Spontaneous Intracranial Hypotension and CSF-Venous Fistulae: from Diagnostic Imaging to Interventional Management

Abstract

Over the past decade, major advances have transformed the understanding and management of cerebrospinal fluid (CSF) leaks leading to spontaneous intracranial hypotension (SIH). Among the most impactful developments has been the identification and characterization of CSF—venous fistulae (CVF), a distinct and still underrecognized cause of SIH. This discovery has prompted a paradigm shift in both diagnostic and therapeutic strategies. Novel imaging techniques—including lateral decubitus CT myelography and digital subtraction myelography—have markedly improved detection rates, while targeted interventions such as transvenous embolization have emerged as safe and effective treatment options. In this review, we first outline the general pathophysiology and imaging principles related to SIH and CSF leaks, followed by a focused and comprehensive analysis of the current evidence managing CVFs. We discuss their diagnostic challenges, therapeutic approaches, and evolving clinical implications, aiming to provide a thorough update for diagnostic and interventional neuroradiologists involved in the care of patients with SIH.

Introduction

Spontaneous intracranial hypotension (SIH) is a condition characterized by low cerebrospinal fluid (CSF) pressure, typically due to a spontaneous CSF leak at the level of the spine, without a known trauma or medical procedure as the cause. The leakage of CSF may be into the epidural space from a pachymeningeal tear, leading to longitudinal epidural collections, in the paravertebral space through leaking nerve root dural diverticulum, or into periforaminal, epidural, paravertebral or intraosseous veins through a CSF-venous fistula (CVF) [1, 2]. In recent years, CVFs have gained increasing attention as they represent a common cause for SIH and are a target for an effective endovascular management [3, 4]. In the following review we will describe the epidemiology, pathophysiology and clinical symptomatology of SIH in general and of CVFs in particular and will then focus on imaging and treatment options for this entity.

Epidemiology and Demographics

Over the past decades, awareness about SIH has increased, particularly since advances in imaging technology shed light on its underlying mechanisms. While improved understanding has contributed to a rise in diagnosed cases, precise incidence and prevalence of this disease are challenging to determine, due to frequent under- or misdiagnosis. Current estimates place the annual incidence at approximately 3–5 per 100,000 individuals, though this may be an underestimation given that many cases are overlooked or confused with other headache conditions, such as migraines or tension headaches [5–7]. SIH most commonly affects adults, with the highest incidence typically observed between 40 and 50 years of age. While SIH is rare in children and older adults, cases have been reported across the age spectrum. Notably, patients with CVFs tend to present at a later age, with reported mean ages around 60 years, suggesting a distinct demographic profile compared to other leak types. There is a clear female predominance, with studies often reporting a female-to-male ratio of 2:1 or greater [8, 9]. This gender disparity might stem from differences in connective tissue characteristics or hormonal factors, though the precise causes remain unclear. While no significant geographic patterns have emerged, higher reported rates in developed regions likely reflect better access to advanced diagnostic tools like MRI, compared to areas with limited medical resources.

Pathophysiology

SIH is inherently “spontaneous,” indicating that it arises without an evident cause such as surgery or major trauma. Still, several potentially predisposing factors have been associated with its development, including minor trauma (such as repeated coughing), anatomical abnormalities (nerve root dural diverticula or spinal bony spurs), and connective tissue disorders—such as Marfan syndrome, Ehlers-Danlos syndrome, or other inherited conditions—that can compromise the strength of the spinal dura, increasing the likelihood of CSF leaks [10, 11]. SIH’s pathophysiology reflects a cascade of events: a dural alteration leads to CSF loss, which lowers intracranial pressure, prompts compensatory vascular and structural changes, and ultimately produces the syndrome’s distinctive neurological symptoms, that will be discussed in the following paragraph.

Spinal CSF leaks have been divided into 3 categories. Spinal CSF leaks are currently grouped into three types based on the anatomic site and mechanism of the leak: Type 1 includes ventral dural tears, Type 2 lateral dural tears, and Type 3 CVFs [12, 13].

Most type 1 leaks are constituted by ventral dural tears (type 1a, 96%) caused by a sharp calcified spur. While early surgical series showed that discogenic microspurs account for the majority of ventral tears in intractable SIH, more recent work demonstrates that the presence of a discogenic spur is present in up to 1/3 of the general population and are not necessarily related to the presence of a leak [14, 15]. A recent surgical series demonstrated how many “spurs” at the tear site are sharp fibrotic calcifications of the ventral dural ligament (Hofmann ligament), connected to the posterior longitudinal ligament. Such spurs were consistently attached to the posterior longitudinal ligament rather than the disc or vertebral body and easily removable at surgery [16]. Very rarely, dural tears can be dorsolateral (type 1b) [17].

