Spontaneous Low Pressure, Low CSF Volume Headaches: Spontaneous CSF Leaks


  • Conflict of Interest: No conflict.
  • Sources of Financial Support: None.

Address all correspondence to B. Mokri, Department of Neurology, Mayo Clinic, Neurology, 200 First Street SW, Rochester, MN 55905, USA.


Spontaneous intracranial hypotension typically results from spontaneous cerebrospinal fluid (CSF) leak, often at spine level and only rarely from skull base. Once considered rare, it is now diagnosed far more commonly than before and is recognized as an important cause of headaches. CSF leak leads to loss of CSF volume. Considering that the skull is a rigid noncollapsible container, loss of CSF volume is typically compensated by subdural fluid collections and by increase in intracranial venous blood which, in turn, causes pachymeningeal thickening, enlarged pituitary, and engorgement of cerebral venous sinuses on magnetic resonance imaging (MRI). Another consequence of CSF hypovolemia is sinking of the brain, with descent of the cerebellar tonsils and brainstem as well as crowding of the posterior fossa noted on head MRI. The clinical consequences of these changes include headaches that are often but not always orthostatic, nausea, occasional emesis, neck and interscapular pain, cochleovestibular manifestations, cranial nerve palsies, and several other manifestations attributed to pressure upon or stretching of the cranial nerves or brain or brainstem structures. CSF lymphocytic pleocytosis or increase in CSF protein concentration is not uncommon. CSF opening pressure is often low but can be within normal limits. Stigmata of disorders of connective tissue matrix are seen in some of the patients. An epidural blood patch, once or more, targeted or distant, at one site or bilevel, has emerged as the treatment of choice for those who have failed the conservative measures. Epidural injection of fibrin glue of both blood and fibrin glue can be considered in selected cases. Surgery to stop the leak is considered when the exact site of the leak has been determined by neurodiagnostic studies and when less invasive measures have failed. Subdural hematomas sometimes complicate the CSF leaks; a rebound intracranial hypertension after successful treatment of a leak is not rare. Cerebral venous sinus thrombosis as a complication is fortunately less common, and superficial siderosis and bibrachial amyotrophy are rare. Short-term recurrences are not uncommon, and long-term recurrences are not rare.


chronic daily headache


cerebrospinal fluid


computed tomography


computed tomography myelography


epidural blood patch




gadolinium myelography




lumbar puncture


magnetic resonance imaging


opening pressure


postural orthostatic tachycardia syndrome


subdural hematoma


spontaneous intracranial hypotension

Historical Background, Terminology, and Objective

In 1891, lumbar puncture (LP) was introduced independently by Heinrich Quincke of Germany (aimed at treatment of hydrocephalus)[1] and W. Essex Wynter of England (aimed at alleviation of pressure in tuberculous meningitis).[2] In 1898, having recognized post-LP headaches, August Beir (a student of Quincke) along with his assistant, August Hildebrandt, performed experiments upon themselves and suffered post-LP headaches.[3] In 1939, Georg Schaltenbrand, a German neurologist, using the term “aliquorrhea” described spontaneous occurrence of a syndrome of orthostatic headache and a few other symptoms associated with low cerebrospinal fluid (CSF) opening pressure (OP).[4] This later came to be known as spontaneous intracranial hypotension (SIH).[5, 6] Modern neuroimaging has revolutionized our understanding of this entity. The original theory of Schaltenbrand that the disorder was due to decreased CSF production has never been substantiated. It is now recognized that almost all cases of SIH result from spontaneous CSF leaks. The overwhelming majority of these spontaneous leaks occur at the spinal level and only rarely from the skull base. In contrast, posttraumatic or postsurgical CSF leaks from the skull base (rhinorrhea, otorrhea) are not rare at all. The first report on pachymeningeal enhancement in intracranial hypotension appeared about two decades ago.[6] In this interval, additional imaging features of the disorder have been recognized, and far more patients are diagnosed than previously.[7-10] A broad clinical spectrum of the disorder has come to be recognized. SIH can no longer be simply equated with postdural puncture headaches.[11] Although the triad of orthostatic headaches, low CSF pressures, and diffuse pachymeningeal enhancement is the classic hallmark of this disorder, the variability is indeed substantial. This includes patients who do not display meningeal enhancement,[12] those who may not have headaches, or patients who may show CSF OPs that are well within normal limits.[13] The core factor in pathogenesis, and the independent variable, is loss of CSF volume; while CSF pressures, clinical manifestations, and magnetic resonance imaging (MRI) abnormalities are variables dependent on the loss of CSF volume (Fig. 1). The term “SIH” no longer appears broad enough to embrace all these variations. Therefore, terms such as “CSF hypovolemia” or “CSF volume depletion” as well as “spontaneous CSF leaks” have appeared in the literature and have been used interchangeably.[6, 14, 15]

Figure 1.

In the syndrome of “spontaneous intracranial hypotension” (SIH), the independent variable is decrease in cerebrospinal fluid (CSF) volume while CSF pressure may be normal. Head magnetic resonance imaging (MRI) may lack pachymeningeal enhancement or may even be normal, and clinical features may show substantial variability while headache may even be absent. The CSF pressures, the clinical manifestations, and the MRI changes appear to be variables dependent on CSF volume.

This review article attempts to outline the broad clinical spectrum of this disorder including substantial headache variability as well as diagnostic approaches and imaging findings including the mechanisms of these findings, etiologic considerations, the treatment options, and expectations from these treatments, as well as various complications in spontaneous CSF leaks.


