LOCALIZATION OF CEREBROSPINAL FLUID LEAKS BY GADOLINIUM-ENHANCED MAGNETIC RESONANCE CISTERNOGRAPHY: A 5-YEAR SINGLE-CENTER EXPERIENCE

Abstract

OBJECTIVE

Intrathecal gadolinium (Gd)-enhanced magnetic resonance (MR) cisternography is a newly introduced imaging method. Two main objectives of this study were to investigate the sensitivity of Gd-enhanced MR cisternography for presurgical localization of cerebrospinal fluid (CSF) leaks in patients with CSF rhinorrhea and to study the potential long-term adverse effects of intrathecal Gd application.

METHODS

Fifty-one patients (19 women; mean age, 36.2 yr) with CSF rhinorrhea were included in the study. A total of 0.5 ml of Gd was injected into the lumbar subarachnoid space. T1-weighted MR cisternographic images were obtained to detect CSF leakage. The patient's neurological states and vital signs were recorded for the first 24 hours after the procedure. Neurological evaluations were repeated 1, 3, and 12 months after the procedure. The patients were followed for at least 3 years with annual neurological examinations.

RESULTS

Gd-enhanced MR cisternography demonstrated CSF leaks in 43 of the 51 patients. The sensitivity of Gd-enhanced MR cisternography for localization of CSF leaks was 84%. Forty-four patients underwent surgery to repair dural tears. Surgical findings confirmed the results of Gd-enhanced cisternography in 43 of the 44 patients who underwent surgery (98%). Eight patients with negative Gd-enhanced MR cisternography had no active rhinorrhea at the time of procedure, and seven of them did not need surgery. None of the patients developed an acute adverse reaction that could be attributed to the procedure. None of the patients developed any neurological symptoms or signs caused by intrathecal Gd injection during a mean follow-up period of 4.12 years.

CONCLUSION

Gd-enhanced MR cisternography is a sensitive and safe imaging method for detection of CSF leaks in patients with rhinorrhea.

Rhinorrhea is defined as the leakage of cerebrospinal fluid (CSF) from the subarachnoid space into the paranasal sinuses and subsequently into the nasal cavity. Most cases of CSF rhinorrhea develop after an accidental and iatrogenic injury to the dura mater, although nontraumatic and spontaneous cases have been reported (16,19). CSF leakage stops spontaneously within a week in approximately 70% of traumatic cases (5,15). In persistent or recurrent cases, the CSF leakage site may be a port of entry for bacteria into the subarachnoid space, subsequently leading to recurrent meningitis. In traumatic cases, the first choice of management is conservative treatment, including bed rest with head elevation, diuretics, and lumbar drainage (16). In patients who do not respond to conservative treatment, surgical repair of the dural tear is indicated. Presurgical radiological identification of the CSF leakage site guides the surgeon during operation, increases the success rate of operation, and decreases the risk of complications. Although anatomic resolutions of radiological imaging modalities have recently improved, presurgical radiological localization of CSF leakage is still a challenging problem for both neuroradiologists and neurosurgeons (13). Thin-slice high-resolution computed tomographic (CT) examination, T2-weighted magnetic resonance (MR) imaging scans of the cranial base, water soluble contrast-enhanced CT cisternography, and radionuclide cisternography are currently used for presurgical localization of CSF leakage in patients with rhinorrhea (8,9,10,17,24). However, these imaging techniques cannot adequately address the problem of localization of CSF leakage. Intrathecal injection of iodinated contrast agents for CT cisternography is not a completely safe procedure. Arachnoiditis, epilepsy, allergic reactions, and rarely intracranial bleeding caused by intrathecal injection of iodinated contrast agents have been reported (22,27).

