Endovascular Treatment of Cerebellar Arteriovenous Malformations: A Single-Center Experience of 75 Consecutive Patients
Keywords: Aneurysm, arteriovenous malformation, cerebellum, embolizationKey Message: The high rate of associated cerebral aneurysms could explain the tendency of initial hemorrhage in cerebellar AVMs and targeted embolization of coexisting cerebral aneurysms should be the first priority.
Cerebellar arteriovenous malformations (AVMs) are congenital blood vessel disorders located of the posterior fossa with an incidence of less than 15%., Rupture of the posterior fossa AVMs can result in more severe clinical manifestations than those of the supratentorial counterparts with equivalent size of hematoma. Cerebellar AVMs present with more aggressive hemorrhage. For untreated brain AVMs, the overall annual rupture rate has been reported to be 2.10-4.12%, according to a recent review of the natural history of brain AVMs.,
The concentration of highly eloquent neurological structures and aggressive natural history of this type of cerebellar AVMs complicates the management of such neurovascular lesions. However, the clinical and angiographic description of cerebellar AVMs is limited and is often conducted with other posterior fossa AVMs, especially with brainstem AVMs. In addition, compared with other infratentorial AVMs, cerebellar AVMs are more likely to be associated with cerebral aneurysms. To our knowledge, the reported overall incidence of a coexisting aneurysm with cerebellar AVMs ranges between 2.7 and 24.5%.,,, In addition, coexisting cerebral aneurysms are regarded as independent risk factors for the subsequent hemorrhage after endovascular or radiosurgical treatment.
Limited information regarding endovascular treatment outcomes of cerebellar AVMs is available in the literature, with most of the studies focusing on reporting the treatment outcomes of cerebellar AVMs combined with other posterior fossa or infratentorial AVMs. However, cerebellar AVMs may be different with respect to treatment outcomes and pose a high risk of treatment-related complications, despite the rapid progress in microsurgery, radiosurgery, and endovascular embolization.
Since the treatment modalities and timing for cerebellar AVMs have not been described clearly in the literature compared with their supratentorial counterparts, we performed this study to determine the safety and effectiveness of endovascular treatment for cerebellar AVMs in our department and to examine the potential factors for predicting poor outcomes.
This study was approved by the Institutional Review Board of Beijing Tiantan Hospital Affiliated to the Capital Medical University. We retrospectively analyzed 75 patients with cerebellar AVMs who received endovascular treatment and were hospitalized in our department between January 2006 and January 2016.
Demographic, angiographic, and treatment data of the patients were retrospectively reviewed. The occurrence of intracranial hemorrhage was defined by the radiologic criteria on the basis of computed tomographic (CT) findings. All cerebellar AVMs were evaluated by preprocedural digital subtraction angiography (DSA).
Patient demographics included age at presentation, sex, and initial presentation. Angiographic features included AVM location, size, angioarchitecture, presence of associated aneurysms, and Spetzler–Martin grade. The associated aneurysms were divided into prenidal and intranidal. Prenidal aneurysms were identified as arising from a major feeding vessel if it was located on an artery directly supplying the AVMs as a feeding pedicle. Intranidal aneurysms were defined as those aneurysms located within the AVMs nidal.
Location of AVM lesion was categorized as eloquent (deep nuclei or cerebellar peduncles) and noneloquent (cerebellar hemispheric or vermis). AVM size was categorized as small (<3 cm) and large (>3 cm). Venous drainage was classified into those with exclusive deep venous drainage and those without exclusive deep venous drainage. Feeding arteries were classified as single and nonsingle patterns.
Considering that the high rate of hemorrhagic presentation and subsequent hemorrhage in patients with cerebellar AVMs may cause high morbidity and mortality in the narrow confines of the posterior fossa, we recommended treatment for almost all cerebellar AVMs; for patients associated with cerebral aneurysms, targeted embolization was recommended. The principle for targeted embolization was to eliminate the AVM-correlated aneurysms or fistula prior to gamma knife surgery (GKS) treatment. Targeted embolization provides a relative safe period for GKS to take effect.
