Various studies have sought to identify which combinations of patient and bAVM morphological and anatomical factors are most closely related to an elevated risk of future haemorrhage [46,50] (Table 1). The risk factors postulated to be associated with highest risk of bleeding appear to be previous rupture and haemorrhage, deep bAVM location or brainstem location, and exclusively deep venous drainage, with rates varying from 0.9% for unruptured bAVMs without a single risk factor to 34.4% for those with all three risk factors [50]. Other factors include the presence of a cerebral aneurysm on a feeding artery and restriction of venous drainage from the malformation [22] (narrowing or occlusion of one or more principal draining veins). The latter is associated with the highest risk when the malformation has only a single draining vein [44].

In a retrospective review of 46 children, Reitz identified small bAVM size and exclusive deep venous drainage as independent risk factors for haemorrhage [43]. This was corroborated in another study evaluating 130 patients from two high volume centres with infratentorial location being an additional risk factor [18]. Another retrospective review of 85 patients confirmed the impact of pattern and number of draining veins with single draining vein and deep venous drainage all associated with higher risk of haemorrhage and worsening outcomes [31]. bAVM associated aneurysms (pre-nidal flow related, intra-nidal, venous) were reported in 16-29% of paediatric bAVMs. In a review of 77 children, it was a risk factor associated with haemorrhage in univariate analysis but not in multivariate analysis [2]. Hoffman and colleagues [29], however in a smaller review of 30 patients noted that haemorrhage rates appeared to be similar with or without aneurysms but that emergent treatment was often required more when patients presented with associated aneurysms. In general however, previous haemorrhage does appear to remain the most significant risk factor in children for re-haemorrhage, especially in the first year in adults, but this risk may appear to be longer in children [3,40].

It would be reasonable to estimate a realistic rate of haemorrhage for an unruptured bAVM without significant risk factors of 1-2% per year and this is mirrored in the America Stroke Association (AHA) guidelines with a synthesis of a combination of pooled patient cohort data, meta-analysis, review article and a randomised control trial [14]. Risk of haemorrhage in a previously ruptured bAVM is estimated at 4.5-4.8% per year.

To guide cumulative risk over a lifetime with regard to rupture and haemorrhage, a number of formulae have been proposed : Lifetime cumulative risk of rupture = 1 - (1 - R)^L; R, yearly risk of rupture; L, years of life at risk (life expectancy).

This formula does rely on assumptions that the yearly risk of haemorrhage remains constant every year and that the risk in any year is independent of the risk in any other year [32]. A simpler calculation assuming an annual risk of haemorrhage of 3% per year is 105-age (years).

With a potential morbidity rate of 33% per haemorrhage (30-day rate of moderate to severe disability with Rankin 3-5) [9], even utilising the lowest annual rupture rate (0.9% per year), this results to a very significant lifetime risk of rupture in children with a cumulative risk of morbidity and mortality among children over time [13,47].

Timing of definitive management of bAVM following haemorrhagic presentation is not standardised in the cerebrovascular community. Hyperacute surgery has advantages of treating the life-threatening intracranial pressure (ICP) complications, early definitive treatment of the bAVM with an immediate reduction in risk of re-haemorrhage, as well as the concept of ‘operating within the neurological deficit attributed to the haemorrhage’. The disadvantage is difficult operative conditions with secondary cerebral oedema and a risk of incomplete obliteration of bAVM due to concealment of the true morphology of the bAVM and vessels and an accurate understanding of its anatomical features.

In a cohort of 102 patients with ruptured bAVMs where the bAVM was definitively treated at least 4-week post haemorrhage [4], the risk of re-haemorrhage was felt to be <1%. The authors therefore supported this paradigm of delaying definitive treatment of a ruptured bAVM for at least 4-week post-haemorrhage with the attendant advantages of reduced cerebral oedema, better visualisation of compressed vessels, more accurate anatomical delineation of the bAVM, superior operative conditions and lower risk of incomplete obliteration and residual bAVM outweighing the <1% risk of re-haemorrhage and its associated morbidity. This was only modified when a high-risk feature such as an associated intracranial aneurysm (IA) was identified where the re-haemorrhage rate was calculated at 11.4% per patient-month.

A larger more recent study assessed via a propensity score-matched analysis, timing of microsurgical resection in 195 pairs of ruptured bAVMs with early (<30 days) or delayed (>30 days) resection following haemorrhage [56]. Short term neurological outcomes were better in the delayed resection group but there was no difference in long-term neurological outcomes, and favourable outcome and long-term disability were similar across both groups. Post-operative seizure, length of hospital stay and post-operative haemorrhage were similar across both groups. Delayed surgery however had a significantly better complete obliteration rate compared to early resection (99% vs. 91.3%; p=0.001).

These studies very much reflect the authors’ own personal paradigm where definitive treatment of a ruptured bAVM is performed in delayed fashion usually at least 6-weeks following haemorrhage. This allows more favourable microsurgical conditions, favourable planes of dissection, reduced complications with cerebral oedema, better delineation of bAVM angiographic morphology and therefore more accurate bAVM grading and risk stratification to design an optimal management strategy to achieve a higher possibility of complete obliteration (Fig. 1). Surgical treatment is primarily geared towards control of ICP complications and if surgical evacuation of intracranial haemorrhage is required, it should be tailored only to removal of haematoma and acute bleeding control. This is also supported by the AHA guidelines on management of bAVMs where small superficial bAVMs can be considered for removal during emergency surgery but resection of larger, deep bAVMs should be deferred for a period of 2 to 6 weeks at least [14], owing to the greater likelihood of neurological morbidity and mortality associated with immediate removal of a ruptured bAVM. High risk features for recurrence haemorrhage such as perinidal or intranidal aneurysms should be addressed immediately, most commonly by endovascular or rarely by surgical means in lieu of complete resection [14].