Type 2 includes lateral dural tears. Contrary to older interpretations (“ruptured/leaking diverticula”), contemporary radiologic-surgical correlation studies show that type 2 leaks are true lateral dural tears, typically at the axilla or shoulder of the nerve-root sleeve, less commonly at the pedicle. These defects often show arachnoid herniation through the tear, mimicking a diverticulum, with CSF escaping at the margins of the dural defect [18, 19]. Type 3 are CVFs, IEare abnormal connections between the subarachnoid space and the venous system, leading to CSF leakage into the venous circulation. This lesion has been first described by Schievink in 2014 and is increasingly recognized as a cause of SIH, accounting for more than 25% of total cases. Furthermore, it has been estimated that CVF account for more than 75% of SIH patients with normal spinal MRI findings [3, 20, 21]. The precise mechanism behind the formation of these lesions remains uncertain. CVFs typically occur along the spinal nerve roots, at sites where spinal arachnoid granulations—presumed fetal CSF resorption pathways—may persist into adulthood; many arise at levels that also harbor meningeal diverticula. This has led to the hypothesis that rupture of a persistent spinal arachnoid granulation can establish an aberrant direct CSF-to-vein channel, with diverticula (and, in select cases, post-traumatic pseudomeningoceles) facilitating fistula formation by bringing the subarachnoid space into closer proximity with paraspinal veins; higher BMI and antecedent intracranial hypertension have also been observed in a subset of patients, potentially predisposing to CVF development [22]. In addition to spontaneous mechanisms, iatrogenic CVFs have now been reported after lumbar puncture and intrathecal contrast injection for myelography [23].

CVFs have been described bilaterally, from the skull base to the sacrum, but they are most often found between T6 and L2 and appear to have a right-sided preponderance; importantly, they can occur synchronously at multiple levels and may recur de novo at new levels. The basis for multiplicity and recurrence remains poorly understood. These features support a thorough diagnostic strategy, including systematic bilateral myelography even after an index lesion is identified and a low threshold to repeat the work-up if symptoms recur post-treatment [24–29].

Clinical Presentation

Symptoms

SIH most commonly presents with symptoms that begin or significantly worsen on becoming upright and improve soon after lying flat. This key feature is secondary to downward traction of intracranial structures such as the dura, blood vessels, nerves and the brainstem. The hallmark symptom is orthostatic headache, which occurs in the majority (up to 90%) of patients [30]. Headaches usually improve within an hour of lying flat. This symptom is specific but not 100% sensitive, as a a minority of patients have non-orthostatic headaches or no headache at all. In addition, the positional nature of the headaches may lessen over time. Thus, when evaluating a patient months or years after the inception of headaches, it’s crucial to ask about the headache’s characteristics shortly after it began [31, 32].

Orthostatic headaches are also a sign of non-spontaneous CSF leaks, i.e. post-dural puncture, surgery or trauma. In these cases, the time and location of the inciting insult—and thus of the spinal CSF leak—are usually identifiable, unlike in SIH, where the leak’s origin is unknown at first presentation. For this reason, when SIH is suspected, it’s critical to ask about any prior lumbar punctures, epidural or spinal anesthesia, or spinal surgeries. These cases have a completely different pathophysiology and management and should be promptly identified in order to avoid long diagnostic pathways or unnecessary treatments [33].

Some individuals with postural tachycardia syndrome or orthostatic hypotension also experience orthostatic headaches, which may be the primary sign of orthostatic intolerance. These possibilities should be explored when imaging appears normal; an active stand test can help screen for them, with autonomic testing providing confirmation. Orthostatic headache can also be confused with motion sensitivity in migraine, and position-related headache in cervicogenic headache, so it is important to confirm that the headache relates to the upright position, rather than to motion or neck position.

Patients with SIH may also complain of a plethora of unspecific symptoms, the most common of which are nausea (50%) and neck stiffness (40%), vertigo, hearing loss, tinnitus, photophobia, fatigue, gait disturbances [2]. What differentiates patients with SIH-related symptoms from those from other causes is that the occurrence of the symptoms is related to orthostasis, with amelioration of symptoms in the supine position. The interval between becoming upright and onset of symptoms varies considerably between patients (minutes to hours) but is usually consistent for any one patient [34].

Complications

Untreated SIH has been associated with a significant number of serious complications, supporting treatment of this condition when a culprit lesion is identified even when symptoms are relatively tolerable for the patient.

Subdural hematomas are a frequent complication of SIH and have been reported in 16–57% of cases, likely related to venous engorgement and traction on bridging veins [35]. Infratentorial superficial siderosis is another clinically important debilitating long-term complication, classically presenting with progressive hearing loss, cerebellar ataxia, and myelopathy. While it is particularly associated with type 1 leaks, it has been increasingly reported in CVFs. Additional reported complications include bibrachial amyotrophy (often associated with ventral leaks and/or siderosis), acquired tonsillar descent that mimics Chiari I, and syringomyelia [36–38]. Very rarely, SIH can lead to profound brain sagging and downward displacement of critical structures like the diencephalon and brainstem resulting in impaired consciousness or even coma or death [39].

Lastly, a rare SIH phenotype presents with progressive behavioral change and executive dysfunction mimicking frontotemporal dementia. Contemporary series indicate that about up to 40% of such patients have an underlying CSF-venous fistula; when the fistula is identified and treated, cognitive deficits can be reversible, underscoring the need for targeted work-up in this subgroup [40–43]. There have also been a few reported cases of ischemic stroke and cerebral venous sinus thrombosis due to extreme forms of SIH [44, 45].

Tables 1 and 2 summarize the clinical symptoms and complications that have been associated with SIH.

Table 1.

Clinical manifestations of spontaneous intracranial hypotension

Symptom	Frequency (%)
Headache	> 95
Orthostatic	95
Non-positional	2–3
Reverse Orthostatic	2
Nausea/Vomiting	50
Disequilibrium	40
Dizziness/vertigo	20–40
Neck Pain/Stiffness	33
Hearing disturbance	33
Tinnitus	25
Fatigue	25
Facial numbness/paresthesia/pressure	15–20
Blurry vision	15–20
Cognitive disturbance	10–20
Upper back pain	10
Low back pain	4
Diplopia	1–2

Table 2.

