Expert Commentary: Role of Surgery for the Management of CSF Leaks

Abstract

Spontaneous intracranial hypotension (SIH) was described nearly 70 years ago, and for years it was essentially equated with post dural puncture headaches (PDPH) (1). The interest in this entity substantially increased after the magnetic resonance imaging (MRI) abnormalities of this disorder were recognized in the early 1990s (2-6), and consequently many more patients were diagnosed than before and a far broader clinical spectrum of this disorder began to become apparent. It also came to be recognized that the overwhelming majority if not all cases of SIH result from spontaneous cerebrospinal (CSF) leaks and that the independent pathogenic parameter is loss of CSF volume (7).

S pontaneous intracranial hypotension (SIH) was described nearly 70 years ago, and for years it was essentially equated with post dural puncture headaches (PDPH) (1). The interest in this entity substantially increased after the magnetic resonance imaging (MRI) abnormalities of this disorder were recognized in the early 1990s (2–6), and consequently many more patients were diagnosed than before and a far broader clinical spectrum of this disorder began to become apparent. It also came to be recognized that the overwhelming majority if not all cases of SIH result from spontaneous cerebrospinal (CSF) leaks and that the independent pathogenic parameter is loss of CSF volume (7).

Although there are many case reports and several series, experience with treatment of this disorder is limited, and there are no controlled studies. Many of the treatments currently utilized in spontaneous CSF leaks are simply adoption of modalities implemented for the management of PDPH. For this we are indebted to our colleagues in anaesthesiology, who have treated headaches following lumbar punctures or inadvertent dural tears after epidural catheterizations for years and have generated a fairly extensive literature in this regard (8).

As experience with spontaneous CSF leaks is increasing, it is becoming more apparent that often, although not always, this entity can not be equated with PDPH (9). This is particularly noteworthy with regard to treatment outcome. Conservative estimates point to a persistent symptomatic relief in nearly three-quarters of patients with post puncture headaches, whereas the response to a second epidural blood patch (EBP) is even higher. The figures are even more impressive if only patients with post diagnostic lumbar puncture headaches are considered (10). This can be contrasted with nearly one-third good response to each EBP in spontaneous CSF leaks (11) and somewhat better response in highly selected cases (12). Why such a discrepancy?

There are at least a few contributory factors. First, the traumatic dural defects after lumbar puncture or epidural catheterization are typically in the posterior aspect of the dura, and this is where the EBP is typically placed. Second, the exact site of the lumbar puncture is known, and with epidural catheterization at least the approximate level of the dural defect is known. Therefore, EBPs in these post procedural dural defects and leaks are typically quite targeted. Targeted EBPs in spontaneous CSF leaks are somewhat more effective than distant EBPs, but not dramatically so (11). Third, and most important, is the complexity of the anatomy of the leak in many patients with spontaneous CSF leaks (13). This will be addressed later in this communication.

EBP has emerged as the treatment of choice for spontaneous CSF leaks when initial conservative and non-invasive measures such as bed rest, caffeine, over-hydration and, most importantly, ‘time’, fail to bring relief. Surgery is considered as the last resort when less invasive measures have failed and when imaging studies have clearly identified the site of the CSF leak. Studies evaluating the efficacy of EBP in spontaneous CSF leaks are rare (11, 12). For surgical treatment, the available information, understandably, is even scarcer (13–15). What is available consists mostly of case reports, small series, personal experiences and related biases. Nevertheless, the situation is not chaotic, as there are several principles, some common sense and some well-documented in the literature. I will attempt to point to some of these.

• A.

  The overwhelming majority of spontaneous CSF leaks occur at the level of spine. It would be quite unusual, although not impossible, for such spontaneous leaks to occur at the skull base level.

• B.

  The anatomy of the leak in many of these patients is complex and different from a simple hole or rent. In spontaneous CSF leaks there are frequently dural defects suggestive of a pre-existing dural weakness (13, 16). Arachnoid herniations are often seen. Sometimes circumscribed dural diverticula are identified at the roots or root axilla (13), and arachnoid herniations may frequently be encountered through the dural defects weeping CSF. These often occur at the level of nerve root sleeves, but sometimes in the anterior aspect of the dura, in which case it may not be so obvious to inspection through laminectomy and posterior approach. Sometimes at surgery, despite encountering the leaked CSF, the exact site of the leakage may not be identified.

• C.

  A substantial minority of patients with spontaneous CSF leaks have stigmata of a disorder of connective tissue matrix (17–19). The borders of the dural defects may be so attenuated or fragile that they may not be suitable for direct suturing and repair, requiring different measures such as packing the site of the leak (13, 14).

• D.

