Abstract

Please see the Editorial Comment by Brent D. Weinberg discussing this article.
CSF–venous fistulas (CVFs), first described in 2014, are an important cause of spontaneous intracranial hypotension. CVFs can be challenging to detect on conventional anatomic imaging because, unlike other types of spinal CSF leak, they do not typically result in pooling of fluid in the epidural space, and imaging signs of CVF may be subtle. Specialized myelographic techniques have been developed to help with CVF identification, but these techniques are not yet widely disseminated. This article reviews the current understanding of CVFs, emphasizing correlations between venous anatomy and imaging findings as well as potential mechanisms for pathogenesis, and describes current imaging techniques used for CVF diagnosis and localization. These techniques are broadly classified into fluoroscopy-based methods, including digital subtraction myelography and dynamic myelography, and cross-sectional methods, including decubitus CT myelography and MR myelography with intrathecal injection of gadolinium. Knowledge of these various options, including their relative advantages and disadvantages, is critical in the care of patients with spontaneous intracranial hypotension. Investigation is ongoing, and continued advances in knowledge about CVFs as well as in optimal imaging detection are anticipated.

HIGHLIGHTS

CVFs are an important and increasingly recognized cause of spontaneous intracranial hypotension with which radiologists should be familiar.
Knowledge of spinal venous anatomy aids in interpretation of imaging of CVFs.
Specialized imaging tests have been developed to help detect CVFs; selecting an appropriate modality requires an understanding of the advantages and disadvantages of each.
CSF–venous fistulas (CVFs), first described in 2014 [1], are a cause of CSF volume depletion that results in spontaneous intracranial hypotension (SIH). Unlike previously recognized cases of SIH caused by dural tears, CVFs do not necessarily cause pooling of fluid in the epidural space, and thus the presence of a leak is more difficult to detect. In recent years, several specialized imaging techniques have been developed to address challenges with CVF detection, but familiarity and practical experience with these techniques are not yet widespread. Spinal venous anatomy may be unfamiliar to many radiologists. However, knowledge of this topic facilitates identification of CVFs.
The objectives of this article are to review the current state of knowledge about CVFs, including anatomic features and the role of various spinal imaging techniques in CVF detection, and to discuss potential mechanisms for the pathogenesis of CVFs.

General Features of CSF–Venous Fistulas

CVFs are abnormal connections between the spinal subarachnoid space and adjacent paraspinal veins that allow unregulated egress of CSF into the venous system. Resultant CSF depletion causes intracranial hypotension. Although this entity was first recognized very recently, it has rapidly emerged as a very common cause of underlying spinal CSF leak in SIH, particularly in the challenging population of patients with no leak seen on initial spinal imaging.
CSF flow in CVFs appears to be unidirectional; CSF flows into the venous system, but to our knowledge no reversal of flow allowing venous blood into the CSF has been reported. Physiologically, this is explained by the observation that CSF pressure is maintained at a level greater than venous pressure [2, 3].
CVFs are typically found in the thoracic spine. The lower thoracic levels between T7 and T12 are the most common locations, although CVFs are frequently encountered at upper thoracic levels as well [47]. Fistulas in the upper lumbar or lower cervical levels are less common but have also been reported [4, 5, 8]. Various case series have reported discrepant results as to whether the right or left side of the spine is more commonly involved, suggesting that if a predilection for laterality exists, it is not strong [4, 7, 8].
Unlike dural tears (the first described cause of intracranial hypotension), most CVFs do not result in pooling of CSF in the epidural space. Only rarely will there be dual pathology in which a CVF coexists with an epidural CSF leak [4, 9]. This lack of epidural fluid accumulation makes CVFs more difficult to detect on conventional anatomic imaging. Rather, a contrast agent specific for the CSF (i.e., a myelographic contrast agent) must be used to detect CVFs. Consequently, conventional spine MRI, a technique often used for detection of epidural CSF leaks, does not detect CVFs.
CVFs are often anatomically associated with a diverticulum of a nerve root sleeve, with the fistulous connection commonly originating from the diverticulum itself; one investigation found that 82% of CVFs arose from such a diverticulum [4]. The association of CVFs with these diverticula highlights the need for careful scrutiny of these structures when SIH is present but no epidural leak is seen. At the same time, one must avoid overinterpreting the significance of diverticula on any spinal imaging if no CVF is seen, because nerve root sleeve diverticula are normal findings that are also found in patients without CSF leak [10]. As an isolated finding, nerve root sleeve diverticula should not be considered an indicator of the diagnosis of SIH more broadly or CVF specifically.

