Introduction

Although cerebral arterio-venous malformations (AVMs) belong to the most frequent cerebral vascular malformations, AVMs of the posterior fossa are relatively rare comprising only between 10 and 15 % [1, 8]. Based on their location near to important neurovascular structures such as the brainstem and the cranial nerves, these patients are at an increased risk for severe morbidity or even mortality due to bleeding of the AVM or additionally caused by the treatment.

Over the last decades, only a small number of larger series with posterior fossa AVMs have been published, including historical surgical series [2, 22, 7] or series treated inconsistently by different treatment modalities such as surgery, radiotherapy or more recently by endovascular therapy [20, 12, 19, 3]. However, treatment methods have changed over the last decades, especially due to technical developments of endovascular procedures, as well as due to more insights about the natural history of AVMs and their risk for bleeding.

The aim of our study was to analyse a sizable series of patients with posterior fossa AVMs, consistently treated in our neurovascular centre by an interdisciplinary team with individualized multimodal treatment concepts.

Methods

Patients

Between 1998 and 2012, a total of 53 patients with AVMs of the posterior fossa were seen in our neurovascular centre. There were 20 males and 33 females with a mean age of 43.8 years (age ranging between 2 and 87 years). Basic demographic data, clinical presentation and outcome were evaluated. Initial clinical presentation on admission was graded according to the Glasgow Coma Scale (GCS). Thirty-nine patients (74 %) presented with intracranial haemorrhage, either from the AVM or an AVM-associated aneurysm. Haemorrhage was defined as the acute onset of clinical and neurological symptoms as well as detection of blood revealed by computed tomography (CT) or magnetic resonance imaging (MRI). Patients with unspecific symptoms leading to the diagnosis of the posterior fossa AVM or incidental AVMs were assigned to the non-haemorrhagic group (n = 14, 26 %).

Neuroimaging studies

Patients were examined with MRI and cerebral digital subtraction angiography (DSA) before treatment. MRI was performed to determine the exact location, eloquence, size and configuration of the AVM. AVMs located in the deep cerebellar nuclei, cerebellar peduncles or brainstem were considered eloquent [24].

Cerebral DSA was performed to assess the angiographic characteristics of the AVM with special consideration of number and type of arterial feeders (e.g. en passant) and absence or presence of associated aneurysms. Aneurysms were divided into feeding artery aneurysms (FAA), intranidal and venous aneurysms.

Characteristics of the AVM nidus (plexiform, high-flow/fistulous) and type of venous drainage were also determined by this procedure. Superficial venous drainage was only assumed in those cases with direct drainage into the straight and transverse sinus [24].

Based on the neuroradiological findings, AVMs were classified according to the original 5-tier and the revised 3-tier Spetzler-Martin grading system (SMG) [24, 25].

Decision-making process and treatment indication

Treatment options were evaluated in an interdisciplinary approach in our neurovascular centre by experienced neurosurgeons, neuroradiologists and, if considered necessary, radiosurgeons. Criteria for treatment decision were clinical and neurological symptoms, type of presentation with or without haemorrhage, with special consideration on the potential source of bleeding (e.g. presence of flow-associated aneurysms) and SMG. For patients, who presented with non-haemorrhagic symptoms, the decision was based additionally on the estimated haemorrhage risk as well as on patients’ will.

With exception of emergencies showing signs of decerebration (n = 4), all patients with acute bleeding were admitted to treatment in the subacute stage, giving the opportunity of clinical stabilization. Mean time interval between bleeding onset and AVM treatment was 11 days, ranging from 1 to 21 days. The primary aim was complete AVM elimination by endovascular procedures, microneurosurgery or in a combined approach with one or repetitive preoperative interventional procedures in order to reduce both the flow in the AVM feeding arteries and the size of the AVM nidus. In cases, in which only incomplete AVM occlusion was assumed, or in those cases with significantly increased surgical and endovascular risk, radiosurgery or conservative therapy was recommended.