For patients presenting with life-threatening hematoma in the posterior fossa, emergency hematoma evacuation was performed by the neurosurgeons in our hospital as per the availability of complete radiologic angiography. For patients without hematoma mass effect and those with only hematoma evacuation, endovascular treatment of the AVMs was deferred for approximately 4–12 weeks after hemorrhagic presentation. For patients with obstructive hydrocephalus, external ventricular drainage was performed before other treatments. Then, a ventriculoperitoneal shunting may be performed if the patients presented with persistent symptomatic hydrocephalus after endovascular treatment.
Although we recommended endovascular treatment as the first-line treatment for most cerebellar AVMs, the ultimate treatment modality and treatment timing were decided by patients' preference.
Procedures were carried out with the patients under general anesthesia. All procedures started with femoral artery sheath placement. Interoperation heparinization was achieved by adding 3000 IU of heparin to 500 mL of 0.9% wt/vol sodium chloride infusion solution which was administered through the guiding catheter during the procedure., Through a transfemoral route, a 6-F guiding catheter was positioned in the vertebral artery supplying the AVM. In general, the microcatheter (Marathon, ev3) was introduced into the feeding pedicle as near as possible to the nidus of the AVM. For the patients harboring associated prenidal aneurysms, coil embolization with or without stent was utilized. Moreover, for patients with intranidal aneurysms, the microcatheter tip was placed near the proximal part of the aneurysm for embolization of the aneurysm and the AVM nidus at the same time.
Superselective angiography via the microcatheter with a 1 mL syringe was performed to evaluate any normal branches, AVM architecture, flow volume, as well as to illustrate the anatomy of the proximal part of the draining vein.
Onyx-18 was preferred in most of the cases, whereas 20% Glubran 2 (GEM, Viareggio, Italy) was utilized in the other cases. Onyx-18 injections were administered under real-time road-mapping to allow easier visualization of the glue cast. Twenty percent Glubran 2 was injected under subtracted fluoroscopy. Once reflux occurred, the injection was terminated and the microcatheter was withdrawn. Control angiography was performed to evaluate the AVM nidus. If there was another feeding pedicle, the above procedure was repeated with a new microcatheter.
The patients were observed in an intensive care unit overnight and induced hypotension was used for 3 days to prevent normal perfusion pressure breakthrough. Multistaged embolizations were performed at intervals of 8–12 weeks. For patients with associated aneurysms, complete occlusion of the aneurysms was our first target.
Patients underwent DSA every 12 months during the first 3 years. When hemorrhage occurred after treatment, it was recorded. Complete AVM occlusion was defined as the disappearance of the nidus and absence of early venous drainage at the time of follow-up DSA. For patients with associated aneurysms, total occlusion also implied complete obliteration of the aneurysms. Near-total occlusion implied more than 90% reduction of the AVM volume. The following parameters were recorded – the number of embolization sessions, the number of feeding pedicles embolized, complete occlusion rate, and procedure-related complications.
Clinical outcome evaluation
The clinical follow-up period was defined as the period from endovascular treatment to the last follow-up. Postprocedure treatment modalities and clinical outcomes were described and analyzed. Postprocedure treatment outcome was evaluated using the following variables: occlusion rate, treatment-related complications, and modified Rankin Scale (mRS) scores at the last follow-up. The occlusion rate was based on postoperative DSA or CTA. The mRS scores at the last follow-up were dichotomized as good outcome (mRS <3) and poor outcome (mRS ≥3).
Univariate analysis (Chi-square test, Fisher's exact test, Student's t-test, or Wilcoxon's two-sample test) including gender, age, AVM location, size of AVMs, associated aneurysms, deep venous drainage, and Spetzler–Martin grade were used to assess the angioarchitectural characteristics of AVMs for patients presenting with or without good clinical outcome. Multivariate logistic regression analysis was performed to identify the independent risk factors associated with poor clinical outcome after the endovascular treatment. A P value of less than 0.05 for a 95% confidence interval (CI) was considered to be statistically significant. Statistical analysis was performed using SPSS version 19.0 (SPSS, Inc, Chicago, IL, USA).
Among the 75 cerebellar AVMs, there were 40 men (53.3%) and 35 women (46.7%). Initial hemorrhagic presentation occurred in 61 patients (81.3%), and 14 patients (18.7%) presented with dizziness, headache, or transmit neurological deficits. Seizure occurred in none of the patients. In this series, the average age at endovascular treatment was 37.7 ± 15.5 years (range, 4–66 years). Of all patients, 44 (58.7%) presented with mRS ≥3 and 31 (41.3%) presented with a mRS <3 prior to the endovascular treatment.