The etiologies of CSF volume depletion are listed in Table 1. The effect of total body water loss (true hypovolemic state) and the role of various types of trauma (eg, cranial, spinal, or sinus surgeries) as well as the impact of CSF shunt overdrainage would seem essentially obvious. However, the most challenging remains the etiology of the spontaneous group, which needs to be addressed in greater depth. More often than not, the exact cause of spontaneous CSF leaks remains undetermined. Nonetheless, significant minorities of patients display clinical or imaging features suggestive of the presence of a disorder of the connective tissue matrix. The evidence for a preexisting dural sac weakness has been increasingly recognized. Many patients have joint hypermobility or have ectatic dural sacs (especially in lumbar and low thoracic regions), multiple meningeal diverticula, or dilated nerve root sleeves (Fig. 2). Dural sac ectasia, meningeal diverticula, and CSF leaks have been noted in Marfan's syndrome,[16, 17] a known heritable disorder of connective tissue matrix involving elastin and fibrillin. Stigmata of heritable connective tissue disorder, including but not limited to Marfanoid features, have been observed in a notable minority of the patients with spontaneous CSF leaks.[18, 19] Single or multiple meningeal diverticula, which are frequently noted in patients with spontaneous CSF leaks, are also seen in certain heritable disorders of connective tissue. Familial occurrence of spontaneous CSF leaks and meningeal diverticula in the setting of familial joint hypermobility and strong family history of aortic aneurysms[20] is yet further testimony to the role of heritable disorders of the connective tissue in causing dural weakness that can lead to CSF leak (Fig. 3).

Figure 2.

Joint hypermobility (left upper), skin hyperelasticity (left lower), and multiple meningeal diverticula (right) in a patient with spontaneous cerebrospinal fluid (CSF) leak, suggestive of an underlying disorder of connective tissue matrix (with permission of Mayo Foundation).

Figure 3.

Familial spontaneous cerebrospinal fluid (CSF) leak and joint hypermobility. Note fingers at rest (left upper) and at contraction (left lower). Joint hypermobility and interphalangeal (IP) joint subluxations are seen in all. P1 and P2: the two sisters with CSF leak. S: the sister who had not developed CSF leak. M: the mother. CT myelogram in the two sisters with CSF leak (right upper and lower). Both show leaking meningeal diverticula (from Mokri B, Ref 20, with permission of Headache).

Table 1. Etiology of CSF Leak, CSF Volume Depletion, or CSF Hypovolemia
  1. True hypovolemic state (reduced total body water)
  2. Traumatic CSF leaks
    1. Definite trauma (MVAs, sports injuries, etc)
    2. Thecal holes and rents from LPs and epidural catheterizations
    3. Spinal and cranial surgeries including skull base and some sinus surgeries
    4. Proximal brachial plexus avulsion injuries, nerve root avulsions
  3. CSF shunt overdrainage
  4. Spontaneous CSF leaks
    1. Undetermined cause
    2. Preexisting weakness of the dural sac, surgical anatomical observations
      1. Meningeal diverticula
      2. Disorders of connective tissue matrix
        1. Marfan syndrome, Marfanoid features
        2. Joint hypermobility
        3. Retinal detachment at young age
        4. Abnormalities of elastin and fibrillin in cultured dermal fibroblasts
    3. Trivial trauma in the setting of preexisting dural weakness
    4. Spondylotic spurs, herniated discs

A trivial previous trauma such as coughing, pulling, pushing, and lifting is sometimes reported in a minority of the patients. It is not unlikely that a combination of a weak thecal sac and a trivial trauma, which normally would have been harmless, might have caused a “spontaneous” CSF leak in some of the patients. Less common in occurrence, a dural tear from a spondylotic spur[21, 22] or disc herniation[23] may cause CSF leaks.

Clinical Features and Related Mechanisms


Headache is the most common clinical manifestation. This is often orthostatic (present when upright and relieved in recumbency). The latency of headache onset or resolution from change in posture classically should be only a few minutes, but in reality, the variability is substantial, and with chronicity, this latency may become even further prolonged. The headache may be throbbing, but more commonly it is not, and is described as a pressure sensation of variable intensity, sometimes quite intense. It is typically, although not invariably, bilateral.[24] It may be bifrontal, occipital, bifrontal-occipital, or holocephalic. Occasionally, it may start as a focal or unilateral headache and evolve into a holocephalic headache if the patient continues to be up and about. The headaches are often aggravated by Valsalva-type maneuvers and occasionally are even triggered by such maneuvers. At this point, it should be emphasized that not all orthostatic headaches are due to intracranial hypotension or CSF leaks (this will be discussed later in this communication), and not all headaches in CSF leaks are orthostatic. The headaches of spontaneous CSF leaks may have a variety of different features:

