1Department of Paediatric Neurology, Great Ormond Street Hospital for Children NHS Trust, London, UK
  2. 2Department of Radiology, Great Ormond Street Hospital for Children NHS Trust, Institute of Child Health, UCL, London, UK
  3. 3Neurosciences Unit, Institute of Child Health, UCL, London, UK
  1. Correspondence to Dr Vijeya Ganesan, The Wolfson Centre, Mecklenburgh Square, London WC1N 2AP, UK; v.ganesan{at}


Objective To describe our experience of treating children with vein of Galen aneurysmal malformation (VGM) in a single UK centre between 2003 and 2008.

Method Retrospective review of case notes and neuroimaging.

Results 33 children were seen (26 neonates, seven infants), of whom 28 underwent endovascular treatment. Four were not offered treatment as they had evidence of severe diffuse brain injury at presentation; treatment was deferred in another who subsequently died. Seven children died (two of whom had endovascular treatment). Of the survivors (all treated), 13 (39%) are neurodevelopmentally intact, seven (21%) have mild neurodevelopmental impairment and the remaining six (18%) have significant neurological impairment. The authors were not able to identify clinical or radiological parameters which strongly predicted outcome. Of note, two children with initially low Bicêtre scores were neurologically intact after successful embolisation.

Conclusion The outlook for children with VGM is significantly better since the advent of endovascular treatment. Decisions about the appropriateness and timing of treatment should be taken by an experienced multidisciplinary team.

Statistics from


Vein of Galen aneurysmal malformation (VGM) is a rare intracranial arteriovenous malformation (AVM) with connections between choroidal arteries and the median prosencephalic vein (the forerunner of the vein of Galen) occurring between 6 and 11 weeks of gestation1 (figure 1A). It is a distinct entity from an AVM draining into an already formed vein of Galen. In the choroidal subtype (which usually presents earlier in life), there are contributions from all the choroidal arteries, with an interposed network, which then opens into the large venous sac, whereas in the mural subtype there are single or multiple direct arteriovenous fistulae within the wall of the median prosencephalic vein.

Figure 1

Neonatal case. Patient 2 developed in utero cardiac failure at 37 weeks gestation. She was born by emergency caesarean section due to fetal distress but was in good condition at birth. She collapsed at 6 h and required high frequency oscillation ventilation and inotropic support including nitric oxide; her Bicêtre score was 6. (A) Pre-embolisation imaging (T2 weighted MRI in axial plane) showed ventricular dilatation and high signal in the white matter, in addition to the vein of Galen aneurysmal malformation (VGM). Post-embolisation MRI (B) shows glue in the VGM, little evidence of residual flow and normal appearances of the ventricles and white matter. The child is neurologically intact and the shunt has been shown to be obliterated on catheter angiography.

Historically, management of VGM was neurosurgical, but outcomes were uniformly poor with some studies reporting 80–100% mortality in neonates.2,,6

Endovascular treatment has transformed this previously bleak outlook and a cure is now potentially achievable. However, many children present as gravely ill newborns with multiorgan failure. In this situation, challenging decisions have to be taken about the appropriateness of high-risk intervention and the consequent likelihood of intact survival. The Bicêtre group, pioneers in this field, have advocated a composite clinical score to assist decision-making (table 1).7 We report our experience of treating children with VGM at our centre.

Table 1

Bicêtre neonatal score2

What is already known on this topic

  • Vein of Galen aneurysmal malformation is a rare intracranial arteriovenous malformation.

  • Endovascular treatment has had a significant impact on outcome.

  • A clinical scoring system is in widespread use to select patients for treatment in the neonatal period.

What this study adds

  • Sixty per cent of children with vein of Galen aneurysmal malformation (VGM) treated in our centre had good clinical outcomes.

  • Clinical and radiological features were not clearly predictive of outcome; in particular, presentation Bicêtre scores were sometimes associated with unpredicted good or bad outcomes.

  • A multidisciplinary approach to decision-making about the appropriateness and timing of treatment is required in children with VGM.

Patients and methods

This was a retrospective review of case notes and imaging studies, undertaken with the approval of the institutional ethics committee. We included all children with VGM (as defined above) seen at our centre between 2003 and 2008 who had been managed by the multidisciplinary team listed as authors. Clinical data included demographic features, clinical presentation and neurological outcome.