Spinal Venous Anatomy and CSF–Venous Fistula Drainage

Knowledge of the normal vertebral venous anatomy is helpful in understanding the drainage patterns of CVFs and in interpreting imaging performed to detect CVFs.
The vertebral venous system is subdivided into three components: the internal vertebral venous plexus (IVVP), the external vertebral venous plexus (EVVP), and the basivertebral veins [11] (Fig. 1). The IVVP is an interconnected network of venous channels that lies within the epidural space of the spinal canal. It surrounds the thecal sac, running longitudinally from the cranium to the sacrum, and receives drainage from the intrinsic venous system of the spinal cord.
Fig. 1 —Illustration of normal vertebral venous anatomy. Internal epidural venous plexus communicates with external epidural venous plexus via intervertebral veins. (Illustration by Lydia Gregg, MA, CMI © 2021 JHU)
The EVVP lies outside of the spinal canal, surrounding the bony vertebral column anteriorly and the posterior spinal elements posteriorly. The EVVP communicates with the IVVP through veins that traverse the neural foramen, known as intervertebral veins. Muscular branches and lateral branches from the paraspinal muscles and intercostal spaces, respectively, also drain into the EVVP [12]. Segmental spinal veins are part of the EVVP and run alongside the midportion of vertebral bodies to anastomose anteriorly with the azygous and hemiazygos venous systems as well as with the longitudinal lumbar veins. Through these pathways, venous blood from the EVVP ultimately drains into the superior and inferior vena cava [13, 14].
The basivertebral veins run through the vertebral bodies themselves, connecting the portion of the IVVP in the anterior epidural space with the EVVP along the anterior surfaces of the vertebral body.
Although the fistulous connection arises from the nerve root sleeve, a CVF may drain into any of these venous components. Drainage into a segmental spinal vein is common, with an elon-gated vein visibly extending anteriorly around the midportion of a vertebral body (Fig. 2). However, drainage into the IVVP may manifest as a contrast-opacified network of veins in the epidural space, appearing either as discrete tubular vascular structures or as a single crescentic plexus, separated from the thecal sac by the epidural fat (Fig. 3). This is especially common in the ventro-lateral spinal canal where the veins adjacent to the nerve roots (where CVFs typically arise) communicate with the rich venous plexus posterior to the vertebral body in the ventral epidural space. Drainage into the IVVP and the adjacent intervertebral veins may take on the appearance of a fine reticular network of veins, as opposed to the single dominant tubular appearance that is characteristic of draining segmental veins. In some cases, venous drainage from CVFs may reflux into lateral or muscular branches of the EVVP located posterior to the neural foramen and surrounding the posterior spinal elements (Fig. 4). Filling of the basivertebral venous plexus within vertebral bodies can also be seen in conjunction with filling of the IVVP or intervertebral veins (Fig. 5).
Fig. 2A —CSF–venous fistula (CVF) drainage into segmental veins.
A, 48-year-old woman with CVF at T10 on right. Axial maximum-intensity-projection (MIP) image from CT myelography performed with patient in decubitus position shows drainage into segmental spinal vein (arrow).
Fig. 2B —CSF–venous fistula (CVF) drainage into segmental veins.
B, 53-year-old woman with CVF at T10 on right. Axial MIP image from CT myelography performed with patient in decubitus position shows drainage into segmental spinal vein (arrow) and eventually into azygos vein.
Fig. 3A —CSF–venous fistula (CVF) drainage into internal vertebral venous plexus (IVVP).
A, 79-year-old woman with CVF at T12 on right. Axial image from CT myelography performed with patient in decubitus position shows drainage into IVVP (arrow).
Fig. 3B —CSF–venous fistula (CVF) drainage into internal vertebral venous plexus (IVVP).
B, 53-year-old woman with CVF at T9 on left. Axial image from CT myelography shows drainage into IVVP (arrow). Contrast material is separated from thecal sac by thin band of epidural fat, indicating its intravascular rather than epidural location.
Fig. 3C —CSF–venous fistula (CVF) drainage into internal vertebral venous plexus (IVVP).
C, 52-year-old man with CVF at T10 on right. Axial image from CT myelography shows drainage into ventral IVVP (arrow).
Fig. 3D —CSF–venous fistula (CVF) drainage into internal vertebral venous plexus (IVVP).
D, 33-year-old man with CVF at T11 on right. Axial image from CT myelography shows drainage into ventral IVVP (arrow).
Fig. 3E —CSF–venous fistula (CVF) drainage into internal vertebral venous plexus (IVVP).
E, Image from dynamic myelography performed with patient in right lateral decubitus position (same patient as in Fig. 3D) shows draining vein (arrow) in IVVP inferior to T11 nerve root, marking presence of CVF.
Fig. 4A —CSF–venous fistula (CVF) drainage into lateral and muscular branches of external vertebral venous plexus (EVVP).
A, 31-year-old woman with CVF at T10 on left. Axial (A) and coronal oblique (B) images from CT myelography performed with patient in decubitus position show reflux of contrast material into lateral branches of EVVP in intercostal space (arrows).
Fig. 4B —CSF–venous fistula (CVF) drainage into lateral and muscular branches of external vertebral venous plexus (EVVP).
B, 31-year-old woman with CVF at T10 on left. Axial (A) and coronal oblique (B) images from CT myelography performed with patient in decubitus position show reflux of contrast material into lateral branches of EVVP in intercostal space (arrows).
Fig. 4C —CSF–venous fistula (CVF) drainage into lateral and muscular branches of external vertebral venous plexus (EVVP).
C, 35-year-old woman with CVF at T11 on right. Axial (C) and coronal oblique (D) images from CT myelography performed with patient in decubitus position show reflux of contrast medium into muscular branches of EVVP (arrows).
Fig. 4D —CSF–venous fistula (CVF) drainage into lateral and muscular branches of external vertebral venous plexus (EVVP).
D, 35-year-old woman with CVF at T11 on right. Axial (C) and coronal oblique (D) images from CT myelography performed with patient in decubitus position show reflux of contrast medium into muscular branches of EVVP (arrows).
Fig. 5A —72-year-old woman with CSF–venous fistula (CVF) at T9 on left illustrating drainage into basivertebral venous plexus.
A, Axial (A) and sagittal (B) maximum-intensity-projection images from CT myelography performed with patient in decubitus position show opacification of basivertebral venous plexus (arrows). Communication between basivertebral venous plexus and segmental spinal veins (arrowhead, A) is seen.
Fig. 5B —72-year-old woman with CSF–venous fistula (CVF) at T9 on left illustrating drainage into basivertebral venous plexus.
B, Axial (A) and sagittal (B) maximum-intensity-projection images from CT myelography performed with patient in decubitus position show opacification of basivertebral venous plexus (arrows). Communication between basivertebral venous plexus and segmental spinal veins (arrowhead, A) is seen.
Rarely, CVFs can also coexist with paraspinal venous or venolymphatic malformations [4, 15]. In such cases, fistulas generally occur at the same level and on the same side as the vascular malformation, with identifiable drainage of the fistula into the malformation (Fig. 6).
Fig. 6A —43-year-old man with CSF–venous fistula (CVF) draining into venolymphatic malformation.
A, Axial T2-weighted MR image shows large venolymphatic malformation involving spinal column, epidural space, and pleural space.
Fig. 6B —43-year-old man with CSF–venous fistula (CVF) draining into venolymphatic malformation.
B, Axial image from CT myelography shows CVF (arrow) extending from left T8 nerve root, with pooling of contrast material in venolymphatic malformation (arrowhead).