The treatment of associated aneurysms was mainly addressed to FAA. Thus, nidal and venous aneurysms were incorporated into the treatment strategy of the AVM itself, due to the fact that bleeding from these lesions could not be distinguished conclusively from rupture of the AVM nidus. In case of bleeding from a ruptured FAA, treatment was primarily directed towards the symptomatic aneurysm by either endovascular or microneurosurgical procedures, alone, or in combination with the AVM treatment. FAAs of patients with haemorrhage from ruptured AVM as well as patients without bleeding were treated based on the estimated risk from both FAA and AVM.

Follow-up

Patients were followed up clinically and radiologically. Mean follow-up was 19 months, ranging from 2 to 96 months. Clinical outcome was assessed using the modified Rankin scale (mRS) in order to achieve a good functional grading. Patients’ outcome was defined as favourable, if the mRS was less than three and unfavourable if equal or higher than three. Most treated patients underwent follow-up DSA to assess the obliteration rate of the AVM or the associated aneurysm.

Statistical analysis

To assess the influence of angiographic characteristics on haemorrhagic presentation, odds ratio (OR) and 95 % confidence interval (CI) were calculated using a univariate analysis. A p value of <0.05 was considered as statistically significant.

Results

Clinical presentation

On admission, 39 patients presented with haemorrhage (74 %) and 11 individuals with AVM-related unspecific symptoms. These symptoms were vertigo (n = 6), temporary sensory (n = 2) and cranial nerve (n = 2) disturbances. One patient suffered from a repetitive syncope. In three cases, AVMs were considered incidental with no symptoms. Epilepsy as a primary symptom was not observed. Seizures following a haemorrhagic event were seen in four patients.

A total of 33 of 39 patients (85 %) presenting with haemorrhage were referred to our neurovascular centre in the acute stage, whereas six patients (15 %) were admitted in a delayed stage between 2 months and 3 years after bleeding (mean 18 months). Among these, 20 patients (51.5 %) were in a good clinical condition with a GCS > 13, including all six patients, presenting in a subacute state. Six patients (15.5 %) presented in an intermediate clinical stage according to a GCS 8–12. Thirteen patients (33 %) were in poor clinical condition with unconsciousness and a GCS of <7. Among these, nine individuals were admitted with a GCS = 3. In four comatose patients, signs of decerebration with unilateral or bilateral dilated and non-reactive pupils were present.

Acute hydrocephalus was found in 22 individuals (56 %). All patients except for one in poor and intermediate clinical condition displayed with concomitant hydrocephalus, compared with only four patients in good clinical stage. Five patients suffered from rupture of a FAA, representing 13 % of all bleedings. The clinical presentation is summarized in Table 1.

Table 1 Clinical presentation of 53 patients with posterior fossa AVM
Full size table

Radiological findings

In 22 out of 39 patients (56 %) with haemorrhage, a cerebellar haematoma was found, which was located in the cerebellar hemispheres in 12 cases (55 %) and in the vermis in ten cases (45 %). Mean maximal haematoma diameter was 31.5 ± 13.0 mm. Seventeen patients (44 %) suffered from subarachnoid haemorrhage (SAH) of the basal and suboccipital cisterns or the fourth ventricle.

AVMs were located in the cerebellum in 44 patients (83 %), in the brainstem in four patients (7.5 %) and in the cerebello-pontine angle in another four individuals (7.5 %). Among patients with cerebellar AVMs, 13 (30 %) were detected in the vermis and 31 (70 %) in the cerebellar hemispheres, including the tonsils and the cerebellar peduncles. In one individual, the AVM occupied nearly the complete posterior fossa. Eloquent location, as defined above, was found in 29 patients (55 %).

Complete angiographic data were available in 49 patients (92 %). Those four patients with signs of decerebration underwent immediate surgery without preoperative DSA but received CT angiography prior to surgery. According to the SMG, AVM size was graded as small in 29 patients (59 %), intermediate in 17 patients (35 %) and large in three patients (6 %). En passant arterial supply was found in 11 cases (22 %). Twelve AVMs (24 %) had only one single feeder and 15 cases (31 %) two feeders. More than two arterial feeders were found in 22 cases (45 %) and bilateral supply in 11 patients (22 %). A high-flow/fistulous nidus component was present in nine cases (18 %).

