Slides
Transcript
Thanks so much. Thanks for the invitation to speak. I’m going to talk about epidural blood patch for CSF leak. I’m taking the non-targeted side of this debate.
Okay, so let’s jump right in. This is a 23-year-old female with orthostatic headache, and I think the T1 sagittal tells most of the story here. This patient has SIH and a spinal CSF leak. So what’s the next step in management? Many of us would want to see a spine MR to see if there’s an epidural collection or not that may inform our myelography, but I think most of us would consider offering this patient an epidural blood patch, regardless of what the spine imaging shows. I’m going to talk about the benefits and perhaps some drawbacks of that.
Any other type of targeted therapy is going to require a localization study as we just heard, most likely either CT myelogram or digital subtraction myelogram, but there are some disadvantages of this as well.
So just before we jump in, we’re going to get into some nomenclature. What we’re talking about are non-targeted but image-guided patches, not blind patches. This is a paper with a scary title published in Headache a few years ago: “Arachnoiditis, a complication of epidural blood patch.” Tim and I actually wrote a letter to the editor that was published, and just to quote from it, “Image guidance is not described in any of the case reports and many reports describe frank intrathecal injection of blood products.” So to my knowledge there are no reports of arachnoiditis or other complications when image guidance is employed. So we suggested perhaps a better title: “Arachnoiditis following epidural blood patch—An avoidable rare complication due to blind technique.” So we’re talking about non-targeted but image-guided, not blind patches.
The image guidance can be fluoroscopy as shown here. When you inject contrast in the epidural space you want to see this intermixing of the epidural fat with your contrast to ensure you’re in the right position. If you have the ability to do a spin or a DynaCT you can confirm that you’re in the right location. In this case it was kind of interesting because you can see the contrast off to one side because of a mediana dorsalis or prominent dorsal dural reflection. CT myelography, or excuse me CT imaging, is also a great guidance technique for blood patches. I prefer CT fluoroscopy. It’s very safe, fast, and accurate.
So let’s go back to our case and think about a decision framework. If we’re going to offer an epidural blood patch that’s non-targeted, there are a few questions we probably want to ask ourselves. How effective is this therapy and what are the risks of the alternative diagnostics or therapeutics? Any type of localization study is going to require dural punctures. There’s some small risk of creating a new leak from that puncture site. Intrathecal contrast is very well tolerated but has a non-zero risk profile. And I think most importantly, there can be significant ionizing radiation exposure from these studies. Some other questions that come to mind, how does the blood patch work and how far does this patch spread?
So taking the radiation issue first – this is a nice paper from Farb’s group published recently. They were comparing the radiation dose to the patient from a DSM versus CTM. They found on average the dose for DSM was lower. The average for CTM was about 20 millisieverts, but you can see quite a wide range in the effective dose of the patient, all the way up to 80 millisieverts. And this is something we should all be aware of as physicians, but it is also something our patients are asking us about more.
This is a paper in JAMA Internal Medicine published earlier this year projecting the number of cancers as a result of diagnostic CT imaging, picked up by the lay press. Certainly an important topic.
This is an outside CT myelogram – few phases of acquisition. I was asked to review it, and I was really struck by this dose report. The DLP is almost 5,000, which translates to an estimated effective dose of nearly 80 millisieverts, which is quite a significant radiation exposure for one study, perhaps a lifetime attributable cancer risk as high as 1%.
So blood patch was first described in 1960 after the observation that a bloody tap resulted in lower rates of post-LP headache. The first report for blood patch for SIH was in the early 80s, but to date there are no randomized controlled studies of blood patching for CSF leak.
How does a blood patch work? There’s an immediate effect from the epidural blood resulting in increased CSF pressure and a subsequent sealing of the puncture site when successful. So essentially, you’re creating a controlled epidural hematoma. This causes a direct increase in epidural and CSF pressure, and CSF pressures as high as 80 centimeters of water have been observed after a 20 cc blood patch. Interestingly, this pressure increase is not sustained if you inject, say, normal saline in the epidural space.
And it’s been demonstrated both on MR imaging and with nuclear medicine studies that the patch spreads quite a number of levels.
