Transcript
Good evening. I'm the only stupid guy in this room who's spent 15 years of trainings, 11 years in residency, and 3 in fellowship. But, anyway, what it gave me is a unique perspective on treating lesions in the brain with radiosurgery, because that's all I've been doing for 25 years now. If Shakespeare was a radiosurgeon, Hamlet would have said, "To treat or not to treat," and Lexell, in Swedish, would have said [foreign language 00:00:29].
But that's really the biggest question, and somebody earlier told me when they sat through tumor boards, the number one question always is, "What are we looking at? Is this treatment effect? Is this recurrent tumor? What should we be doing?" And I'm actually very delighted that the speakers before me have multiple platforms for radiosurgery delivery.
And like them, I've used protons, I've used CyberKnife, I have a stereotactic LINAC, and, of course, I have a Gamma Knife. So, the nice thing you see is cross-platform integration. No matter how you deliver, the good tool is a good tool. And I think Elements brings to. So I want to show you why we incorporated Elements, why we incorporated contrast clearance into our clinical workflow. And these are my colleagues who actually are in the process of analyzing and writing a lot of this stuff up.
And that's our program. I've been doing this for a long time, 12,000-plus patients in my personal series. Roswell Park is a 20-year-old program, nearly 6,000 patients treated with the Gamma Knife. And the bulk of what we do today is basically metastatic disease. So that's the number one thing we treat. And we understand where radiosurgery now plays in the bigger management of brain mets, and certainly gets called in to treat recurrent GBMs, gets called in to treat brain mets, prima facie, recurrence, residuals, and a whole variety of other indications, interplaying with surgery, interplaying with immunotherapy and everything else.
The unspoken part is the underutilization of radiosurgery, which is why these docs are so critical. Both optimization of delivery, efficiencies in contouring, efficiencies in planning and delivery are critical if we're going to truly make a dent. Even in the United States right now, utilization of radiosurgery is very, very low. And it has demographic, bare, and socioeconomic differences. So, even in renal cancer, where we know that conventional whole-brain radiotherapy has very little to do for the patient, radiosurgery is underutilized, particularly in populations that don't have adequate coverage.
But this is how Elements is now integrated into our workflow. On the IMRT side in my primary brain tumor practice, every patient's pre and post-op MRs are auto-segmented in Elements and sent over to the Eclipse planning system because that's the planning system we use for treatment. We also use Elements to delineate the tumor cavities and tumor beds because three-dimensional delineation is actually very good within Elements, better than GammaPlan, certainly better than even an Eclipse. So I find that it's a very useful add-on for me.
In the world of the Gamma Knife, it is interspersed between our MRI stereotactic acquisition in the morning, and our delivery so that it auto-segments all the critical structures for me. We review them, and actually we find, and Elekta agrees, that the most high-fidelity transfer of contours between two GammaPlan occurs from Elements. They're not distorted. They don't come out as clean from MIMS. They certainly don't come out as clean from Varian. It becomes problematic because you're looking at volumes differently. As you showed in your study, the volume segmentation grids are different. So you have to be very careful. But for small volumes, we find that using the Brainlab Elements really makes things work very well for us.
And then, ultimately, the big piece of the puzzle here is the decision-making when treatment results are being evaluated. And so we play a big part in this. And just so you understand the dynamics of doing contrast clearance analysis, most of you are probably familiar with contrast clearance, and we'll come to that in a minute, but the logistics of doing it are not simple.
You have to keep the patient in the MRI scanner or in the MRI waiting area for about 60 to 90 minutes, put them back on the same scanner that they were scanned on, do the same exact sequence they had before. Requires a lot of compliance from radiology and radiology tech. And from my point of view, I need a helper in the clinical world who does look at the follow-ups, does the TRAM analysis, and brings to my attention what's going on with the patient. And that's my physician assistant.
So she was the first one credited in the list of people there because she spends the time deciding who needs a TRAM analysis. We have built TRAM into our standard workflow. So, within our EMR, there is a request for MRI brain, and then there's a request for MRI brain with TRAM. And that tells them that they need to keep the patient there. And usually, what I'm still struggling with is how to preempt and get TRAM on someone who may or may not have adverse radiation effect. Unfortunately, that's an unsolved problem.
So what happens typically is you suspect adverse radiation effect, and you do the next MRI, usually about four weeks later, and then you plan a TRAM study. So there is a slight delay. If I'm really concerned, I will do it quickly, but it's unlikely I'll catch every patient on follow-up who's already come and gone from radiology long before I figure out that they needed a TRAM evaluation.
