Spinal dural arteriovenous fistulae (sdAVFs) are rare (5-10 per million per year), often misdiagnosed vascular lesions most commonly found in the thoracolumbar spine [1-7]. These lesions commonly present with progressive myelopathy in male (approximately 80%) patients who are usually in their 7^th decade of life [3-6,8]. The underlying pathophysiology is presumed to arise from venous congestion resulting from direct arterial connections from radicular arteries and radicular veins inside the dura [3,9,10]. In fact, Takai et al. used intraoperative indocyanine green video angiography and histology in a series of 7 patients to demonstrate that sdAVFs arise from multiple meningeal vessels draining intradurally into a single intradural vein that becomes modified to handle arterial inflow [10]. The purpose of this paper is to review the literature regarding presentation, pathophysiology, workup, imaging characteristics, management, challenges, and future directions associated with spinal dural arteriovenous fistulae.

Presentation, Diagnostics, Imaging, Associated Challenges

Delayed diagnosis is a known problem with sdAVFs, with most studies reporting a median time to diagnosis from symptom onset of approximately 15-18 months or longer [1,2,5,6,8]. Patients are often misdiagnosed with degenerative disc disease, spinal tumors, and peripheral nerve pathologies, with up to 22% undergoing unnecessary laminectomy or other procedures, with worsening of symptoms [3,8]. One study of 153 patient with sdAVFs by Muralidharan et al. found that severe neurogenic bladder was associated with delayed diagnosis [8].

Appropriate imaging is essential for diagnosis. T2-weighted MRI is very sensitive, with conus hyperintensity noted in as many as 95% of cases, secondary to orthostasis [6,8,11]. Flow voids and mass effect are less common [6]. However, few patients may be without T2 signal changes [6,8]. Furthermore, extent of T2 signal does not appear to correlate with severity of disability [8]. In addition, medullary signal changes do not appear to correlate with functional status following treatment [12]. One case by Kim et al. demonstrates the natural radiographic progression of sdAVFs, including ascending and worsening spinal cord edema with increasing enhancement suggestive of infarction [7]. Following MRI, contrast-enhanced MR angiography can often be used to accurately detect and localize sdAVFs, reducing the radiation dose and volume of contrast needed in subsequent confirmatory angiography [11,13-15]. In the majority of sdAVF cases, diagnosis is confirmed by diagnostic spinal angiography (DSA) [3,8,16]. Meanwhile, intra-arterial cone beam CT angiography has demonstrated significant diagnostic value, even relative to DSA, and ability to delineate local anatomy and exact shunt location [17-19].

Cao et al. used results from a series of 71 patients to derive a 4 item score (AVFS) that predicts angiography confirmed sdAVFs [20]. Age ≥ 50 years, length of intramedullary lesion ≥ 5 levels, dilated perimedullary vessels, and subcervical location were the 4 predictors [20]. An AVFS threshold score of 3 had a sensitivity of 85% and specificity of 95% for predicting sdAVFs [20].

It is important to distinguish sdAVFs from spinal epidural arteriovenous fistulas (seAVFs). One study by Kiyosue et al. showed that 49% of seAVFs were misdiagnosed as sdAVFs [21]. Both lesions often present with progressive myelopathy in middle-aged men. They also present similarly on MRA and angiography, with similar T2 hyperintensity involving the conus, but seAVFs feature an epidural venous pouch at the fistula level and may sometimes have multiple arterial feeders [16,22]. However, location and shunting anatomy may help with diagnosis, as sdAVFs are more often found in the thoracolumbar spine and shunt medial to the medial interpedicle line into a bridging vein, while seAVFs are more often found in the lumbar and sacral spine, and more commonly shunt ventrally into a perimedullary vein [21,22]. Furthermore, Kiyosue et al. found seAVFs to be more associated with a history of spinal surgery/injury [21]. A summary of differentiating features between the two lesions is presented in Table 1.