  1. Nonorthostatic lingering chronic daily headache (CDH) or head pressure sensation.
  2. Lingering CDHs or cervical or interscapular pain, or both, preceding the orthostatic headaches by days or weeks.
  3. CDHs that follow orthostatic headaches by months or longer – “transformed orthostatic headaches.” These sometimes may still carry a vague and rudimentary orthostatic component.
  4. Acute thunderclap-like onset mimicking a subarachnoid hemorrhage[25] with the orthostatic headaches to follow. Patients with this type of headache at onset may present to an emergency room with an understandable fear of a catastrophic event. Finally, when the diagnosis is established and the acute pain has settled, the orthostatic features of the headaches come to be recognized.
  5. A paradoxical postural headache sometimes may be encountered. These headaches are present in recumbency and are relieved in an upright position.[26]
  6. Sometimes, especially in slow-flow leaks or leaks that have been transformed to slow flow by chronicity or as the result of epidural blood patches (EBP), a second-half-of-the-day headache can be seen.[27] These headaches, with clear or not so clear orthostatic features, are absent in the morning and usually begin by late morning or early afternoon and increase in severity if the patient continues to be up and about.
  7. Although Valsalva-type maneuvers typically aggravate the headaches of CSF leaks, sometimes exertional headaches in isolation are the only type of headache that is reported by patients with CSF leaks.[28]
  8. Intermittent CSF leaks, not surprisingly, would lead to intermittent headaches, which may appear and disappear for variable periods of time.
  9. Sometimes patients with documented CSF leaks and with the typical MRI abnormalities may have no headaches at all, in other words: “acephalgic form.”[29]

Sinking of the brain, and the resultant traction on pain-sensitive suspending structures of the brain, is thought to be the main cause of the orthostatic headaches in CSF leaks. Dilatation of the cerebral veins and venous sinuses may also be a participatory mechanism and, in some situations, perhaps even the dominant mechanism.

Some patients with stubborn orthostatic headaches, in recumbency, may report an earlier and a more effective relief in certain positions or postures, such as Trendelenburg position,[30] or by lying prone with the head dropped somewhat at the edge of the bed. It has been demonstrated that CSF OP is significantly higher in prone than in lateral decubitus position.[31]

Clinical Features Other Than Headaches.

Headache, the most common clinical manifestation of spontaneous CSF leaks, is often (although not always) associated with one or more of a variety of other manifestations listed in Table 2. Sometimes one or more of these may be the dominant clinical feature or, more rarely, the only clinical manifestation. Occasionally, headache may be completely absent. In the past two decades, increasing reports of various, and sometimes unexpected, manifestations of spontaneous CSF leaks have appeared in the literature.

Table 2. Nonheadache Clinical Features of CSF Volume Depletion
  1. †The level of spine pain should not be assumed to correlate with the level of the leak. ‡Often described as distant, muffled, echoed, or “under water.” §Typically associated with elongated ventral extra-arachnoid fluid collection at cervical spine level often extending more caudally to the thoracic or even lumbar region.
  • Spinal pain (neck, interscapular, less commonly lower back), sometimes orthostatic
  • Nausea with or without emesis (often orthostatic)
  • Diplopia, horizontal and due to unilateral or bilateral 6th cranial nerve palsy[71] (more common)
  • Diplopia due to 3rd or 4th cranial nerve palsy or both[72, 73] (much less common)
  • Cochleovestibular manifestations (tinnitus, change in hearing, dizziness)
  • Photophobia, visual blurring
  • Upper limb numbness, paresthesias
  • Gait unsteadiness[74]
  • Facial numb feeling, vague paresthesias, or rarely weakness
  • Change in level of consciousness (ie, encephalopathy,[75] lethargy, stupor,[76] coma[77])
  • Personality change, memory decline, apathy, frontotemporal dementia (FTD)-like picture[78, 79]
  • Movement disorders: choreiform,[80] parkinsonism,[81] torticollis, tremor
  • Bibrachial amyotrophy§ (hanging arm syndrome, bilateral hand muscle weakness, and atrophy) mimicking motor neuron disease[62]
  • Galactorrhea[82]
  • Meniere-like syndrome[83]
  • Upper limb radiculopathy[84]
  • Trouble with bowel or bladder control[85]

Traction or compression is suspected to be the involved mechanism of various cranial nerve palsies in these patients. Cochleovestibular manifestations may result from traction or compression of the 8th cranial nerve or decrease in pressure of the perilymph, or both. Other manifestations have been similarly attributed to traction, compression, or displacement of various related structures including different lobes of the brain, brainstem or mesencephalon, pituitary stalk, or nerve roots.[32] Gait disorder and incontinence have been attributed by some researchers to spinal cord congestion. These attributions, however, are to be considered as proposed rather than proven mechanisms.


LP and CSF Analysis.

In the early years of MRI detection of pachymeningeal thickening, many patients were subjected to multiple CSF examinations in search of inflammatory, infectious, or neoplastic disease. Many lessons were learned including the substantial variability in the CSF findings in different patients with CSF leaks as well as in each individual patient who had undergone multiple spinal taps on multiple occasions in the setting of symptomatic active CSF leaks.

  1. CSF OP is low in the large majority; but in a significant minority, perhaps in about one fourth of patients, it is within normal limits. The OP is uncommonly atmospheric and rarely is even negative.
  2. Color is often clear and only sometimes xanthochromic. Note that difficult and blood-tinged taps are not uncommon considering the very low pressure in some of the patients and presence of dilated epidural venous plexus in many (see spinal MRI findings and Table 4).
  3. Protein concentration may be normal or high. Values up to 100 mg/dL are not uncommon and concentrations as high as 1000 mg/dL have been reported.[33]
  4. Leukocyte count may be normal but a lymphocytic pleocytosis of up to 40 cells/mm2 is not uncommon, and values over 200 cells/mm2 have been reported.[8]
  5. Erythrocyte count may also be normal or high, and at times quite high. Low CSF pressures and struggle to secure CSF flow to measure the OP and obtain fluid can increase the likelihood of traumatic tap. The associated congestion of the epidural venous plexus will also increase the incidence of blood-tinged CSF.
  6. Glucose concentration, cytology, ad bacteriology should all be normal.