Radionuclide cisternography is another invasive imaging method used to detect CSF leakage sites (24). Although radionuclide cisternography has sensitivity similar to CT cisternography, its poor spatial resolution limits its value in the precise presurgical localization of CSF leakage. Thin-slice, high-resolution CT scanning is a noninvasive imaging method that has a sensitivity of 71% for detection of bony defects in patients with CSF leakage (24). Compatible clinical and CT findings may eliminate the need for further invasive imaging methods. However, in some cases, CT images cannot prove that the observed fracture line is the actual site of CSF leakage (13,24). The actual site of CSF leakage may differ from fracture lines observed on thin-slice CT images. Moreover, multiple fractures on the walls of paranasal sinuses adjacent to CSF-containing spaces at the cranial base may lead to inconclusive examinations. Congenital or traumatic bony defects in the absence of an accompanying meningeal defect may result in a false-positive diagnosis by high-resolution CT examination.

MR cisternography is another noninvasive imaging method used in the evaluation of patients with CSF rhinorrhea. The presence of a continuous T2-weighted high signal extending from the subarachnoid space into the paranasal sinuses is considered a sign of CSF leakage in MR cisternography (8,9,10). However, nonspecific mucosal inflammation in paranasal sinuses may mimic the appearance of a CSF leak and lead to a false-positive diagnosis (11,13).

Gadolinium (Gd)-enhanced MR cisternography is a newly introduced imaging method that can be used for the detection of CSF leakage in patients with CSF rhinorrhea (7,13). Although intrathecal use of gadopentetate dimeglumine (Magnevist; Schering, Berlin, Germany) has not been approved by the United States Food and Drug Administration, preliminary clinical studies in relatively small patient groups have recently reported the success of Gd-enhanced MR cisternography in the localization of CSF leakage in patients with rhinorrhea (3,13,29). The patient groups in these preliminary reports were not large enough to measure the sensitivity of Gd-enhanced MR cisternography for detection of CSF leaks. In the current study, we studied a relatively large series of patients who were clinically suspected to have CSF rhinorrhea. We aimed to measure the sensitivity of Gd-enhanced MR cisternography for presurgical localization of CSF leak in patients with rhinorrhea. In previous studies, the patients were clinically followed for 9 to 12 months after Gd-enhanced MR cisternography (3,13,29). Although no side effect that could be attributed to intrathecal Gd application was observed during follow-up periods in previous studies, a study with a longer follow-up period is needed to evaluate the safety of the procedure. In our study, we followed our patients for 3 to 6 years to investigate the potential long-term adverse effects of intrathecal Gd application.

PATIENTS AND METHODS

Patients

Fifty-one patients (19 women, 32 men) who presented with persistent or intermittent rhinorrhea for more than 2 weeks were included in the study. The mean age of the patients was 36.2 ± 10.3 years (range, 19–61 yr). CSF rhinorrhea was posttraumatic in 59%, iatrogenic in 23% (after a cranial base surgery or endoscopic application), and spontaneous in 18% of the patients. The laboratory analysis of rhinorrhea fluid revealed a high glucose (>30 mg/ml) and protein content (>45 mg/dl) in all cases. A β2 transferrin test was performed and confirmed CSF leaks in 41 (80%) of the patients. Two patients (4%) had at least one attack of bacterial meningitis. None of the patients had any clinical sign of meningitis at the time of intrathecal injection.