  It is essential to locate the site of the leak through neurodiagnostic imaging studies before surgery is undertaken.

• E.

  Pain at different levels of the spine in not uncommon in spontaneous CSF leaks. It might be tempting to use the level of the spinal pain as a clinical indicator of the level of the leak. However, in the majority of instances these may prove to be false localizing signs (16). Furthermore, retrospinal C1-2 fluid collection, sometime seen on spine imaging, does not indicate the site of the leak (20, 21).

• F.

  Dilated nerve root sleeves and meningeal diverticula are common in patients with spontaneous CSF leaks. A dilated nerve root sleeve or a meningeal diverticulum, even when large, may not necessarily be the site of the CSF leak (7, 24). Furthermore, many such patients may have more than a single dilated nerve root sleeve or meningeal diverticulum, or both. The size of the abnormality, although important and helpful, may not always or necessarily favour the larger lesion over the smaller ones as the potential site of the leak. This again emphasizes the fundamental importance of demonstrating the site of the leakage and avoiding presumptions before the patient is subjected to surgery.

• G.

  Fast-flow CSF leaks may lead to rapid and substantial extravasation of contrast on the usual computed tomography (CT)-myelographic studies with accumulation of extra-arachnoid fluid over several spinal levels, rendering the task of locating the exact site of the leakage virtually impossible (22). In such instances, dynamic CT-myelography (23) will often be helpful to locate the site or sites of CSF leakage.

• H.

  Slow-flow CSF leaks can present a different challenge (22). Here the leak may be so slow that by the time of CT scanning not enough contrast has extravasated to allow detection. In these instances, delayed CT re-imaging after 3–4 h may help. Another approach, if CSF pressure is low enough, is to elevate the CSF pressure to levels of upper normal by intrathecal (IT) injection of saline. MR-myelography (spine MRI following IT injection of gadolinium with or without elevating low CSF pressure by IT saline injection), again with the option of increasing the CSF pressure to upper limits of normal, is yet another technique with encouraging results.

• I.

  With ¹¹¹indium radioisotope cisternography (RIC), the most common and reliable finding in spontaneous CSF leak is absence or paucity of radioactivity over the cerebral convexities at 24 h. Therefore, in patients with uncertain clinical or MRI manifestations and when CSF opening pressure has not been low or has been borderline, this is an invaluable diagnostic modality to differentiate between patients with CSF leak who present with atypical and diagnostically uncertain manifestations and those without CSF leaks. The test also provides an additional opportunity to measure the CSF opening pressure. There are, however, several misgivings and pitfalls:

When should surgery be considered? This should be in cases where (i) symptoms are severe enough to warrant surgical intervention, (ii) the site or sites of the leak have been identified, and (iii) symptoms have been refractory to less invasive measures.

A review by Schievink et al. (14) of 10 patients with SIH and CSF leak who had undergone surgical treatment at the Mayo Clinic between 1992 and 1997 consisted of seven women and three men (mean age 42.3 years, range 22–61 years). Preoperative neurodiagnostic studies had shown meningeal diverticulum in two patients, a complex of diverticula in one patient, and focal leak alone in seven patients. On surgical exploration only one of these seven patients showed a clear source of CSF leak in the form of two actively leaking meningeal diverticula in the axilla of two nerve roots, T10 and T11. The exact site of CSF leak could not be identified in the remaining six patients despite careful exploration of the dura and multiple nerve roots in all cases, and despite intradural exploration in three patients and intradural injection of saline or dye in two patients. Of note is that extradural CSF was encountered in all of the patients at the level suggested by the preoperative imaging studies except for two of the three patients who had been explored in sitting position for leaks at the cervical spinal level. Even then, the dura was noted to be very laxed, pointing to low CSF pressure and ongoing leak, as determined by preoperative neurodiagnostic studies. There were areas of root sleeve dural attenuation in three patients, but there were no convincing active leaks in these areas. Overall, three patients were treated by ligation of the meningeal diverticula and the remaining seven by packing the epidural space with muscle and gelfoam and in some also fibrin glue. Despite the fact that the exact site of the CSF leakage was not identified at surgery in six of these 10 patients, after surgery all experienced complete relief from their headaches, without recurrence at follow-up (mean duration of follow-up 19 months, range 3–58 months). The surgical approach was thought to be effective and with low morbidity.

This study pointed to the fact that surgery in spontaneous CSF leaks is not always straightforward. Furthermore, there are frequently instances where extradural CSF is encountered at surgery, but despite careful exploration the site of the leak may not be found. The surgeon is then left with the sole option of packing the epidural space with muscle, gelfoam soaked in patient's own blood, and sometimes also fibrin glue, and hoping for the best. Fortunately, this seemed to work.