High Spetzler–Martin grade (IV or V) AVMs occurred in 5 patients in this series, whereas 70 patients presented with Spetzler–Martin grade (I–III). Eloquent AVMs were found in 19 (25.3%) patients. The AVM size was <3 cm in 49 (65.3%) patients and ≥3 cm in 26 (34.7%) patients. Deep venous drainage was noted in 25 (33.3%) patients, and superficial in 50 (66.7%). Feeding arteries were single in 43 (57.3%) of the patients and 2 or more in 32 (42.7%).
A total of 44 (58.7%) patients harboring 63 cerebral aneurysms were found in our series, which included 14 (22.2%) intranidal aneurysms and 49 (77.8%) prenidal aneurysms. A total of 14 patients presented with multiple cerebral aneurysms. Regarding the 49 prenidal aneurysms, 25 were located in the cerebellar posterior inferior artery, 16 in the superior cerebellar artery, and 11 in the cerebellar anterior inferior artery. The mean size of aneurysms was 4.13 ± 1.6 mm (range 1–8 mm). None of the patients experienced rebleeding prior to the endovascular treatment.
Embolization results of AVMs and associated aneurysms
We performed 97 embolization sessions in 75 patients, with 12 patients treated for two sessions and 10 patients for three sessions. A mean of 2.5 (range, 1–6) feeding pedicles were embolized per patient. One feeding pedicle was embolized in 56 patients, two feeding pedicles in 13 patients, three feeding pedicles in 3 patients, five feeding pedicles in 2 patients, six feeding pedicles in 1 patient.
Complete occlusion of intranidal and prenidal aneurysms were preferred in 44 patients with associated aneurysms [Figure 1] and [Figure 2]. All the patients with associated aneurysms were treated within 2 months after initial hemorrhage (32 within the first month and 12 in the second month). All intranidal aneurysms were occluded by endovascular treatment with liquid embolic agent (Onyx 12; glubran 2). Prenidal aneurysms were all treated by using coils without stent.
For the 44 patients with cerebral aneurysms, AVMs nidus was treated at the same time in 23 patients and 21 in multistage processes. The immediate angiography performed after the procedure showed total occlusion in 32/75 (42.7%) patients, 99–90% occlusion in 31/75 (41.3%) patients, and <90% occlusion in 12/75 (16.0%) patients.
Of 32 completely occluded patients, 12/32 were using with glubran 2 and 20/32 with Onyx. Additional radiosurgery was performed in 33 patients at 1–24 months following embolization, which included 15 patients with <90% occlusion, 10 patients with 99–90% occlusion and 8 patients with <90% occlusion.
Endovascular treatment-related complications occurred in 10 patients (13.3%), of which 7 patients suffered ischemic symptoms, 2 patients suffered hemorrhagic complications, and 1 patient suffered catheter entrapment.
Regarding 7 patients with ischemic symptoms, 5 patients suffered minor headache and dizziness after the embolization. Two patients suffered infarction of cerebellum, as revealed in the postprocedural MRI. All the symptoms were relieved after conservative treatment.
Catheter entrapment was encountered in 1 patient, and the microcather was reserved in the vessel in a young patient with AVM. Perforation of the feeding pedicle by the guidewire occurred in 2 patients, and vessel perforation was treated immediately by injection of Onyx. All these perforations occurred without clinical consequences.
Ten patients (10/75, 80%) presented or developed hydrocephalus. Most hydrocephalus (8/10; 89.9%) developed due to AVM hemorrhage before endovascular treatment of the AVM. In the other 2 patients, hydrocephalus developed during the postoperative period after embolization. External ventricular drainage to alleviate life-threatening hydrocephalus was performed in 3 patients, and in the other 7 radiologically mild hydrocephalus were left untreated. Overall, 2 patients required a permanent ventriculoperitoneal shunting. No complications occurred during the management of hydrocephalus.
Angiographic and clinical follow-up results
Overall, after a mean follow-up period of 73 months (48–168 months), two patients (2/75, 2.7%) with 99–90% occlusion died of severe intracranial hemorrhage during the follow-up period of 6 months and 15 months. DSA or CTA follow-up was available for 50 patients (66.7%), which showed total occlusion in 35 patients, 99–90% occlusion in 3 patients and <90% occlusion in 2 patients. Finally, good outcome (final mRS score <3) was achieved in 81.3% (61/75) in this series. The overall mortality rate was 2.7% (2/75).