Pathogenesis of CSF–Venous Fistulas

The event or sequence of events that precipitates the formation of a CVF is currently unknown. Nevertheless, the result is the creation of a connection between the subarachnoid space of the spinal nerve root and an adjacent vein. It is therefore worthwhile to consider the anatomic relationships that exist between the spinal subarachnoid space and paraspinal veins.
The vertebral venous plexus is an extensive valveless network of epidural veins that surrounds the thecal sac and the spinal nerve roots and helps dampen fluctuations in CSF pressure during normal physiologic processes, such as a change in posture or during coughing, sneezing, or the Valsalva maneuver [11]. Numerous animal studies have suggested that this venous network contributes to reabsorption of CSF via spinal arachnoid granulations [16, 17]. In humans, it is known that arachnoid granulations are involved in CSF resorption along the intracranial dural venous sinuses, but arachnoid granulations have also been found to be widely distributed along spinal nerve roots [18, 19]. It has been estimated that spinal routes of CSF absorption account for approximately 20% of total CSF absorption, and that rate may be higher in the upright position because of increased pressure gradients across spinal arachnoid granulations [20].
Spinal arachnoid proliferations (subdivided into smaller arachnoid villi and larger arachnoid granulations) vary in size, with some smaller forms penetrating into but not through the dura, others extending into the epidural space, and larger arachnoid granulations projecting through the dura and into adjacent epidural spinal veins [18, 21] (Fig. 7). In human cadaveric specimens, Kido et al. [18] found that these arachnoid proliferations are most commonly found in the thoracic spine, in particular the mid to lower thoracic spine. This distribution seems to parallel the distribution of CVFs observed by multiple investigators [4, 7, 22]. The size of the spinal arachnoid proliferations varies considerably, with the largest reported to be up to 0.5 mm3 and connected to the subarachnoid space by a well-defined endothelial-lined neck [18]. Most spinal nerve roots harbor at least one arachnoid proliferation. Occasionally, proliferations arise directly from the meninges of the spinal cord rather than the nerve root sleeve [18].
Fig. 7 —Illustration of relationship between spinal arachnoid proliferations and vertebral venous plexus. Under normal conditions, spinal arachnoid granulations project through epidural space into adjacent veins. Rupture of these arachnoid granulations has been proposed as possible mechanism for formation of CSF–venous fistulas. (Illustration by Lydia Gregg, MA, CMI © 2021 JHU)
Given the known relationship of spinal arachnoid granulations to the paraspinal venous plexus, and given the fact that these arachnoid granulations occur predominantly along the nerve roots of the thoracic spine, it has been suggested that rupture of a granulation may be the inciting event in the creation of a CVF [4].
Alternatively, it is conceivable that the fistula arises between the subarachnoid space of the thecal sac, connecting the subarachnoid space to the extrinsic venous system that drains the spinal cord parenchyma, with CSF flowing into radicular veins and ultimately into the epidural venous plexus. The fact that surgical nerve root ligation is effective in treating CVFs suggests that for this alternative theory to be correct, the site of the fistula needs to be along the nerve root rather than in the thecal sac proper, because fistula formation proximal to this site would not be addressed by ligation of the more distal nerve root. Ultimately, these theories of pathogenesis are speculative, and more investigation is needed to prove causality conclusively.

Imaging

Detection of CVFs can be challenging, and the optimal methods for detecting CVFs are still being explored. Since the initial description of CVFs, modifications to existing imaging techniques and novel imaging techniques have been developed to aid detection. This review addresses the current state of evidence regarding imaging of CVFs. It is likely that further investigations into current techniques and innovations that lead to new imaging techniques will continue to be forthcoming given the pace of innovation in this area.
Broadly, the available imaging techniques for detection of CVFs can be divided into those based on fluoroscopy and those based on cross-sectional imaging.

Patient Selection and the Role of Brain MRI

Previous reviews have described the clinical presentation, diagnostic criteria, and imaging selection in patients with SIH [22, 23]. All patients with SIH optimally should first undergo initial spine imaging that is capable of detecting epidural fluid collections that would indicate an epidural leak of fluid [23]. Patients with SIH who have no identifiable epidural fluid collection on spine imaging (whether spine MRI or CT myelography [CTM]) should be suspected of harboring a CVF. Although one might speculate that radiographically occult epidural leaks might be present in patients with SIH who have negative spinal imaging, there is very little empirical evidence for this supposition, and the available literature suggests that CVFs are the actual cause of SIH in the large majority of these cases [24], an observation supported by the authors' experience.
The large majority of cases of SIH due to CVFs are associated with abnormalities on brain MRI. One study found at least one sign of SIH on contrast-enhanced brain MRI in 20 of 22 (91%) patients with confirmed CVF. Of the individual signs, venous distention (seen in 77% of cases) and dural enhancement (seen in 73%) were most common [4]. Similarly, in an investigation of patients with CVF diagnosed using digital subtraction myelography (DSM) performed with the patient in the decubitus position, 20 of 21 patients (95%) had at least one sign of SIH present on brain MRI [24].
Recently, two investigations addressed the question regarding the diagnostic yield of decubitus DSM in patients with orthostatic headaches but without clear evidence of SIH on brain or spine MRI, with conflicting conclusions. The first investigation found a CVF in six of 60 patients (10%) with a clinical suspicion for SIH but negative brain and spine MRI [25]. A second contemporaneous study [26] did not find any CVF in patients with normal findings on both brain MRI and spine MRI who underwent decubitus DSM, although the cohort of patients with negative imaging was smaller in this second study compared with the first study.
CVF detection requires expertise limited to relatively few referral centers, invasive myelographic techniques using ionizing radiation, and possibly up to two consecutive sessions of general anesthesia. It will thus be important to develop strategies to better predict which patients with normal MRI are most likely to benefit from specialized evaluation.