Just to give you an idea, the other piece of Elements that we use a lot of, and I know this is not necessarily the primary audience for it, but it's interesting for you to see what we're doing with the Elements piece using tractography. So you can see the dentatorubro tract beautifully delineated. I keep touching the wrong button, sorry. Beautifully delineated with crossing fibers, the capsular fibers. So when we place our thalamotomy target, we are now getting our tractography upfront. And even though we can't truly use it for targeting, we're gonna use it in the follow-up. And there's some early data that shows that if you get a successful thalamotomy lesion, you will see disruption of the DRT fibers. So that's another piece of what the whole package does for a busy Gamma Knife practice.
We treat about 700 patients a year with our Gamma Knife. The average number of mets treated in a patient is going up slowly, but I still think it stays around 3.8 to 4 mets per patient. The largest number can vary quite a bit. We did a patient with 26 brain mets last week. The most I've done in my career is 44, but this was a patient who had craniospinal as a child, and I did not want to give them whole-brain radiotherapy. So we sat for 4 hours and we took care of 44 spots, but we did keep the 3-dose cloud.
And there is a comment to be made there. It is easy to dispense with the fact that the 3-gray cloud is the only place where you successfully see a differentiation between techniques. But if you look at Dan Trifiletti's study from UVA, looking at leukoencephalopathy, two years out, the predictor for long-range leukoencephalopathy survivors was the 3-gray integral volume. And so 3 gray does matter. Ten and 12 will predict short-term complications. Three gray will protect your patient's brain long-term. So I think both are critical. And as we evolve in parallel techniques, we should keep both ideas in mind, and not discard any one number as less relevant. Inter-planning system differences for those volumes can be tricky, but important to keep in mind.
Then why do we worry about this adverse radiation effect? Because we know it occurs. I think it's about 5%. You'll see my patients have about 30 cases in here to show you as early results, and that represents an at-risk patient volume of roughly 600 patients that I was working with. So right about the number that was reported early on showing from San Francisco, that's about a 5% risk of adverse effects when you do radiosurgery focally for lesions.
And we've looked at...The interesting thing about adverse effect, roughly three times a year, I'll go in and operate on somebody who we've done radiosurgery and then had adverse effect. There's mass effect, and you need to go and decompress it. What do we find? Interestingly, we did not have clearance analysis at this point, so we were looking at it on MRI and guessing, is this tumor? Is this necrosis? But because it was big enough, it was taken out. And for the most part, that's what radiation necrosis should look like, fibrinoid hyalinized tissue, lots of vascular bizarre-looking vessels, but that's what it should look like. But what it often looks like is this, a mixture of largely necrotized tissue with viable tumor cells. And it's very interesting when you start doing contrast analysis for metastatic disease, you start to see those mixed patterns.
So, what is contrast clearance analysis? It really is taking your vessel clearance of contrast, which is a timeline that decides how quickly the contrast gets into a big vessel and gets out of the venous system. And that calibration can then help you decide how the contrast is flowing through tumor tissue and surrounding brain. And that's really the differentiator. The basic idea is tumors are highly vascular. If they're viable, contrast goes in quickly, comes out not as quickly as it comes out of a vessel, but reasonably quickly flushes out. On the other hand, if it's damaged brain, there's radionecrosis, contrast flows in, and then stains the brain and stays put. So if you take a delayed scan, you will see a blush of contrast in areas where there is potentially radiation damage and radiation necrosis.
So that's the basic principle. You can get that from Brainlab. And this is what we find. Let's start where this was designed. This study, this whole idea came from the GBM world. We treat a number of GBMs, I'd say about 30 a year, radiosurgically, for recurrences after they've had conventional therapy. And here's an example of a patient who was treated. Left panel shows you the treated area. Right panel is the follow-up, along the primary isodose where we treat it. But if you look at the...and you can see on the TRAM study, pretty much radiation effect. A little bit of blue, but not any more blue than rest of the brain. But if you go one slice higher, in this patient, the area of concern was here, outside of our treatment field.