Possible complications of untreated intracranial hypotension

Complications	Frequency (%)
Subdural hematoma/collections	16–57
Infratentorial siderosis	4
Bibrachial amyotrophy	< 1
Frontotemporal dementia like syndrome	< 1
Reduced consciousness/coma	< 1
Syringomyelia	< 1
Stroke	< 1
Cerebral venous sinus thrombosis	< 1

Diagnostic Approach

The symptoms described above represent the clinical syndrome of SIH, regardless of the underlying etiology. It is therefore essential to evaluate these patients using a structured and standardized diagnostic approach to accurately identify and classify the causative abnormality. Notably, measuring opening pressure via lumbar puncture is not a reliable method for diagnosing CSF hypotension. Intracranial pressures in these patients are known to be variable, rendering this test poorly sensitive [46, 47]. Confirmation of clinical suspicion, as well as precise characterization and localization of the underlying lesion—whether a CVF, epidural collection, or another rare type of CSF leak—requires a systematic, multimodal imaging strategy guided by the clinical presentation. The diagnostic work-up typically begins with MRI of the brain and spine, followed by targeted myelographic studies when further investigation is warranted. Figure 1 displays a proposed diagnostic algorithm for patients presenting with SIH.

Fig. 1.

Proposed diagnostic algorithm for evaluation of patients presenting with spontaneous intracranial hypotension

MRI Brain

Contrast-enhanced brain MRI should be the first investigation performed in the diagnostic algorithm for SIH, as at least one characteristic sign is present in more than 80% of patients [48]. A large meta-analysis found that of the individual signs, pachymeningeal enhancement is present in > 70% of cases, followed by venous sinus engorgement (57%), effacement of the pre-pontine cistern and reduced mammillopontine distance (“brain sagging”, 43%) and pituitary gland enlargement (38%). Subdural fluid or blood collections were concomitantly present in 43% of patients [25, 30]. These signs are not exclusive of SIH, and the differential diagnosis is broad. Thus, their interpretation needs to be contextualized in each patient’s clinical scenario (Fig. 2a–d).

Fig. 2.

Brain and spine MRI findings of spontaneous intracranial hypotension in different patients. (a Contrast-enhanced (CE) volumetric T1-weighted axial brain image showing diffuse pachymeningeal thickening, b T2-weighted coronal brain image demonstrating bilateral subdural fluid collections, c CE volumetric T1-weighted sagittal image depicting brainstem sagging with a reduced mammillopontine angle and effacement of the prepontine cistern, along with pituitary enlargement (arrow) and engorgement of the right transverse sinus (arrowhead), d T2-weighted axial image of the thoracic spine demonstrating nerve roots dura diverticula, the most frequent location of CSF-venous fistulae, e Fat-saturated T2-weighted sagittal image of the thoracic spine demonstrating a prominent posterior epidural collection in a patient with severe intracranial hypotension. Note the sharp delineation of the posterior margin of the thecal sac. This collection was caused by an anterior dural tear from a degenerative cervical osteophyte)

A quantitative assessment tool known as the SIH Bern score has been developed to assess the likelihood of a spinal CSF leak in patients suspected of having SIH and to help clinicians pre-determine appropriateness of more advanced, often invasive work-up. This score focuses on six radiological SIH-related characteristics: venous sinus distention (2 points), pachymeningeal enhancement (2 points), effacement of the suprasellar cistern (≤ 4 mm, 2 points), subdural fluid collections (1 point), effacement of the pre-pontine cistern (≤ 5 mm, 1 point), and reduced mammillopontine cistern (≤ 6.5 mm, 1 point). The score ranges from 0 to 9, stratifying patients into low (0–2), intermediate (3–4), or high (5–9) probability categories for the presence of a spinal CSF leak [20]. Table 3 lists the typical imaging findings of SIH and their associated score. The Bern Score is highly specific but less highly sensitive, as up to 20% of patients with symptomatic and confirmed leaks may exhibit minimal or no abnormalities on brain MRI, resulting in low scores. Further to that, it does not reliably correlate with the severity of clinical symptoms, such as headache intensity. Lastly, sensitivity of the Bern Score can be influenced by the timing of imaging relative to symptom onset. As the duration of a CSF leak increases, some brain MRI findings may become less conspicuous, potentially leading to lower Bern Scores despite the ongoing presence of a leak [49]. A low Bern score should thus not be considered a contraindication to further imaging investigations in patients with clinically convincing SIH syndrome. While the original validation of the Bern score was based exclusively on patients exhibiting high-flow leaks characterized by the presence of a spinal longitudinal extradural CSF collection (SLEC), a recent systematic review corroborated its high negative predictive value in CVFs [50]. Recently, a report showed that in patients with orthostatic headache and normal brain and spine MRI an optic nerve sheath diameter < 4.4 mm or perioptic subarachnoid space < 1.0 mm were associated with ~50% likelihood of detecting a CVF on digital subtraction myelography (DSM) versus < 2% above these cutoffs. These measurements increase after surgical ligation, supporting physiologic basis of this observation [51]. This new observation can be a useful adjunct to the preoperative evaluation of a patient in cases of borderline clinical syndromes. Deep learning models for determining the likelihood of a CVF based on brain MRI findings are currently under development [52].