Radioisotope Cisternography.

Indium-111 is the radioisotope of choice. It is introduced intrathecally (IT) via an LP and its dynamics are followed by sequential scanning at various intervals of up to 24 or even 48 hours. Normally after 24 hours, though often earlier, ample radioactivity can be detected over the cerebral convexities while no activity outside the dural sac is noted, unless there has been inadvertent injection of part of the radioisotope extradurally or if some of the IT-injected radioisotope has extravasated through the dural puncture site. In CSF leaks, the following should be expected:

  1. The radioactivity should not extend much beyond the basal cisterns, and therefore, at 24 or even at 48 hours, there is paucity of activity over the cerebral convexities (Fig. 4A,B).[34-36] Although an “indirect evidence,” this is the most common and most reliable cisternographic abnormality in active CSF leaks. This is particularly helpful when the clinical and MRI findings are atypical, insufficient, or unconvincing and, therefore, leaving the clinician with a fundamental uncertainty about the diagnosis.
  2. Presence of parathecal activity as a “direct evidence” of leak pointing to the level or the approximate site of the leak (Fig. 4B,C), unfortunately, is far less commonly noted than paucity of activity over cerebral convexities. Of note, meningeal diverticula – if large enough – may appear as foci of parathecal activity and sometimes may not be reliably distinguished from actual sites of leak. Computed tomography myelography (CTM) is frequently needed to advance the workup appropriately, not only to enable this differentiation but to confirm the actual site of the leak. Meningeal diverticula may or may not be the actual site of the leak even when they are large.
  3. Early appearance of radioactivity in the kidneys and urinary bladder (in less than 4 hours vs 6-24 hours) is a fairly common “indirect evidence,” indicating that the IT-introduced radioisotope has extravasated and entered the venous system quickly with subsequent early renal clearance and early appearance in the urinary bladder. This finding, however, is of limited reliability and can be affected by partial extradural radioisotope injection or perhaps even more commonly by extravasation of IT-injected radioisotope from the dural puncture site back to the epidural tissues. This is identical to the mechanism involved in postdural puncture headaches.
Figure 4.

Indium-111 radioisotope cisternography in spontaneous cerebrospinal fluid (CSF) leak. A and B: 24 h images. (A) Normal; (B) Paucity of activity over the cerebral convexities at 24 h in a patient with spontaneous CSF leak. Cervical (C) and thoracic (D) parathecal activity.

MRI Abnormalities of Head and Spine and Their Related Mechanisms.

MRI has truly revolutionized our understanding of SIH. Head and spine MRI abnormalities of CSF leaks and CSF hypovolemia are listed in Tables 3 and 4 and also illustrated in related figures. It is important to explore the mechanisms of these imaging abnormalities in the setting of decreased CSF volume. In doing so, the principles of Monro-Kellie doctrine[37] need to be considered. In the core of this doctrine exists the following principle: “with intact skull, sum of volume of brain plus volume of CSF plus volume of intracranial blood is constant, and therefore decrease or increase in one will result in increase or decrease in one or both of the remaining two.” In decreased CSF volume such as CSF leaks (Fig. 5), given that the brain is essentially nonexpandable, it is the increase in intracranial blood volume that has to compensate for decrease in CSF volume. With meningeal venous hyperemia, there is diffuse pachymeningeal enhancement (leptomeninges, in contrast to pachymeninges, have blood brain barriers and therefore do not enhance). Engorgement and enlargement of cerebral venous sinuses and pituitary gland are also part of this compensatory hyperemia. Another volume compensatory phenomenon is collection of subdural fluids (Figs. 6 and 7).

Figure 5.

Schematic demonstration of effect of cerebrospinal fluid (CSF) volume loss on magnetic resonance imaging (MRI) findings. Normal (left row); after CSF volume loss (right row). Upper panels (coronal views at the level of sella-pituitary) show some of the changes of volume compensation (B) in response to CSF hypovolemia including diffuse meningeal hyperemic thickening, subdural fluid collections (in green), engorged cerebral venous sinus (blue), and engorged pituitary. Also noted are sinking of the brain including decrease in perichiasmatic cistern, flattening of the optic chiasm, and decrease in ventricular size (“ventricular collapse”). Middle panels (sagittal views) show descent of the cerebellar tonsils and brainstem, increase in anterior-posterior diameter of the brainstem, crowding of the posterior fossa, and decrease in prepontine cistern and smaller fourth ventricle, noted in (D) compared with (C). Lower panels show changes at the spine level including engorgement of the epidural venous plexus and extra-arachnoid fluid collection (green) (with permission of Mayo Foundation).

Figure 6.

Head magnetic resonance imaging (MRI) in cerebrospinal fluid (CSF) leak – CSF hypovolemia. Upper panels show T1-weighted gadolinium-enhanced coronal images at the level of sella-pituitary (A) during active leak; (B) after surgical treatment of a leaking meningeal diverticulum. Note diffuse pachymeningeal enhancement (upper arrows), enlarged pituitary, flattening of the optic chiasm, smaller perichiasmatic cistern, and lateral ventricles in (A); all resolved after cessation of the leak in (B). (C) Shows descent of the cerebellar tonsils below the foramen magnum. (D) Shows descent of the cerebellar tonsils and brainstem, increase in AP diameter and deformity of brainstem, crowded posterior fossa, obliteration of prepontine cistern and flattening of anterior pons all related to the sinking of the brain (A and B from Mokri B, Ref 52, with permission of Mayo Foundation).