Cohen-Gadol et al. studied 13 patients who had undergone surgery at the Mayo Clinic for spontaneous CSF leaks between 1994 and 2003 (13). There were 13 patients (four men, eight women and a 12-year-old girl). The mean age was 40 years (range 12–62 years). Preoperative neurodiagnostic studies had revealed the sites of CSF leakage in all patients and included a single site in eight patients, two sites in three patients and multiple sites in two patients. During surgery the exact site of CSF leak was not found in four patients. Although in one of these four patients extradural CSF was encountered, still no site of dural disruption was visible. Of the remaining nine patients, closure of a single meningeal diverticulum was achieved in one patient. In the remaining eight patients, the zones of dural attenuation and the related dural fragility were such that they did not seem to yield to suturing and primary repair. In one of these patients, ligation of two non-appendicular nerve roots resulted in cessation of the leak. In the rest, epidural packing with muscle, fibrin sealant and gelfoam soaked in patient's own blood were used in an attempt to seal the leak. Postoperative follow-up with a mean duration of 20.5 months revealed durable resolution of symptoms in eight patients, substantial improvement in three patients and only transient improvement in two patients.

The results of Cohen-Gadol et al.'s study is less striking than those of the earlier study of Schievink et al., probably at least partly due to a gradual shift of referrals to the tertiary centres from more straightforward and typical cases to more complex and atypical cases, including treatment failures.

One of the important aspects of the study by Cohen-Gadol et al. was careful analysis of the surgical anatomy of these spontaneous spinal CSF leaks and clear and detailed demonstration of the pre-existing regional dural abnormalities in a substantial percentage of the patients. The frequent observation consisted of zones of absent or highly attenuated dura in the nerve root sleeves or sometimes in the anterior aspect of the dura with herniations of the underlying arachnoid, leaking CSF. The size or extent of these arachnoid herniations may vary from very small to clear diverticular formations. Such dural abnormalities are yet another testimony to the presence of an underlying disorder of connective tissue matrix in many of these patients (13, 14, 19).

Details of the neurosurgical techniques applied in order to achieve cessation of the leak in the spontaneous spinal CSF leaks are outlined by my neurosurgical colleagues in the studies of Schievink et al. (14) and Cohen-Gadol et al. (13), as well as in a review by Maher et al. (15).

Sometimes all efforts, including surgery, fail to stop the leak. The rate of such disappointing results has not yet been determined, but they are fortunately infrequent (16).

Recurrent CSF leaks may occur after weeks, months, or years. The exact rate of recurrence also has not been determined, but it should be modest. Those patients with multiple meningeal diverticula and with disorders of the connective tissue matrix are likely to be at higher risk for recurrence of leak at single or multiple sites (29).

Subdural haematomas, unilateral or bilateral, are among the complications of CSF leaks. These are frequently small and may be asymptomatic, but sometimes become large and symptomatic, cause notable mass effect and shift the midline structures. Overall, many, although not all, subdural haematomas in spontaneous CSF leaks can be managed by repair of the CSF leak without need to evacuate the haematoma. Sometimes enlarging symptomatic subdural haematomas call for neurosurgical intervention. However, the risk of recurrence of the subdural is higher if the CSF leak is left untreated (16). In active spontaneous CSF leaks, sometimes further sinking of the brain may occur following craniotomies.

CONCLUSION

1. 1.

   SIH often, if not always, results from spontaneous CSF leak. In an overwhelming majority of the patients these leaks are at the level of the spine.

2. 2.

   The anatomy of these leaks is often complex, with pre-existing zones of dural defects or attenuation, ptobably related to a disorder of connective tissue matrix, and with CSF-weeping herniations of the underlying arachnoid, often at the level of nerve root sleeves and sometimes in the anterior aspect of the dura. Also common are meningeal diverticula of various sizes, often at nerve root sleeves or nerve root axilla.

3. 3.

   It is essential to locate the site or sites of the CSF leak preoperatively with appropriate neurodiagnostic techniques. It is also essential that these studies are chosen properly and executed and interpreted accurately.

4. 4.

   Surgery should be considered when patients are sufficiently symptomatic to warrant surgery, when site or sites of the leak have been clearly demonstrated preoperatively, and when symptoms have been refractory to less invasive measures.

5. 5.

   Surgery in well-selected cases is often, although not always, associated with good outcome with durable resolution of symptoms in approximately two-thirds of patients and with substantial improvement in one-quarter.

6. 6.

   Recurrences may occur after weeks, months or years. The exact rate or pattern of recurrence has not been determined, but the risk of recurrence is likely to be only modest. Those patients with evidence of disorders of connective tissue matrix, pre-existing dural defects and meningeal diverticula are probably at higher risk.