Treatment Options and Outcomes

Treatment options for sdAVFs include surgical disconnection, endovascular embolization, or a combination of both. As mentioned above, the fistula is typically located in the dural layer, near the dural sleeve of the nerve root. The goal of surgical treatment is intradural disconnection of the draining vein with coagulation of the dural fistula. Due to advances in spinal angiography, endovascular embolization has been gaining traction as a treatment option. In cases in which there is recurrence after an initial attempt at endovascular embolization, a combinatorial approach may be used.

The typical surgical method for a spinal dural AVF involves a posterior approach with a laminectomy, hemi-laminectomy, or a laminoplasty. After a dural opening is performed along the appropriate nerve root, the arterialized medullary vein is identified and coagulated with bipolar forceps, before being disconnected. Typically, there is no stripping of the enlarged pial veins, and intra-operative monitoring can be used to ensure the fistula has been eliminated [23]. In patients with both intradural and extradural drainage, disconnection of both the intradural and extradural venous drainage or complete excision of the fistula is indicated [24]. Complications of surgical intervention include cerebrospinal fluid leakage. Rates of complications don’t appear to vary between surgery and endovascular embolization [25].

Endovascular embolization is another commonly used approach. Polyvinyl alcohol is no longer commonly used as an embolic agent due to high fistula recurrence rates. The liquid substance N-butyl cyanoacrylate (NBCA) is commonly used as an embolic agent. Over the past few years, the embolic agent of choice has shifted to Onyx. If the initial embolization does not work, patients may require a repeat embolization or open surgical intervention. Recurrence can occur if the embolized vessels become recanalized or if collateralization develops. In one 2014 study, one patient treated with glue embolization experienced recanalization of the fistula, as the glue settled in the feeding artery and did not reach the fistulous connection or the draining vein. Another patient in the same study also had recanalization of the fistula when he experienced arterial spasm during Onyx embolization, which prevented the Onyx from reaching the vein [26]. However, of note, in a recent systematic review and meta-analysis, Onyx embolization had higher odds of failure and recurrence compared to NBCA [27].

Initial occlusion rates are often significantly better with surgery compared to endovascular embolization, leading to less recurrence [25,27-29]. Furthermore, in rare DSA-occult cases, surgical management is the preferred option [30]. Nonetheless, surgical and endovascular management of sdAVFs significantly improves symptoms, even for those with severe preoperative disability, as assessed by the Aminoff and Logue Scale (ALS) [3,12,28,31,32]. However, there is concern that delayed diagnosis may lead to less symptomatic improvement [2].

Several studies have tried elucidating the pathophysiology and molecular pathways that may be involved in the pathogenesis of spinal dural arteriovenous fistulas (sdAVFs), although there is no clear consensus on a specific mechanism. However, two case reports suggest that mutations in CCM genes (cerebral cavernous malformation genes) can also lead to pathogenesis of spinal cavernous malformations at large. A 33 yo Korean male developed spinal cavernous malformations in cervical and thoracic spinal cords with a heterozygous nonsense mutation in the PDCD10/CCM3 gene [33]. The patient’s daughter also had an identical mutation that was initially asymptomatic, but presented with seizures three months later, suggesting a familial inheritance pattern [33]. CCM3 mutations have been noted to be uncommon in cerebral cavernous malformations, suggesting that this mutation may help in differentiating SVADFs from CCMs [34]. Another case report showed a heterozygous frameshift mutation in the KRIT1/CCM1 gene, leading to cavernous malformations in the cervical, thoracic, and lumbar spinal cord [35]. While these reports did not elucidate a particular molecular pathway, CCM studies have suggested several possible pathways. The general principle of developing CCMs rely on the Knudsonian two-hit mechanism, which suggest the loss of one allele due to a germline mutation in all cells, followed by a second somatic mutation in certain cells, initiating CCM lesions [36,37]. This may be applicable for sdAVFs as well. Regarding the CCM family at large, mutations in one of the three CCM genes can lead to loss of function of certain proteins, leading to endothelial barrier dysfunction and hyper-permeable blood vessels, causing leakage and bleeding [38]. KRIT1/CCM1, CCM2, and PDCD10/CCM3 are all involved in several individual signaling pathways, but can also associate with each other, forming trimeric complexes that converge several signaling pathways [38]. CCM2 in this complex is known to inhibit MEKK3 in normal physiology; the inhibition or downregulation of CCM2 can lead to hyper angiogenesis. Thus, mutations in CCM1 and CCM3 may disrupt this complex, leading to the MEKK3 pathway. Another possible pathway is that KRIT1 and PDCD10 both affect Notch signaling. Thus, when PDCD10 is downregulated, this leads to the upregulation of BMP6 and ERK1/2 transcriptional activity. The ultimate result is hyper angiogenesis, once again [39-41]. CCM genes are also implicated in several other signaling pathways, but must be further elucidated to study the specific pathogenesis implications in sdAVF, as opposed to relying on molecular pathways seen in cerebral cavernous malformations.

