Discussion

The two cases reported above are strikingly similar in their clinical presentation, lesional anatomy and morphology, and treatment technique/result. Embryologically,1 ,2 the craniocervical junction represents a transition point from a spinal intersegmental arterial organization to a carotid-based system. In the first few weeks of development, each of the 31 somites is associated with one pair of intersegmental arteries from the dorsal aorta. While the temporary connections between the first six cervical segments and the dorsal aorta go on to regress, the vertebral arteries originate as the C7 intersegmental branches. With regression of the other cervical intersegmental branches, the longitudinal axis of the definitive vertebral arteries within the foramen transversarium results from the coalescence of the first six cervical intersegmental arteries.

In contrast, at the craniocervical junction, development is instead organized along the caroticovertebral anastomoses—namely, the trigeminal, otic, hypoglossal, and proatlantal intersegmental arteries. These temporary connections between the carotid arteries and the plexiform longitudinal neural arteries go on to regress, and the longitudinal neural arteries unite to form the definitive basilar artery.

Thus, for arteriovenous lesions affecting the upper vertebral artery, distinct embryologic pathways of coalescence and regression between the cervical intersegmental arteries and the longitudinal neural arteries may be related to developmental anatomic variants, as has been emphasized in previous reports.3 However, the mid and lower cervical longitudinal axis of the vertebral artery represents purely intersegmental development, and is thus thought less likely to be a congenital lesion originating from anatomic variants. One potentially relevant anatomic variant involving the lower vertebral artery is that of intraforaminal entrance at C44 rather than C6; in these cases, the ascending cervical artery (which develops from the C4 segmental artery) is thought to supply the lower ‘vertebral’ artery, rather than the C7 intersegmental artery. However, neither of our patients manifested this variant.

The similarity between the two cases presented here is marked: in both, school-age children manifested a clearly new and non-subtle symptom—a machine-like holosystolic murmur. The likelihood is thus very high that the arteriovenous lesion was either entirely of recent origin, or else lay dormant from birth with undetectably low flow, and developed a dramatic flow increase soon before diagnosis. As the AVF was located in both cases precisely at the mechanical pivot point where the vertebral artery entered the foramen transversarium suggests that the cause may be related to shear or torsion on the artery at the site. In the second case, given the distant history of ventricular septal defect repair, the possibility of an iatrogenic central venous catheter-related fistula could also be considered.5

The distinction between upper and lower vertebral arterial development underlies the most well-known classification of vertebrovertebral AVF (VVAVF), due to Rodesch and Lasjaunias.6 However, they characterized lower vertebral AVF as typically low flow, with large paraspinal collecting venous pouches; this is clearly a different entity from the two cases described here, both of which showed high flow. Other authors3 have proposed that traumatic lesions tend to affect the lower vertebral artery, whereas spontaneous lesions tend to characterize the upper vertebral AVF, though this distinction has not been clearly proved. Overall, no left- or right-sided predominance has been clearly demonstrated in the current literature; however, individual reports have occasionally suggested otherwise.

In addition to the above distinction between upper and lower vertebral artery AVF, venous drainage is an important factor. The term VVAVF has been broadly used to characterize extradural AVFs in which the arteriovenous shunt occurs between the vertebral artery and any adjacent venous structure, including jugular,7 suboccipital,8 paraspinal,9 and epidural veins.3 When epidural venous drainage is present, more often than not, it occurs indirectly via retrograde drainage from the primary paraspinal venous drainage of the fistula. True single-hole vertebral-epidural spinal AVFs, in which the vertebral artery fistulizes directly with the epidural venous plexus,10 are extremely rare. Normal connections between the intradural and extradural spinal venous systems may be characterized by sharp angulations that naturally prevent retrograde extradural-to-intradural venous drainage11 or harbor an otherwise poorly understood anti-reflux mechanism;12 however, in the setting of an AVF, retrograde intradural drainage into the perimedullary or intramedullary spinal veins may occur and must be identified or excluded as part of the diagnostic evaluation.