Figure 7.

Head magnetic resonance imaging (MRI) in cerebrospinal fluid (CSF) leak – CSF hypovolemia. (A) and (B) Subdural fluid collections. (C) Bilateral subdural collections with signal characteristics different from the CSF (arrows), likely related to increased protein concentration or blood-tinged fluid. (D) Bilateral subdural hematomas with mass effect on underlying brain (arrows). (E) Enlarged pituitary and obliteration of the perichiasmatic cistern. (F) Engorged cerebral venous sinuses.

Table 3. Head MRI Abnormalities in CSF Leaks
  1. †The most common and most reliable head MRI abnormality in spontaneous CSF leaks. ‡Cephalad opening of aqueduct of Sylvius as seen in midline sagittal views. §On midline sagittal view, line drawn from anterior tuberculum sellae to the point of junction of straight sinus, inferior sagittal sinus, and the great vein of Galen.
  1. Diffuse pachymeningeal enhancement: uninterrupted, non-nodular, can be thick or thin, no leptomeningeal abnormality
  2. Descent (“sagging” or “sinking”) of the brain
    1. Descent of cerebellar tonsils at or below the foramen magnum (may mimic type I Chiari)[86]
    2. Descent of the brainstem and mesencephalon, occasionally without descent of cerebellar tonsils to or below foramen magnum
    3. Increase in anteroposterior diameter brainstem resulting from distortion of the brainstem
    4. Descent of iter below the incisural line§[9]
    5. Obliteration of prepontine or perichiasmatic cisterns
    6. Crowding of the posterior fossa
    7. Flattening of the optic chiasm
    8. Flattening of the anterior pons
  3. Subdural fluid collections, typically hygromas, infrequently hematomas
  4. Enlargement of the pituitary (may mimic pituitary tumor or hyperplasia)[87]
  5. Engorged cerebral venous sinuses
  6. Decrease in size of the ventricles (“ventricular collapse”)
Table 4. Spine MRI Abnormalities in Spontaneous CSF Leaks
  1. aMay or may not be the actual site of leak, even when the diverticulum is large, although larger diverticula may be more prone to be the site of the leak.
  1. Extra-arachnoid fluid collections (often extending along several spinal levels)[88-90]
  2. Extradural extravasation of fluid (extending to paraspinal soft tissues)
    1. May identify the level of the leak (ie, cervical, thoracic or lumbar), not uncommon
    2. May identify the actual site of the leak, uncommon[91]
  3. Meningeal diverticula,a single or multiple, various sizes, any level of spine
  4. Spinal dural enhancement[92]
  5. Engorgement of spinal epidural venous plexus

Similar changes are noted in spine MRI (Table 4) including dural enhancement and extra-arachnoid fluid collections. However, at the spine level, in contrast to the skull, there exist the epidural space with adipose and soft connective tissue and the epidural venous plexus. Therefore, with CSF volume depletion the dural sac can collapse somewhat, and this will result in engorgement and prominence of epidural venous plexus, yet another spine MRI abnormality of CSF leaks (Fig. 8).

Figure 8.

Spine magnetic resonance imaging (MRI) in cerebrospinal fluid (CSF) leak – CSF hypovolemia. (A) Spinal dural enhancement (arrows). (B) Meningeal diverticula and dilated nerve root sleeves (can be seen in T2-weighted and even better on highly T2-weighted images as in this case after intrathecal injection of gadolinium – the so-called MR myelography). (C) Engorgement of epidural venous plexus. (D) and (E) Extra-arachnoid fluid collections.

Sinking of the brain is another consequence of CSF leak. On head MRIs, this is manifested by a decrease in size of the ventricles (“ventricular collapse”), descent of the cerebellar tonsils, descent and distortion of the brainstem, obliteration of some of basal cisterns, flattening of the optic chiasm, or crowding of the posterior fossa. Descent of iter below the incisural line, an indication of descent of the brainstem, may be seen in the absence of any obvious descent of the cerebellar tonsils.[9] Iter is the cephalad opening of the aqueduct of Sylvius as seen in the midline sagittal MRI views. Incisural line is the line drawn from anterior tuberculum sellae on midline sagittal image to the junction of straight sinus, inferior sagittal sinus, and the great vein of Galen.

In reviewing head MRIs of patients with spontaneous CSF leaks, this author has been helped the most (although not exclusively) by T1-weighted midline sagittal image and gadolinium (Gd)-enhanced coronal image through sella and pituitary. The former is helpful to look for descent of cerebellar tonsils, descent and deformity of brainstem, and location of the iter. The latter typically shows the pachymeningeal enhancement well and enables assessing the size of pituitary, the appearance of the optic chiasm, and the perichiasmatic cistern. It will also provide information on size and shape of the lateral ventricles and perhaps relative size of the superior sagittal sinus. This absolutely does not mean that other views or sequences should not be obtained or carefully studied, as they can be quite helpful, but it is intended to suggest that an ideal study for CSF leak or CSF hypovolemia should include these images.