Gd-enhanced MR Cisternography

We performed Gd-enhanced MR cisternography to confirm and localize CSF leakage in all patients. Gd-enhanced MR cisternographic images were obtained with a 1.5-T MR scanner (Symphony; Siemens Medical Systems, Malvern, PA). Before intrathecal Gd injections, nonenhanced MR images were obtained. Nonenhanced MR sequences included coronal T1-weighted images (repetition time [TR], 524 ms; echo time [TE], 14 ms; number of excitations [NEX], 2; slice thickness, 3 mm), sagittal T1-weighted images (TR, 524 ms; TE, 14 ms; NEX, 2; slice thickness, 3 mm), axial T1-weighted images (TR, 524 ms; TE, 14 ms; NEX, 2; slice thickness, 3 mm), coronal fat-suppressed T1-weighted images (TR, 524 ms; TE, 14 ms; NEX, 2; slice thickness, 3 mm), sagittal fat-suppressed T1-weighted images (TR, 524 ms; TE, 14 ms; NEX, 2; slice thickness, 3 mm), and axial fat-suppressed T1-weighted images (TR, 524 ms; TE, 14 ms; NEX, 2; slice thickness, 3 mm). After obtaining nonenhanced MR images, we performed a lumbar puncture by using a 21-gauge needle at the L4–L5 level under sterile conditions. We collected 5 ml of CSF before contrast injection. Undiluted gadopentetate dimeglumine (0.5 ml, Magnevist; equals 250 μmol of gadolinium; corresponds to a dose of 0.17 μmol/g brain if the average weight of a human brain is accepted as 1400 g) was injected into the subarachnoid space (3,13). Then, the CSF obtained was reinjected to fill the dead space of the needle to ensure that the entire 0.5 ml of contrast medium was injected into the subarachnoid space. Patients were placed in a prone postion at 30 to 40 degrees, with their head down for 15 minutes after withdrawal of the needle to maximize accumulation of contrast medium in the basal cisterns and to provoke its passage into the CSF fistula. T1- and T1-weighted fat-suppressed MR cisternographic images with the same parameters as previously described were obtained with the patient prone and with their chin up. If there was insufficient intracranial contrast medium for analysis, the imaging study was repeated 30 minutes later. Two experienced neuroradiologists (KA and SS) evaluated the Gd-enhanced MR cisternographic examinations by comparing nonenhanced and enhanced MR images. In the evaluation, an extension of hyperintense Gd-enhanced CSF from cerebral cisterns into the paranasal sinuses or the nasal cavity on Gd-enhanced MR cisternographic images was considered as the positive finding of CSF leakage. In cases with an accumulation of contrast agent in the paranasal sinuses without demonstration of CSF leak itself, MR scanning was repeated without delay.

All patients were hospitalized for 24 hours after the procedure. Neurological examinations and vital signs of the patients were recorded before intrathecal injection; 30, 60, 120 and 180 minutes; and 24 hours after the procedure. Neurological evaluations were repeated 1, 3, and 12 months after Gd-enhanced MR cisternography. After the 1-year clinical evaluation, the patients were followed annually for at least 3 years. The mean follow-up period was 4.12 ± 0.84 years (range, 3–6 yr). During the annual evaluations, the patients were asked about the development of any new neurological symptoms unrelated to the primary pathology leading to rhinorrhea. Detailed neurological examinations were also performed at the annual follow-up examinations. Any difference between preprocedural and follow-up neurological findings was sought. The study was approved by the local Human Subject Committee. Informed consent was obtained from all patients.

RESULTS

MR images obtained after intrathecal Gd injection showed optimal enhancement of the subarachnoid space in the cerebral cisterns and sulci in all patients. Gd-induced enhancement of CSF provided an excellent contrast among the subarachnoid space, calvarial bones, and paranasal sinuses on T1-weighted and fat-suppressed MR images (Figs. 1 and 2).

Gd-enhanced MR cisternographic images demonstrated CSF leakage in 43 of the 51 patients (84%). The CSF leakage site was in the ethmoidal region in 26 (60%) patients, in the posterior wall of the frontal sinus in six (14%) patients, and in the superior wall of the sphenoid sinus in 11 (26%) patients (Figs. 1–4). Gd-enhanced MR-cisternographic examinations were negative in eight (16%) of the 51 patients. None of the patients with negative Gd-enhanced MR cisternography had active rhinorrhea during intrathecal injection. Surgical closures of CSF leaks were performed in all 43 patients with positive Gd-enhanced MR cisternography. CSF leaks with a large cranial base defect and the ones that were related to penetrating trauma or located in the frontal sinus were primarily managed with open surgery. Endoscopic repair of CSF leakage was preferred in spontaneous, iatrogenic cases and in patients with sphenoethmoidal CSF leakage. One patient with negative Gd-enhanced MR cisternography had experienced two previous attacks of bacterial meningitis. Therefore, this patient also underwent surgery. A CSF leakage in the ethmoidal region was identified and repaired successfully in the operation (false-negative case).