Table 3.

Imaging findings of spontaneous intracranial hypotension

Finding	Frequency (%)	Bern Score Points
Pachymeningeal thickening/enhancement	70–80	2
Venous sinus engorgement	55–65	2
Subdural collections	35–50	1
Pituitary enlargement	40–60	–
Brain sag	30–40	–
Narrow suprasellar cistern (< 4.0 mm)	–	2
Narrow prepontine cistern (< 5.0 mm)	–	1
Reduced mammilopontine distance (< 6.5 mm)	–	1

MRI Spine

Whole-spine MRI without contrast with heavily T2-weighted myelographic sequences should be performed in all patients undergoing evaluation for SIH as an essential preliminary step prior to more advanced contrast-based imaging techniques. Myelographic sequences are a critical portion of the exam, as its primary purpose is to assess the presence of a SLEC—identifiable as a hyperdense ventral and/or dorsal fluid collection along the thecal sac on T2-weighted images (Fig. 2e). The presence of a SLEC indicates active CSF extravasation into the epidural space, most commonly due to types 1a or 2 leaks. Very rarely, a SLEC can be caused by a type 1b (dorsal) leak or by a sacral dural tear (present in up to 6% of SLECs [17, 53]). Dural tears are the most common cause of SIH, and a SLEC is seen on spinal MRI in ~70% of cases. When no extradural collection is present, a CVF is more likely. Thus, SLEC status guides the workup: if present, the focus should be on localizing a dural tear; if absent, on investigating for CVF [19, 54].

Other MRI spine findings can provide supportive clues. As discussed earlier, CVFs are frequently found in anatomical continuity with nerve root sleeve diverticula, with most fistulas—up to 82% in some studies—originating directly from these outpouchings. This strong association underscores the importance of thoroughly evaluating nerve root sleeve diverticula in patients with SIH who do not exhibit an extradural fluid collection. However, it is equally important to recognize that these diverticula can be incidental findings. They are commonly observed in individuals without any evidence of a CSF leak. Therefore, in the absence of direct visualization of a CVF, the presence of a nerve root sleeve diverticulum alone should not be taken as diagnostic of SIH or considered sufficient to infer the existence of a fistula [55, 56]. Spinal MRI may also demonstrate prominent epidural venous plexuses, reflecting venous engorgement secondary to low CSF pressure—though this is a nonspecific finding. Overall, spinal MRI plays a pivotal role in distinguishing between high-flow, overt leaks and subtle or “occult” leaks. In the latter group—characterized by the absence of a SLEC—further evaluation with specialized myelographic techniques is required to identify a CVF.

CT Myelography

Conventional CT myelography is no longer part of the standard workup of CSF leaks. Its main utility (showing a SLEC) can now be achieved noninvasively with heavily T2-weighted MRI myelography. In cases where a SLEC is identified on MR myelography, DSM (see below) plays a key role in identifying the level of leakage. A CT myelogram including thin-section sharp reconstruction images obtained immediately after DSM can help identify potentially causative bone spicules and correlate their location with the SLEC. For type 2 leaks, lateral-decubitus myelographic studies provide final localization. In CVFs, conventional CTM is usually unrevealing; the “hyperdense paraspinal vein” sign may occasionally be seen but is insensitive and increasingly obsolete compared with modern lateral-decubitus DSM/dynamic techniques [57, 58].

Digital Subtraction Myelography

Prone DSM plays a key role in accurately confirming and localizing dural tears in patients with SLECs by dynamically visualizing the active leakage of intrathecal contrast into the epidural space. This capability is especially beneficial for precisely identifying the spinal level of subtle or intermittent dural defects, thereby facilitating targeted dural repair. Prone DSM is typically not employed for diagnosing or characterizing lateral dural leaks or CVFs [54, 57], though lateral dural leaks may be identified on post-DSM CT myelography.

Lateral Decubitus Imaging Studies

Lateral decubitus myelographic studies such as DSM and CTM are now the established gold standard for visualization of lateral CSF leaks and CVFs. In these procedures, the patient is placed in a lateral decubitus position with the right or left side down and the hips raised with respect to the head to create a gravity gradient in a caudocranial direction. The patient is maintained in this position during and after intrathecal contrast injection. This allows contrast to distribute only on one side of the thecal sac, promoting preferential filling of nerve root sleeves and dural ectasias, increasing the chance of visualization of contrast escape into the extradural space or in a venous structure by preferential gravitational outflow on the dependent side [59]. To enhance conspicuity of CVFs during lateral-decubitus myelography, many authors suggest to pre-pressurize the thecal sac with preservative-free saline before contrast instillation. Transiently increasing the subarachnoid-venous pressure gradient can augment contrast egress through small fistulas and convert equivocal or occult studies to being diagnostic; a recent multi-institutional series reported definitive visualization of subtle CVFs only after saline pressurization, supporting this simple adjunct in suspected fistulas [60].