Fluoroscopy-Based Techniques

Digital subtraction myelography—Initial descriptions of CVFs were based on fluoroscopy performed with digital subtraction for a patient in the prone position on a tilting table [1]. This technique, known as DSM, was developed to better characterize the site of origin of rapid CSF leaks into the epidural space. It offered the primary benefit of high temporal resolution, which allowed the leak to be identified and precisely localized immediately as contrast material entered the epidural space [27].
In the first description of CVF in the setting of SIH, biplane DSM identified direct filling of paraspinal veins in a series of three patients with brain MRI findings of SIH, two of whom had no concurrent epidural leakage of CSF [1]. Using this DSM technique with prone positioning, subsequent investigations identified CVFs in 19% of patients with SIH who had no identifiable epidural CSF leak [7].
In later research, decubitus positioning, a modification first reported to be useful during evaluation of CVF with CTM [28], was found to further increase the diagnostic yield of DSM [22, 24]. In an investigation of 23 patients with SIH and no visible epidural leak who underwent either prone or decubitus DSM, prone DSM revealed a CVF in 15% of patients, whereas decubitus DSM identified a CVF in 74%, for a fivefold increase in diagnostic yield [24]. Decubitus positioning has now been adopted as a standard technique when DSM is performed for CVF detection (Fig. 8).
Fig. 8A —Examples of CSF–venous fistula (CVFs) seen on digital subtraction myelography (DSM).
A, 50-year-old woman with CVF at T10 on left. Unsubtracted DSM image shows drainage into intervertebral vein (arrow) and branches of external epidural venous plexus (arrowhead).
Fig. 8B —Examples of CSF–venous fistula (CVFs) seen on digital subtraction myelography (DSM).
B, 38-year-old woman with CVF at T8 on right. Subtracted DSM image shows CVF originating from nerve root sleeve diverticulum (arrowhead) with drainage into internal vertebral venous plexus (arrow).
Fig. 8C —Examples of CSF–venous fistula (CVFs) seen on digital subtraction myelography (DSM).
C, 74-year-old woman with CVF at T2 on right. Subtracted DSM image shows complex CVF with drainage into intervertebral veins (arrow) and branches of external vertebral venous plexus.
Performance of decubitus DSM requires myelographic contrast material to flow in a cranial direction after it is injected into the lumbar thecal sac [29]. This can be accomplished by placing the patient on a tilting table and tilting the head down or by using foam wedges to elevate the hips; the former method provides more precision in controlling the degree of tilt. Contrast material is rapidly injected into the thecal sac, and digital subtraction imaging is performed while the contrast material is actively migrating cranially along the lateral thecal sac, to capture real-time filling of the spinal nerve root sleeves and any CVF. To minimize misregistration caused by respiratory motion between the mask image and the myelographic images, general endotracheal anesthesia may be used with suspension of respirations during imaging [24, 29]. For patients who are able to cooperate with breath-holding, DSM may also be performed without general anesthesia. General anesthesia or appropriate sedation also helps to avoid patient discomfort due to headache from myelo-graphic contrast medium extending intracranially.
Using this technique, each side of the thecal sac must be examined separately because of the deleterious effect that residual intrathecal contrast material would have on additional digital subtraction runs. Furthermore, the total amount of iodinated contrast material that can be delivered intrathecally in a 24-hour period is limited, generally to a maximum dose of 3 g of iodine [30]. As a result, if no CVF is identified on one side of the thecal sac, the procedure must be repeated on the contralateral side on a subsequent day; DSM is thus typically scheduled as a 2-day procedure. An important limitation of DSM is that the length of the spine that can be imaged in a single run depends on the FOV of the fluoroscopy unit [29].
Dynamic myelography—An alternative to DSM is real-time fluoroscopy performed without digital subtraction, a procedure referred to as dynamic myelography [31]. This technique can identify CVFs, although the diagnostic yield of dynamic myelography has not been as well studied as DSM. In this technique, contrast material is injected into the thecal sac with the fluoroscopy table initially tilted into the feet-down position. The needle is removed, and the table is then slowly tilted into the head-down position to promote a controlled cranial migration of contrast material [4, 32]. Intermittent pulsed fluoroscopy and spot images are used to assess contrast migration and evaluate for CVFs, respectively [31] (Fig. 9). Unlike DSM, a distinct benefit of dynamic myelography is that it does not require rapid contrast injection, and image acquisition is not limited to a single run.
Fig. 9A —Dynamic myelography for confirmation and localization of CSF–venous fistula (CVF).
A, 42-year-old woman with CVF at L1 on left. Axial image from CT myelography (A) shows small linear area of contrast opacification (arrow) lateral to left L1 nerve root, interpreted as possible CVF. Fluoroscopic image (B) from dynamic myelography performed with patient in left lateral decubitus position confirms and more clearly shows CVF, with drainage of contrast material into network of veins in internal vertebral venous plexus (IVVP) (arrow) and more peripheral filling of branches of external vertebral venous plexus (arrowhead).