In all patients with positive Gd-enhanced MR cisternography who underwent surgery, CSF leakage observed on Gd-enhanced MR cisternographic images was confirmed during surgery. Dural repair was successful in 39 out of 44 of these patients. However, a second operation was needed in four of the operated patients in whom recurrent CSF leakage could be closed successfully. Seven posttraumatic patients with negative Gd-enhanced MR cisternography were treated with conservative treatment. CSF leakage was stopped without a need for surgery in all patients who were followed conservatively. If we considered all the patients in this study (operated and nonoperated), the sensitivity of Gd-enhanced MR cisternography for detection of CSF leaks was calculated as 84%.

Of the 51 patients, 12 (24%) had a headache within 12 hours of the intrathecal injections. In all cases, the headaches responded to oral analgesic treatment and subsided within 24 hours after intrathecal Gd injection. Within 24 hours after the procedure, none of the patients developed an acute adverse reaction (other than headache) such as seizure, change in behavior or consciousness, development of any focal neurological sign, or any type of allergic reaction. Vital signs of all patients were normal and remained stable during the 24 hours after intrathecal injection. There was no difference between preprocedural and postprocedural neurological findings within 24 hours of follow-up. At the annual follow-up examinations, none of the patients demonstrated any neurological symptoms or signs that could be attributed to intrathecal Gd injection. In all patients, there was no difference between the preprocedural findings and the findings recorded at the annual neurological examinations.

DISCUSSION

Rhinorrhea is etiologically classified into traumatic and nontraumatic types (spontaneous). Head trauma is the most common cause of rhinorrhea, complicating 2% of all head injuries (5,13,15). Traumatic rhinorrhea develops after a blunt or penetrating head injury. Iatrogenic CSF leaks may develop after neurosurgical or endoscopic procedures of the paranasal sinuses. Most traumatic CSF leaks stop spontaneously. Therefore, the first line of treatment for traumatic CSF leakage is conservative treatment, including bed rest with head elevation, diuretics, and lumbar drainage (16). Surgical repair of dural tears is indicated in persistent or recurrent leaks. The chance of spontaneous closure in nontraumatic CSF leaks is much lower. Therefore, surgical treatment is preferred as a first choice in the management of nontraumatic rhinorrhea (16).

Presurgical radiological definition of the CSF leakage site is important in both traumatic and nontraumatic cases (13,16,19). It confirms CSF leakage and helps surgical planning. Precise localization of CSF guides the surgeon and increases the surgical success rate. Diagnostic procedures used in the evaluation of CSF leakage in patients with rhinorrhea have evolved during the past two decades. Currently, the most widely accepted method for evaluating patients with a suspected CSF leak is the combination of high-resolution CT scanning with a contrast-enhanced CT cisternography. Water-soluble iodinated contrast agent is injected into the subarachnoid space for CT cisternography, which is not an entirely safe procedure. Intrathecal iodinated contrast injection may cause severe allergic reactions, seizure, and even intracranial hemorrhage (22,27). Moreover, the combination of high-resolution CT images with CT cisternography doubles the radiation dose to which a patient is exposed. In cases with a delayed opacification of subarachnoid cisterns, repeated CT examinations increase the radiation dose even more. The sensitivity of CT cisternography for detection of CSF leakage is approximately 72 to 81% (4,5). Sensitivity of CT cisternography is even reduced for slow-flow CSF leakage (9).