Deep venous drainage was identified in 19 patients (39 %), whereas superficial drainage was seen in the majority of cases (n = 30, 61 %). Venous aneurysms or ectasias were seen in four patients (8 %).

A total of 17 FAAs were found in 12 of 49 patients (24 %), with four patients carrying multiple aneurysms. In two patients, an additional intranidal aneurysm was found and one patient harboured a single intranidal aneurysm only. Most frequent location for FAA was the posterior inferior cerebellar artery (PICA) in eight cases (67 %). Rupture from FAA was found in five patients. Thus, 40 % of patients with FAA suffered from aneurysm rupture.

Whereas the AVM size (<3 cm) was associated with a statistical higher risk for haemorrhage (p = 0.04), no other angiographic features showed an influence on clinical presentation or haemorrhage. An overview of the patients’ radiological characteristics in relation to clinical presentation (haemorrhagic vs. non-haemorrrhagic) is shown in Tables 2 and 3.

Table 2 Angiographic characteristics of 49 patients with complete angiography in correlation to clinical presentation
Full size table
Table 3 Univariate tests of angiographic characteristics on incident haemorrhage in 49 patients with posterior fossa AVMs
Full size table

Treatment

Forty-two patients (79 %) were treated by either embolization, surgical resection, with or without preoperative embolization or radiosurgery. Among this group, 31 patients initially presented with haemorrhage and 11 patients without.

Twelve patients received surgical resection with preoperative embolization. Mean time between endovascular treatment and surgical AVM excision in patients with haemorrhage (n = 8) was 5.6 days (range 2 to 16 days), compared with 14.5 days (range 1 to 32 days) for patients without haemorrhage (n = 4).

Sixteen patients underwent surgery alone without preoperative endovascular treatment, including those four cases with emergency surgery due to life-threatening haemorrhagic presentation.

Endovascular embolization alone was performed in 12 patients, including three patients with sole coiling of the associated aneurysms and without further treatment of the AVM itself. Two of these patients suffered from rupture of the aneurysm. Reasons for leaving the AVM untreated were a poor posthaemorrhagic condition (mRS = 4) without recovery after 19 months in one patient and a significantly increased risk for treatment in the other two cases.

Two patients without bleeding were submitted to stereotactic radiosurgery, because of an increased treatment related risk for both endovascular therapy and surgery.

Eleven patients did not receive any treatment. Three patients died from primarily devastating bleeding within 2 days and 1 week, respectively. Four patients refused therapy, although treatment was recommended. Among these, three patients initially presented with haemorrhage. In four cases, the AVM was classified as untreatable and observation only was recommended, with two patients presenting with haemorrhage and two patients without. An overview of all treatment modalities is shown in Table 4.

Table 4 Treatment modalities for 53 patients with posterior fossa AVM
Full size table

Complications

In 12 cases (29 %), major therapy-related complications occurred. Three patients with haemorrhage sustained rupture of the AVM during endovascular embolization with a need of urgent decompressive suboccipital craniectomy, evacuation of the haematoma and excision of the remaining AVM. Among these patients, one patient died and one patient made a good recovery (mRS = 2). The third patient, initially presenting in a poor clinical state (GCS < 7), did not recover and remained in an unfavourable condition (mRS = 4). One patient with complete endovascular obliteration developed progressive cerebellar swelling 2 days after embolization and underwent secondary suboccipital decompression. This patient made an excellent recovery (mRS = 1). Another patient in intermediate condition after initial haemorrhage suffered from midbrain ischemia after embolization and remained in highly disabled clinical condition (mRS = 5) at the time of the last follow-up.