Now, if we did not have contrast clearance, we would have assumed this is new disease. GBMs are like that, anyway. So, what did the contrast clearance analysis show? It confirms that that's new tumor. Now, what it helps me do is I can obviously treat the new area, but when I'm going to go back and treat this patient, I'm gonna look at all of my previously treated field. And if there are actually persistent blue spots in there, then I'd be tempted to go back and re-radiate them. Particularly because they're within the field of prior treatment, I'm not worried about the real estate in that area, so we'll end up doing it like that. And that's exactly what we do.
In terms of brain mets, you see a very similar picture. A normally treated brain met, once responded, will show very little change on a contrast clearance analysis. Most will show this combined pattern, a little bit of radiation effect, a little bit of residual tumor. And if you do...for example, here, if you looked at this scan and had no contrast clearance available, you would say, "This is radiation effect. Let's follow the patient, maybe give them some steroids." Interestingly, if you do the clearance analysis, you still see the persistent tumor. Make this about six times bigger and force a surgery, you will get that microscopic picture I showed you earlier, a mixture of dead tissue with viable tumor cells present in it. And I think this technique is pretty sensitive. It picks up this pattern.
Our goal now is, for everybody undergoing a resection of a met that undergoes adverse effect, we will perform TRAM before we take them to surgery so we can provide you a comparison between the surgical resected specimen, which we will try to do on-block, and mark geographically so we can match it to the radiograph, and show you exactly what areas corresponded to viable tumor. It's a tall order, but we're hoping we do that for you.
We also use our TRAM studies serially, not just to decide what's radiation necrosis, but to what happens to them. So here's an example of a patient treated for 11 lesions, but here's the 2 that were of interest. And you can see one hour of delayed contrast already shows some blush, and the TRAM shows you a mixed response. So we followed this patient. As we go along, it becomes clear the frontal lesion is just showing radiation effect and response even though the MRI contrast shows a ring-enhancing area there. But there is failure in the posterior area. So that patient, while all other reasons have disappeared, has a persistent blue uptake in the area back there, and we retreated that. So we did not wait for this to present itself as a thing. We actually went back and retreated this patient.
We also intervene, of course, sometimes. Here's a patient who, seven years after breast cancer diagnosis, suddenly developed brain mets, a lot of them. And over the course of her 3-year management by me, and she's alive and well with no brain mets now, I have treated 33 brain mets in her. She did not get whole-brain, she didn't want it, and what she developed was a side effect, adverse radiation effect only in the cerebellum. So this is interesting for you to see because this is actually treated area.
Since she was having a little unsteadiness of gait, we gave her Avastin. We usually use steroids, but sometimes steroids and Avastin. Two cycles of Avastin do wonders for these patients. And here is the follow-up with Avastin, serial TRAM imaging, and now you're almost six months out. Avastin was given early on. That Day 0 refers to the Avastin. And so, you could see that you can see the response, you can follow it, and it correlates nicely with the response on the adverse radiation effect side.
What have we done so far? This is where you're gonna think I'm crazy. See that AVM lurking in there? Yeah. We did a TRAM study on an AVM, too. And I was just talking to one of the colleagues who was involved in developing this technique, and it was very interesting what I found, but I'll show you that. But for the most part, these are brain mets. Those are our radiologists' interpretation. They're wonderful people. Most of the time, they comment by not commenting. Twenty-nine patients with adverse radiation effect, possibly. Only seven actually had words that said, "This could be treatment effect." So we're gonna educate them.
But until we do that, this is what TRAM showed us. Eighteen of them had changes suggestive adverse radiation effect, mixed response in about eight, and three showed recurrent tumor, of which we ended up retreating two of the patients. One of them got operated, but two of the others got re-radiosurgery. So, not only is the initial treatment changing the management post hoc, and how you manage these patients to keep their brain controlled, we'll change if we have more information.
Why did we treat the AVM? Well, here, do the AVM study. So we treat a lot of AVMs. Now it's down to 30 a year. But at one time, at Steiner, I used to treat nearly 200 a year. So that's because of all the endovascular stuff. But we have all these patients with beautiful results, and yet, with 90% obliteration rate, we still have about 30% of our patients showing adverse T-2 changes early in follow-up. And I was very curious what that would look like on TRAM. It's difficult ask because, honestly, calibration of TRAM is to vascular flow, which is really high in an AVM. So that area should become silent.
