Slides
Transcript
Thank you for that wonderful introduction. My name is Angelique Tay. I used to be a resident with Dr. Schievink, and I’m here to talk about the systematic review of negative brain MRIs in patients with CSF-venous fistulas.
I remember when I first started liking CSF leak and I went up to Dr. Schievink when I was a budding neurosurgeon and I asked him, “Dr. Schievink, what is your diagnostic pathway for these patients?” and he goes, “Angelique, it’s all in here.” So I went to do my research and try to figure out my own diagnostic pathway for this, and it kind of spurred a lot of research with him.
We all know this, we’ve been talking about it for the past two days, but patients by and far usually show up with orthostatic headache. They get an MRI brain with and without. And if there’s a positive brain MRI, then we go for MR myelogram. If there’s extradural CSF, we do a prone DSM, we find the type 1. And then if there’s no extradural CSF, then we do lateral DSM and we’re looking for a fistula. And by now that’s pretty obvious.
But the question I had when I was more junior was how many patients with positive brain MRIs actually are found to have CSF leaks? And there’s a lot of literature out there right now and mostly basically 75% of the time you either find a ventral leak or a CSF-venous fistula either with a DSM or a dynamic CTM.
But then my question was what if the brain MRI is negative? What do we do with those patients? We go for MR myelogram, and if there’s extradural CSF, then we do a prone DSM and we find a type 1 leak, and that’s very obvious now. In the literature, we think about 20% of patients with a negative brain MRI and extradural CSF leak end up having a negative brain MRI, and that’s not surprising.
But then what about those without the extradural CSF fluid collection on MR myelogram? How likely will you find a CSF-venous fistula?
We went through five databases: PubMed, Embase, Web of Science, Scopus, and Cochrane, and we searched CSF-venous fistula and spontaneous intracranial hypertension, DSM, and brain MRI. We found 586 papers. After the exclusion criteria, we took out 37 papers for positive brain MRIs or an intermediate or high probability on Bern or Mayo scores, or if there were no CSF-venous fistula patients, or if no brain MRI was reported, or if no leak type was discussed, and if the CSF-venous fistula was identified first before they looked at the brain MRIs, or if there were overlapping patient cohorts, or the DSM or dynamic CTM was not done in lateral position, so considered not to have [inaudible] imaging.
We resulted with nine papers. In those papers, what did they consider a positive brain MRI? Well, typically I went with the SEEPS scores: subdural fluid collections, enhancement of the pachymeninges, engorgement of the venous structures, pituitary hyperemia, and sagging of the brain. And then four of the papers went with the Bern score.
And I think by now we all know that one. Subdural hematomas, pachymeningeal enhancement, venous distension sign, and then brain sagging with the measurements of MPD, suprasellar cisterns, and prepontine cistern, and then there’s the Mayo score that’s more recent. That one looked at the pachymeningeal enhancement, venous distension sign, the pituitary enlargement, and also suprasellar cistern effacement, and so cerebellar tonsillar herniation. A couple more points in there that the Bern score doesn’t really have is the dura enhancement in the cerebral aqueduct iter below the incisural line.
So what did we find? We found in the papers—four papers that looked at MRIs and categorized them as entirely negative or positive. There are 252 patients that are entirely negative, and 27 of them were positive, 10% of that. The other thing that is noted there is that a lot of these papers did not use the same imaging modality as their gold standard. So there’s a lot of lateral DSMs, decubitus CTMs, cone beam, photon counting. I’m not a radiologist, so I’d defer to them to say which one’s better. I think that’s still in controversy.
And again, that’s 10% to 11% of the patients that were considered negative brain MRIs who were found to have CSF-venous fistulas. For the Bern score studies, four papers, 82 patients total that had a low probability score, and 15 of those patients—so 18% of the patients—were found to have CSF-venous fistulas. And then if you look at the Mayo score, 109 of those patients had a low probability according to that score, and 38 patients were found to have CSF-venous fistulas. There’s an asterisk there because it was actually 43 patients, but the categorization of whether those patients had CSF-venous fistulas or type 1 was not sure for five of those patients. So we did the conservative measuring, just removed those five patients.
And ultimately what we found is if you look here that the negative predictor value of these scores and studies. If you took the four papers with the MRI categorized based off of entirely negative or any positive signs, if there was an entirely negative MRI, 90% of those patients did not have a CSF-venous fistula, so the predictive value was pretty high, 89%. For a Bern score it was 81%, and then for the Mayo score it was 65%.
So what questions arise from this data? Well, there’s two outliers here that we look at and why there are these differences between an entirely negative brain MRI and a Bern or Mayo score of 0 to 2.