Fig. 9B —Dynamic myelography for confirmation and localization of CSF–venous fistula (CVF).
B, 42-year-old woman with CVF at L1 on left. Axial image from CT myelography (A) shows small linear area of contrast opacification (arrow) lateral to left L1 nerve root, interpreted as possible CVF. Fluoroscopic image (B) from dynamic myelography performed with patient in left lateral decubitus position confirms and more clearly shows CVF, with drainage of contrast material into network of veins in internal vertebral venous plexus (IVVP) (arrow) and more peripheral filling of branches of external vertebral venous plexus (arrowhead).
Fig. 9C —Dynamic myelography for confirmation and localization of CSF–venous fistula (CVF).
C, 59-year-old man with CVF at T9 on right. Axial image from CT myelography (C) shows contrast opacification (arrowhead) of IVVP, indicating presence of CVF. Fluoroscopic image (D) from dynamic myelography performed with patient in right lateral decubitus position shows CVF draining into segmental spinal vein (arrow), definitively localizing site of CVF.
Fig. 9D —Dynamic myelography for confirmation and localization of CSF–venous fistula (CVF).
D, 59-year-old man with CVF at T9 on right. Axial image from CT myelography (C) shows contrast opacification (arrowhead) of IVVP, indicating presence of CVF. Fluoroscopic image (D) from dynamic myelography performed with patient in right lateral decubitus position shows CVF draining into segmental spinal vein (arrow), definitively localizing site of CVF.
The principal downside of dynamic myelography is the lack of subtraction of background tissues, which can make detection of subtle CVFs more difficult. To our knowledge, no studies have been published that compare the diagnostic yield of DSM and dynamic myelography for CVF detection. However, most referral centers use DSM rather than dynamic myelography, and in our opinion it is likely that DSM is more sensitive as a screening technique if no particular site of CVF is already suspected.
Dynamic myelography does not require general anesthesia because no subtraction mask is used. Furthermore, levels of interest can be evaluated in multiple obliquities and over a longer period, allowing a more extensive evaluation. The effects of different phases of respiration can be assessed, which may impact CVF detection in some cases. A preliminary investigation of the effect of respiratory phase on CVF detection found that the conspicuity of CVFs varied in different phases of respiration [33]. In that study, inspiration was associated with greater CVF visualization (Fig. 10).
Fig. 10A —60-year-old man with CSF–venous fistula (CVF) at T2 on right illustrating effect of respiratory phase on CVF visualization. (Adapted with permission of the American Society of Neuroradiology from Respiratory phase affects the conspicuity of CSF–venous fistulas in spontaneous intracranial hypotension. Amrhein TJ, Gray L, Malinzak MD, Kranz PG. AJNR Am J Neuroradiol 2020 Sep; 41(9):1754–1756; permission conveyed through Copyright Clearance Center, Inc.)
A, Image from dynamic myelography performed during maximal inspiratory breath-hold shows extensive draining venous branches (arrows) arising from T2 nerve root.
Fig. 10B —60-year-old man with CSF–venous fistula (CVF) at T2 on right illustrating effect of respiratory phase on CVF visualization. (Adapted with permission of the American Society of Neuroradiology from Respiratory phase affects the conspicuity of CSF–venous fistulas in spontaneous intracranial hypotension. Amrhein TJ, Gray L, Malinzak MD, Kranz PG. AJNR Am J Neuroradiol 2020 Sep; 41(9):1754–1756; permission conveyed through Copyright Clearance Center, Inc.)
B, Images obtained during same dynamic myelography examination performed after patient was asked to perform Valsalva maneuver in expiration. Early image (left) and late image (right) obtained during Valsalva maneuver show progressive decrease in drainage of CVF draining branches (arrowhead).
Fig. 10C —60-year-old man with CSF–venous fistula (CVF) at T2 on right illustrating effect of respiratory phase on CVF visualization. (Adapted with permission of the American Society of Neuroradiology from Respiratory phase affects the conspicuity of CSF–venous fistulas in spontaneous intracranial hypotension. Amrhein TJ, Gray L, Malinzak MD, Kranz PG. AJNR Am J Neuroradiol 2020 Sep; 41(9):1754–1756; permission conveyed through Copyright Clearance Center, Inc.)
C, Image obtained after release of Valsalva maneuver during repeat maximal inspiratory breath-hold shows refilling of draining venous branches (arrow).
In general, we find that dynamic myelography is a suitable alternative to DSM as a problem-solving tool for the focused examination of a specific spinal level that is suspected of harboring a CVF based on prior spine imaging. However, we find techniques such as DSM or CTM to be more practical for screening large areas of the spine when the presence of a CVF is unknown. Dynamic myelography is also more operator dependent than other techniques, and although it is a valuable tool, this learning curve may make it better suited to practices in which referral for SIH imaging is more common.