Because intrathecal contrast is cleared into the vascular system over time, bilateral examinations are typically performed on separate days with at least a 24-hour interval to allow near-complete clearance. Some authors have proposed a single-day protocol: administer a first intrathecal bolus and image in one lateral decubitus position, then reposition to the opposite decubitus and repeat after a second bolus. While this approach can expedite detection of type 2 leaks, its diagnostic accuracy for subtle CVFs is not yet established [61]. Because intrathecal contrast is cleared slowly from the CSF, renal collecting-system opacification on delayed CT has been proposed as an indirect marker of early CSF-to-venous shunting. Renal contrast excretion at approximately 60 min after intrathecal administration has high specificity but low sensitivity for the presence of CVF and may warrant greater attention to subtle findings or prompt repeat myelography [62–64]. Given logistical constraints and the lack of an established threshold distinguishing controls from SIH, routine delayed upper-abdominal CT is not recommended. With appropriate workflow and dose management, a delayed (~45–60-minute) renal CT may be a useful adjunct in SLEC-negative, high-clinical suspicion cases. Regardless of the modality used, small CVFs may be present but not detectable, prompting repeated examinations in cases of very convincing clinical syndromes or of early opacification of the renal pelvises. Missed visualization of the lesion may be secondary to limitations in spatial resolutions that are inherent to any modality and to the potential intermittent nature of the CSF-venous flow. In a recent study which investigated the presence of β‑trace protein in venous blood collected at the fistula site at the time of endovascular embolization, it was found that protein levels were higher in most but not all patients, suggesting a non-continuous flow of CSF towards the bloodstream [65]. Therefore, diagnostic persistence needs to be tailored on each patient’s clinical and imaging scenario.

Lateral Decubitus Digital Subtraction Myelography

Lateral decubitus DSM has been the first modality widely utilized for visualizing lateral CSF leaks and in particular CVFs [57, 66]. The exam is entirely performed in an angiography suite with the patient secured on a tilting table in lateral decubitus. After gaining intrathecal access via a fluoroscopically guided lumbar puncture, contrast is injected, and rapid-sequence fluoroscopic images are obtained and digitally subtracted to highlight any amount of contrast laterally escaping the thecal sac. The exam is dynamic, with the operator injecting contrast sequentially, maneuvering the table and adjusting its tilting to image the different spinal segments. The lateral decubitus modification of DSM has dramatically improved the detection of CSF-venous fistulas and is now the most studied technique when a CVF is suspected [21, 54].

On lateral decubitus DSM, the hallmark finding for a CVF is the early appearance of contrast in a paraspinal vein with no significant epidural pooling, visualized as a serpentine contrast-opacified vein emanating from a nerve root shortly after intrathecal injection, indicating direct CSF-venous shunting. The exact level is then inferred from bony landmarks on fluoroscopy, but post-DSM cone beam CT can be performed on the table to clarify the precise anatomy. Delayed cone-beam CT images can also be obtained to ascertain the presence of contrast in the renal collecting system [59, 67]. A recent study reporting the temporal characteristics of CVFs using DSM found that on average, the fistula appeared 9.1 s (range, 0–30 s) after intrathecal contrast reached the spinal level. Imaging protocols and techniques should be optimized to ensure that image acquisition captures the moment of contrast passage through the fistula [26]. The advantages of DSM are its high spatial and temporal resolution and specificity. The limitations are that it is a test that relies significantly on operator experience and requires an angiography table that can be tilted. Also, accuracy of subtraction imaging in the cervicothoracic regions may be reduced by respiration motion artifacts and superimposition of different osseous structures (shoulders, arms) making the exam less sensitive in these segments.

Lateral Decubitus CT Myelography

Lateral decubitus CTM has emerged as a high-yield technique for detecting CVFs and type 2 leaks. The entire procedure can be performed in the CT scanner without the patient changing decubitus and preferably utilizing CT fluoroscopy to streamline the process. After gaining intrathecal access under CT guidance, contrast is instilled, and a helical CT scan of the entire spine is performed. The acquisition is sometimes repeated with apnea or resisted inspiration to a) obtain images with different intra-abdominal to intrathecal pressure gradients and b) to visualize the dynamic behavior of contrast at different time points. Resisted inspiration reportedly yields a higher fistula detection rate [68]. CTM has reached detection rates comparable or in some cases superior to those of DSM and the exam can be completed by delayed acquisition for the renal collecting system as described above [69, 70]. The advantages of lateral decubitus CTM are that it uses widely available CT technology and provides high spatial resolution to pinpoint the exact vertebral level and anatomy of the fistula, including an outline of the tributary vein and its three-dimensional course in large volume lesions. It is less technically demanding and operator dependent than real-time DSM. The disadvantages include significant exposure to ionizing radiation, and potential misdiagnosing foci of hyperdensity representing small areas of attenuation noise as subtle CVFs. The exam also requires a significant utilization of CT time, especially if delayed kidney imaging is performed, which may limit the utilization of the technique in some contexts. Furthermore, timing is crucial—a very small or momentary fistula might still be missed if contrast does not escape during the scanning window [71]. Figure 3 depicts different types of CVFs and of their drainage patterns on CTM.

Fig. 3.

