Article Text

MRI with intrathecal gadolinium to detect a CSF leak: a prospective open-label cohort study
  1. Ludo J Vanopdenbosch1,
  2. Peter Dedeken1,
  3. Jan W Casselman2,
  4. Stephan A P A Vlaminck3
  1. 1Department of Neurology, AZ Sint Jan Brugge Oostende AV, Brugge, Belgium
  2. 2Department of Radiology, AZ Sint Jan Brugge Oostende AV, Brugge, Belgium
  3. 3Department of ENT Surgery, AZ Sint Jan Brugge Oostende AV, Brugge, Belgium
  1. Correspondence to Dr Ludo J Vanopdenbosch, Department of Neurology AZ Sint Jan Brugge Oostende AV, Ruddershove 10, Brugge 8000, Belgium; ludo.vanopdenbosch{at}


Objectives The techniques currently used to detect a cerebrospinal fluid (CSF) leak are an indium radionucleotide scan and a CT scan with intrathecal iodinated contrast agent. They have a low spatial and temporal resolution and are unpleasant for the patient. This open-label prospective observational cohort study was designed to investigate the feasibility, success ratio, complications and therapeutic consequences of MRI with gadolinium administered by lumbar puncture to detect a CSF leak.

Methods Patients were selected with either confirmed liquorrhoea, recurrent bacterial meningitis, or symptoms, and MRI findings of spontaneous intracranial hypotension. High-resolution T1 weighted MRI with fat suppression of the spinal column at 1 h and of the brain at 6 and 24 h postinjection of 0.5 ml of gadolinium were performed.

Results 27 patients were included. The clinically suspected CSF leak was found in six of eight patients with liquorrhoea, three of five patients with recurrent bacterial meningitis and nine of 14 patients with spontaneous intracranial hypotension. The procedure was easy to perform and generally well tolerated. One patient developed streptococcal meningitis in the hours following the procedure but recovered completely with antibiotic treatment. 17 of 18 patients in whom a dural defect was found underwent surgery. All patients became symptom-free after closure of the dural leak.

Conclusions Spinal cord and brain MRI after intrathecal gadolinium injection is an easy-to-perform and accurate technique for detection of a dural defect with excellent anatomical detail.

  • intracranial hypotension
  • cerebrospinal fluid
  • magnetic resonance imaging
  • CSF
  • headache

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The diagnosis of a cerebrospinal fluid (CSF) leak can be suspected in patients with orthostatic headache or a sweet-tasting clear nasal drip. Headache attributed to low CSF pressure is characterised by a dull headache that worsens within 15 min of sitting or standing and that can be associated with tinnitus, hyperacusis, neck stiffness, photophobia and nausea. Nasal CSF drip can be confirmed with high specificity and sensitivity by beta-trace protein detection. Spinal CSF leaks can occur after lumbar puncture, complicating spinal surgery or epidural infiltrations. Cranial CSF leaks can be a complication of head trauma, surgery to the ear and nose or neurosurgical interventions. Idiopathic intracranial hypotension or spontaneous intracranial hypotension was recently reported to be caused by spinal epidural leaks. Cranial MRI with intravenous gadolinium can show meningeal thickening and contrast enhancement, downward displacement of the cerebellar tonsils, subdural collections, engorgement of veins and pituitary hyperaemia. The treatment of a CSF leak usually consists of pain medication and prolonged bed rest. A blood patch is successful in up to 75% of patients.2 With a blood patch, 20–50 ml autologous blood is injected into the lumbar epidural space. The mode of action is not known. Refractory spinal and cranial CSF leaks require surgery. The techniques currently used to detect and localise a dural defect with CSF leak are an indium radio nucleotide scan or a CT scan with intrathecal iodinated contrast agent. They have significant limitations in spatial and temporal resolution, and important side effects with meningeal irritation, large volumes to be injected, imaging artefacts and irradiation dosage. Intermittent leaks can be missed with both techniques.2 MRI with intrathecal gadolinium was reported to be safe and a promising technique to detect CSF leaks.3–5 Intravenous administration of gadolinium is rarely complicated with nephrogenic systemic fibrosis. There is no information on whether this could be an issue with intrathecal administration. The intravenous dosage usually is 20 ml, while only 0.5 ml is injected intrathecally.


