Epidemiology

Classically, arteriovenous malformations (AVMs) were thought to be strictly congenital lesions that arise during the third week of gestation secondary to disordered embryogenesis where primordial vascular channels fail to differentiate into mature arteries, capillaries, and veins, forming direct arteriovenous shunts without intervening capillary beds [1, 7, 19, 22]. However, reports of the novo formation and recurrence of AVMs after treatment exist and raise questions about the true epidemiology and natural history of these complicated vascular anomalies [1, 20, 42].

Hashimoto et al. tested the hypothesis that AVMs are dynamic lesions and found an approximately sevenfold increase in the number of non-testing endothelial cells in AVMs when compared with endothelial cells of normal cortical vessels. Furthermore, AVM vessels from younger patients showed a trend to increase Ki-67 index, providing some evidence of increased endothelial cell turnover in AVMs [12]. Other authors have documented increased vascular endothelial growth factors as possible mediator of angiogenesis in AVM development [19, 36].

Prevalence of cerebral AVMs in the overall population is less than 1%, and despite that most of these are congenital, only 3–19% of those treated for AVMs are children [7, 19, 23]. Most occur sporadically although there are documented familial cases and several syndromes associated with vascular malformations. In particular, children with Osler-Weber-Rendu disease have a 7.9% risk for developing a symptomatic AVM and potentially multiple AVMs. Wyburn-Mason syndrome may present with a visual pathway or midbrain AVM and ipsilateral facial nevi. Even in the absence of a recognized syndrome, multiple AVMs occur more commonly in children, and unlike AVMs in the adult population, pediatric AVMs have a predilection for the posterior fossa.

Natural history

Most cerebral AVMs are diagnosed at age 20–40 years, and the most frequent clinical presentation of AVMs in both adults and children is hemorrhage. Children, however, present with hemorrhage more commonly than adults (75 ± 80% vs. 50 ± 65%) [7, 14, 19, 23, 24]. Seizure is the second most common presentation in children and is seen in about 15% of pediatric patients. Infants with large vascular malformations may present with symptoms caused by mass effect from a large draining vein or varix. Infants can also present with hydrocephalus secondary to high venous pressures created by the AVM precluding an adequate gradient to allow cerebrospinal fluid (CSF) absorption across the arachnoid granulations. This hydrodynamic effect of CSF space suggests that hydrocephalus enlargement may be managed with the treatment of the AVM rather than direct CSF diversion.

The natural history of AVMs has not been determined specifically for children, and some studies suggest that the annual risk of hemorrhage in children is higher than that of adults (2–4% vs. 1–3% per year) [16, 22, 34]. Each hemorrhagic event carries a 5–10% risk of mortality and up to 50% risk of morbidity [10, 16, 44]. Some authors report that the related morbidity and mortality of hemorrhage in pediatric AVMs is even higher than that of adults [22, 24, 31, 34]. This may be related to the increased incidence of lesions located in the deep basal ganglia and brainstem in children [14].

The significant risk of hemorrhage associated with pediatric AVMs, the resultant morbidity, and the longer life expectancy of children necessitate aggressive treatment of these lesions. Furthermore, the lower incidence of comorbidities in children and the plasticity of the developing CNS often allow for more aggressive therapeutic intervention when compared to the adult population. Treatment often employs a combination of therapies which may include microneurosurgery, radiation, and endovascular embolization. The use of intraventricular t-PA in the treatment of intraventricular hemorrhage seems to be effective; however, randomized clinical trials are required [40, 41].