And I've only one case, and it's a disclaimer, do not try this at home. I'm not endorsing it. I did it because I was curious, and I know the Brainlab guys are back and laughing. They already know I'm a renegade. But here is your flow voids. Interestingly, at this point, the T-2 is not showing any flare change, but look at the changes on TRAM. Now, this, I don't understand, and I'll have to work with the group in Israel to understand what I'm really seeing here. But this is very interesting to me. It has to be related to the pathology because it's following the white matter track, going right up against the ventricle. But this is something just as a curiosity factor. So stay tuned, there's more to come, and we're gonna keep doing this.
The take-home part of this is we've integrated Elements into a non-LINAC-based radiosurgery platform to make up for what we can't do with some of the software that we already have. It enhances our capabilities. It has really opened a window into what we can do with follow-up, and we'll have many more cases because we have now incorporated it into our flow system. And once it's in the workflow, it's gonna be done on a routine basis. And I still think, like you have the GBM study, we need some histology studies to prove if mets are doing exactly what GBMs do. But we'll get there. So, hopefully, by the next ASTRO, or maybe the one after that, we'll have some more robust data to show you. But this is our early experience, and I just thought that it would be interesting for the audience to hear. Thank you.
But that's really the biggest question, and somebody earlier told me when they sat through tumor boards, the number one question always is, "What are we looking at? Is this treatment effect? Is this recurrent tumor? What should we be doing?" And I'm actually very delighted that the speakers before me have multiple platforms for radiosurgery delivery.
And like them, I've used protons, I've used CyberKnife, I have a stereotactic LINAC, and, of course, I have a Gamma Knife. So, the nice thing you see is cross-platform integration. No matter how you deliver, the good tool is a good tool. And I think Elements brings to. So I want to show you why we incorporated Elements, why we incorporated contrast clearance into our clinical workflow. And these are my colleagues who actually are in the process of analyzing and writing a lot of this stuff up.
And that's our program. I've been doing this for a long time, 12,000-plus patients in my personal series. Roswell Park is a 20-year-old program, nearly 6,000 patients treated with the Gamma Knife. And the bulk of what we do today is basically metastatic disease. So that's the number one thing we treat. And we understand where radiosurgery now plays in the bigger management of brain mets, and certainly gets called in to treat recurrent GBMs, gets called in to treat brain mets, prima facie, recurrence, residuals, and a whole variety of other indications, interplaying with surgery, interplaying with immunotherapy and everything else.
The unspoken part is the underutilization of radiosurgery, which is why these docs are so critical. Both optimization of delivery, efficiencies in contouring, efficiencies in planning and delivery are critical if we're going to truly make a dent. Even in the United States right now, utilization of radiosurgery is very, very low. And it has demographic, bare, and socioeconomic differences. So, even in renal cancer, where we know that conventional whole-brain radiotherapy has very little to do for the patient, radiosurgery is underutilized, particularly in populations that don't have adequate coverage.
But this is how Elements is now integrated into our workflow. On the IMRT side in my primary brain tumor practice, every patient's pre and post-op MRs are auto-segmented in Elements and sent over to the Eclipse planning system because that's the planning system we use for treatment. We also use Elements to delineate the tumor cavities and tumor beds because three-dimensional delineation is actually very good within Elements, better than GammaPlan, certainly better than even an Eclipse. So I find that it's a very useful add-on for me.
In the world of the Gamma Knife, it is interspersed between our MRI stereotactic acquisition in the morning, and our delivery so that it auto-segments all the critical structures for me. We review them, and actually we find, and Elekta agrees, that the most high-fidelity transfer of contours between two GammaPlan occurs from Elements. They're not distorted. They don't come out as clean from MIMS. They certainly don't come out as clean from Varian. It becomes problematic because you're looking at volumes differently. As you showed in your study, the volume segmentation grids are different. So you have to be very careful. But for small volumes, we find that using the Brainlab Elements really makes things work very well for us.
And then, ultimately, the big piece of the puzzle here is the decision-making when treatment results are being evaluated. And so we play a big part in this. And just so you understand the dynamics of doing contrast clearance analysis, most of you are probably familiar with contrast clearance, and we'll come to that in a minute, but the logistics of doing it are not simple.
You have to keep the patient in the MRI scanner or in the MRI waiting area for about 60 to 90 minutes, put them back on the same scanner that they were scanned on, do the same exact sequence they had before. Requires a lot of compliance from radiology and radiology tech. And from my point of view, I need a helper in the clinical world who does look at the follow-ups, does the TRAM analysis, and brings to my attention what's going on with the patient. And that's my physician assistant.