This is a study using radiolabeled blood with technetium, and they followed the patch to see how far up and down the spine it spread. A couple of interesting observations: the first was the patch spreads quite far. In one patient they documented that a high lumbar patch spread 14 spinal segments. The other observation was that there appears to be a linear correlation between the volume of blood injected and the spread. Similar observations on an early MRI study just showing that the blood patch you give, even if it’s high lumbar or low thoracic, will spread quite a number of levels.
The persistent effect of a patch is from clot formation. That clot gives mass effect on the thecal sac and has been shown on MR to resorb pretty rapidly, but you’ll get a thin adherent clot to the dura that persists for some time. So after the injection the blood coagulates, and you have fibroblasts that invade the clot, transforming it into a durable scar and sealing the leak when successful.
In post-dural puncture headache this is highly efficacious. A large case series of over 500 patients – 75% of patients had complete success after the blood patch and only a 7% failure rate. We published a patch in a patient as young as three years old post tethered cord release who didn’t have to go back for subsequent surgery.
There’s a case series on doing blood patches even in postoperative leaks with pseudomeningocele formation. In SIH it’s a little more controversial and it most likely depends on the leak type. There’s a meta-analysis I’ll talk about a little bit more that reported a single epidural blood patch in SIH was successful in 64% of cases. That’s not bad. But other studies have called this into question, with a failure rate up to 52%. There’s a paper by Wu looking at dural tears, and they found that if you have a larger volume blood patch and a shorter segment SLEC, the patch appears to be more effective.
Here’s a 2021 JAMA meta-analysis. I’m just going to quote from the paper here. “The use of targeted epidural blood patches gave similar success rates compared with non-targeted epidural blood patches.” Based on their review of the existing literature, they weren’t able to demonstrate a benefit for the targeted patches. You can see the overlapping error bars here.
More recently, published just this year, another meta-analysis looking at exactly this question. They looked at nearly 800 articles, and seven were included. Just to quote from the paper, “Targeted and non-targeted epidural blood patches are both effective treatments for SIH… The analysis suggests that non-targeted patches may be considered a viable initial approach regardless of the identification of the leak.”
So there are a couple different ways of thinking about this. Some authors have suggested trying one or two non-targeted lumbar or low thoracic blood patches and then a targeted patch in case of failure. Others have recommended myelography or at least heavily T2-weighted MR myelography in all cases and then a targeted patch as the first-line treatment.
Okay, let’s take a look at all the randomized controlled data for targeted blood patches. And of course, this doesn’t exist. So we’re looking at retrospective data. In this regard, the issue with the retrospective data is that the patient populations are heterogeneous. There are lots of different leak types that are pooled together. The location of the patch and the volume varies across studies. Some describe fibrin, some blood, some both. And the reports before 2014 don’t include consideration of CSF-venous fistula. But perhaps the treatment guidance should differ in a community practice versus a tertiary referral CSF leak center.
What is a targeted patch after all? I mean, here’s a transforaminal patch that I did, very targeted to a ventral dural defect in the center. Nice paper by Dr. Mamlouk and Callen in Radiology published recently showing direct puncture of the SLEC for patching. So this is clearly highly targeted to the leak site.
How about this though? This is an Italian group that described placement of an epidural glidewire and 4 French catheter with injection of up to 70 cc’s of epidural blood all up and down the spinal axis. I’m not sure if this is targeted therapy.
All right. So in summary, I want you to try a non-targeted epidural blood patch for the following reasons. It’s the first-line treatment for post-dural puncture headache. It may result in a clinical cure for SIH. It may obviate the need for additional imaging with associated risk and, in particular, ionizing radiation exposure. Two separate meta-analyses have shown no benefit of targeted blood patch over non-targeted, and the blood patch spreads over multiple levels, perhaps in a linear relationship to the blood patch volume. Thanks very much for your attention.
Controversies in SIH: Dr. Timothy Amrhein on Directed Treatment
Okay, great. Thank you very much, Levi. I get the honor of doing the counter argument, which is targeted patching. It’s always difficult to go after Levi, especially since he’s our journal editor, so I may not get any of my papers published if I’m too aggressive here.
So I’m going to take the targeted side. I have no relevant disclosures. Use of fibrin glue is off label.
So what is patching? What I tell my students is that patching can be conceptualized as spackling a hole in a wall. And I would argue it makes sense to put the spackle over the hole rather than randomly on the bottom of the wall.
