Cerebrospinal fluid (CSF) leaks are a challenging condition characterized by the loss of CSF, leading to severe orthostatic headaches and other debilitating symptoms. Diagnosis and management require a multifaceted approach involving clinical evaluation, imaging, and various treatment modalities to improve patient outcomes and quality of life.
Key points
- •
Cerebrospinal fluid leaks significantly impair quality of life due to severe orthostatic headaches and associated symptoms.
- •
Accurate diagnosis often requires a combination of clinical assessment and advanced imaging techniques.
- •
Management includes conservative measures, epidural blood patches, and surgical interventions.
- •
Posttreatment recovery varies, with some patients experiencing rebound intracranial hypertension or persistent symptoms.
Introduction
Intracranial hypotension is a syndrome caused by a spinal cerebrospinal fluid (CSF) leak and clinically characterized by the presence of an orthostatic headache and vestibulocochlear symptoms. Leaks can be iatrogenic, for example, following an intentional or unintentional dural puncture, resulting in either an acute or a chronic postdural puncture headache, or spontaneous, resulting from a ventral or lateral dural tear or CSF–venous fistula (CVF). A standardized nomenclature has been developed to aid in discussion of spontaneous leak types: type 1 CSF leaks are caused by a dural tear located ventral to the spinal cord (type 1a) or posterior/lateral to the spinal cord (type 1b). Type 2 CSF leaks are associated with simple (type 2a) or complex (type 2b) meningeal diverticula. Type 3 CSF leaks are CVFs. Type 4 CSF leaks are of indeterminate origin. Unlike spinal CSF leaks, cranial CSF leaks typically do not clinically manifest with intracranial hypotension; this is thought to be related to the intracranial compartment being above the zero-pressure point in the craniospinal fluid column. Intracranial leaks are most commonly associated with craniofacial trauma, intracranial and sinus surgeries, as well as idiopathic intracranial hypertension (IIH), which can result in chronic thinning of the skull base. Cranial leaks are often accompanied by CSF rhinorrhea, otorrhea, headaches (less orthostatic in nature), auditory and vestibular symptoms as well as meningitis. Interestingly, cases of patients with IIH and then subsequently developing a spinal CSF leak have been described, and IIH has been implicated as a potential pathogenic mechanism in CVF development. Thus, a spinal CSF leak should be considered in patients with clinically or radiographically suspected IIH who develop a sudden change in character to their symptoms.
Cause and epidemiology
Spontaneous spinal CSF leaks often have no identifiable cause but may be associated with connective tissue disorders such as Marfan syndrome or Ehlers-Danlos syndrome. The incidence of spontaneous CSF leaks is estimated to be between 3.8 and 5 per 100,000 per year. ,
Secondary leaks can result from spinal surgeries, trauma, or lumbar punctures (intended or unintended). The incidence of secondary CSF leaks is not clear but posttraumatic and postsurgical leaks are generally more easily identified by treating providers. It is estimated that approximately 500,000 lumbar spine surgeries are done each year in the United States, while a 9% incidence of dural tear in lumbar surgery for a common back problem has been reported. , Lumbar surgery alone, thus, could cause 45,000 spinal leaks per year. Most of these leaks, however, are successfully repaired at the index surgery and do not become chronic leak problems.
Dural punctures are more interesting to consider, and many patients do not recall a history that would indicate this risk without specific questioning.
It is estimated that more than 10 million epidural steroid injections are done yearly in the United States, with inadvertent dural puncture complicating this procedure in up to 2.7% of cases, equating to 270,000 cases yearly. When considering epidural injections done for childbirth, this adds another 2.8 million procedures annually with a reported inadvertent dural puncture rate of 0.7% to 1.5% of which 60% to 80% suffer acute postdural puncture headache and 30% report new-onset chronic postpuncture headache. This adds another tens of thousands of potential leak cases. Four hundred thousand diagnostic (intended) lumbar punctures are also performed annually, with up to one-third suffering from postdural puncture headaches.
It is unclear what the incidence of symptomatic intracranial hypotension resulting from all these dural punctures is, but the potential numbers are staggering particularly when considering that symptoms can fluctuate or onset in a delayed fashion, making them hard to relate back to the dural puncture itself. For instance, a case report was published recently, detailing symptomatic intracranial hypotension associated with an arachnoid bleb caused by an inadvertent dural puncture from labor epidural 12 years before symptom onset. Another 6 years transpired before diagnosis.
Clinical presentation
Patients typically present with orthostatic headaches, which are exacerbated by standing and alleviated by lying down. However, it is not uncommon for patients to lose the clear postural component to their headaches over time. Other common symptoms may include neck pain, tinnitus, hearing loss, nausea, diplopia or blurry vision, and cognitive disturbances, which patients frequently describe as “brain fog.” The variability in symptoms often leads to misdiagnosis or delayed diagnosis.
Pathogenesis
Intracranial hypotension has classically been described as low pressure due to an imbalance in production versus absorption of CSF. On a practical level this is almost always related to a leak or loss of CSF rather than inadequate production. Low intracranial pressure causes the brain to sag with resultant traction on pain-sensitive structures, leading to typical postural headaches and related symptoms. However, many reported cases of symptomatic intracranial hypotension actually have “normal” opening pressures when measured by lumbar puncture, leading some to prefer the term “intracranial hypovolemia.” , In a recent meta-analysis only 67% of patients with spontaneous intracranial hypotension (SIH) had low opening pressures, whereas 32% fell in the normal range and 3% reportedly had high pressure. In addition, several studies have demonstrated that most of the patients with CVF have normal or elevated opening pressures. The exact location of the leak can be challenging to identify, complicating diagnosis and treatment. Diagnosis requires a high index of suspicion and involves a variety of imaging studies.
Diagnostic approach
In our CSF leak program at the University of Colorado the following algorithm has been adopted ( Fig. 1 ). We have previously described this in detail elsewhere and have made slight modifications, adding our clinical workflow. A specific “CSF leak” clinic has been established where a detailed history is obtained, which includes recording all headache and other associated symptoms and documenting all potential risk factors (eg, any history of needle punctures/injections).