CTM thus far is the most accurate study for demonstrating the exact site of the spinal CSF leakage.[32] Similar to radioisotope cisternography, it also provides an opportunity to measure the CSF OP at the time of dural puncture. In addition to its accuracy in revealing the site of the leak, it can show meningeal diverticula, dilated nerve root sleeves, extra-arachnoid fluid collections, and extra dural egress of contrast into the paraspinal tissues (Fig. 9). Because some of the leaks can be rapid (fast flow) or slow (slow flow), each may present special diagnostic challenges:

  1. When leaks are fast flow, after the preliminary myelogram and before the patient is taken for the subsequent computed tomography (CT) scanning, already so much of the CSF (and therefore of the contrast) has leaked that it spreads across many spinal levels; therefore, it becomes essentially impossible to locate the exact site of the leak. In an attempt to overcome this obstacle, one strategy would be to bypass the initial myelogram and proceed with CT scanning right after the IT contrast injection, utilizing a high-speed multidetector spiral CT which allows obtaining many cuts in a short period of time. This technique, referred to as “dynamic CTM,”[38] as well as its variation (hyperdynamic CTM) and digital subtraction myelography[39, 40] often have enabled us to overcome the significant difficulties we had in determining the site of the high-flow leaks.
  2. Slow-flow leaks provide an opposite challenge. Even by the time of the postmyelogram CT scanning, as the result of the slowness of the flow of the leak, still not enough contrast has extravasated to allow detection. Obtaining a delayed CT after 3-4 hours may enable the detection of the site of the leak. Gadolinium myelography (GdM) (spine MRI after intrathecal injection of Gd)[41] may also be helpful but, unfortunately, not as much as initially hoped. Nevertheless, GdM remains a useful test. IT injection of Gd contrast is an off-label use and should be reserved for highly selected patients who are substantially symptomatic, have high clinical suspicion of CSF leak, and have demonstrated no leak on CTM.[42]

Overall, the detection of the site of the slow-flow leaks not infrequently can remain problematic and sometimes quite frustrating for the patient and the physician.

Figure 9.

Myelography, computed tomography (CT) myelography in cerebrospinal fluid (CSF) leak. (A) Meningeal diverticulum noted in myelogram (arrow). (B) Meningeal diverticulum seen in CT myelogram (arrow). (C) Extradural extension of the leak in the paraspinal soft tissues (A and B from Mokri B, Ref 52, with permission of Mayo Foundation).


Here, the focus will be on management of spontaneous CSF leaks rather than postsurgical or post-traumatic ones. For spontaneous spinal CSF leak, a variety of treatment modalities have been tried (Table 5). The efficacy of caffeine or theophylline is unpredictable. Some patients may report benefit while others may not. Overall, no substantial improvement is to be expected. Some patients, but not most, report definite improvement with corticosteroids. When it occurs, the improvement is often partial and hardly durable. Besides, considering the potential side effects from its chronic use, corticosteroid treatment hardly seems to be a long-term solution. Traditionally, bed rest and increased fluid intake have been advocated, mainly based on long-practiced recommendations regarding post-LP headaches. Epidural saline infusion[43] has produced marginally unpredictable results but the experience has not been extensive. It can be tried with limited expectations in some of the patients who have failed repeated EBPs and when surgery is not an option. Even then, a sustained relief would seem unlikely. Similarly, experience with epidural infusions of colloids such as dextran[44] has been quite limited. Intrathecal infusion of fluid[45] has been tried when urgent volume replacement has been a treatment objective, such as stupor or coma related to sinking of the brainstem. It is not difficult to predict that, as long as such infusions continue, the patients with CSF hypovolemia may note improvement. However, after cessation of infusion, a sustained improvement, although possible, would seem unlikely. With prolonged epidural and intrathecal infusions, risk of infection will be a serious consideration.

Table 5. Treatment of Spontaneous CSF Leaks/CSF Volume Loss
  1. Conservative measures
    1. Bed rest (those with substantial orthostatic headaches remain reclined much of the time anyway)
    2. Coffee
    3. Hydration (actually overhydration since most patients are not dehydrated)
    4. Time
  2. Medications
    1. Analgesics
    2. Caffeine
    3. Theophylline
    4. Corticosteroids
  3. Abdominal binder
  4. Epidural injections of:
    1. Homologous blood (epidural blood patch, “EBP”)
      1. Targeted
      2. Distant at lumbar level or bilevel, “blind EBP”
    2. Fibrin glue
    3. Fibrin glue and blood
  5. Surgical repair or the leak
  6. Other measures in special situations
    1. Intrathecal fluid injection (volume replacement)
    2. Epidural saline infusion
    3. IV saline infusions
    4. Epidural infusion of dextran

Excess use of vitamin A may cause increased intracranial pressure,[46] and decreased blood concentration of vitamin A has been reported in “spontaneous” intracranial hypotension.[47] Recent scant and anecdotal observations have invited attention to potential utility of vitamin A as an adjunct in the management of SIH. Further observations are needed; and indeed, if effective, the optimal dosing needs to be determined as excess use of vitamin A can cause several toxic effects.[48]