We included patients 18 years or older with nasal liquorrhea confirmed by beta-trace protein, recurrent bacterial meningitis or symptoms and MRI findings of spontaneous intracranial hypotension. When a spinal leak was suspected, a lumbar epidural blood patch was performed and, if needed, repeated up to three times. A lumbar puncture was performed at lumbar level L3–4 with a 22-Gauge atraumatic needle, and 0.5 ml of gadolinium (Magnevist, Schering, Germany) was injected. The patient was positioned in the Trendelenburg position for 30 min. From February 2005 to January 2006, MRI scanning was performed with a 1.5 T Philips Achieva (Best, The Netherlands) and from then onwards with a 3 T Philips X-series Quasar Dual Achieva. The full spine was scanned at 1 h postinjection with high-resolution T1 with fat suppression with coronal and sagittal slices and axial slices through regions of interest. Thin T1 with fat suppression coronal slices of the head including all paranasal sinuses, the middle ear and mastoid bone were taken at 6 and 24 h postinjection. All scans were performed and reported by the same neuroradiologist (JWC). The local ethics committee reviewed and approved this study design. Written informed consent was obtained from all patients.


Between February 2005 and January 2009, 27 patients were included: eight with confirmed nasal CSF drip, five with recurrent bacterial meningitis and 14 with spontaneous intracranial hypotension (see online table). The average age was 45 (18–73), 18 patients were female, nine male. In six out of eight patients with confirmed nasal CSF drip, a dural leak was detected. This leak was through the ethmoid bone in all, mainly at the olfactory recessus and lamina cribrosa. A CSF leak was found in three out of five patients with recurrent meningitis. Two of these leaks were through the ethmoid bone in close association with the olfactory bulb. One patient had a leak through the roof of the petrous bone into the middle ear. In nine of 14 patients with spontaneous intracranial hypotension, we found a CSF leak. All nine were spinal epidural leaks mainly at the cervicodorsal junction (figure 1). The intrathecal injection of 0.5 ml of gadolinium by lumbar puncture was well tolerated. No new headaches occurred. We did not observe drowsiness, behavioural disturbances, attention deficit or epileptic seizures. One patient with liquorrhea developed meningitis 12 h after intrathecal gadolinium injection presenting with rapidly worsening headache, fever and obtundation. CSF obtained within 2 h of the first symptoms contained 29 600 white blood cells/mm3, with a 93% neutrophilic formula. CSF lactate was 14.2 meq/l and glucose 0 mg/dl. Treatment with 2 g of ceftriaxon and 1 g of vancomycin intravenously was started promptly. Streptococcus pneumoniae was cultured in CSF, and the patient was treated for 14 days with intravenous ceftriaxon (2 g) twice daily with full recovery. Documented nasal CSF leaks were closed by the ENT specialist with endoscopic nasal surgery using fibrin glue (Tissucol, Baxter Healthcare, Deerfield, Illinois). The spinal epidural CSF leaks except one were closed by the neurosurgeon. One spinal epidural leak resolved spontaneously. The neurosurgeon observed dural tears associated with minor osseous protrusions or a herniated thoracic disc. In some cases, an artificial dural patch was applied. All patients who underwent surgery had a complete resolution of symptoms.

Figure 1

Spinal epidural cerebrospinal fluid leak. Axial (A, B) and sagittal (C) 4 mm thick T1-weighted images of the thoracal spinal column with fat suppression following intrathecal gadolinium injection. (A) A disc-osteophyte protrusion is causing a disruption of the anterior wall of the dural sac, and gadolinium is leaking through a defect in the dura mater (arrow) into the anterior epidural spaces on the left and right side. (B) At an adjacent level, one can see another disc herniation, but the signal intensity of the intradural gadolinium is much higher than the gadolinium-cerebrospinal fluid intensity in the surrounding epidural collection (arrow), indicating that the entry of gadolinium in the epidural space is not situated at this level. (C) On the sagittal image, a Dubbel fluid space can be seen anterior to the medulla: intrathecal gadolinium is separated from the gadolinium in the epidural collection by the dura mater (arrowheads).