Cross-Sectional Techniques

CT myelography—Shortly after initial descriptions of CVFs were reported using DSM, analogous findings on CTM were described [31]. Increased attenuation of the paraspinal veins was observed after intrathecal administration of iodinated contrast material, a finding that indicated the presence of a CVF. This observation, referred to as the hyperdense paraspinal vein sign, has been found to correlate with CVFs seen on fluoroscopy-based imaging studies performed for the same patients [4, 31]. Subsequent work showed that attenuation values greater than 70 HU within a paraspinal vein after intrathecal injection of myelographic contrast material were specific for the presence of a CVF [4].
Lateral decubitus positioning during CTM scanning facilitates the detection of CVFs (Fig. 11). An investigation involving patients undergoing CTM for evaluation of CVFs found an increase of more than 500% in attenuation in the draining veins of CVFs with decubitus positioning compared with prone positioning, greatly increasing conspicuity of the hyperdense paraspinal vein sign [28]. The decubitus position was observed to increase the concentration of myelographic contrast material in the thecal sac over the site of the fistula (a 300% increase in attenuation compared with prone CT). This increased concentration of contrast material was deemed likely to be a primary factor in increasing the visibility of the CVFs, but there also appeared to be an effect of gravity that was independent of the concentration of contrast material. The diagnostic yield of decubitus CTM for CVF detection was found to be 50% in patients with SIH who had no epidural fluid collection seen on spine MRI [8].
Fig. 11A —44-year-old woman with spontaneous intracranial hypotension illustrating improved CSF–venous fistula (CVF) visualization in decubitus positioning.
A, Axial image from CT myelography obtained at T10 level with patient in prone position shows no definite CSF leak or CVF.
Fig. 11B —44-year-old woman with spontaneous intracranial hypotension illustrating improved CSF–venous fistula (CVF) visualization in decubitus positioning.
B, Axial image from subsequent CT myelography examination performed with patient in left lateral decubitus position and obtained at same level as image in A shows venous filling into lateral (arrow) and muscular (arrowhead) branches of external vertebral venous plexus not seen on myelogram obtained with patient in prone position (A), confirming presence of CVF.
In decubitus CTM, contrast material ideally would be injected with the patient in the lateral decubitus position without rolling the patient to distribute contrast material before scanning, to maximize contrast concentration over the nerve roots. Since this evaluates only one side of the spine, the injection must be repeated in the contralateral decubitus position. Promoting the correct amount of cranial migration of contrast material while simultaneously achieving rapid CT after injection of contrast material can be challenging, and techniques to reliably facilitate this are still under development. However, simply turning the patient into the decubitus position immediately after prone CTM and rescanning the patient is sufficient to identify some additional cases of CVF. The additional diagnostic yield of each of these two types of decubitus scanning compared with prone CTM is unknown, although available evidence justifies the use of some form of decubitus scanning during CTM if no rapid epidural leak is seen in a patient with SIH.
On CTM, the EVVP was the most frequent location where the hyperdense paraspinal vein sign was identified, with the paravertebral segmental spinal veins being the most common signs, and the muscular branches were a less frequently seen location [4]. Of importance, this same series of CVFs evaluated with CT found that in approximately one-third of cases, the hyperdense vein was seen exclusively within the spinal canal, filling the IVVP. In our experience, venous filling in this location can be easily overlooked or mistaken for epidural rather than intravascular contrast material, and therefore the IVVP should be scrutinized when searching for CVFs, particularly around nerve root sleeves and in the ventrolateral spinal canal.
Filling of paraspinal veins can be subtle, with only small areas of venous opacification seen or subtle differences in attenuation appearing between veins draining a fistula and normal adjacent veins. One study examining CTM performed for patients with SIH before widespread recognition of CVFs found that, in retrospect, seven of 101 myelograms that were thought at the time to reveal no abnormality actually showed evidence of a CVF that was not identified on the original interpretation [6].
Detection of subtle CVFs is further facilitated by careful attention to CT technique, including performing scanning immediately after intrathecal injection of contrast material, using a submillimeter slice thickness, and using inspiratory breath-hold technique while scanning [23]. Even small delays in scanning can cause CVFs to be missed; in our experience, delayed imaging is never helpful in CVF detection and should be avoided.
As with dynamic myelography, filling of CVFs on CTM may be augmented by different phases of respiration [33]. In many cases, maximum inspiration produces the greatest augmentation of CVF filling. Determining how these observations should be used to guide imaging protocols for CVF detection using CTM, DSM, or other modalities remains the subject of ongoing investigation.
An indirect sign of the presence of a CSF leak in general and perhaps CVF in particular is early opacification of the renal collecting system on postmyelography CT. Two studies found that the presence of early renal contrast opacification after myelography was seen in some patients with SIH but was not seen in any control patients [34, 35]. In one of these investigations that assessed CTM examinations performed approximately 20–40 minutes after intrathecal injection of contrast material, the sign was more commonly observed among patients with CVFs compared with those with epidural CSF leaks [34]. In the other investigation that assessed CTM performed approximately 60 minutes after intrathecal injection of contrast material, renal contrast opacification was seen in seven cases, but only one of those cases was attributable to CVF [35]. Regardless, when early opacification of the renal collecting system is seen in the absence of an epidural leak, a CVF should be suspected. A similar phenomenon has also been observed on nuclear medicine cisternography (a technique now generally considered obsolete in the workup of SIH) in which early radiotracer activity over the kidneys and bladder was considered to be an indirect sign of CSF leakage, albeit one subject to false-positive and false-negative findings [36, 37].
Spine MRI—When brain imaging shows evidence of SIH, spine MRI is used by many centers as first-line spinal imaging to evaluate for fluid collections that would indicate the presence of an epidural CSF leak. The presence of SIH on brain imaging with no visible epidural fluid collection on spine imaging may therefore indirectly suggest that the source of CSF loss is attributable to a CVF. Direct identification of a CVF, however, relies on identification of shunting from the CSF to the venous system rather than on static anatomic abnormalities such as pooled epidural fluid. As a result, conventional spine MRI cannot directly detect CVFs [5].
Preliminary research showed that MR myelography using off-hlabel intrathecal injection of gadolinium-based contrast agents was able to detect at least some CVFs [38]. The sensitivity of MR myelography using off-label intrathecal injection of gadolinium-based contrast agents compared with other techniques, such as DSM or CTM, has not yet been convincingly established, and this technique is not commonly used at most major referral centers at present.

Comparison of Spinal Imaging Techniques for CVF Detection

Optimal techniques for detection of CVFs are still being developed. Comparison of imaging modalities for detecting CVFs has so far been limited to retrospective reports of CVFs confirmed using one imaging technique but not seen on previous imaging; such investigations are subject to selection bias that favors the modality used as the reference standard. To our knowledge, no head-to-head studies have been performed that directly compare the sensitivity and specificity of CVF detection using more than one imaging modality, which limits the ability to make broad generalizations about whether any technique is superior for CVF detection.
Utilization of different imaging techniques is currently largely driven by the availability of imaging equipment and operator preference. Qualitative comparison of the techniques, however, is summarized in Table 1.
TABLE 1: Comparison of Imaging Modalities Used for CSF–Venous Fistula (CVF) Detection
Category and TechniqueEquipmentAdvantagesDisadvantagesRadiation Dose
Fluoroscopy-based techniques    
ߓDSMAngiography table, preferably with table-tilt capabilityWell-studied technique
Good diagnostic yield compared with conventional CTM
Lower radiation dose than CTM
Two-day procedure
May require general anesthesia to avoid respiratory motion misregistration
Limited by FOV of detector
Approximately 5–10 mSv [39]
ߓDynamic myelographyC-arm fluoroscopy or angiography table with table-tilt capacityDoes not require general anesthesia
Less sensitive to respiratory motion
No FOV limitation
Allows for multiple obliquities
Allows examination of different respiratory phases
Background structures may obscure small CVFs
Optimal for confirmation of suspected CVF rather than screening for CVF
Dependent on operator experience
Variable (depending on anatomy examined, typically less than DSM)
Cross-sectional techniques    
ߓCTM with patient in decubitus positionMDCT scanner with CT fluoroscopy capabilityWide availability of equipment
Does not require general anesthesia
May evaluate both sides of the spine in single procedure
Can evaluate entire spine
Diagnostic yield not as well studied as that of DSM
Higher radiation exposure than DSM
Highly technique dependent
No tilting table capability on CT
Approximately 20–40 mSv [39, 40]
ߓMR myelography with intrathecal gadoliniumMRI scannerDoes not require immediate imagingUncertain diagnostic yield
Requires off-label intrathecal injection of GBCA
Potential neurotoxic effects of GBCA if overdosed
None

Note—DSM = digital subtraction myelography, CTM = CT myelography, GBCA = gadolinium-based contrast agent.