Initial presentation and evaluation of AVMs in children

As previously mentioned, hemorrhage is the most common presentation in pediatric AVMs, and in the absence of known trauma, a child with a new intracranial hemorrhage is presumed to have an underlying vascular anomaly until proven otherwise. The primary goal of an initial presentation is to stabilize the child with respect to airway management and hemodynamic resuscitation if necessary. A non-contrast CT scan is usually obtained by emergency physicians prior to being transferred to tertiary care centers or being seen by the pediatric neurosurgeon. A simple non-contrast head CT will provide the caregiver with important information regarding the location of the hemorrhage, and in some cases, dilated vessels and calcification may be seen to indicate the presence of an underlying vascular lesion. Most centers are now capable of performing CT angiograms as part of the initial management which can further define underlying lesions (Fig. 1). The clinical status of the patient then determines whether emergent neurosurgical intervention is needed for intracranial pressure (ICP) management. Further workup is often necessary prior to intervention when the patient is neurologically and hemodynamically stable. Basic ICU care with strict blood pressure control utilizing arterial line monitoring and antiepileptic drug administration if necessary will assist in controlling ICP and decreasing re-hemorrhages rates. Further workup is utilized to better define and evaluate the malformation including magnetic resonance imaging (MRI), magnetic resonance angiography (MRA), CT angiography, and lastly formal cerebral angiography [19, 21, 34]. Although angiography remains the gold standard, MRI with MRA is often performed as it can provide significant information about the architecture of the feeding vessels, nidus size, draining vein, location, and age of blood products, all done in a non-invasive manner. A formal catheter angiogram, however, best defines the anatomy of these lesions, often requires general anesthesia, and includes the risks of invasive catheterization. Conventional angiography should be considered in children presenting with an ICH of unclear origin or in the case of a strong diagnostic indication, given the inherent procedural risk [17, 45]. In a series of 241 pediatric patients, Burger et al. showed that diagnostic cerebral angiography can be performed in children with extremely low periprocedural complication rates [5]. With small infants or children, the contrast load must be taken into account and may require additional staging of interventions.

Fig. 1
figure 1

Patient with history of drug abuse presented with severe headache. The head CT demonstrated subarachnoid hemorrhage. a, b The patient subsequently had an angiogram which revealed a right posterior lateral choroidal AVM draining into the vein of Galen with an intranidal aneurysm (white arrow). c There is no evidence of residual vascular malformation after endovascular embolization demonstrated 8 years later

Preoperative embolization

Angiography with embolization continues to grow as a part of the armamentarium used in treatments for AVM; however, complete obliteration of AVM by embolization remains low. Estimates of angiographic cure from embolization range from 7% to 14%, with one large study by Frizzel et al. that analyzed 1,246 patients from the literature reported a cure rate of only 5% [8]. Morbidity and mortality of embolization procedure for AVM range from 7% to 12% and 1% to 6%, respectively [4]. Deruty et al. reported an overall complication rate of 25% after embolization [6]. Success rates of embolization in the literature are affected by AVM architecture, the location, and the predilection for the surgical management of lower SM grade lesions. Newer embolic materials such as Onyx (eV3, Ca) and its success rates are not clearly reflected in the literature [16]; therefore, more studies on the use of Onyx with long-term follow-up are needed to confirm its long-term safety and efficacy [39]. While, in many cases, embolization alone does not seem to be a standard therapy, it has proven to be an effective adjunct therapy to surgical management and radiation treatment of AVMs. Partial embolization of an AVM, which can be achieved in an estimated 54–90% of attempted cases, provides the surgeon with better hemostasis and less risk of clinically significant blood loss during surgery (Figs. 2 and 3). Reducing intraoperative blood loss is critical especially in children who have less reserve secondary to small circulating blood volumes and less ability to compensate for blood loss when compared with adults.

Fig. 2
figure 2

A 5-year-old male with intracerebral hemorrhage secondary to a right frontal temporal AVM. a There is a large hematoma in the superior temporal gyrus consistent with hemorrhage of an AVM. There is minimal secondary compression of the right temporal horn. b Central decreased signal intensity on FLAIR imaging consistent with deoxyhemoglobin of an evolving hematoma. The circumferential high signal is consistent with surrounding edema. There is a small lacunar infarct in the right basal ganglia. c MRA imaging demonstrates no arterial nidus feeding the right temporal AVM

Fig. 3
figure 3

a The angiogram revealed a right frontal temporal AVM draining into the vein of Galen. b Status post embolization and resection of the right frontal temporal vascular malformation. There is no evidence of residual vascular malformation 6 months later