Different types of CSF-venous fistulae (CVFs) and drainage patterns visualized on CT myelography (CTM). (a, b Sagittal and coronal reconstructions of a left lateral decubitus CTM showing a prominent CVF opacifying the left T9 intersegmental vein (arrows), draining into the hemiazygos vein, c, d Axial and sagittal images from the same CTM demonstrating a punctate area of enhancement in the left T4–T5 periforaminal space (arrow). This finding was absent in a subsequent CTM performed 3 min later (not shown), suggesting a CVF at this level, e, f Right lateral decubitus CTM axial images showing a high-flow CVF draining directly into the basivertebral plexus, with intraosseous venous outflow into the right T7 intersegmental vein, g Left lateral decubitus CTM axial image depicting a thin, posteriorly directed CVF at the level of a dural ectasia involving the left L1 nerve root. This drainage pattern is relatively uncommon, as most CVFs drain anteriorly, h Right lateral decubitus CTM axial image demonstrating a linear area of enhancement in the T1–T2 periforaminal space, consistent with a high thoracic CVF)

These limitations are less relevant when CTM is performed on photon-counting CT, which provides superior spatial resolution and improved contrast-to-noise ratio, allowing detailed visualization of tiny paraspinal veins and subtle fistulous connections that may be missed with conventional CT methods. Preliminary findings suggest that photon counting CT demonstrates superior sensitivity and equal specificity compared to traditional CTM, particularly in challenging cases with very small fistulas adjacent to bony or degenerative structures or to prior embolic material in recurrent or de novo cases. The technique is also promising in the accurate delineation of dural tears, possibly waiving the need for localization by DSM. Utilization of this technique is currently limited by the scarce availability of photon counting CT machines [72–74].

Treatment Options for CVFs

Conservative Therapies

Conservative approaches, including Trendelenburg position, hydration, caffeine and abdominal binders typically play a supportive role in managing CVFs. They are particularly useful in stabilizing patients with severe symptoms that are too unstable to undergo necessary diagnostic tests [39].

Epidural Blood Patch

Current proposed guidelines suggest performing epidural blood patching (EBP) in all patients with SIH before advanced imaging studies, based on the rationale that early intervention may resolve symptoms without the need for extensive diagnostic evaluation [75]. This technique is often successful for SIH from small dural tears, as in post-lumbar puncture CSF leaks, but demonstrates limited efficacy for CVFs due to the nature of their pathology, which bypasses epidural space accumulation, making the traditional sealing mechanism of EBP less effective [76]. A notable variability exists in clinical practice regarding the timing and utilization of EBP in managing SIH. In the authors’ opinion, empirical EBPs should not precede lateral decubitus DSM or CTM in cases with high suspicion for a CVF, as these advanced imaging modalities provide critical diagnostic clarity and a treatment target, especially when conventional imaging fails to identify the leak source [48, 54, 77, 78]. EBP may be considered a first-line treatment when clinical presentation is suggestive of SIH but imaging findings are inconclusive. A positive clinical response can support the diagnosis of a CSF pressure-related disorder and justify further investigation or repeat imaging.

Targeted Transforaminal Patch with Fibrin Glue

Targeted transforaminal patching with fibrin glue (with or without added blood) has been used for a long time as an intervention for CVFs, particularly as an alternative to surgery before endovascular embolization became an alternative modality of treatment. Unlike standard EBPs, which rely on the spread of blood within the epidural space to seal dural defects, this targeted approach involves the precise delivery of fibrin glue directly to the site of the fistula. This is supposed to obliterate the lesion by either exerting extrinsic compression and scarification on the draining vein or by inducing intrinsic thrombosis within it [79].

The procedure utilizes CT guidance to ensure accurate needle placement, targeting specific anatomical landmarks such as the cyst-vein junction, paravertebral wall, or directly into a meningeal diverticulum. In some cases, multiple needles may be employed to enhance the likelihood of successful occlusion, especially when the fistulous tract is complex or extensive. Prior to the injection of fibrin glue, small amounts of air or blood diluted with contrast material may be injected to confirm proper needle positioning and anticipated spread of the adhesive. Post-procedural imaging is then performed to verify the distribution of the fibrin glue and assess the immediate outcome of the intervention (Fig. 4). Some authors propose direct injection of fibrin glue within the dural root diverticulum that gives rise to the CVF. A few targeted clinical studies have demonstrated favorable outcomes with this technique. The largest multi-institutional study reported complete clinical improvement in approximately 60% of patients and partial improvement in an additional 34.5% following CT-guided fibrin glue occlusion of CVFs. Interestingly, the authors found that the effectiveness of the intervention may depend heavily on whether the injected material followed the specific venous drainage route of the fistula. In cases where the glue spread through the epidural space—similar to how it would in a traditional transforaminal steroid injection—success was more likely only if the fistula drained into the internal epidural venous plexus. If this was not the case, as in most CVFs, which drain into paraspinal veins, then a more tailored approach was recommended, with needle placement adjusted to access the fistula’s actual anatomical course—whether lateral, anterior, or posterior to the foramen. Interestingly, in this study duration of symptoms correlated inversely with treatment efficacy. Complications were relatively rare; however, beyond transient radicular pain or asymptomatic pulmonary embolism, epidural blood patch with fibrin glue carries a small but clinically important risk of hypersensitivity up to anaphylaxis (0.5% per procedure; 0.95% per patient), typically after re-exposure, with higher odds reported in younger patients, those with lower BMI, and when inadvertent intravenous injection occurs [80–82].

Fig. 4.