The Bicêtre score at presentation was calculated for children presenting in the neonatal period based on data in the case records. The Bicêtre score is a composite score made up of indices of function of five organ systems: cardiac, cerebral, respiratory, renal and hepatic. The maximum score is 21. The Bicêtre group recommend that if the score is <8 embolisation is not offered, if the score is 8–12 emergency embolisation is undertaken and if >12 embolisation is delayed (table 1).7

All radiology studies were reviewed by S Brew. Information regarding outcome was obtained from neurological examination, developmental history and examination at the latest follow-up appointment. We categorised outcome into one of three categories: intact (no neurological or developmental impairments), mild impairment (neurological or cognitive impairment not precluding age appropriate function), and significant impairment (neurological or cognitive impairment precluding age appropriate function) or death. Endovascular treatment comprised transarterial embolisation with n-butyl-2-cyanoacrylate glue (Histoacryl; Aesculap, Tuttlingen, Germany), usually from a femoral approach but occasionally from an axillary approach in the presence of iliac artery occlusion.


Thirty-three children with VGM were managed by us between 2003 and 2008 (see table 2 for summary). Five had previously been treated elsewhere. Twenty-six children (13 male, three born preterm) presented within the first month of life (median 1.5 days, range 1–28 days) and comprise the neonatal group, and seven children (four male) presented later, at a median age of 3 months (range 11 weeks to 18 months) and comprise the non-neonatal group. In the 29 children in whom cerebral angiography was undertaken, VGM was of the choroidal subtype in 19 and mural subtype in 10.

Table 2

Summary of clinical and imaging findings

Neonatal group (n=26)

Clinical presentation in these children was most commonly (in 18 children) with high output cardiac failure on the first day of life (range 1–21 days); other presentations included macrocephaly (n=4), seizures (n=2) and vomiting (n=1). One baby was initially asymptomatic but then developed fulminant cardiac failure at 1 month of age. Of the 22 who had had catheter angiography, the VGM subtype was choroidal in 14 and mural in eight.

Eleven babies had had an abnormal antenatal ultrasound, of whom seven had an antenatal diagnosis of VGM (with cardiac failure and ventriculomegaly in one). Another three babies had evidence of cardiac failure in utero. One child was thought to have a large interhemispheric cyst on antenatal ultrasound scan but the diagnosis was revised to VGM on postnatal brain imaging. Twelve babies were delivered by caesarean section (six as an emergency for fetal distress). Their mean birth weight was 3.2 kg (range 2.5–4.2 kg) and mean head circumference was 35 cm (33.2–37.1 cm). The median Apgar score at 1 min was 8 (mean 7, range 1–10). The median neonatal Bicêtre score at presentation was 12 (range 3–21). Thirteen babies required ventilation for between 5 days and 4 weeks. Thirteen required inotropic support for treatment of cardiac failure.

Echocardiogram reports were available in 23 babies; the commonest abnormal finding was dilatation of the right side of the heart (in 15) with suprasystemic pulmonary arterial pressure or persistent pulmonary hypertension of the newborn in seven (five of whom subsequently died). Cardiac abnormalities included patent foramen ovale in eight (isolated in one), persistent ductus arteriosus in 10, atrial septal defect in five (isolated in two) and ventricular septal defect in one.

Twenty-one neonates had endovascular treatment as previously described (including one coil embolisation carried out at another centre). A median of two embolisations were undertaken in each child (range 1–6) at a median age of 13 weeks (range 1 day to 5 years 9 months). The median age at first embolisation in this group was 5 days (1 day to 10 months). Procedural complications occurred in nine children. These were mostly technical: intraventricular haemorrhage in three, glue escape into the venous sinuses in two, microcatheter adherence in two (with arterial rupture causing haemorrhage and subsequent death in one), interhemispheric haemorrhage and venous ischaemia from cortical venous thrombosis in one and cardiac arrest (resuscitated) probably related to general anaesthesia in another infant.

Secondary problems related to VGM included hydrocephalus (post-embolisation) in five babies (treated with third ventriculostomy in three and temporary external ventricular drainage (EVD) in two). Another baby had a ventriculoperitoneal shunt inserted prior to embolisation, which subsequently occluded and required revision. Seven neonates developed haemorrhage, which included interhemispheric, intraventricular and subarachnoid haemorrhages. Two required subsequent temporary EVD insertion. Venous outflow tract obstruction occurred in three babies (left transverse sinus thrombosis in two, jugular bulb stenosis in one). Iliac artery occlusion was observed in two babies and was thought to be related to repeated femoral artery puncture and catheterisation. One child (patient 16) developed repeated episodes of aseptic meningitis related to infection of the embolic material 2 years after his last endovascular procedure and has had surgical evacuation and ongoing courses of intraventricular antibiotics.