Predictors of poor outcome after endovascular treatment of AVM
As shown in [Table 1], the clinical and angiographic characteristics of the patients with good outcome and those with poor outcome were compared at the final follow-up. [Table 2] demonstrates the multivariate logistic regression analyses for the predictors of poor neurologic outcome.
Univariable logistic analysis showed the patient age, Spetzler–Martin grade, preprocedure mRS, size of AVM, and eloquent location were significantly associated with the poor outcome after endovascular treatment of AVM (P < 0.1).
Patients with good neurologic outcome were more likely to be younger than those with poor outcome (35.9 ± 15.1 vs 45.3 ± 15.4, P = 0.041). As demonstrated in [Table 2], patients with poor outcome at the final follow-up were more likely to have a higher preprocedure mRS score (mRS score ≥3), however, the difference did not reach statistical significance (P = 0.094). In addition, eloquent AVM are more likely to present with poor outcome (P = 0.0001). Patients with poor outcome seemed to be more likely to harbor a larger AVM (AVM size ≥3 cm) in comparison with those with good outcome (P = 0.006). Patients with good outcome and those with poor outcome did not differ in gender, AVM characteristics (patterns of arterial supply and venous drainage), and associated aneurysms.
Multivariate logistic regression analysis showed that patient age (OR, 1.097; 95% CI, 1.020–1.180), AVM size (OR, 8.783; 95% CI, 1.322–18.332), and eloquent location (OR, 10.772; 95% CI, 1.852-22.659) were significantly independent predictors of poor outcome. After adjusting for other factors, multivariate logistic regression analysis showed patient age (OR, 1.086; 95% CI, 1.098–1.182), AVM size (OR, 9.072; 95% CI, 1.164–20.703), and eloquent location (OR, 9.209; 95% CI, 1.557-35.481) were still the significantly independent predictors of poor outcome. Eloquent location was the strongest independent predictor for poor outcome, increasing the risk of poor outcome more than 9.2 fold, followed by size of AVM (approximately 9.1 fold) and patient age (1.1 fold).
Compared with the supratentorial counterparts of AVMs, cerebellar AVMs are complex neurovascular lesions and pose an increased risk of bleeding as well as increased morbidity due to their presence in the narrow confines of the posterior fossa proximity to many eloquent structures. Unlike other brain AVMs, cerebellar AVMs are more likely to present with hemorrhage. In our series, 61 patients (81.3%) presented with a bleeding event and 44 patients (58.7%) presented with cerebral aneurysms. Compared with nonhemorrhagic patients, associated cerebral aneurysms were more commonly detected in hemorrhagic patients (41/61, 67.2% vs 3/14, 26.2%). Our results are compatible with previous reports on cerebellar AVMs, which revealed a rate of bleeding presentation of 72–92% in cerebellar AVMs with an incidence of a coexisting aneurysm ranging 2.7–24.5%.,, In 2014, Bowden et al. reported an associated aneurysm in 25% (16/63) and hemorrhage in 67% (43/63) of cerebellar AVMs patients. In 2015, Robert et al. reported a cohort of 59 patients with cerebellar AVMs, with a hemorrhagic presentation for 81.4% (48/59) and an associated aneurysm in 45% (25/59) of the patients. In 2016, Zhu et al. analyzed the angioarchitecture of 142 cerebellar AVMs, and found that 115 patients (115/145, 81.0%) presented with hemorrhage and 42 (42/113, 29.6%) with associated prenidal or intranidal aneurysms.
The specific mechanism underlying the high rate of associated aneurysms remains unclear and ambiguous. A previous study showed that hemodynamic may play an important role in the development and progression of aneurysms in patients with AVMs. Coexisting aneurysms are usually flow-related and increasing blood flow can result into high wall shear stress, and hence, distention and degeneration of arterial walls., Radiographic studies found that wall shear stress is directly proportional to flow velocity, and cerebral blood flow was faster in the cerebellum than in other regions, which could induce a heavier vascular damage.,
Although complete surgical removal provides the benefit of early hemorrhage protection, it is associated with a higher risk profile., In 2009, the Toronto study group showed that the diagnosis of coexisting aneurysms is a significant risk factor for hemorrhage of cerebellar AVMs. Similarly, in 2012, Kano et al. also found that aneurysms management may reduce the risk of re-hemorrhage during the follow-up period.