The first thing about these outliers is that there’s no true gold standard for detection of CSF-venous fistula. Aside from the DSM, different techniques for the dynamic CTM have been evaluated in the literature. There’s differences in how these images are captured: the timing like contrast medium injection, the type of anesthesia done, sedation versus general anesthesia for clearer images, whether or not the saline is injected for positive pressurization prior to myelography, and whether respiratory maneuvers are employed during image capture.
One of those outliers had an interesting technique of combining modalities to increase diagnostic yield using DSM and cone beam CTM. They had 100 patients; 15 were excluded for extradural CSF or patients that didn’t meet SIH criteria in the ICDH criteria and did not undergo both imaging modalities, and they were left with 85 patients. 38 of the patients were found to have a CSF-venous fistula after DSM only, so about a 45% diagnostic yield. If they also had cone beam CTM after, the diagnostic yield increased to 69%.
And how many of these patients actually had a negative brain MRI? Well if you took those 85 patients and split them up, 25 of those patients had negative brain MRI and six were found to have a CSF-venous fistula, so about 24%. Four of those patients were only identified after they had both imaging modalities.
So it kind of raises questions about maybe we should be combining these. And then the other question we had was why are there differences between entirely negative brain MRIs and Bern or Mayo scores of 0 to 2? Well, first the Bern score was actually developed to predict the likelihood of finding an epidural leak on spine imaging, not likely entirely applicable to CSF-venous fistula patients who don’t usually have those SLECs.
The Mayo score was developed with only patients without SLECs, but they also could include a dural tear with a small localized extradural fluid collection. But I think the major difference is the fact that having a low Bern or Mayo score of 1 or 2 is not equivalent to entirely negative MRI, and not all the points in these scales are equal. And what do I mean by that?
There was a recent paper by Kang et al. who looked at 90 epilepsy patients, and they chose epilepsy patients because they have MRIs equivalent to SIH patients that were not clinically suspected for SIH. The average Bern score was actually 1.5 and the average Mayo score was 0. And what’s interesting is if they stratified these patients by low, intermediate, or high, the intermediate there were actually 20 patients with the intermediate score. So where are these points coming from? And we looked at that. The points are actually coming from the measurements of the prepontine cistern, the suprasellar cistern, and the minimal pontine distance. So maybe in patients with SIH we should be putting more weight on the subdural fluid collections or the pachymeningeal enhancement and/or venous sinus engorgement that isn’t found in these patients without SIH.
So back to the original question – what about these patients with a negative brain MRI? If there’s no extra dural fluid collection, what’s the likelihood of finding a CSF-venous fistula? So in the end we found that if there was an entirely negative MRI, 89% chance the patient does not have a CSF-venous fistula. If it was a Bern score of 0 to 2, it’s 81% chance, and if it was a Mayo score of 0 to 2, it’s 65% chance that the patient did not have a CSF-venous fistula.
However, while this is a very good negative predictive value with an entirely negative MRI, if there’s still high clinical suspicion of SIH, we would still recommend proceeding with a lateral dynamic CTM or lateral DSM.
So when do we proceed? Again, the high clinical suspicion, and as mentioned earlier by Dr. Schievink, the presence of meningeal diverticula on spine imaging. It was shown in previous papers that 20% of the patients positive for meningeal diverticuli were found to have a CSF-venous fistula. But if they had no diverticuli, they didn’t have a CSF-venous fistula.
The other thing we look at is the reduced optic nerve sheath diameter, which could increase the chances of finding a CSF-venous fistula leak in a patient with a negative brain MRI. We found that if you did a cutoff of an optic nerve sheath diameter of less than 4.4 mm and a perioptic subarachnoid space of less than 1.0 mm, you found CSF-venous fistulas in about 50% of those patients. And if they had values above that, less than 2% of those patients would have a CSF-venous fistula. So if optic nerve sheath diameter was considered a positive MRI, there’s a chance that the negative predictive value of a negative MRI would actually go up.
So this is what I ended up with in my diagnostic algorithm after talking to Dr. Schievink. If there is no extradural CSF fluid collection in a negative brain MRI, then if there’s a high clinical suspicion or optic nerve sheath diameter of less than 4.4 mm or a perioptic subarachnoid space of less than 1 mm and meningeal diverticuli, then we would proceed with a lateral DSM or dynamic CTM.
And thank you. These are my references, and I especially wanted to thank everyone who helped me with this, which was Dr. Madhavan, Dr. Kranz, Dr. Maya, and Dr. Shievink. Thank you.