The median length of follow-up in the neonatal group was 2 years (range 6 months to 6 years). Seven children died: five without endovascular treatment, one of multiorgan failure despite technically successfully embolisation and another as a consequence of intraprocedural arterial rupture. One of the five untreated children who subsequently died (of heart failure) presented in good condition with a Bicêtre score of 21 (and was therefore not offered immediate treatment) and had no evidence of cardiac failure on echocardiography. Delayed embolisation was planned but he decompensated at 1 month of age and died at home. The other four children were not offered treatment as brain imaging showed evidence of severe diffuse parenchymal brain injury.

Table 3 summarises the data on the relationships between key clinical and imaging features and clinical outcome. In the seven children who died (two embolised), the Bicêtre score at presentation was <8 in four; however one child had a score of 21 at presentation. Two babies with Bicêtre scores at presentation of 6 (figure 1, case 2) and 8 were treated (as they did not have evidence of parenchymal brain injury) and are neurologically intact. Four neonates with evidence of diffuse ischaemic parenchymal injury at presentation were not treated and thus died as a direct result of the VGM (see table 2).

Table 3

Clinical and imaging findings in relation to outcome (n=33)

Non-neonatal group (n=7)

The median age at presentation in these children was 3 months (range 11 weeks to 18 months). Clinical manifestations were macrocephaly (n=4), cardiac failure (n=2) and an abnormal gait in one child (see figure 2, for example). Six had choroidal and one had mural type VGM.

Figure 2

Infant case. Patient 30 presented at 18 months of age with irritability, macrocephaly, developmental regression and episodic raised intracranial pressure. Pre-embolisation imaging (A, CT scan; B, T2 weighted MRI, axial plane) showed a large VGM with ventricular dilation, apparent reduction in white matter bulk and calcific changes in the basal ganglia. Catheter angiography (C, internal carotid artery injection, lateral projection) shows the sac of the VGM with multiple arterial feeders. Post-embolisation MRI (D) shows glue in the VGM and improvement in the ventricular and white matter appearances. Final catheter angiography (E, internal carotid artery injection, lateral projection) shows no evidence of a residual arteriovenous shunt. Clinically there was a dramatic improvement in the patient's level of alertness, cognition and developmental progress following embolisation and she now has just mild developmental delay.

The median number of embolisations in each child was three (range 1–7), undertaken at a median age of 20 months (range 3 months to 14 years). The median age of first embolisation was 4 months (range 3 months to 8 years). Procedural complications occurred in three and included post-embolisation intraventricular and thalamic haemorrhage in one, glue escape into the venous sinuses in two and iliac artery occlusion in another child.

Secondary problems related to VGM included hydrocephalus (not resolving after endovascular treatment) in four infants (requiring ventriculoperitoneal shunt in two and third ventriculostomy in two). One infant had insertion of a ventriculoperitoneal shunt prior to embolisation. One child developed an intraventricular haemorrhage and another developed jugular bulb stenosis.

Clinical and radiological factors and outcome

Although our ability to make meaningful statistical comments about this small retrospective dataset is limited, we were not able to identify any significant differences in age at presentation (neonate vs non-neonate), presence of cardiac failure, presentation Bicêtre score, pulmonary hypertension, VGM subtype, parenchymal injury at presentation and residual arteriovenous shunt between the good and poor outcome groups (χ2 or Fisher's exact test, p>0.05 for all) (see table 3).


Despite encouraging outcomes in the majority of children with VGM seen by our group, with intact survival or mild impairment in 60% of children, a fifth (all in the neonatal period) died (usually without treatment). Our study is limited by the lack of prospectively collected standardised outcome data and by the relatively small sample size, which limited our ability to comment on predictors of outcome. A further limitation is that many children from the neonatal group are still relatively young and therefore impairments may emerge over time. However, given the still widely held misperception that VGM has a universally grim outcome, we feel our observations add to the growing body of literature reporting on the improved outcomes in the era of endovascular treatment.

