Cranio-cervical junction arteriovenous fistulas (CCJ-AVFs) are uncommon vascular shunts that occur in the complex vasculature of the craniocervical region[1, 2]. A recent multicenter cohort study in Japan proposed a classification system comprising five types of CCJ-AVF[6, 7], however, the categories was too complex for clinical application [8–13]. Therefore, further investigation involving independent cohorts and previous cases is necessary to determine the diagnosis and clinical presentation of CCJ-AVF.
Clinical Presentation
Subarachnoid hemorrhage and myelopathy are the primary presentations of untreated CCJ-AVFs and can significantly contribute to morbidity[14]. However, the natural history of CCJ-AVF remains poorly described[1]. Our present study suggests that the risk of neurological deterioration in untreated CCJ-AVF is 15–20%. Aneurysms are observed in 35% of cases and are an independent risk factor for subarachnoid hemorrhage. While unlike the intracranial saccular aneurysmal SAH, the SAH of CCJ-AVF was usually in a benign course (12% with mRS > 2 in our study). In contrast, myelopathy is associated with a higher ratio of neurological disability (30%) in untreated CCJ-AVF. The present study suggests that descending spinal venous drainage is a risk factor for myelopathy.
Angioarchitecture characteristics
The angioarchitecture of CCJ-AVF is typically complex. The dura mater surrounding the foramen magnum is supplied by the anterior and posterior meningeal branches of the vertebral artery, as well as the meningeal branches of the anterior and posterior spinal arteries (APA and PSA), the dorsal meningeal branch of the meningohypophyseal trunk arising from the intracavernous segment of the internal carotid artery (ICA), and the posterior inferior cerebellar artery (PICA). This complexity may be attributed to anatomical variations in the distal branches of the vertebral artery and the variable development of embryonic metameric vessels between the level of the hypoglossal artery and proatlantal artery[15, 16]. The origin of feeding arteries varied as was shown in our study, with no feeders of majority (10–35% of each category) but over 50% of multiple feeders, and 15–30% of bilateral feeders or dual-feeding by carotid artery and vertebral artery branches. Thus, a six-vessel angiogram with or without superselective angiography may be recommended to evaluate the entire angioarchitecture of CCJ-AVF. In addition, standard terminology would be helpful for comparing across different studies, given the frequent variations in vertebral artery branches at the craniocervical region. The V3 segment, which is most closely related to the foramen magnum, extends from the foramen in the transverse process of the atlas to the site of passage through the dura mater.
A recent angioarchitecural classification of these AVFs was developed primarily based on a presumed shunt location and its association with nerve rootlets, dura and medulla-spinal structure, and eventually proposed the ‘radiculomeningeal AVF’ aimed to overcome the difficulties in taxonomy of feeding arteries[5]. However, during the development of this classification based on multicenter data, disagreements among investigators still occurred in most cases in determining the specific AV shunts location because of the complicated arrangements of vessels, the limited number of cases with superselective angiography, and the lack of information about the neural structures on angiogram[5]. These challenges may hinder the clinical application of this classification and suggest the impracticality of taxonomy based on feeding artery and shunt location in real-world practice.
According to Hiramatsu et al [5], the draining veins of craniovertebral junction arteriovenous fistulas (CCJ-AVFs) can be classified into three types: predominantly subdural venous drainage, predominantly epidural venous drainage, and a combination of both sub- and epidural venous drainage. Although CCJ-AVF lesions can involve tangled vessels, multiple feeders, aneurysmal dilation, and venous varices, they lack an arteriovenous malformation-like nidus and are classified as arteriovenous fistulas.
Angioarchitecture associated with presentation and neurological status
Previous studies have shown that most patients with thoracolumbar arteriovenous fistulas (AVFs) commonly present with slowly progressive myelopathy, and descending venous drainage is associated with congestive myelopathy[17]. Another retrospective study has demonstrated that feeding via vertebral artery dural branches and drainage through perimedullary veins are associated with venous hypertension myelopathy[17, 18]. In our study, we have also identified anterior or posterior spinal vein/perimedullary vein as an independent risk factor for myelopathy. Although the natural history of myelopathy remains unclear, our findings suggest that the obliteration of the AVF led to improvements in MRI findings in 75% of patients, while 44% remained dependent on myelopathy with a median follow-up of 27 months. Furthermore, more than one-quarter of patients had age-related delayed neurological decline beyond the 1-year follow-up. Although our study did not involve treatment modality and outcomes, myelopathy may still be recommended as an indication for treatment of CCJ-AVF in patients with unfavorable neurological status.
Some authors have mentioned that ascending drainage route into an intracranial vein and venous varices were considered the causes of SAH[19]. ecent Japan multicenter cohort studies have identified the inclusion of the anterior spinal artery (ASA) as the feeder and aneurysmal dilatations as the risk factors associated with hemorrhagic presentation[5]. Our study results support those of Hiramatsu et al., indicating that the presence of an associated aneurysm, rather than an intracranial vein or venous varices, appears to be the major factor in CCJ-AVF hemorrhage. Therefore, in cases of CCJ-AVF with associated aneurysm, microsurgery or neurovascular treatment may be recommended to reduce the risk of future hemorrhage and improve treatment outcomes. Radiological features may be useful in identifying patients with a high risk of hemorrhage, thereby facilitating clinical decision-making in the management of CCJ-AVF.
Study limitation
This study was limited by the sample size of the consecutive cohort. Although the pooled data were added, the potential selection bias might contribute to an overestimated rate of favorable outcome. In contrast, the analysis of clinical presentation might be less affected. Secondly, the present angioarchitectural features were not quantitative to evaluate the hemodynamics of the fistula, future studies with quantitative angiographic analysis might be helpful to identify more features that could help in case assessment.