Migraine headache is postulated to have a vascular aetiology and it is therefore unsurprising that bAVM has been recognised as a potential cause. It has been noted that migrainous cephalgia is often ipsilateral to radiological bAVM [25]. Interestingly, it appears that an occipital location for bAVM is particularly associated with headaches [17,33,39]. We have observed similar phenomena in our paediatric patients - Fig. 2 describes a 14-year-old girl who presented with intractable migraine-like headache ipsilateral to an unruptured parieto-occipital bAVM whose headaches resolved after microsurgical resection. Involvement of a paediatric neurologist with an interest in headache in the paediatric neurovascular service is very helpful when dealing with such patients.

Papilloedema - pathological swelling of the optic nerve head due to intracranial hypertension - is an important ophthalmological condition which is untreated can lead to optic neuropathy and visual failure. Papilloedema has been rarely reported as a sequela of unruptured bAVM in children [20], however we have observed it in three of our patients - all teenagers with large unruptured parietal bAVMs (Fig. 3). The pathophysiology is presumably related to chronic venous hypertension due to intracranial arterio-venous shunting. In all three of our patients, the papilloedema resolved after surgical or radiosurgical treatment of the bAVM. We recommend baseline ophthalmology assessments for all children presenting with bAVM.

Large bAVMs may shunt blood preferentially through the nidus resulting in vascular ‘steal’ of blood from eloquent brain regions - resulting in clinical symptoms and even radiological volume loss. This is a controversial concept, but seems to be more accepted as a phenomenon in children with bAVM than with adults, perhaps due to the relationship to the developing brain [35,36,45]. We have seen several cases of children presenting with progressive lateralised neurological deficits, even developing white matter loss radiologically, as a result of often large, deep bAVM (Fig. 4). Treatment in these children is challenging, as surgical resection is often not feasible due to the high-grade nature of the bAVM. Consideration of haemodynamic augmentation (e.g., aspirin, hydration or even pial synangiosis) may be necessary and requires careful evaluation and discussion in a multi-disciplinary setting.

Risks associated with treatment of bAVMs must be balanced against the risks associated with no treatment [46]. The goal of treatment with bAVM is preservation of neurological function, with complete elimination of the bAVM. In the context of ruptured bAVMs, besides emergency life-saving and ICP measures, most physicians would recommend definitive treatment to prevent re-haemorrhage and its associated morbidity and mortality [23].

Treatment of unruptured bAVMs has garnered significant controversy following publication of results of ARUBA, a multicentre randomised trial of treatment versus medical observation of bAVMs in adults [37,38]. A significant risk of stroke and death was associated with intervention led to early termination of the trial with superiority of observation demonstrated. A number of criticisms [23] of this trial include 1) lack of defined subgroup analyses with data not analysed according to the type of intervention, patient specific factors that predict risk associated with an intervention, or the anticipated risk of rupture on the basis of the characteristics of the lesion [44]; 2) the trial did not address surgical outcome for patients with Spetzler-Martin (SM) grade 1 and 2 bAVMS who are the best candidates for surgery [6,48]; 3) mean follow-up was only 33 months, a period too brief to assess long-term risk of rupture or complete therapeutic effects of radiosurgery; 4) high percentage of partially treated cases; and 5) outcomes of different but aggregated management options did not align with either results of previous retrospective studies or experience reported by neurovascular specialists that regular treat unruptured bAVMs [8,28,30,34,42].

A more recent study also identified an increased in incidence of ruptured bAVM since 2014 (13.3% vs. 34.4%; p<0.001) as well as rates of in-hospital mortality (2% vs. 7.6%; p<0.001) and multivariable regression analysis noted that the postARUBA trial era was independently associated with both bAVM rupture and in-hospital mortality and hypothesised to be reflective of a paradigm shift to conservative and non-interventional management strategies in patients with unruptured bAVM post-ARUBA [15].

A recent single-centre study assessed outcome of 107 conservatively managed unruptured bAVMs in adults that were comparable to the patients in the medical management arm of the ARUBA study and noted an overall annual haemorrhage risk of 2.7% in the observation of period with 17% of patients having at least one haemorrhage [16]. The cumulative 1-, 5-, and 10-year overall haemorrhage rates were 3%, 11.3%, and 15.3% respectively with a long-term follow-up related mortality of 8% (compared to ARUBA’s 0% mortality rate) and was felt to reflect a real-life cohort and to be an unacceptable mortality rate among conservatively managed bAVMs. The authors supported active treatment options recommended to patients with unruptured bAVMs taking into consideration lesion-specific risk factors and identifying predictors for bAVM rupture to facilitate patient selection for active treatment.

Several factors support treatment of unruptured bAVM in children including : 1) very high lifetime risk of rupture with associated morbidity/mortality taking into consideration all formulae and rupture risk; 2) higher mortality from bAVM rupture in children; 3) higher plasticity and regenerative capacity of paediatric cerebral vasculature [7]; 4) results of ARUBA have no relevance to treatment of children, as they were excluded from the trial and the length of follow-up was too short to be of any utility in children [24]; and 5) currently there are no randomised trials in the paediatric literature supportive conservative bAVM management.

In the authors’ institution, each bAVM is assessed by a specialist neurovascular multidisciplinary team consisting of two specialist paediatric neurovascular neurosurgeons (AHDS/GJ), three paediatric neurovascular neurologists and two paediatric neurointerventional neuroradiologists and management decisions made on a case-by-case basis.