If clinical history raises concern for SIH, brain MRI with/without contrast and total spine MRI, “leak protocol,” are first obtained, and a visit in our CSF leak clinic is arranged. If no confirmatory brain or spine findings are identified but high clinical suspicion for SIH remains, the case is discussed in our weekly multidisciplinary conference attended by neuroradiology, neurosurgery, neurology, and other partners (eg, neuro-ophthalmology). Based on consensus, the patient is considered for empirical patching versus dynamic computed tomography myelography (dCTM). If brain imaging shows concerns for SIH but no epidural fluid collection or bleb is identified in the spine, dCTM is planned with particular attention to leaking cysts versus CVFs. When epidural fluid is identified on total spine MRI, dCTM is completed with patient positioning based on expected leak site. Treatment is generally implemented from least to most invasive, reserving surgery for cases not responding to patching techniques. When possible, patching procedures are offered following the diagnostic study. Further discussion of the imaging modalities follows.
Magnetic Resonance Imaging
MRI of the brain, especially with intravenous contrast, is a critical tool in diagnosing CSF leaks. Notable findings include diffuse pachymeningeal enhancement, subdural collections, pituitary engorgement, dural venous sinus engorgement, and sagging of the posterior fossa structures. Infratentorial superficial siderosis is an important finding to recognize and is usually associated with chronic SIH ( Fig. 2 ). Additional less common MRI findings in SIH include dural venous sinus thrombosis, thickening of the calvarium, subarachnoid hemorrhage, and spinal pial vascular engorgement.

The Bern score is a probabilistic scoring system that assigns point values to the brain MRI findings of SIH and can be used to estimate the yield of localizing a leak or CVF on subsequent myelography. MRI images depicting this scoring system are shown in Fig. 3 A–D . Scores of 0 to 2 represent a low probability, scores of 3 to 4 an intermediate probability, whereas scores of 5 to 9 represent a high probability. Interestingly, the severity of MRI brain findings does not seem to correlate with symptom severity. Some proportion of patients, estimated up to 20% in one meta-analysis, may have a normal MRI of the brain concurrent with a spinal CSF leak. Importantly, brain MRI findings may normalize over time, despite the persistence of a leak and symptomatology ; this highlights the importance of prompt neuroimaging close to symptom onset.

Although downward cerebellar tonsillar displacement is not part of the Bern score, it is important to differentiate this critical finding of SIH from tonsillar ectopia seen in Chiari I. Although both conditions may present with displacement of the cerebellar tonsils inferior to the foramen magnum, in SIH this is due to downward herniation of the brain due to loss of CSF, whereas in Chiari 1, this is because the skull surrounding the posterior fossa is abnormally small, causing displacement of the tonsils, without brain sag. Fig. 4 A, B shows MRI scans of a patient with SIH before and after surgery for her CVF, showing the dramatic improvement in cerebellar and brainstem sagging.