Several other groups have reported on the outcomes for children with VGM, following the advent of modern endovascular treatment. The late Pierre Lasjaunias reported data from 317 children seen at Bicêtre Hospital, Paris, between 1981 and 2002. Of these, 216 children were treated, 23 of whom died. This group included 140 neonates, 88 of whom were treated with embolisation; seven others untreated were lost to follow-up. Of 193 survivors, 143 (74%) were intact, 30 (15.6%) were moderately and 20 (10.6%) were severely impaired.6

Fullerton et al reported outcome in 27 children with VGM (21 neonates) embolised in San Francisco. No data were provided on those who were not offered treatment. Four of 27 (15%) children died (4/21 neonatal presentations). Of the 23 survivors, 12 were intact and two had mild impairments.8

Heuer et al reported outcome in 13 patients with VGM (seven neonates, eight choroidal type) between 1994 and 2007. Eleven were treated endovascularly, of whom two neonates died, five were intact and one had severe neurological impairment. The outcomes in the five children treated after 2 weeks of age were better than those in children treated earlier.9

Gupta et al reported the cases of 21 children and four adults with VGM treated between 1983 and 2003. Fifteen patients had endovascular treatment; limited follow-up data are provided on 13 of them. Two patients with delayed milestones prior to endovascular treatment improved developmentally after treatment.10

More recently, Geibprasert et al11 reported the outcome of 25 children (20 neonates, eight choroidal type) seen over the last 10 years. Ten had parenchymal brain injury (focal encephalomalacia in seven) and seven had parenchymal calcification. Fifteen had embolisation, six were managed conservatively (three of whom had spontaneous thrombosis) and four had treatment withheld. After a mean follow-up of 3.8 years, nine (36%) children had died and nine were neurologically intact; one child had severe neurological impairment. Factors predicting adverse outcome were neurological symptoms at presentation, neonatal score <12, parenchymal changes especially if focal or parenchymal calcification, arterial steal and more than two groups of arterial feeders. There was no relationship between outcome and subtype of malformation.11

Thus the literature reflects the rarity of VGM and the relatively limited experience at individual centres, other than the large series from Bicêtre.

In our cohort, the highest rate of adverse outcomes was in the neonatal period. Reported neonatal mortalities range from 8% to 63% compared with around 10% in children presenting later. Good clinical outcomes are reported in 25–73% of neonatal presentations compared with 67–100% of those presenting later.12,,18 In Lasjaunias' series, of the 45 neonates not offered treatment, 25 had encephalomalacia on brain imaging, 17 had a Bicêtre score2 of <8 and three pregnancies were terminated following antenatal diagnosis. Twelve of the 88 neonates died, representing half of all the deaths in this series.6

Endovascular treatment can be life saving, but there has been a reluctance to offer it to children who are predicted to have a very poor quality of life should they survive. Treatment is extremely challenging in critically ill babies and thus would ideally be deferred. The Bicêtre score is aimed at assisting decision-making about the appropriateness and timing of endovascular treatment in neonates with VGM. Two of the children in our group who survived intact after treatment had initial Bicêtre scores of 6 and 8. Another with an initial score of 21 decompensated and died suddenly prior to the planned embolisation (as also previously reported by Cherif et al19) and a further child had a reduction in score from 21 to 8 overnight in hospital. These observations highlight the dynamic nature of the Bicêtre score; clearly when the score is calculated is an important factor and the score, like the patient's clinical status, can alter rapidly. Decision-making about the appropriateness of endovascular intervention in sick neonates with VGM is akin to that relating to other high-risk paediatric interventions. Thus, our practice has been to undertake multidisciplinary assessment of the child (involving paediatric neurology, neurosurgery, intensive care, cardiology, fetal and neonatal medicine and the interventional neuroradiologist) and to offer treatment to all neonates other than those with evidence of diffuse, irreversible brain injury (infarction/porencephaly), rather than being overly rigid about applying the Bicêtre score. Without available robust data on predictors of outcome, it is difficult to recommend helpful alterations to the Bicêtre score.

Ventricular dilatation was a common finding in our patients, likely reflecting a combination of impaired cerebrospinal fluid (CSF) resorption due to venous hypertension and a degree of volume loss in the white matter.

CSF diversion, without reduction of the arteriovenous shunt, may result in acute subdural haemorrhage and possibly reduced parenchymal perfusion (due to reversal of flow in the medullary veins) and poor outcomes.20,,22 Thus the initial management of symptomatic or radiologically progressive hydrocephalus in VGM should be embolisation. Hydrocephalus treatment even after embolisation, has been associated with a high risk of complications.20

Although venous hypertension in VGM is largely as a result of the high flow arteriovenous shunt, there is a significant incidence of jugular bulb stenosis or venous sinus thrombosis, which may further exacerbate venous hypertension.11


At least half of the children with VGM who receive endovascular treatment will survive intact. We were unable to identify clinical or radiological factors which were strongly predictive of outcome. A low Bicêtre score was not universally associated with poor outcomes. Decision-making about the appropriateness and timing of treatment mirrors discussions relating to other high-risk interventions in paediatrics and multidisciplinary assessment must be the way forward.



  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Great Ormond Street Hospital Ethics Committee.

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

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.