Conclusion

CVFs are now recognized as an important cause of spinal CSF leak that leads to SIH, but identification can be difficult because of unfamiliarity with their imaging manifestations and the dynamic nature of CVF filling during myelography. In recent years, imaging techniques have been developed that improve detection of CVFs, and knowledge of these various options, including their relative advantages and disadvantages, is critical in the care of patients with SIH. Given the rapid pace of innovation and discovery in this field, continued advances in our understanding of CVFs and their optimal imaging detection are anticipated.

References

1.
Schievink WI, Moser FG, Maya MM. CSF–venous fistula in spontaneous intracranial hypotension. Neurology 2014; 83:472–473
2.
Marmarou A, Shulman K, LaMorgese J. Compartmental analysis of compliance and outflow resistance of the cerebrospinal fluid system. J Neurosurg 1975; 43:523–534
3.
Potts DG, Gomez DG, Shaw DD. Cranial and spinal cerebrospinal fluid absorption and the clearance of water-soluble myelographic contrast media: a review. Invest Radiol 1985; 20(Suppl):S51–S54
4.
Kranz PG, Amrhein TJ, Gray L. CSF venous fistulas in spontaneous intracranial hypotension: imaging characteristics on dynamic and CT myelography. AJR 2017; 209:1360–1366
5.
Duvall JR, Robertson CE, Cutsforth-Gregory JK, Carr CM, Atkinson JL, Garza I. Headache due to spontaneous spinal cerebrospinal fluid leak secondary to cerebrospinal fluid–venous fistula: case series. Cephalalgia 2019; 39:1847–1854
6.
Clark MS, Diehn FE, Verdoorn JT, et al. Prevalence of hyperdense paraspinal vein sign in patients with spontaneous intracranial hypotension without dural CSF leak on standard CT myelography. Diagn Interv Radiol 2018; 24:54–59
7.
Schievink WI, Moser FG, Maya MM, Prasad RS. Digital subtraction myelography for the identification of spontaneous spinal CSF–venous fistulas. J Neurosurg Spine 2016; 24:960–964
8.
Mamlouk MD, Ochi RP, Jun P, Shen PY. Decubitus CT myelography for CSF–venous fistulas: a procedural approach. AJNR 2021; 42:32–36
9.
Kumar N, Diehn FE, Carr CM, et al. Spinal CSF venous fistula: a treatable etiology for CSF leaks in craniospinal hypovolemia. Neurology 2016; 86:2310–2312
10.
Kranz PG, Stinnett SS, Huang KT, Gray L. Spinal meningeal diverticula in spontaneous intracranial hypotension: analysis of prevalence and myelo-graphic appearance. AJNR 2013; 34:1284–1289
11.
Carpenter K, Decater T, Iwanaga J, et al. Revisiting the vertebral venous plexus: a comprehensive review of the literature. World Neurosurg 2021; 145:381–395
12.
LaBan MM, Wilkins JC, Szappanyos B, Shetty AN, Wang AM. Paravertebral plexus of veins (Batson's): potential route of paraspinal muscle metastases as imaged by magnetic venous angiography—a commentary. Am J Phys Med Rehabil 1997; 76:517–519
13.
Groen RJ, du Toit DF, Phillips FM, et al. Anatomical and pathological considerations in percutaneous vertebroplasty and kyphoplasty: a reappraisal of the vertebral venous system. Spine 2004; 29:1465–1471
14.
Santillan A, Nacarino V, Greenberg E, Riina HA, Gobin YP, Patsalides A. Vascular anatomy of the spinal cord. J Neurointerv Surg 2012; 4:67–74
15.
Schievink WI, Maya MM, Moser FG, et al. Spontaneous spinal CSF–venous fistulas associated with venous/venolymphatic vascular malformations: report of 3 cases. J Neurosurg Spine 2019; 32:305–310
16.
Pollay M. The function and structure of the cerebrospinal fluid outflow system. Cerebrospinal Fluid Res 2010; 7:9
17.
Chen L, Elias G, Yostos MP, Stimec B, Fasel J, Murphy K. Pathways of cerebro-spinal fluid outflow: a deeper understanding of resorption. Neuroradiology 2015; 57:139–147
18.
Kido DK, Gomez DG, Pavese AM Jr, Potts DG. Human spinal arachnoid villi and granulations. Neuroradiology 1976; 11:221–228
19.
Tubbs RS, Hansasuta A, Stetler W, et al. Human spinal arachnoid villi revisited: immunohistological study and review of the literature. J Neurosurg Spine 2007; 7:328–331
20.
Edsbagge M, Tisell M, Jacobsson L, Wikkelso C. Spinal CSF absorption in healthy individuals. Am J Physiol Regul Integr Comp Physiol 2004; 287:R1450–R1455
21.
Welch K, Pollay M. The spinal arachnoid villi of the monkeys Cercopithecus aethiops sabaeus and Macaca irus. Anat Rec 1963; 145:43–48
22.
Farb RI, Nicholson PJ, Peng PW, et al. Spontaneous intracranial hypotension: a systematic imaging approach for CSF leak localization and management based on MRI and digital subtraction myelography. AJNR 2019; 40:745–753
23.
Kranz PG, Gray L, Malinzak MD, Amrhein TJ. Spontaneous intracranial hypotension: pathogenesis, diagnosis, and treatment. Neuroimaging Clin N Am 2019; 29:581–594
24.
Schievink WI, Maya MM, Moser FG, et al. Lateral decubitus digital subtraction myelography to identify spinal CSF-venous fistulas in spontaneous intracranial hypotension. J Neurosurg Spine 2019 Sep 13 [published online]
25.
Schievink WI, Maya M, Prasad RS, et al. Spontaneous spinal cerebrospinal fluid–venous fistulas in patients with orthostatic headaches and normal conventional brain and spine imaging. Headache 2021; 61:387–391
26.
Kim DK, Carr CM, Benson JC, et al. Diagnostic yield of lateral decubitus digital subtraction myelogram stratified by brain MRI findings. Neurology 2021; 96:e1312–e1318
27.
Hoxworth JM, Patel AC, Bosch EP, Nelson KD. Localization of a rapid CSF leak with digital subtraction myelography. AJNR 2009; 30:516–519
28.
Kranz PG, Gray L, Amrhein TJ. Decubitus CT myelography for detecting subtle CSF leaks in spontaneous intracranial hypotension. AJNR 2019; 40:754–756
29.
Kim DK, Brinjikji W, Morris PP, et al. Lateral decubitus digital subtraction myelography: tips, tricks, and pitfalls. AJNR 2020; 41:21–28
30.
U.S. FDA. NDA 18-735; ISOVUE (iopamidol) injection. Drugs@FDA website. www.accessdata.fda.gov/drugsatfda_docs/label/2017/018735s057lbl.pdf. Revised March 2017. Accessed April 20, 2021
31.
Kranz PG, Amrhein TJ, Schievink WI, Karikari IO, Gray L. The “hyperdense paraspinal vein” sign: a marker of CSF–venous fistula. AJNR 2016; 37:1379–1381
32.
Piechowiak EI, Pospieszny K, Haeni L, et al. Role of conventional dynamic myelography for detection of high-flow cerebrospinal fluid leaks: optimizing the technique. Clin Neuroradiol 2020 Aug 26 [published online]
33.
Amrhein TJ, Gray L, Malinzak MD, Kranz PG. Respiratory phase affects the conspicuity of CSF–venous fistulas in spontaneous intracranial hypotension. AJNR 2020; 41:1754–1756
34.
Kinsman KA, Verdoorn JT, Luetmer PH, Clark MS, Diehn FE. Renal contrast on CT myelography: diagnostic value in patients with spontaneous intracranial hypotension. AJNR 2019; 40:376–381
35.
Behbahani S, Raseman J, Orlowski H, Sharma A, Eldaya R. Renal excretion of contrast on CT myelography: a specific marker of CSF leak. AJNR 2020; 41:351–356
36.
Kranz PG, Luetmer PH, Diehn FE, Amrhein TJ, Tanpitukpongse TP, Gray L. Myelographic techniques for the detection of spinal CSF leaks in spontaneous intracranial hypotension. AJR 2016; 206:8–19
37.
Mokri B. Radioisotope cisternography in spontaneous CSF leaks: interpretations and misinterpretations. Headache 2014; 54:1358–1368
38.
Chazen JL, Talbott JF, Lantos JE, Dillon WP. MR myelography for identification of spinal CSF leak in spontaneous intracranial hypotension. AJNR 2014; 35:2007–2012
39.
Nicholson PJ, Guest WC, van Prooijen M, Farb RI. Digital subtraction myelography is associated with less radiation dose than CT-based techniques. Clin Neuroradiol 2020 Aug 17 [published online]
40.
Dobrocky T, Mosimann PJ, Zibold F, et al. Cryptogenic cerebrospinal fluid leaks in spontaneous intracranial hypotension: role of dynamic CT myelography. Radiology 2018; 289:766–772