To our knowledge, endovascular treatment is rarely curative on its own as a primary modality for AVMs, combined with the SRS, having been shown to be effective in 75.2% of cases. Recently, several studies have reported their experience of managing cerebellar AVMs with radiosurgery, and found SRS could result in successful nidus obliteration and favorable outcome in most patients with cerebellar AVMs., However, in our department, we proposed a targeted embolization principle for cerebral AVMs, which was aiming for AVM-related aneurysms (intranidus and hemodynamic) and making any residual nidus more convenient for SRS. Therefore, the recommended therapeutic strategy is to obliterate the associated aneurysm first or simultaneously because of the lower annual bleeding rate of an AVM compared with an aneurysm and because the morbidity and mortality of an aneurysm rupture is higher than that of an AVM. Hence, all the associated cerebral aneurysms in our series were given the first priority and all were successfully embolized.
During the long-time follow-up period, only 2 patients with partial occlusion developed hemorrhage after the endovascular procedure and died. In our series, good outcome was achieved in 81.3% of the patients during the follow-up period. After adjusting for other factors, multivariate logistic regression analysis showed that increasing patient age (OR, 1.086; 95% CI, 1.098–1.182), the size of AVM (≥3 cm) (OR, 9.072; 95% CI, 1.164–20.703), and eloquent location (OR, 9.209; 95% CI, 1.557–35.481) were the significantly independent predictors for poor outcome.
In our series, we found that eloquent AVM location and AVM size (≥3 cm) are two significant predictors for poor outcome after AVM embolization. However, Spetzler–Martin grade was not a significant independent predictor for poor outcome. Because only 5 patients presented with Spetzler–Martin grade IV–V, such limited cases may result in the negative conclusion.
According to the Spetzler–Martin grading system, eloquent cerebellar AVMs were those involving the cerebellar peduncles or deep cerebellar nuclei. Despite eloquent location predicting poor neurologic outcome, good neurologic outcome was still achieved in 40% of the patients with eloquent cerebellar AVMs. Hence, for patients with eloquent location and AMV size of ≥3 cm, careful treatment planning and more detailed analysis of angioarchitecture of the AVMs are needed. Further, high pre-procedural mRS score (an mRS score >3) was not significantly associated with poor functional outcome in our series. In the study reported by da Coast et al., poor clinical outcome was associated with poor initial mRS (P < 0.0001). In another study on posterior fossa AVMs reported by Yang et al., the authors found that lower pretreatment mRS score was positively associated with good functional outcome. Similarly, a recent study reported by Nisson PL et al. revealed that the pre-surgical neurological status was the risk factor for poor outcome.
These findings suggest that a poor initial mRS before treatment predicts a poor neurologic outcome. However, in our series, good outcomes could still be achieved in 75% of the patients with poor pre-procedure mRS scores. We still recommend early endovascular treatment for patients with poor initial mRS. In addition, our results suggest that increasing age predicts a poor neurologic outcome after cerebellar AVM treatment. Because cerebellar AVMs are congenital vascular malformation, increasing age may provide more opportunities of subsequent hemorrhage or coexisting with other systematic disease. From this standpoint, we recommend early treatment of cerebellar AVMs after initial diagnosis.
There were two main limitations of our study. First, our study was a retrospective analysis at a single center, and therefore, subject to the inherent biases of this study design. Second, longer follow-up and multicenter experience should be collected to draw any definitive conclusion.
Endovascular treatment of cerebellar AVMs is safe and feasible. The high rate of associated cerebral aneurysms could explain the tendency to initial hemorrhage in cerebellar AVMs, and targeted embolization of coexisting cerebral aneurysms should be the first priority. Increasing patient age, eloquent AVM location, and the size of the AVMs are independent predictors of poor outcomes after endovascular treatment of cerebellar AVMs.
Financial support and sponsorship
This work was supported by the the National Natural Science Foundation of China, grant number (81901197), “the fifth 333” high-level talent training project of Jiangsu Province, “the fifth 521” high-level talent training project of Lianyungang City, and the scientific research project of Lianyungang Municipal Health Commission in 2018.
Conflicts of interest
There are no conflicts of interest.
[Figure 1], [Figure 2]
[Table 1], [Table 2]