So she was the first one credited in the list of people there because she spends the time deciding who needs a TRAM analysis. We have built TRAM into our standard workflow. So, within our EMR, there is a request for MRI brain, and then there's a request for MRI brain with TRAM. And that tells them that they need to keep the patient there. And usually, what I'm still struggling with is how to preempt and get TRAM on someone who may or may not have adverse radiation effect. Unfortunately, that's an unsolved problem.
So what happens typically is you suspect adverse radiation effect, and you do the next MRI, usually about four weeks later, and then you plan a TRAM study. So there is a slight delay. If I'm really concerned, I will do it quickly, but it's unlikely I'll catch every patient on follow-up who's already come and gone from radiology long before I figure out that they needed a TRAM evaluation.
Just to give you an idea, the other piece of Elements that we use a lot of, and I know this is not necessarily the primary audience for it, but it's interesting for you to see what we're doing with the Elements piece using tractography. So you can see the dentatorubro tract beautifully delineated. I keep touching the wrong button, sorry. Beautifully delineated with crossing fibers, the capsular fibers. So when we place our thalamotomy target, we are now getting our tractography upfront. And even though we can't truly use it for targeting, we're gonna use it in the follow-up. And there's some early data that shows that if you get a successful thalamotomy lesion, you will see disruption of the DRT fibers. So that's another piece of what the whole package does for a busy Gamma Knife practice.
We treat about 700 patients a year with our Gamma Knife. The average number of mets treated in a patient is going up slowly, but I still think it stays around 3.8 to 4 mets per patient. The largest number can vary quite a bit. We did a patient with 26 brain mets last week. The most I've done in my career is 44, but this was a patient who had craniospinal as a child, and I did not want to give them whole-brain radiotherapy. So we sat for 4 hours and we took care of 44 spots, but we did keep the 3-dose cloud.
And there is a comment to be made there. It is easy to dispense with the fact that the 3-gray cloud is the only place where you successfully see a differentiation between techniques. But if you look at Dan Trifiletti's study from UVA, looking at leukoencephalopathy, two years out, the predictor for long-range leukoencephalopathy survivors was the 3-gray integral volume. And so 3 gray does matter. Ten and 12 will predict short-term complications. Three gray will protect your patient's brain long-term. So I think both are critical. And as we evolve in parallel techniques, we should keep both ideas in mind, and not discard any one number as less relevant. Inter-planning system differences for those volumes can be tricky, but important to keep in mind.
Then why do we worry about this adverse radiation effect? Because we know it occurs. I think it's about 5%. You'll see my patients have about 30 cases in here to show you as early results, and that represents an at-risk patient volume of roughly 600 patients that I was working with. So right about the number that was reported early on showing from San Francisco, that's about a 5% risk of adverse effects when you do radiosurgery focally for lesions.
And we've looked at...The interesting thing about adverse effect, roughly three times a year, I'll go in and operate on somebody who we've done radiosurgery and then had adverse effect. There's mass effect, and you need to go and decompress it. What do we find? Interestingly, we did not have clearance analysis at this point, so we were looking at it on MRI and guessing, is this tumor? Is this necrosis? But because it was big enough, it was taken out. And for the most part, that's what radiation necrosis should look like, fibrinoid hyalinized tissue, lots of vascular bizarre-looking vessels, but that's what it should look like. But what it often looks like is this, a mixture of largely necrotized tissue with viable tumor cells. And it's very interesting when you start doing contrast analysis for metastatic disease, you start to see those mixed patterns.
So, what is contrast clearance analysis? It really is taking your vessel clearance of contrast, which is a timeline that decides how quickly the contrast gets into a big vessel and gets out of the venous system. And that calibration can then help you decide how the contrast is flowing through tumor tissue and surrounding brain. And that's really the differentiator. The basic idea is tumors are highly vascular. If they're viable, contrast goes in quickly, comes out not as quickly as it comes out of a vessel, but reasonably quickly flushes out. On the other hand, if it's damaged brain, there's radionecrosis, contrast flows in, and then stains the brain and stays put. So if you take a delayed scan, you will see a blush of contrast in areas where there is potentially radiation damage and radiation necrosis.
So that's the basic principle. You can get that from Brainlab. And this is what we find. Let's start where this was designed. This study, this whole idea came from the GBM world. We treat a number of GBMs, I'd say about 30 a year, radiosurgically, for recurrences after they've had conventional therapy. And here's an example of a patient who was treated. Left panel shows you the treated area. Right panel is the follow-up, along the primary isodose where we treat it. But if you look at the...and you can see on the TRAM study, pretty much radiation effect. A little bit of blue, but not any more blue than rest of the brain. But if you go one slice higher, in this patient, the area of concern was here, outside of our treatment field.