EBP is now recognized as the treatment of choice in those patients who have not responded to the initial trial of conservative management.[49] EBP works via two separate mechanisms: (1) the immediate effect related to volume replacement by compression of the dural sac (decreasing the volume of the container); (2) sealing of the dural defect, which may be delayed from the first one. Therefore, it is not uncommon to note an initial quick response in connection with the first mechanism, recurrence of symptoms within merely a day or two, and then a gradual and often variable improvement after several days. Variability is, however, substantial. The efficacy of each EBP is about 30%.[50] A previous EBP failure should not be taken as a signal that a subsequent EBP will fail. Indeed, many patients may require more than one EBP and some have required several. At times, a cumulative effect from multiple EBPs may be noted. Similarly, a previous success will not guarantee success of a future EBP. The efficacy of EBP in post-LP headaches is far more impressive. Here, the first EBP often will cause durable relief in about 90%, and a second EBP brings relief in almost all of the remaining patients.[51] Even in inadvertent dural tears from epidural catheterizations, the efficacy of response to EBP is superior to that of spontaneous CSF leaks. There are several reasons for this discrepancy: (1) in post-LP leaks, the EBP is typically targeted right at the site of the leak or very close to, while this is not the case with spontaneous leaks; (2) in spontaneous CSF leaks, the site of most of the leaks is at the nerve root sleeves or nerve root sleeve axilla as opposed to the post-LP where the leak site is in the posterior aspect of the dura. The site of the leak in spontaneous CSF leaks is mostly at levels above the lumbar spine where most of the epidural block patches are placed. Therefore, the odds are that many of these will be nontargeted and distant from the site of the leak. (3) The dural defect in spontaneous CSF leaks, as opposed to post-LP leak, often is not a simple hole or rent instead it is frequently a preexisting zone of attenuated dura with or without associated diverticula where an unsupported arachnoid may finally give way and ooze CSF from one or more sites. Surgical anatomical observations[52] have clearly identified such defects in many patients who have ended up with surgery.

In one study, impressive results from lumbar EBP were reported when the patients were premedicated with acetazolamide 250 mg, at 18 hours and at 6 hours before the EBP, with the patients at 30-degree Trendelenburg position from 1 hour prior to the EBP, during the procedure, and for 24 hours after the procedure.[53] We have not tested this protocol yet.

Sometimes, when EBPs fail, epidural injections of fibrin glue or fibrin glue followed by blood may help.[54] We have not succeeded in the method of mixing the two together before the injection,[55] as the mixture will have a pasty and noninjectable consistency.

Surgery in well-thought-of cases is effective and can be tried when less invasive measures (such as EBP) fail. It needs to be recognized that the findings at surgery are not always straightforward.[56] Sometimes the surgeon may encounter extravasated CSF but may not be able to locate the exact site of the leakage. The surgeon may then proceed to pack the area with blood-soaked gel foam, muscle, etc, and hope for the best.[8] Sometimes dural defects may be seen that have markedly attenuated and fragile borders. These may not yield to suturing and would require different reinforcing techniques.[52] Furthermore, some patients may have CSF leaks from more than one site and at different levels. It is strongly emphasized that thorough preoperative neurodiagnostic studies should be conducted to identify the actual site of the leak before surgery is undertaken.

The fundamental purpose of the surgery in the treatment of CSF leaks is to stop the leak. Other rarely practiced surgical undertakings are expected to have a convincing rationale and more than anecdotal proof of efficacy or durability.

Complications of Spontaneous CSF Leaks

Subdural Hematoma.

Subdural hematomas (SDH) may be noted right from the start or may complicate a subdural hygroma. They may be thin and asymptomatic but can be large with enough mass effect to compress the underlying brain and cause midline shift. If symptomatic and growing, surgical intervention will become necessary.[57, 58] Vigilant postoperative neurosurgical care and follow-up is important as creating a skull defect may violate the Monro-Kellie principle and lead to more sinking of the brain.[59] It is prudent to have the issue of the leak also addressed at some point along with the treatment of SDH.

Rebound Intracranial Hypertension.

Rebound intracranial hypertension is sometimes encountered after successful treatment of the leak by EBP or surgery.[60] The incidence of this phenomenon is likely higher than is thought as some cases are asymptomatic or only minimally symptomatic. Sometimes the clinical presentation is dramatic enough to even cause florid papilledema. Most of these patients return to their physicians thinking that they have recurrence of the leak. This condition, fortunately, is often self-limiting but can take a frustratingly long time even though acetazolamide may help with the symptoms. At this juncture, it should be noted that occasionally one might encounter a patient with previously diagnosed or undiagnosed pseudotumor cerebri who has self-decompressed through a weak area of dura. This may lead to the syndrome of intracranial hypotension or CSF hypovolemia. When such leaks are successfully treated, the manifestations of pseudotumor will reappear. Acetazolamide can help, but a few patients have finally ended up with shunting procedures (B. Mokri, unpublished data).

Cerebral Venous Sinus Thrombosis.

Fortunately, as a phenomenon, this is very uncommon. In patients with active CSF leaks, when headache characteristics change in a short period, it is prudent to look for unexpected events and surprises. This complication will often call for anticoagulant therapy.[61]

Bibrachial Amyotrophy.

Bibrachial amyotrophy is seen in connection with extra-arachnoid fluid collection, typically in the ventral aspect of the cord in the cervical region that often extends to the thoracic and even lumbar levels. There is weakness and atrophy at a few sequential myotomal distributions of upper limbs with only mild asymmetry resembling and mimicking motor neuron disease,[62] especially when the sensory symptoms are curiously absent or at best minimal.

Superficial Siderosis.

Although a rare occurrence, it can be a remote complication of spinal CSF leaks[63, 64] or CSF leak from brachial plexus injury and nerve root avulsion.[65] In superficial siderosis associated with CSF leaks, frequently extra-arachnoid elongated fluid collections are seen typically ventral to the cord and similar to the fluid collections seen in bibrachial amyotrophy.

Natural History and Outcome.