Stereotactic radiosurgery

Recently, there are numerous publications about the effectiveness of radiosurgery in the treatment of pediatric AVMs. Unfortunately, there is a wide range of results in terms of overall obliteration rates, ranging from 30% to 80%, likely due to differences in XRT dosing, AVM characteristics, study design, and the use of multimodal therapy confounding these results [28, 35]. Predictive factors for successful management of AVM using stereotactic radiosurgery (SRS) also differ in the literature, but smaller lesions, lower SM grades, younger age patients, and higher radiation doses have all been reported as positive predictive factors [25, 27, 28, 33, 35, 47].

One of the most significant advantages in SRS may lie with its low incidence of morbidity and mortality. One large study of 100 pediatric patients with >36 months follow-up demonstrated an overall morbidity rate of 6.7% and a 5% permanent neurological deficit rate and death of one patient due to recurrent hemorrhage [28]. It should be noted that neurological deficits after SRS are often associated with lesions located in the eloquent cortex, deep basal ganglia, or brainstem that would presumably result in significantly higher morbidity if surgically approached. However, as with all newer therapies, long-term outcome data with regards to recurrence rates remain to be seen, and delayed complications from radiation therapy such as cognitive impairment and secondary malignancy are well-known complications of radiation therapy in children.

It should also be noted that about one third of patients undergoing SRS will have radiation-induced changes on follow-up MRI images, but only in a small percentage that these radiological changes correspond to clinical signs or symptoms [11, 29, 43, 46]. Radiation-induced malignancies following external beam radiotherapy are well documented; however, there are some reports of this complication following radiosurgery of AVMs [2, 18, 32].

Unique to SRS versus other forms of treatment for AVM is the delay of radiographic obliteration after treatment. Complete obliteration is often not seen until 2–3 years post treatment, with average delay to obliteration (DO) of 36 months [7, 14, 21]. This rate is similar to that found in adults and there is a direct relationship with DO and the size of the treated lesion [25]. The exact incidence of re-hemorrhage during this period is unknown, and re-hemorrhages are a documented complication of SRS with some authors reporting a 6% re-hemorrhage rate. There is some evidence, however, as that shown by Pollock et al. that radiosurgery offers at least some protection against hemorrhage even if complete obliteration of the AVM is not achieved [26]. But since the rate of hemorrhage is low to begin with, it is unclear if this represents a return to the natural history of the lesion.

Many AVMs encountered will be amenable to either surgery or SRS. The advantages of SRS fall in the lower procedural morbidity and mortality when compared to surgery. However, complete obliteration rates in surgically amenable lesions are generally higher than those achieved with SRS, and there is no lag time in radiographic cure after complete resection. Significant rates of mortality have been reported in young patients experiencing hemorrhage from an AVM in the cerebellum during the latency interval; therefore, surgery should be seriously considered on those patients [33]. For these reasons, most surgeons would recommend surgical treatment of lower grade, accessible AVMs over SRS therapy. For more complex, surgically inaccessible lesions or when other factors prohibit surgical intervention, SRS remains an effective tool for AVM treatment.

Surgery

When amenable, the surgical resection of an AVM remains the fastest and most complete method [15, 19, 28]. Complete excision remains the gold standard of treatment if it can be performed with acceptable rates of morbidity [3] (Fig. 4). Moreover, immediate surgical treatment is indicated in patients with intracerebral hemorrhage and associated progressive neurological deficits. The SM scale for AVMs was developed using data from adult patients; however, multiples studies have shown that is applicable to the pediatric population as well. In their series of 20 pediatric AVMs of SM grades I–III, Kiris et al. reported an 89% radiographic obliteration rate with 5% morbidity and 5% mortality [19]. Similarly, a larger study including both adults and pediatric patients with SM grades I–III by Schaller and Schramm reported a 98.4% cure rate with 3.2% morbidity and no mortality, concluding that in experienced hands microsurgery may provide superior results compared to either embolization or radiation therapy when analyzed independently [31]. Sanchez-Mejia et al. observed superior outcomes in children who underwent microsurgical AVM resection compared with adults, indicating that neural plasticity may augment surgical tolerance and recovery in children [30].