So what are some of the technical considerations? We’re talking about non-targeted, which is a shotgun approach where we just put the blood in there and have no concern for where the leak is necessarily—we hope we cover it—versus targeted. And targeted is directed at the leak. So the fundamental difference here is for targeted patching, you have to be able to do all these myelographic techniques that Marcel, Lal, and others have shown.
So let’s talk about spread. Levi brought up some really good points about spread and that radionuclide study. So for a non-targeted patch, let’s just assume here that this patient has a leak or fistula up in the upper thoracic spine, and you do your lumbar epidural blood patch and hope that you get spread cranially. There’s a high chance, I think, that you don’t get spread and you end up with a persistent leak.
The targeted patch. The same patient, if they have a leak in this location, we’re going to take the needle and we’re going to place it directly to this spot and we’re going to cover that area, and the patient’s going to be happy. That’s my argument.
Let’s look at epidural spread. He showed that paper that had the radionuclide study, and the radionuclide studies have a pretty low spatial resolution, right? And so I think what they’re doing is they’re assessing maybe cranial and caudal spread, but not ventral spread, not lateral spread. And so what do we know from the literature? Most of the literature is with planar imaging under fluoro where it looks at contrast spread during epidurograms, but what happens in that case is you don’t have axial imaging so you can’t actually assess spread to the ventral surface or to the neuroforamen, which is where our leaks occur.
So from a paper that Peter and I put together a while back now, where we looked at epidural spread after cervical epidurals. What we found was that contrast actually didn’t spread to the ventral epidural space except or at where the needle is or maybe one level more superior or inferior. And you can see in going into the foramen – we only had a mean of about two levels of spread to cover that area. So my argument is that targeting does matter for delivery of your patching material. Some degree of targeting must be necessary. And again the ventral epidural spread really only happened at the needle or the adjacent needle, so I think that was statistically significant.
So what about that radiation dose argument? I think that’s a good one. We want to limit our radiation dose, right? And that’s the ALARA principle – as low as reasonably achievable. I totally agree. We need to be careful about our radiation dose for our patients. That is a core principle of radiology.
But the counterargument, I think, to this is actually one of disability. Patients with SIH have severe disability. This is Dr. Friedman’s team’s paper where the quality of life was really bad, right? People are really suffering. They actually have scores that are worse than end-stage cancer patients and AIDS patients.
So if we think about quality of life adjusted years, that matters, right? There is a balance between the potential theoretical minimal risk of stochastic effects versus quality of life, right? And I think we need to consider that when we’re considering what the radiation dose is doing for these patients.
Let me just talk about a typical Duke case. This is not an abnormal unique case, but we see this a lot, right? Classic symptoms, all the stuff. They definitely have SIH based on the brain imaging and significant disability for years. And look at all of the exams that the patient had prior to coming to a center where we were able to find the leak, do targeted patching, or cure the patient. All these different imaging exams and six non-targeted blood patches across three hospital systems. So my counterargument is that’s an awful lot of radiation, and it’s because we didn’t do the imaging to find the leak and then do targeted patching. So the best way to reduce the radiation dose is to find the leak, fix the leak, and limit the number of exams.
This is the first, I think, substantial paper that’s been published on patching in SIH, and it’s by Sencakova. It’s probably the most widely cited paper. It ends up in a lot of our review articles from 2001. It was a chart review over 12 years. It was retrospective case series. And for SIH criteria they didn’t really specify, they documented that they were spontaneous leaks only. So we don’t know what type of leaks they were. The follow-up was just review of records of correspondence, and outcomes were subjective or not really specified—patient got better. They had 54 patients with CSF leaks, and 21 underwent epidural blood patches. Four were excluded without follow-up. What they found is that epidural patches at the level of the leak, which are targeted, did better than non-epidural blood patches at the level of the leak. That was reaching close to significance there. So that’s our first thing. We’re going to build out this table together here.