Axial CT images from a targeted periforaminal blood patch with fibrin glue injection for treatment of a CSF-venous fistula at the right T11 nerve root. Contrast was injected intrathecally to outline the subarachnoid space. (a Needle positioning in the right T11 periforaminal space via a lateral approach. A small dural root ectasia—corresponding to the known fistula site—is noted, b Post-blood patch image showing obliteration of the ectasia and spread of blood and fibrin glue into the right posterior epidural space, with partial effacement of the adjacent thecal sac)

Endovascular Embolization

The advent of transvenous embolization using liquid embolic agents like Onyx (an ethylene-vinyl non-adhesive alcohol copolymer) marked a paradigm shift in the treatment of CVFs. First described in 2021, this technique involves catheterizing the draining vein, or the draining portion of the epidural plexus and injecting the embolic material to occlude the fistulous connection (Fig. 5). Numerous studies have demonstrated excellent clinical outcomes [83].

Fig. 5.

Images from the endovascular embolization of the left T9 CSF-venous fistula (CVF) shown in Fig. 2a, b. (a Digital subtraction angiography (DSA) of the azygos system highlighting inflow from multiple intersegmental veins (ISVs), with the left T9 ISV (arrow) identified as the target, b DSA of the hemiazygos vein obtained via access through a lumbar vein originating from the inferior vena cava. Due to the steep cranial orientation of the left T9 ISV, direct catheterization from above was unfeasible, necessitating a caudal approach through anastomotic venous connections, c Overlap of the DSA from panel a with a single-shot acquisition showing the microcatheter advanced into the left T9 ISV via the caudal approach (arrow), d, e Anteroposterior and lateral single-shot images following Onyx embolization of the left T9 ISV. Embolic material is visible in the epidural and basivertebral venous plexi at the target level, extending cranially and caudally. Onyx is also seen in the venous collectors surrounding the T9 nerve root dural ectasia (arrows))

The largest studies to date, reporting on 100 patients who underwent endovascular embolization for CVFs reported 95% of patients experiencing either complete symptom resolution or significant improvement at a mean follow-up of 15 months. Of 96 patients presenting with headache, 75% achieved full relief, while many also saw improvements in tinnitus, cognitive dysfunction, and other neurologic symptoms. These results are comparable to those obtained with fibrin glue patching [80]. Complete resolution was most observed in patients who were male, had no prior history of headaches, and presented with typical SIH imaging features such as pachymeningeal enhancement or venous sinus engorgement. Again, those with a shorter duration of symptoms prior to treatment were significantly more likely to fully recover. Conversely, partial responses were associated with longer symptom duration, female sex, and a history of primary headache disorders. In this series, no significant neurological complications were reported, with the most described adverse event being back pain (in as many as one in three patients) and, in one case, radiculopathy that caused paresthesias for one year. Paraspinal venous perforations and small Onyx pulmonary emboli were reported without clinical sequelae. Dimethyl sulfoxide injection in paraspinal veins can be painful for the awake patient thus, while some groups perform the procedure under conscious sedation, we recommend use of general anesthesia. This may limit the feasibility of this intervention in some settings, where percutaneous fibrin glue patching may be the preferred approach. This type of intervention also poses the risks of high radiation doses, as do other spinal vascular interventions [84].

Recurrence of symptoms did occur in 17% of patients, primarily due to the formation of new fistulas. Many of these new fistulas developed within two vertebral levels of the initial lesion. This may be explained by at least two theories: a) these fistulae were present but initially undetected; b) increased intrathecal pressure following successful treatment of the first CVF may create a de novo CVF in patients with some form of decreased collagen resistance. Nonetheless, retreatment using embolization or, in rare cases, surgery, was generally successful. When compared to surgical series, embolization appears to offer comparable, if not superior, rates of complete symptom resolution, with a more favorable safety profile. These findings support transvenous embolization as a durable and generalizable treatment strategy for CVFs, while also highlighting the importance of early diagnosis and careful post-treatment monitoring for rebound intracranial hypertension (RIH) and recurrence [85].

Early post-treatment effects can be observed on brain MRI within 24 h of embolization [86]. In one study, the median Bern score decreased significantly from 6 pre-treatment to 3.5 at 24 h and further to 2 at 3 months. These positive changes correlated with clinical amelioration. SIH-specific MRI findings also showed varying rates of improvement: pachymeningeal enhancement regressed in over 50% of patients, venous sinus engorgement in approximately 40%, and other MRI findings in about 30% at 24 h post-embolization [73].

Surgical Ligation

Surgical treatment is an established treatment for type 1 and type 2 CSF leaks. Similarly, surgical ligation remains a definitive treatment option for type 3 leak, CVF, particularly useful for fistulas resistant to embolization or anatomically unsuitable for endovascular approaches [76, 88]. Surgical intervention and goals of treatment vary depending on the type of CSF leak encountered. For type 1 ventral CSF leak, a midline incision with laminectomy is most performed, with repair undertaken in a transdural approach to patch the defect. This may be done with synthetic surgical adjuncts such as Duragen or Tachosil. Autologous muscle graft is also frequently employed which leads to mechanical plugging of the dural defect, which also promotes a local inflammatory response to help aid in the permanent obliteration of the dural defect. Variations to this approach are minimally invasive hemilaminotomy and paramedian transdural approach for direct repair, as well as endoscopic dural repairs [89, 90].

Type 2 leaks are best treated with local reduction of the arachnoid herniation into the dural defect, with subsequent clip ligation of the nerve sleeve. This may be done in conjunction with other surgical adjuncts to promote dural closure, such as muscle graft or Duragen synthetic grafting [18].