Three individuals suffered from rebleeding after primary surgery. Cause for postoperative haemorrhage in one case was subjacent coagulopathy, due to known hepatic dysfunction. The second patient developed perioperative coagulopathy. Due to significant clinical deterioration, both patients died 15 and 16 days after surgery. The third patient, harbouring a large symptomatic AVM of the cerebellar hemisphere without haemorrhage, died 2 days after surgery because of incomplete subtotal nidus resection and fatal rebleeding. One patient sustained a cardiac arrest with a need of resuscitation 2 days after treatment and did not recover (mRS = 5). Two patients with haemorrhage and good initial clinical state suffered from uni- and bilateral PICA infarction after AVM resection, leading to a mRS 4 and mRS 5 on discharge. One patient with haemorrhage and good initial clinical state showed severe neurological deficits after uneventful surgery and died 2 months after surgery during rehabilitation, due to a pulmonary embolism. Thus, treatment related morbidity was 12 % (n = 5) and mortality 12 % (n = 5), respectively.

Clinical outcome

Of those patients with either treatment of the AVM or an associated aneurysm (n = 42), 25 patients (60 %) were in good or favourable state at the last follow-up, according to a mRS = 2 or less. Among this group, 16 patients showed an excellent recovery (mRS 0–1, 38 %) including five patients initially presenting in a poor and one patient in an intermediate state.

Thirteen patients (31 %) remained in a poor clinical condition with a mRS ≥ 3 in eight cases due to a poor initial state (GCS ≤ 7). However five patients (12 %) sustained worsening of their clinical condition after treatment, due to major complications, as described above.

Of all 53 patients with an AVM of the posterior fossa, 14 patients remained in poor clinical state with severe neurological deficits and eight patients died, accounting an overall morbidity of 26 % and mortality of 15 %.

Radiological outcome

Among the 42 treated patients, 37 patients received radiological follow-up evaluation with DSA. Complete AVM elimination was achieved in 30 individuals (81 %). In four cases, the AVM was incompletely treated. Two patients with subtotal AVM embolization (>95 % embolization rate) refused further treatment, although radiosurgery or surgery was recommended. One patient with initial bleeding remained in poor clinical condition (mRS = 5) after partial embolization (50 % embolization rate), not capable for further surgery. Another patient with haemorrhage, treated with endovascular embolization, followed by surgical subtotal nidus excision, received a postoperative stereotactic radiosurgery.

In all cases of SAH from rupture of an associated aneurysm (n = 5), the bleeding source was treated completely by coiling of the aneurysm or occluding the parent vessel. Three patients received additional treatment of the AVM, whereas two patients did not undergo AVM treatment, as mentioned above. Overall, 10 out of 12 patients with associated aneurysms received additional aneurysm obliteration. Radiological outcome details are shown in Table 5.

Table 5 Obliteration rates depending on treatment modality
Full size table

Illustrative cases

Case 1

This 16-year-old girl was admitted in comatose state with a GCS of 3 without signs of decerebration. Cranial CT revealed a large left cerebellar haemorrhage with a maximum diameter of 45 mm and consecutive hydrocephalus. Initially, a ventricular drainage was inserted and the patient was transferred to our hospital. Subsequent MRI and DSA demonstrated a left para-vermal cerebellar AVM with a diffuse nidus configuration and a size of 20 mm. Arterial feeders arose from the left superior cerebellar artery (SCA) and the left anterior inferior cerebellar artery (AICA). No associated aneurysms were detectable, and venous drainage was directly into the left sigmoid sinus. According to the 3-tier SMG, the AVM was classified as grade II. A multimodal treatment strategy in the subacute stage was decided, with initial endovascular intervention, followed by surgical excision, if necessary. Partial embolization was performed 12 days after bleeding via the left SCA, whereas further embolization via the left AICA was not feasible, because of diffuse and multiple en passant feeders. Thus, the patient underwent uneventful microsurgical resection 2 days after embolization, via a left suboccipital trans-cortical approach in sitting position. Postsurgical angiography demonstrated complete AVM elimination. Postoperatively, the patient recovered continuously with broadly regained daily independence 3 month after rehabilitation, moderate gait-ataxia and diplopia (mRS = 2) (Fig. 1).