In 1985, Di Chiro et al. (7) introduced the first intrathecal use of gadopentetate dimeglumine that would increase the diagnostic accuracy of CSF leakage and eliminate radiation exposure in the evaluation of rhinorrhea. After this initial report, several animal studies were conducted to investigate the safety of intrathecal application of gadopentetate dimeglumine (14,20,26). Ray et al. (20) reported that intraventricular injection of gadopentetate dimeglumine above a threshold dose of 5 μmol/g brain produced neurotoxicity in rats, which resulted in severe motor disturbances and behavioral changes. However, the neurotoxic effect of gadopentetate dimeglumine did not develop below a dose of 3.3 μmol/g of brain. Other animal studies showed that intrathecal gadopentetate dimeglumine injection at relatively low doses did not cause any pathological changes in the central nervous system (14,26). The feasibility of intrathecal Gd-enhanced MR cisternography in the detection of CSF leaks was tested in an animal model of traumatic rhinorrhea. In this animal study, Ibarra et al. (12) stated that Gd-enhanced MR cisternography is a promising imaging technique for the evaluation of posttraumatic CSF leakage. After the success of these animal model studies, the first clinical intrathecal use of gadopentetate dimeglumine was reported in two patients with meningial carcinomatosis (23).

In 1999, Zeng et al. (29) reported a prospective human study investigating the safety and clinical use of intrathecal gadopentetate dimeglumine application for enhanced MR myelographic and cisternographic examinations. They injected 0.2 to 1.0 ml of gadopentetate dimeglumine, corresponding doses of 0.07 to 0.36 μmol/g of brain, for Gd-enhanced MR myelographic and cisternographic examinations. No significant neurological abnormalities were observed during a 9- to 15-month follow-up period. In a multicenter human study, Jinkins et al. (13) used intrathecal gadopentetate dimeglumine at a dose of 0.17 μmol/g of brain in the evaluation of CSF leakage in 15 patients with rhinorrhea. They did not report any neurological or electrophysiological abnormalities during a follow-up period of 6 to 12 months. Thus, they stated that Gd-enhanced MR cisternography using gadopentetate dimeglumine at a dose of 0.17 μmol/g of brain is a safe and feasible technique in confirming the presence and defining the localization of CSF leakage.

A recent study demonstrated the safety of low-dose intrathecal Gd application in pediatric patients (18). Another recent article reported the findings of intrathecal Gd–induced encephalopathy in humans, which was caused by an accidental high-dose intrathecal Gd injection (2). In this unfortunate case, the patient developed severe neurological symptoms such as dysarthria, blurred vision, nystagmus, ataxia, and somnolence after the accidental intrathecal injection of a very high dose of Gd (20 ml of gadopentetate dimeglumine; 7 μmol/g of brain). Behavioral disturbances and psychotic symptoms followed the focal neurological symptoms. However, most of the symptoms resolved within 2 weeks. Two months after intrathecal Gd application, neurological examination revealed only a very mild ataxia.

In our study, we reported successful application of Gd-enhanced MR cisternography for presurgical localization of CSF in a large series of patients with rhinorrhea. We followed the clinical neurological states of our patients for a mean period of 4.12 years. To our knowledge, our study had the longest follow-up period compared with previous studies of intrathecal gadopentetate dimeglumine application (1,3,13,18,21,25,28,29). During this relatively long follow-up period, we did not observe any adverse effects in our patients (other than a transient headache) that could be attributed to intrathecal gadopentetate dimeglumine injection with a dose of 0.17 μmol/g of brain. The headache observed in our patients was minor and responsive to oral analgesics. It was likely related to lumbar puncture. When we consider our long-term follow-up results along with the results of previous reports, we can state that low-dose (0.17 μmol/g of brain) Gd-enhanced MR cisternography is a safe procedure for the evaluation of patients with CSF rhinorrhea.

In our study, CSF leakage was confirmed by surgery in all patients with positive Gd-enhanced MR cisternography. Among 44 operated patients, Gd-enhanced MR cisternography did not detect the CSF leak in only one patient; this was the only false-negative case among the operated patients. However, there were no false-positive cases in our study. The sensitivity of Gd-enhanced MR cisternography was 84% in our study. Because CSF leak may be intermittent, the success of cisternographic procedures depends on the timing of the examination. All of our patients with negative Gd-enhanced MR cisternography did not have active rhinorrhea during the procedure and did not show recurrence during the follow-up period. If we consider only the operated patients, Gd-enhanced MR cisternography successfully localized the CSF leakage in 43 (98%) of the 44 operated patients. Because of ethical considerations, it would be impossible to conduct a prospective study to compare the sensitivities of CT cisternography and Gd-enhanced MR cisternography for detection of CSF leakage in the same patient population.