Furthermore, to study protein changes in sdAVFs, one study looked at histology and pathologic changes under sustained high vascular pressure (abbreviated as sdAVF-DV or sdAVF-draining vein) [42]. The study compared sdAVF-DV to superior temporal artery and vein, utilizing immunohistochemistry as well as Masson and H&E staining to look at extracellular matrices, cell distribution, inflammation, and more. Overall, 95 proteins were upregulated and 303 were downregulated in SDAVF-DV, compared to controls [42]. Overall conclusions include smooth muscle cell dysfunction and inflammation. Decreased smooth muscle contractile fibers, induced by long-range venous hypertension stretching in sdAVF-DV, lead to smooth muscle dysfunction. Histological examination showed an artery-like structure that develops in sdAVF-DV, with lower piezo-1 expression [42]. Piezo-1 is expressed in embryonic endothelial cells, activated by fluid shear stress. Thus, the loss of piezo-1 leads to reduced ability for endothelial cells to alter their alignment when subjected to shear stress [43]. In addition, collagen and elastin distributions were also altered, with decreased collagen 4 and 6, while CD45+ cells and COX-1 were increased [42]. When looking at the formation of sdAVFs, inner vessel wall inflammation might lead to insufficient extracellular matrix and trigger changes in pathological proteins [42]. Further analysis is needed to elucidate mechanisms of sdAVF formation, as opposed to histological changes seen with sdAVF progression, although studying these molecular pathways can shed light on the sdAVF formation as well.

Lastly, as discussed above, one of the greatest challenges in treating sdAVF is timely diagnosis [44]. However, no specific predictive biomarkers have been identified thus far. Current screening strategies for sdAVFs rely on MRI imaging with or without enhancement [45]. One study analyzed cerebrospinal fluid (CSF) samples, which pinpointed overexpression of APOB, C4BPA, and C1QA. APOB and C4BPA are involved in acute inflammatory responses and complement activation, suggesting that activation of these pathways might be yet another additional molecular mechanism for venous hypertensive myelopathy [44]. In conclusion, while several molecular mechanisms and possible biomarker targets have been suggested, further study is needed to elucidate pathways in sdAVF formation and possible biomarkers.

A summary of implicated molecules and pathways in the pathophysiology of sdAVF is presented in Table 2.

The above paper describes presentation, imaging characteristics, treatment paradigms, and pathogenesis of spinal dural arteriovenous fistulae. Limitations of this study include nonsystematic collection of all relevant studies and possible selection bias due to the paper not being a systematic review. In conclusion, remember to consider the diagnosis of sdAVF in middle aged men who present with gait changes and ascending motor and sensory deficits.

sdAVF: spinal Dural Arteriovenous Fistulae; sdAVF-DV: spinal dural Arteriovenous Fistula-Draining Vein; seAVF: spinal epidural Arteriovenous Fistulae; DSA: Diagnostic Spinal Angiography; ALS: Aminoff and Logue Scale; CCM: Cerebral Cavernous Malformation; CSF: Cerebrospinal Fluid