Fig. 1
figure 1

Case 1; a Initial CT scan showing a large right cerebellar haemorrhage with compression of the fourth ventricle. b–d Preinterventional MRI and cerebral DSA revealing a left 20 mm cerebellar para-vermal AVM with arterial supply from the left SCA and AICA, draining into the left sigmoid sinus. e After partial embolization, DSA demonstrates a significant reduction of flow and nidus size, but a still present early draining vein (arrow). f–h Postoperative CT scan after surgical resection of the residual nidus and controlangiography, verifying complete AVM elimination; 219 × 124 mm (300 × 300 DPI)

Full size image

Case 2

This 45-year-old neurologically intact female patient complained of chronic headache, ear and neck pains, combined with intermittent nausea. After an orthostatic dysregulation, cranial MRI was performed, showing an AVM of 35 mm in the left cerebellar hemisphere. DSA confirmed a plexiforme AVM of the left para-mediane cerebellar hemisphere. This AVM SMG III [25] was supplied by feeders from the left SCA, AICA and PICA, with venous drainage into the left transverse and left inferior petrous sinus. Additionally, an associated distal bilobar saccular PICA aneurysm was present, with a maximum diameter of 8 mm. The situation was discussed extensively, and recommendation was given towards treatment, which was in accordance with the patient’s strong will for treatment.

Endovascular treatment led to complete coil embolization of the aneurysm with preservation of the parent PICA and in the same setting to nearly 50 % occlusion of the AVM nidus, thus, obliterating the caudal portions of the AVM. Further embolization was judged as not possible due to en passant supply from the AICA and SCA. The patient tolerated this intervention without any problems and subsequent microsurgical resection was accomplished 16 days later. Complete AVM resection was proven by postoperative angiography. At the last follow-up 5 years after treatment, she showed no deficits and a full activity at daily life and work (Fig. 2).

Fig. 2
figure 2

Case 2; a, b Initial angiogram in a patient without bleeding and unspecific symptoms showing a plexiforme AVM of the left paremediane hemisphere, feed by branches of the left SCA, AICA and PICA. An additional FAA of the left PICA was found (arrow). c–e Postinterventional angiogram and MRI after partial AVM embolization and coil occlusion of the FAA (arrow). f–h Postoperative DSA and MRI after subsequent surgery, showing complete AVM resection; 173 × 91 mm (300 × 300 DPI)

Full size image

Discussion

Clinical presentation

The natural history of posterior fossa AVMs is still poorly understood. However, if these lesions become symptomatic, their treatment remains challenging despite significant technological developments over the last two decades. Due to the specific anatomical conditions of the posterior fossa with the presence of highly eloquent neurovascular structures as the cranial nerves and the brainstem, these patients are at an increased risk for severe neurological symptoms and even mortality.

In contrast to supra-tentorial lesions, haemorrhage and subsequent hydrocephalus was the leading clinical presentation for posterior fossa AVMs in our series, which was also reported by others [11, 13, 21]. On admission, 74 % of our patients presented with haemorrhage, most commonly as an intraparenchymal haematoma (56 %). This observation is different to previous reports, emphasizing subarachnoid haemorrhage as the most common presentation in 61–90 % [9, 10, 26]. Haemorrhage and a resulting tight posterior fossa [28, 4] with consecutive brainstem compression and hydrocephalus might be one reason why nearly 50 % of our collective presented in a poor or intermediate neurological condition with impaired consciousness and severe neurological deficits. Similar observations were made by the Toronto Brain Vascular Malformation Study Group [3], accounting a bleeding rate of 62 % and neurological deficits in 48 %.

Interestingly, in our series, only a small AVM nidus size correlated with an increased risk for bleeding, which has been previously reported by others [23, 15]. Despite these findings, we could not identify further potential angiographic features to estimate future haemorrhage. Although we found FAA in 29 % of patients presenting with haemorrhage compared with only 14 % in the non-haemorrhagic group, this finding failed to show statistical difference.

Treatment indication and modalities

With the ongoing development in treatment techniques over the last years, an interdisciplinary board for treatment indication and decision processes nowadays is mandatory. The risks of treatment have to be balanced with the natural history and the risk of bleeding. Treatment appears justified for symptomatic patients with bleeding and potentially treatable lesions [16]. However, therapy indications have to be thoroughly reviewed in patients with non-specific symptoms or incidental AVMs. Reflecting a more conservative approach not only for non-symptomatic posterior fossa AVMs, the number of untreated posterior fossa AVMs constantly grew from about 6–9 % in earlier reports [2, 7] up to 18–37 % in more recent publications [3, 26], with 21 % of untreated lesions in our cohort.