Once a spinal CSF leak is suspected by either clinical presentation and/or brain MRI findings, MRI of the spine should be performed to assess for an epidural fluid collection. Most type 1 and type 2 leaks will present with a fluid collection, whereas CVFs will not. An MRI protocol consisting of T2-weighted, fat-saturated, three-dimensional (3D) imaging can aid in the localization and characterization of spinal epidural fluid collections and has shown noninferiority to conventional CT myelography (cCTM) for this purpose. As a result, we do not perform cCTM at our institution in the evaluation of SIH.
Myelography
cCTM, whereby iodinated contrast is injected into the CSF via a lumbar puncture, allowed to diffuse evenly throughout the spinal subarachnoid space, and imaged after a substantial delay, lacks the temporal resolution to precisely localize the site of a spinal CSF leak, as by the time a patient is imaged, contrast is diffused throughout any epidural collection, obscuring the exact site of the dural defect. Further, CVFs eluded detection until their first description in 2014 because cCTM does not typically identify CVF. To achieve this localization, dynamic myelography, performed either fluoroscopically using digital subtraction (DSM) or under CT using a dynamic technique (dCTM), must be performed. DSM benefits from superb temporal resolution, high spatial resolution, and a slightly lower radiation dose, but suffers from motion artifacts, often requiring general endotracheal anesthesia to minimize respiratory motion, superimposition artifacts, particularly at the cervicothoracic junction, and can only study one side of the spine per session. dCTM on the other hand can achieve very good temporal resolution when multiple scans are performed in succession, can achieve high spatial resolution when submillimeter imaging slices are used, and can study the entire spine in 1 day, including multiple injections/contrast runs in a single session, but suffers from a higher relative radiation dose. Whether or not DSM or dCTM is used depends on institutional resources and available expertise. One study evaluating the performance of DSM versus dCTM in the localization of CVF found a slightly higher yield of dCTM.
The technique for dynamic myelography differs depending on the type of leak suspected. When a ventral dural defect is suspected, the patient is placed in the prone Trendelenburg position, and rapid imaging of the entire spine is performed immediately following contrast injection. When a lateral dural defect is suspected, the patient is placed in the lateral decubitus Trendelenburg position, and rapid imaging of the entire spine is performed immediately following contrast injection. When a CVF is suspected, the patient is placed in the Trendelenburg lateral decubitus position, and injection is performed with the goal of filling all nerve root sleeves and meningeal diverticula with dense contrast, as most CVFs arise from the nerve root sleeve complex ( Fig. 5 A–C ). Additional maneuvers such as saline pressure augmentation, resisted inspiration, and multiphase imaging have been described as adjunct maneuvers to increase the yield of CVF detection.

MRI myelography, whereby a small amount of gadolinium-based contrast is injected and imaged after a delay with multiplanar T1 fat-saturated imaging, is an off-label use of gadolinium-based contrast but has been described as an adjunctive tool to localize both slow CSF leaks and CVFs. , This can be concurrently performed during dCTM or DSM, with dynamic imaging initially acquired and with subsequent transfer of the patient to the MRI suite for delayed gadolinium myelography.
Management
Management of CSF leaks includes conservative measures, epidural blood/fibrin patches, and surgical interventions.
Conservative Management
Initial management often includes bed rest, increased fluid intake, and caffeine to promote CSF production and alleviate headaches. Patients are advised to avoid activities that increase intrathoracic pressure, such as heavy lifting and straining. These measures can provide symptomatic relief while awaiting spontaneous resolution of the leak.
Epidural Blood and/or Fibrin Glue Patching
Epidural blood patching (EBP) involves injecting the patient’s blood into the epidural space in effort to promote clot formation and sealing of the dural tear. This can be done empirically without exact localization of the leak site or in a targeted fashion at the level of an identified leak. Targeted patches often use fibrin glue rather than or in addition to autologous blood. Strict activity restrictions are often maintained for several days to weeks after these procedures. Patching has variable success rates and may need to be repeated. Studies have reported efficacy of EBPs in 36% to 90% of cases, depending on the underlying cause and location of the leak.
Surgical Intervention
Surgery is indicated when conservative and interventional measures fail. Techniques include direct dural repair, placement of patches, and ligation of CVFs. Surgical outcomes are generally favorable, but long-term follow-up is essential to monitor for recurrence.
Surgical treatment of ventral dural defects first used various extradural approaches. These techniques were used for degenerative conditions such as herniated discs in the cervical and thoracic spines. In the cervical spine exposure required discectomy and more often corpectomy to gain enough working space for localization and repair. In 1998 Vishteh and colleagues, reported a case of a ventral leak at C5-6, which was repaired using a standard anterior cervical discectomy and fusion approach. In the thoracic region, classically thoracotomy followed by corpectomy or an alternative approach such as costotransversectomy was required. The premise behind these extensive anterior approaches was that potentially harmful manipulation of the spinal cord could be avoided. Unfortunately, these procedures were attended by high complication rates and disappointing success rates.
Over the past 10 to 13 years posterior transdural approaches have become standard. , Laminectomy is carried out at the appropriate level as determined by intraoperative fluoroscopy. As most of these defects are in the thoracic spine it is often helpful to have a fiducial marker placed by the radiologist at the time of the localizing diagnostic study ( Fig. 6 E ). The operative microscope is used, and the dura is opened dorsally and retracted with stay sutures. The dentate ligament is divided, and the spinal cord is gently mobilized to identify the leak site. Dural defects are identified as oval slits measuring 5 to 8 mm in length, with calcified microspurs frequently identified protruding into the opening. The spur is removed with micro curettes or rongeurs. The dura can then be repaired in several ways. When possible, primary suture repair is accomplished. More often, patching techniques are used, given the difficulty in suturing in close proximity to the spinal cord. Patches can be autologous fat or muscle graft harvested from the same incision or dural substitutes. Patch grafts are supplemented with fibrin glue or other dural sealants. The dorsal durotomy is then closed in standard fashion.