The second paper is from 2011. This is by Cho et al. in Neurology. This was a retrospective case series in a chart review again over 11 years. Here they did use ICHD-2 criteria to diagnose SIH, and they found 56 patients who had non-targeted or fluoro-guided epidural blood patching. They used blood, and the decision to target was based on physician preference, and it led to this. They basically actually had a comparative study here where they had 31 patients that were targeted and 25 that were non-targeted. Their outcomes were again subjective, whether or not the patient got better as defined by chart review with variable follow-up. There were different time points too, which also matters, right? The results found though that 87% of patients that had targeted patches had a positive response to this, and 52% only with non-targeted, and that was a statistically significant difference, that’s advocating for targeted patching. But some of the limitations of this paper I would say, first I think is CSF leak localization. So if we look at this paper, this is actually the figure that they selected to show for a CSF leak, right? And this is actually the false localizing sign at C1-2 that Wouter and Marcel and the Cedars group published a long time ago. That calls into question whether or not they were actually targeting leaks or if they knew where the leaks were at that time. And of course this was back in 2011, so we’ve all learned a lot since then. The other thing is they didn’t report on subtypes. And so if there were non-targeted cases where they didn’t see the leak, were those actually fistulas? Were they actually targeting? And so I think it calls into question some of the conclusions that can be drawn from the data here.
So we’re going to throw that into our table and move on to our next study, which was He et al., and this was a chart review from 2018. They used ICHD-2 criteria at the time and had 165 patients that had targeted patches with blood only. What I thought was interesting about this paper is that 90% of the cases had multiple CSF leaks. I don’t know what everybody else’s experience is, but that seems a little bit suspect to me. So I kind of wonder whether or not they were identifying the CSF leaks and knew which were the actual leak, were we actually targeting. They did show a couple of images here and this is one of the leaks that they pointed out. I can’t see it. I’m not sure where that leak is, if that’s the leak. And then this is the false localizing sign that was shown at the C7-T1 level. And that was Wouter and Marcel again. You guys keep publishing these false localizing things. It’s great. Okay, so what do we see here? We had in that targeted group, they said that there was an 87.8% symptomatic relief after one patch, and so they were strongly advocating for this targeted approach.
So, these are some of the data that we’re building out in this table, and then we’re going to move on to a more recent study. This is Pagani-Estévez. This is part of the Mayo group. This was a fairly detailed and complicated paper where there were 202 patients, a whole bunch of blood patches. The SIH criteria were an expert neurologist reporting. In this case, it was Jeremy Cutsforth-Gregory, so I kind of believe it, but it wasn’t using ICHD criteria. The outcomes efficacy definition wasn’t specified, but they did say there were set time points or blocked ranges for the time. They evaluated some of the patient procedure characteristics as predictors of a successful patch, and they found greater likelihood of blood patching efficacy with larger volumes, which we see across several studies, multi-level patching, and specifically targeting. Targeting was statistically significant here. Reduced likelihood when it was non-targeted, and so I think this is another paper and maybe even stronger evidence paper that’s advocating for the targeted approach from 2019.
This paper in 2024 is more recent, and one thing I’ll say is that if you look at the literature, there’s been a rapid ramp-up in the number of publications in our SIH field, and I think that’s a credit to everybody in this room. I think a lot of good things are coming, and the evidence is going to be building rapidly over the next decade or so. In this paper, they looked at 20 patients with a retrospective cohort study and identified the leak in 80 of them. So 20% they didn’t know where it was, and it was kind of interesting. They did these mini interlaminar surgical approaches and ended up placing two epidural catheters around where they thought the leak was, one above and one below, and then patched through those catheters, which I had never heard of before, but effectively this is quite targeted, theoretically. They found significant improvement in 90% of cases, and so this is another paper advocating for targeted approaches.
This one by Kong et al. is actually our first prospective randomized controlled trial, so that’s pretty cool. It’s 68 patients, they had a couple of different arms. So this was interesting. I think what they were doing is they were presenting this new catheter-assisted approach. They had a targeted group, a non-targeted group, and then these catheter-assisted patients. There was a randomization schema. 55% of them responded when it was non-target, 76% when it was targeted, and within their novel technique they said it was a 95% response rate. But there does appear to be some suggestion that targeting was helpful in this case. Some of the notes about this, it doesn’t seem like they categorized the leak subtypes or used myelography for localization. So I’m not exactly sure how they knew where the leak was.