Type 3 CVF repair typically involves targeted laminectomy or hemilaminectomy with medial facetectomy at the site of the fistula, with exposure of the affected nerve sleeve and associated vascular structures. Macroscopically, this will appear as enlarged, engorged veins surrounding a nerve sleeve. Microsurgical identification and isolation of the neurovascular structures is critical, followed by surgical ligation of the fistulous connection. This may be accomplished with surgical clip ligation coupled with cauterization and disconnection of the venous structures from the nerve sleeve. This surgical approach achieves a rate of cure, though it carries inherent surgical risks such as infection, postoperative pain, local CSF leak and pseudomeningocele, and necessitates extended recovery time as compared with less invasive treatment approaches [88, 91].

Rebound Intracranial Hypertension

RIH has emerged as a key post-treatment complication with rates between 19 and 28% of patients experiencing it to some degree, depending on the series. Occurrence rates are similar between embolization, fibrin patching and surgery. RIH more frequently affects those with partial or no symptom resolution and middle-aged women more than men (21% vs. 9%, respectively) [71, 75]. RIH typically presents as a new-onset, frontal or peri-orbital headache that worsens when lying down, often accompanied by nausea, visual disturbances, and, rarely, papilledema. These symptoms are opposite to the orthostatic headaches’ characteristic of SIH and are often immediately recognized as different by the patients [76]. Though frequently self-limiting, RIH can persist for weeks or even months in some individuals. Diagnosis is primarily clinical, as objective measures like opening pressure or imaging findings can be inconsistent, except in more extreme cases.

The pathophysiology of RIH is not completely understood, particularly why it happens inconsistently. It has been theorized that some patients are in a state of elevated craniospinal elastance, akin to what is observed in idiopathic intracranial hypertension or compensated SIH. In these conditions, compensatory mechanisms such as venous engorgement and reduced venous outflow adjust to chronic CSF hypovolemia. Once the CSF leak is sealed, these mechanisms may persist, leading to a pressure imbalance. Venous sinus stenosis is also more frequently seen in patients who develop RIH. Individual differences in venous physiology, dural compliance, and the presence of venous outflow restriction may explain why some patients develop RIH while others do not [42, 77, 78].

Treatment of RIH generally begins with conservative measures such as head elevation. Pharmacologic management with carbonic anhydrase inhibitors like acetazolamide is common and often effective in reducing CSF production and alleviating symptoms. A typical starting regimen for acetazolamide mirrors that used in idiopathic intracranial hypertension, beginning at 250 to 500 mg twice daily with the option to adjust the dose based on clinical response. The medication is generally well tolerated at doses up to 4000 mg per day. In clinical practice, Schievink and colleagues have administered 500 mg orally two to three times daily, with some patients receiving up to 3000 mg per day to control symptoms [79]. Lumbar puncture and high-volume CSF drainage are reserved for severe or refractory cases, as most RIH episodes resolve within a few weeks to three months. Anecdotally, it has been reported that RIH can lead to transverse sinus stenosis, which may need stenting to be effectively treated [personal communication]. There is emerging interest in pre-emptive strategies, such as acetazolamide pre-treatment before certain procedures in high-risk patients, although more research is needed to validate these approaches. Importantly, RIH is often considered a marker of successful CSF leak treatment rather than a complication indicating treatment failure [97].

The role of the glymphatic system—a network involved in cerebral fluid clearance—in symptoms resolution and in development of RIH remains uncertain. A small study using diffusion tensor imaging along perivascular spaces in three SIH patients with CVFs found reduced glymphatic flow pre-treatment, with improvement following embolization. Notably, patients with more substantial symptom relief also showed greater increases in glymphatic flow. While these preliminary observations suggest a potential relationship between glymphatic function and SIH recovery, further research is needed to clarify the relevance of glymphatic disruption in the disease’s pathophysiology [98]. Despite the current reliance on empiric and symptomatic treatment, understanding of RIH remains incomplete, and RIH remains an evolving area within the broader context of CSF leak diagnosis and management.

Conclusions

Over the past decade, the field of CSF leak diagnosis and treatment—particularly in the context of CVFs—has advanced substantially, driven by both clinical insight and technological innovation. Improved imaging techniques, including lateral decubitus CTM and DSM, have dramatically increased diagnostic sensitivity, while minimally invasive therapies such as transvenous embolization have reshaped the treatment landscape, offering effective and durable outcomes for many patients. Despite these gains, critical questions remain unanswered. The precise etiopathogenesis of CVFs is still not fully understood, and the optimal diagnostic and therapeutic algorithms have yet to be clearly defined or standardized. Several clinical challenges also continue to complicate care. These include the management of RIH following fistula closure, strategies for treating patients with multiple concurrent fistulas, and approaches for those with a high clinical suspicion for CVFs but negative imaging studies. As our ability to identify and treat these complex cases improves, so too must our frameworks for diagnosis, treatment selection, and long-term follow-up. Continued research into pathophysiology of CSF dynamics and of the venoglymphatic system, prospective validation of diagnostic protocols, and the development of tailored treatment strategies will be essential in moving the field toward more standardized, effective, and patient-centered care.

Abbreviations

CSF

cerebrospinal fluid;

CTM

CT myelography

CVF

CSF-venous fistula

DSM

digital subtraction myelography

EBP

epidural blood patch

RIH

rebound intracranial hypertension

SIH

spontaneous intracranial hypotension

SLEC

spinal longitudinal extradural CSF collection

Conflict of interest

E. Orru, N. Patel, J. Pace and T. Krings declare that there are no competing interests.