However, if we compare our results with the previously reported sensitivies of CT cisternography, Gd-enhanced MR cisternography seems to be more sensitive in the detection of CSF leakage in patients with rhinorrhea (4,5). In our previous study, Gd-enhanced MR cisternography detected CSF leakage in some of the patients whose CT cisternographic examinations had revealed no leak. Gd-enhanced MR cisternography has certain advantages over CT cisternography. High viscosity of iodinated contrast agent that is used for CT cisternography may impair its free distribution in the subarachnoid space and its passage into slowly flowing CSF fistulae. Accumulation of a certain amount of iodinated contrast agent is needed to attenuate x-rays and to enhance CSF fistulae on CT cisternographic images. However, gadopentetate dimeglumine distributes easily in the subarachnoid space and enhances CSF fistulae. In addition, a much lower volume of Gd is needed to enhance CSF in Gd-enhanced MR cisternography compared with the volume of iodinated contrast agent used in CT cisternography. In some cases, a thin line of contrast in the CSF leakage site on CT cisternographic images may not be differentiated from adjacent bony structures, which have a CT density value similar to that of the contrast agent diluted with CSF. Fat-suppressed T1-weighted Gd-enhanced MR cisternographic images increase the conspicuity of contrast medium leakage due to saturation of medullar bone fat in cranial base and nasal cavity.

CONCLUSION

Gd-enhanced MR cisternography is a sensitive imaging method for the detection of CSF leakage in patients with rhinorrhea. This method has a sensitivity value of 84%. Lack of any adverse effect in our long-term follow-up study demonstrated the safety of Gd-enhanced MR cisternography for presurgical investigation of CSF leakage in patients with rhinorrhea.

REFERENCES

COMMENTS

This long-term follow-up of patients who have had gadolinium (Gd)-enhanced magnetic resonance (MR) cisternography found no ill effects of intrathecal injection of Gd and a high degree of sensitivity in detection of the site of leakage.

Robert G. Grossman

Houston, Texas

Aydin et al. convincingly demonstrate the safety of intrathecal Gd-enhanced MR cisternography. There are several theoretical advantages of this technique over contrast-enhanced computed tomographic (CT) cisternography, including reduced risk of arachnoiditis and reduced exposure to x-rays. The authors do not compare MR cisternography with CT cisternography so we do not know which technique is more sensitive or specific. Such a study was recently reported by Goel et al. (2) in which patients had both CT and MR cisternography and MR cisternography was clearly more sensitive. The fallibility of CT cisternography was attributed to the density of the contrast material being similar to that of bone and the high viscosity of iodinated contrast material, which may limit passage across the defect. It should be noted that in higher doses, Gd is toxic when administered in the cerebrospinal fluid, and the term “gadolinium encephalopathy” has been used to describe this scenario (1). In our experience, once the site of the defect has been localized radiographically, the use of intrathecal fluorescein is extremely helpful during endoscopic endonasal repair (3,4). It is likely that MR cisternography will be extremely useful in certain patients with cerebrospinal fluid leaks, although dehiscences of the bone will be difficult to recognize.

Theodore H. Schwartz

New York, New York

The authors describe the use of intrathecal Gd-enhanced MR cisternography for the definition of cerebrospinal fluid leaks in 51 patients. They were able to identify the site of leak in 43 of 51 patients (84%). Although the technique is not new, most of the previous articles are found in the neuroradiology and radiology literature. Thus, the value of this article is to further inform the neurosurgical readership regarding the technique. There have been reported complications with the administration of intrathecal GD, although it was reassuring that none have been reported with the patients studied here.

This study does not compare CT cisternography in the same group of patients, but the reported sensitivity for leak detection using MR cisternography reported here compares favorably with previous reported CT cisternography data. CT scanning, however, provides complementary information, as it is easier to correlate site of leak with a bone defect.

William T. Couldwell

Salt Lake City, Utah