Considering all VVAVF, the most common presenting sign of a vertebral-epidural spinal AVF is an objective bruit,13 with other recognized clinical presentations including radiculopathy due to local mass effect from epidural venous compression,14 myelopathy related to retrograde intradural venous drainage/congestion,15 subarachnoid or epidural hemorrhage,16 high-output congestive heart failure,9 and vertebrobasilar insufficiency secondary to arterial steal.17 The natural history of direct lower vertebral-to-epidural single-hole AVF is unknown, though given the presence of very high flow into a capacious venous collector, the eventual development of high-output cardiac failure would not be unexpected. In our cases, treatment was offered because of the new-onset murmur, the presumed theoretical risk of lesion and clinical progression, and the relatively low expected risk of curative intervention.

In adults, VVAVFs most commonly result from spontaneous or iatrogenic trauma, or are associated with connective tissue diseases such as neurofibromatosis15 or fibromuscular dysplasia.18 In children, VVAVFs have been reported to present either as a symptomatic ‘high-flow’ congenital lesion in the very young19 or as an asymptomatic ‘low-flow’ spontaneous lesion in the older pediatric age groups.13 ,20 As mentioned above, our two cases do not fit this previously described pattern of lower vertebral AVF. Careful review of prior publications discloses similar cases of high-flow lower vertebral-to-epidural single-hole AVF, though not previously identified as a unique subtype. For example, although Goyal et al10 described the standard categorization into upper vertebral high-flow versus lower vertebral low-flow AVF, figure 2 in their manuscript depicts a case in a 3-year-old patient morphologically identical to our two cases, presenting with new-onset bruit, treated similarly by them with retrograde venous-to-arterial coil embolization.

Surgical treatments for VVAVFs include radical fistula resection/trapping, clip ligation of the fistulous point, vertebral artery sacrifice, and surgical decompression for spinal canal hematoma or compressive venous ectasia.14 ,15 ,21 ,22 Given the morbidity and technical demands of surgery, including the potential for significant blood loss and the need for wide operative exposure of the upper or lower cervical regions, and in light of continued improvements in endovascular technology, embolization has emerged as a primary approach for treatment of these fistulas. Detachable latex balloons23–25 (not available in the USA) and coils7 ,8 ,17 ,19 have the longest track record for this purpose, although more recently, liquid embolic agents26 ,27 have also been used for spinal AVF embolization. However, unlike coils, which can be retrieved before detachment if the pattern of deployment appears unfavorable, liquid embolic agents are not retrievable after injection, and pose the potential risk of distal embolization through the epidural plexus, or proximal embolization into the vertebral artery. Moreover, close proximity of radiculomedullary arterial supply to the AVF feeder may preclude safe use of a liquid embolic agent, as illustrated in our first case. Additionally, the ability to monitor the pattern of coil deployment before detachment was advantageous, because it allowed us to perform control angiography and ensure that the coils were not prolapsing into the parent vertebral artery. Covered stenting could also have been considered. However, given the associated risks (dual antiplatelet therapy, in-stent stenosis/thrombosis, endoleak, bending/kinking of the stent with motion, etc), and because this treatment would occlude only the arterial inflow without truly occluding the arteriovenous shunt, we felt that direct embolization at the arteriovenous connection was preferable.

The POD is a fairly new hybrid coil designed to achieve occlusion in relatively large arteries, such as the carotid or vertebral artery. The initial loops of the device constitute an ‘anchoring’ segment designed to grip the vessel wall by virtue of a stiffer nitinol core and a helical shape; the contrasting deployment of these initial loops in a POD as opposed to a traditional detachable coil was vividly illustrated in our second case. After the initial anchoring loops are deployed, the device transitions to a softer ‘packing’ segment, designed to create a dense cross-sectional plug, as highlighted in a recent swine model feasibility study.28 In our second case, this device was successfully deployed at the most distal aspect of the proximal epidural venous outlet in order to form a scaffold for retrograde coiling and ensure no distal migration of the coils that were subsequently deployed to occlude the fistula. To our knowledge, this case represents the first reported human use of the POD.

Finally, while some consider the use of a liquid agent mandatory to achieve definitive closure of AVF, both cases discussed here were treated entirely with detachable coil-like devices. Patient 1 is an active athlete who has remained fully engaged in the 6 years since her treatment, so while shear stress may be involved in initial development of lower vertebral AVF, at least in this case, continued torsion of the neck and upper extremities did not result in recurrence after definitive AVF closure.