This prospective observational cohort was set up to study the feasibility, success ratio, complications and therapeutic consequences of MRI with intrathecal gadolinium. We did not make a direct comparison with iodinated contrast CT myelography or cisternography with radiolabelled indium, since it was found unethical to subject patients to three sequential intrathecal contrast injections. Positive predictive values and sensitivity therefore cannot be calculated. MRI with intrathecal gadolinium proved to be easy to perform. The contrast ratio and detail of the scans were excellent with superior spatial and temporal resolution. The gadolinium remains in the cranial cavity over 24 h permitting detection of intermittent leaks. We found a leak in 18 of 27 carefully selected patients (67%). The intrathecal injection of gadolinium was well tolerated. No behavioural disturbances, seizures or new headaches occurred. It should be noted that the majority of the patients had orthostatic headaches at presentation, and no worsening of headache was reported. One patient developed bacterial meningitis in the hours after the injection; he was treated with antibiotics and made a full recovery. The cultured S pneumoniae suggests a nasal entry through the documented ethmoidal CSF leak rather than a complication of the intrathecal injection with gadolinium. Seventeen of 18 patients in whom a dural defect was found successfully underwent surgical closure of the defect and became symptom-free. All patients with spontaneous intracranial hypotension in which a leak was found had the same MRI finding of spinal epidural collections (figure 1). The gadolinium-enhanced CSF leak through a dural defect was usually seen at the cervicodorsal junction in association with a small herniation of an intervertebral disc. Recent publications have reported similar findings.2 6–9 In a minority of patients, we could not detect a CSF leak, which is also in concordance with other observations.10 The fact that a spinal epidural leak being the cause of spontaneous intracranial hypotension was only reported in recent years makes us believe that MRI is superior to CT with intrathecal iodinated contrast or radiolabelled indium cisternography.


I thank my colleagues O Deryck, G Vanhooren, M Van Zandijcke and K Verhoeven, for useful comments and correcting the manuscript, and R Chahal, for language assistance. Patients were referred by: P Staels, Department of Neurosurgery, S Vlaminck and R Kuhweide, Department of Nose-ear-throat, AZ Sint Jan Brugge Oostende, Brugge; J Delbecq, Department of Neurology, Regionaal Ziekenhuis Sint Maria, Halle; M De Weweire, Department of Neurology, Stedelijk Ziekenhuis, Roeselare; T Van Havenbergh, K De Smet, Department of Neurosurgery, T Somers, Department of ENT, Sint Augustinus GZA Ziekenhuis, Antwerpen; D Hemelsoet and K Paemeleire, Department of Neurology, Universitair Ziekenhuis, Gent; I Van Walleghem, Department of Neurology, OLV van Lourdes Ziekenhuis, Waregem, M Vandenheede, Department of Neurology, CHC Clinique de l'Espérance, Montégnée; J Caekebeke, Department of Neurology, OL Vrouw Ziekenhuis, Aalst, P Soors, Department of Neurology, Virga Jesse Ziekenhuis, Hasselt, P Leuridan, Department of Neurology, Sint Rembert Ziekenhuis, Torhout, Belgium.


Supplementary materials


  • Presented at the annual meeting of the American Academy of Neurology, advances in headache integrated neuroscience platform session, Seattle, 29 April 2009. The methods and first three patients were published in a Dutch journal.1

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the Commissie voor ethiek AZ Sint Jan Brugge Oostende AV, Ruddershove 10, 8000 Brugge, Belgium.

  • Provenance and peer review Not commissioned; externally peer reviewed.