Fig. 4
figure 4

a Patient with a giant parietal arteriovenous malformation measuring approximately 9 × 5 cm. b A complete microsurgical resection of the AVM was performed with excellent results

Surgical technique

Planning for complete surgical excision of parenchymal pediatric AVMs involves preoperative angiography, frameless stereotaxy, and coordination with neuroanesthesia. At our center, many patients undergo several stages of preoperative embolization. Due to the fragile hemodynamics of small children with limited intravascular reserves, an arterial line, a central venous line, and EEG leads all are placed in anticipation of possible blood loss and the need for burst suppression for cerebral protection. Mild intraoperative hypothermia is often used for the same reason. The child is positioned in either an adult or pediatric three-pin head fixation, depending on the age and weight, or on a simple padded headrest in the case of an infant. We routinely use a horseshoe headrest for children under 2 years of age and at least three-pin fixations for slightly older children. Frameless navigation cannot be performed in children too young for skull fixation. These children are also not candidates for functional MRI or intraoperative corticography, making a complex resection in a young child a more complex undertaking [37].

The opening is not different than that for any pediatric craniotomy with the one caveat of being cognizant of the possible meningeal vessels contributing to the vascular malformation. Microsurgical dissection of the AVM is performed under the operating microscope and proceeds circumferentially around the lesion in the place between the parenchyma and abnormal vasculature. Care is taken to identify feeding arteries and draining veins.

It is utmost of importance during surgery to be certain that feeding vessels planned for sacrifice are not feeding normal parenchyma and that all of the draining veins are not sacrificed until the AVM’s arterial supply has been extirpated completely. A careful review of preoperative angiography is imperative to identify these important components of the angioarchitecture. These lesions are often extremely fragile and subject to hemorrhage with even minor manipulation. As the surgeon works in a circumferential manner, feeding arteries are sacrificed and the nidus is bipolared shrinking the lesion and controlling blood loss. Telfa padders are placed in the dissection plane between AVM and the parenchyma. This protects the surrounding brain from injury and allows the surgeon to develop an understanding of the mass of the lesion. Teflon-coated, irrigating bipolar cautery is essential along with topical hemostatic agents and patience.

Complication avoidance

Hemorrhage is certainly the most feared complication before, during, and after AVM resection. Intraoperative blood loss, even if not from frank uncontrollable hemorrhage, is crucial to limit in small children whose hemodynamic reserves are limited. Hemodynamic shock can ensue after one-quarter of a child’s blood volume has been lost, an amount that is very small in infants and small children. Blood volume in milliliters can be roughly calculated by multiplying the child’s weight in kilograms by 80, and thus, a 10-kg 1-year-old child can go into shock after losing as little as 200 ml of blood. Preoperative embolization when possible should be preferred to assist an intraoperative blood loss control.

Although avoiding hemorrhagic complications is the responsibility of the surgeon, the anesthesia team must alert the surgical team to blood loss, and more importantly, to any hemodynamic consequences, including tachycardia or hypotension, which may necessitate temporary cessation of dissection or abortion of the case entirely. Attempts to minimize blood loss during the scalp opening, craniotomy, and dural reflection must be rigorous, as must meticulous circumferential dissection and avoidance of premature sacrifice of draining veins. In conjunction with preoperative embolization, these techniques will help avoid catastrophic hemorrhage.

Aside from hemorrhage, the major complications involve injury or resection of surrounding parenchyma, either through direct resection or, more commonly, due to sacrifice of vessels that partially supplied those territories. Attention to the preoperative MRI and angiogram will help define the limits of resection along with intraoperative stereotaxy and visual guidance. Resecting unnecessary cortex is often balanced against leaving behind residual AVM. The surgeon should not hesitate to perform intraoperative angiogram if there is concern about leaving residual AVM that may necessitate further treatment.