The majority of patients make a complete recovery with conservative management (ie, hydration, analgesics, caffeine, theophylline, and especially time), while some may require epidural injections of blood or fibrin glue, or even surgery. Recurrence may occur as addressed below. The outcome of the recurrences should not be expected to differ significantly from the initial episode. Sometimes all therapeutic attempts fail and the patients remain frustratingly symptomatic and work disabled. Fortunately, such cases are only a small minority. Not uncommonly, with time or with therapeutic attempts, patients' symptoms may decrease to the point that they will be asymptomatic most of the time, can work and do most of their usual activities, but will have such manifestations as CDH, Valsalva-induced headaches, or headaches in the second half of the day. In these cases, likely a low-grade slow-flow leak persists[27] and may continue for variable periods of time, even years.

Recurrence of CSF Leaks

These can occur with variable frequency and with variable intervals from the previous leak, ranging from weeks to years, sometimes from the same site and sometimes from a different site. Data on surgical patients[66] may not be applicable to all patients with spontaneous leaks as the large majority do not come to surgery and likely have a different course and outcome. Accurate data are not available but it is possible, although not formally studied or proven, that those with disorders of connective tissue matrix might be at a somewhat higher risk for the recurrence.

Orthostatic Headaches without CSF Leak

Orthostatic headaches are the hallmark of CSF leaks. However, as discussed earlier, not all headaches of CSF leaks are orthostatic and also not all orthostatic headaches are due to CSF leaks.

Orthostatic headaches without CSF leak may be seen in connection with several other conditions:

  1. Postural orthostatic tachycardia syndrome (POTS): In some of the patients with POTS, an orthostatic headache can be the prominent, or one of the prominent, clinical features of the disorder.[67]
  2. After surgery for Chiari malformation: A small minority of patients who have undergone decompressive surgery for Chiari malformation may develop an orthostatic headache without any CSF leak.
  3. The “syndrome of the trephined”: Sometimes patients, who have undergone large decompressive craniectomies for massive life-threatening cerebral edema, should they survive the life-threatening event, may complain of orthostatic headache that can be severe. Sometimes these headaches, along with the residual deficits from the original injury, can create substantial disability. Such patients sometimes show drastic improvement after cranioplasty.[68]
  4. Increased compliance of the dural sac,[69] especially in those with generous lumbar dural sacs and stigmata of disorders of connective tissue matrix.
  5. Headache is the most common symptom of colloid cysts of the third ventricle, a rare tumor comprising less than 0.5% of brain tumors. Although these lack any particular outstanding features, they can be present when standing and relieved by lying down.[70]


From this extensive review, several conclusions can be drawn:

SIH almost always results from spontaneous CSF leaks. The past theories of increased CSF absorption or decreased CSF production have never been substantiated. The disorder is far more common than was believed only one or two decades ago.

The overwhelming majority of spontaneous CSF leaks occur at the level of the spine, particularly the thoracic spine. Spontaneous leaks at the skull base do occur but only rarely.

Spontaneous CSF leaks can no longer be equated with postpuncture headaches. There is considerable variability in clinical presentations, imaging findings, and CSF findings including CSF pressures that can be within normal limits. CSF volume depletion (CSF hypovolemia) rather than decreased CSF pressure appears to be the pathogenetic core as the independent variable. CSF pressures, clinical manifestation, and MRI abnormalities are variables dependent on the CSF volume. The term “SIH” no longer appears broad enough to embrace all of these variables. Terms such as CSF volume depletion or CSF hypovolemia have appeared in the literature and have been used interchangeably with spontaneous CSF leak.

The anatomy of spontaneous CSF leaks is often complex and different from a simple hole or a rent. It is typically not the same as what is encountered in CSF leaks resulting from LP, epidural catheterization, or craniospinal surgeries.

Clinical stigmata of disorders of connective tissue matrix can be seen in a significant minority of the patients with spontaneous CSF leaks. This very likely plays a role in the weakness of the dural sac, formation of meningeal diverticula, and pathogenesis of the disorder.

Not all headaches in spontaneous CSF leaks are orthostatic and not all orthostatic headaches result from CSF leaks.

Sometimes after treatment of CSF leak, whether by EBP or surgery, a rebound increased intracranial pressure may occur, which is often self-limiting but sometimes may require treatment.

The rate of CSF leakage in spontaneous CSF leaks may vary considerably. Fast-flow and slow-flow leaks each present special diagnostic challenges. Novel diagnostic techniques have been quite helpful in locating the site of the leak in fast-flow leaks. Locating the site of slow-flow leaks remains challenging.

EBP has emerged as treatment of choice when initial conservative measures including time have failed. These may be targeted or blind (presumed distant from an undetermined leak site) or single level or bilevel. Epidural injection of fibrin glue also has utility in selected cases. Combined EBP and fibrin glue injections have also been tried but it needs special considerations.

Surgery aimed at stopping the leakage is often undertaken when less invasive measures (such as EBP) have failed. It is essential to determine the site of the leak by appropriate imaging before surgery is undertaken.


The author thanks Mrs. Lori Lynn Reinstrom, Research Administrative Assistant, Mayo Clinic-Rochester, for her excellent editorial assistance and Mr. John V. Hagen, Senior Medical Illustrator, Mayo Clinic-Rochester, for his talent in preparation of the artwork (Fig. 5).

Statement of Authorship

Category 1

  • (a)Conception and DesignBahram Mokri
  • (b)Acquisition of DataBahram Mokri
  • (c)Analysis and Interpretation of DataBahram Mokri

Category 2

  • (a)Drafting the ManuscriptBahram Mokri
  • (b)Revising It for Intellectual ContentBahram Mokri

Category 3

  • (a)Final Approval of the Completed ManuscriptBahram Mokri