Information & Authors

Information

Published In

American Journal of Roentgenology
Pages: 1418 - 1429
PubMed: 34191547

History

Submitted: May 7, 2021
Revision requested: May 21, 2021
Revision received: May 28, 2021
Accepted: June 14, 2021
Version of record online: June 30, 2021

Keywords

  1. CSF–venous fistula
  2. CT myelography
  3. digital subtraction myelography
  4. spinal CSF leak
  5. spontaneous intracranial hypotension

Authors

Affiliations

Peter G. Kranz, MD
Department of Radiology, Division of Neuroradiology, Duke University Medical Center, DUMC Box 3808, Durham, NC 27710
Linda Gray, MD
Department of Radiology, Division of Neuroradiology, Duke University Medical Center, DUMC Box 3808, Durham, NC 27710
Michael D. Malinzak, MD, PhD
Department of Radiology, Division of Neuroradiology, Duke University Medical Center, DUMC Box 3808, Durham, NC 27710
Jessica L. Houk, MD
Department of Radiology, Division of Neuroradiology, Duke University Medical Center, DUMC Box 3808, Durham, NC 27710
Dong Kun Kim, MD
Department of Radiology, Division of Neuroradiology, Mayo Clinic, Rochester, MN
Timothy J. Amrhein, MD
Department of Radiology, Division of Neuroradiology, Duke University Medical Center, DUMC Box 3808, Durham, NC 27710

Notes

Address correspondence to P. G. Kranz ([email protected]).
P. G. Kranz, L. Gray, and T. J. Amrhein are unpaid members of the medical advisory board of the Spinal CSF Leak Foundation, and P. G. Kranz and L. Gray are unpaid members of the medical advisory board of Spinal CSF Leak Canada. The remaining authors declare that they have no disclosures relevant to the subject matter of this article.

Metrics & Citations

Metrics

Citations

Export Citations

To download the citation to this article, select your reference manager software.

Articles citing this article

View Options

View options

PDF

View PDF

PDF Download

Download PDF

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media