In the immediate postoperative setting, several complications may arise. Postoperative hematomas, either from poor resection cavity hemostasis or from residual AVM, are possible surgical emergencies requiring operation. Of less obvious but equally dangerous concern is swelling or hemorrhage in the surrounding brain due to redistribution of vascular flow. Blood that was being directed through a very high flow malformation is now redirected into smaller vessels. The parenchyma supplied by these vessels is not only used to such high flow, but also it likely has been the victim of vascular steal from the adjacent malformation. When the hyperperfusion of that tissue outstrips the autoregulatory mechanisms of the vessels, infarct, edema, and hemorrhage, all may result over a period of hours to days.

Seizures also may occur postoperatively, and care must be taken to ensure strict control of blood pressure in these instances. It has been calculated that new onset seizures occur in more than 10% of children after AVM resection, although less than half of them will require chronic seizure control. Vasospasm, stroke, and vascular thrombosis are rare but reported complications in children.

In summary, the postoperative patient’s complications will depend almost entirely on the extent of resection and the ability of the brain to compensate for new blood flow dynamics. However, the intensive monitoring of these children with arterial lines and frequent neurological checks will detect any complication early and allow prompt intervention.

Recurrence of pediatric AVMs

Recurrence of AVMs after angiographically proven obliteration is low with reports ranging from 1.5% to 5.5% [4]. Most recurrences (69%) occur in children and latency between and recurrence is often quite delayed with the longest reported interval lasting 19 years. Pathogenesis of recurrent AVMs remains a mystery. It is unclear if recurrent lesions occur de novo or arise from angiographically occult residual after treatment. There is evidence that both occur. Akimoto et al. reported a case of a symptomatic de novo AVM appearing ectopically 17 years after total resection of other two AVMs [1]. Numerous theories about the natural growth of AVMs and possible reasons for recurrence exist. It has been theorized that growth occurs secondary to hemodynamic stress on dysplastic vessels of an AVM with recruitment of collateral vessels, destruction of surrounding neural tissue through recurrent hemorrhage may create space for the dysplastic vessels to grow into, and others propose the shunt itself promotes angiogenesis [13, 20, 38]. Other authors suggest AVMs are dynamic lesions that contain “hidden compartments” or “reserve nidus” outside the main nidus that are picked up during routine angiography as they initially have low or absent blood flow [9, 33].

Outcomes

The exact basis for judging neurosurgical success in children with AVMs may differ from series to series, but the most important factors are prevention of further hemorrhage, control of seizures, and avoidance of neurological decline. On the basis of these criteria, it now appears appropriate to quote a 95% success rate, with angiographic obliteration rates nearly matching that. The rate of severe morbidity is close to 10%, and the mortality rate is approximately 5% depending on the size of the series and the average grade of the lesions resected.

Di Rocco et al. suggest that of all prognostic factors, neurological status may be the best predictive factor of good outcome [7]. Children with devastating hemorrhages may eventually recover to a degree not possible in adults with similar hemorrhages and neurological grades. In one study, more than 50% of comatose children who survived a ruptured AVM had a good functional outcome. With respect to seizure control, more than half of the children who present initially with seizures are seizure-free without antiepileptic medication after resection. Early surgery and a younger age prognosticate better long-term seizure control in children. Because of the possibility of AVM recurrence, children must be followed for several years after treatment for AVM irrespective of the mode of treatment. We favor yearly MRIs due to ease and non-invasiveness when compared with angiography, but angiography must be performed within the first several years after resection to confirm obliteration. We also perform an angiogram at a 5-year interval from a suspected cure to evaluate for subclinical recurrence.

Over the last 25 years, the management of pediatric AVMs has evolved into a multidisciplinary art requiring the close interaction of neurological surgeons, interventional neuroradiologists, and radiation oncologists as well as neuroanesthesiologists and pediatric intensivists. To optimize the neurological and angiographic outcomes, each child presenting with an AVM must have an individualized treatment paradigm designed with those goals as primary endpoints. As embolization, surgery, and SRS become further refined and indication for each is better defined, we will be able to offer children with these lesions an increasingly excellent change for a complete recovery and angiographic cure.