Taking three historical surgical series together [2, 22, 26], only six of 65 treated patients (9 %) underwent an endovascular intervention of their AVM as a single or a combined procedure, thus, endovascular approaches displayed only a marginal role. Kelly and co-workers [12] recently presented a larger series of multimodally treated AVMs of the posterior fossa, reflecting a more interdisciplinary treatment strategy. The growing role of endovascular interventions can also be pointed out by other publications, emphasizing endovascular embolization as an important additional treatment modality, either alone or as part of a multimodal treatment concept for the majority of patients [3, 18].

We only considered radiosurgery alone in two carefully selected cases, thus leading to a 100 % obliteration rate within 3 years. Both individuals harboured a small AVM involving the brain stem. There are several studies dealing with radiosurgery for deep-seated AVMs of the posterior fossa. Nagy et al. [19] presented their large number of 160 pons/medulla sited AVMs with considerable obliteration rates up to 70 % after 4 years in those malformations, smaller than 4 cm3. Adverse radiation effects and clinical outcome were though correlated with AVM size, leading to 50 % permanent morbidity in larger AVMs. In contrast, Kelly et al. [12] reported only 29 % occlusion rates after 3 years, including only SMG III–V and thus certainly larger AVMs. They finally recommended radiosurgery only for brain stem AVMs, whereas all other lesions of the posterior fossa should undergo an interdisciplinary interventional approach.

Clinical outcome

Excellent and good clinical results were reported in up to 67–81 % of treated patients even for earlier series [2, 3, 7, 12]. These results are slightly different from our findings with a favourable outcome in only 60 %. Despite improvement of surgical and endovascular techniques, different patient characteristics and a high number of patients in a poor clinical condition in our cohort might explain this discrepancy. There are numerous reports indicating the initial neurological state as a major predictive factor for clinical outcome for patients with spontaneous posterior fossa bleedings as well as for bleedings due to cerebrovascular malformations [4–6, 27, 29]. Although the impact of haemorrhage from a ruptured AVM or spontaneous cerebellar haemorrhage is different, the local effects with brainstem compression and hydrocephalus are similar, leading to a high number of initially poor-grade patients and thereby to a poor clinical outcome in 31 % in our cohort. Other authors suggested that patients with cerebellar haemorrhage and a comatose state should not be considered for intensive therapy [14]. However, we observed 11/19 patients (58 %) with initially poor or intermediate clinical state improving in clinical condition and six patients (32 %) even making an excellent recovery. Therefore, intensive therapy in our opinion might be justified for these patients. Da Costa et al. [3] reported similar results with 66 % of patients improving in clinical state at the last follow-up. An overview of some series and their results is presented in Table 6.

Table 6 Overview of previous studies on treatment modalities and outcome
Full size table

Despite successful treatment results in most cases, utmost attention has to be drawn to those patients, suffering from severe treatment complications. Major treatment related complications causing permanent and severe neurological disability or death occurred in ten patients (24 %); among these, six patients initially presenting in a good neurological condition. Although the rate of treatment complications in literature is comparable to our results, ranging between 17 and 20 % [2, 12, 26, 22], this issue must be stressed in particular, since data of the ARUBA trial have been recently published [17].

These circumstances underline previous assumption, that “in view of the rarity of these lesions to obtain the optimal outcome, these patients should be treated in a specialized neurovascular centre” [26].

Conclusion

Posterior fossa AVMs are complex and rare lesions, requiring thorough treatment strategies by an experienced interdisciplinary neurovascular treatment board. A multimodal treatment sequence with endovascular embolization, surgical resection, radiotherapy and conservative treatment in selected cases nowadays represents the gold standard for these lesions. The initial neurological condition seems to be crucial in terms of clinical outcome and patients are at high risk for morbidity and mortality, caused by both the impact of haemorrhage and treatment. Therefore, treatment indication in particular for those patients without bleeding remains difficult and needs further to be clarified.