And then our next one that we’re going to fill in here is Dr. Callen and Lal and this great team, many of whom are in the room. There’s 56 patients. These are the lateral tears. They did CT-guided targeted patching, right? And response rates: 37% complete response, 48% partial response, and 16% no improvement. And so I’m giving the targeted side, so I’m going to add those numbers together and say this was an 83% response rate to targeted patching. Right? So there’s some advocacy here, I think, for the targeted patching working.
And then I’m drawing a yellow line here because I’m going to choose two papers from the literature for non-targeted patching and say why it doesn’t work. And one of them is the Cedars group here. And so this is Dr. Martin and Dr. Schievink, who’s the senior author on this. So a retrospective cohort of 94 patients. They did use ICHD criteria to find the SIH patients. They did two patches in a non-targeted fashion at these levels, one in the lower thoracic spine, one at the lumbar spine, with a large volume of blood. They did exclude a whole bunch of patients. So what happened to them? Nobody knows, right? So it could have been that they all got better, it could have been that they all got worse. And they defined success as no need for further treatment. And 28% didn’t need any further treatment after the first patch. And you can see the numbers there for the second and the third.
So I think in general here what we’re seeing is that there’s not a significant response rate to non-targeted patches, particularly compared to the papers below that yellow line.
And then finally Eike and our wonderful group here in Freiburg and Bern with Tomas and Jürgen Beck published this retrospective cohort study. These are patients that were all SLEC positive, so type 1 or type 2 leaks, when they did non-targeted patches down in the lumbar spine. The response rates were 29% imaging success at the best case scenario is what they wrote in that paper. Right. So that’s not a great number. 36% had symptom improvement long term, but 57% eventually required surgery. So at least in this subset of CSF leak types, it seems like non-targeted patching is not particularly helpful.
So we’ll fill out the top of our table there. And I think the end summary here is I would say the evidence is suggesting that maybe it’s targeted. Now I may also be selecting particular papers from the literature to support my argument, but we won’t go there.
What is the state of the evidence though? This is where I want to kind of transition away from our tongue-in-cheek discussion between me and Levi about targeted versus non-targeted. What I really want to talk about is this sort of a shared component to this where we’re going to talk about the state of the evidence, and this is a systematic review and evidence map that we put together at Duke and Mayo a while back, and some of the stuff from this I think is really interesting.
So what we’re seeing here is that basically these are the different types of studies with randomized controlled trials and higher levels of evidence at the top, and each of these little dots represents a paper. The size of the dot is the number of patients in there, and what we see is that there’s actually quite few papers on targeted patching. There just wasn’t a lot at that time. The other thing that Levi had mentioned before is there are no randomized controlled trials. There are very few prospective cohort studies, and I really actually view this as an opportunity for everybody in this room. There are tons, there’s a whole white space for us to fill, so I think this is going to be a really exciting decade or two ahead as we’ve all figured this out together.
In terms of patching materials, almost all the studies use blood. Very few use fibrin alone. If I go back to that, it’s only like one or two dots where it’s only fibrin glue, and there’s often some mix where it’s a little unclear about whether it was blood or fibrin in particular patches.
The other thing is for the SIH diagnosis method, we found that about 40% of them were not specified or subjective. So, we weren’t entirely sure that the patients actually had SIH, and you couldn’t determine whether it met criteria for SIH in greater than 40% of the papers.
I think another really big issue and challenge is that 77.7% of the cases, or papers rather, didn’t report the leak subtype. So if we all think that fistulas do or do not respond to patches or may respond differently compared to type 1 leaks or lateral tears, that’s going to be critical, and that wasn’t present in many of the papers. I think there are some fundamental challenges that we’ve got here and opportunities for improvement.
One other thing here is that most of the papers out there, about 85% or so, just report subjective outcome measures at variable time points, which limits the interpretation you can make from the data, and very few use subsequent imaging after the treatment to determine whether or not there has been successful treatment—only 61%.
So we did attempt to submit an NIH grant to run a targeted versus non-targeted randomized controlled trial, and many in this room were part of that. It was a whole group of us, and basically what I got back from the NIH was that most data suggest that both approaches work, and actually there were some reviewer comments that said, well, targeting is clearly better, this is not an important question. I would argue countering to that is how could you be so sure with what we’ve got so far, right? From My Cousin Vinny. And I think we have a long way to go and a lot of opportunity for us to help figure this out together. So thanks a lot. I appreciate everybody’s time.