Now, if we did not have contrast clearance, we would have assumed this is new disease. GBMs are like that, anyway. So, what did the contrast clearance analysis show? It confirms that that's new tumor. Now, what it helps me do is I can obviously treat the new area, but when I'm going to go back and treat this patient, I'm gonna look at all of my previously treated field. And if there are actually persistent blue spots in there, then I'd be tempted to go back and re-radiate them. Particularly because they're within the field of prior treatment, I'm not worried about the real estate in that area, so we'll end up doing it like that. And that's exactly what we do.
In terms of brain mets, you see a very similar picture. A normally treated brain met, once responded, will show very little change on a contrast clearance analysis. Most will show this combined pattern, a little bit of radiation effect, a little bit of residual tumor. And if you do...for example, here, if you looked at this scan and had no contrast clearance available, you would say, "This is radiation effect. Let's follow the patient, maybe give them some steroids." Interestingly, if you do the clearance analysis, you still see the persistent tumor. Make this about six times bigger and force a surgery, you will get that microscopic picture I showed you earlier, a mixture of dead tissue with viable tumor cells present in it. And I think this technique is pretty sensitive. It picks up this pattern.
Our goal now is, for everybody undergoing a resection of a met that undergoes adverse effect, we will perform TRAM before we take them to surgery so we can provide you a comparison between the surgical resected specimen, which we will try to do on-block, and mark geographically so we can match it to the radiograph, and show you exactly what areas corresponded to viable tumor. It's a tall order, but we're hoping we do that for you.
We also use our TRAM studies serially, not just to decide what's radiation necrosis, but to what happens to them. So here's an example of a patient treated for 11 lesions, but here's the 2 that were of interest. And you can see one hour of delayed contrast already shows some blush, and the TRAM shows you a mixed response. So we followed this patient. As we go along, it becomes clear the frontal lesion is just showing radiation effect and response even though the MRI contrast shows a ring-enhancing area there. But there is failure in the posterior area. So that patient, while all other reasons have disappeared, has a persistent blue uptake in the area back there, and we retreated that. So we did not wait for this to present itself as a thing. We actually went back and retreated this patient.
We also intervene, of course, sometimes. Here's a patient who, seven years after breast cancer diagnosis, suddenly developed brain mets, a lot of them. And over the course of her 3-year management by me, and she's alive and well with no brain mets now, I have treated 33 brain mets in her. She did not get whole-brain, she didn't want it, and what she developed was a side effect, adverse radiation effect only in the cerebellum. So this is interesting for you to see because this is actually treated area.
Since she was having a little unsteadiness of gait, we gave her Avastin. We usually use steroids, but sometimes steroids and Avastin. Two cycles of Avastin do wonders for these patients. And here is the follow-up with Avastin, serial TRAM imaging, and now you're almost six months out. Avastin was given early on. That Day 0 refers to the Avastin. And so, you could see that you can see the response, you can follow it, and it correlates nicely with the response on the adverse radiation effect side.
What have we done so far? This is where you're gonna think I'm crazy. See that AVM lurking in there? Yeah. We did a TRAM study on an AVM, too. And I was just talking to one of the colleagues who was involved in developing this technique, and it was very interesting what I found, but I'll show you that. But for the most part, these are brain mets. Those are our radiologists' interpretation. They're wonderful people. Most of the time, they comment by not commenting. Twenty-nine patients with adverse radiation effect, possibly. Only seven actually had words that said, "This could be treatment effect." So we're gonna educate them.
But until we do that, this is what TRAM showed us. Eighteen of them had changes suggestive adverse radiation effect, mixed response in about eight, and three showed recurrent tumor, of which we ended up retreating two of the patients. One of them got operated, but two of the others got re-radiosurgery. So, not only is the initial treatment changing the management post hoc, and how you manage these patients to keep their brain controlled, we'll change if we have more information.
Why did we treat the AVM? Well, here, do the AVM study. So we treat a lot of AVMs. Now it's down to 30 a year. But at one time, at Steiner, I used to treat nearly 200 a year. So that's because of all the endovascular stuff. But we have all these patients with beautiful results, and yet, with 90% obliteration rate, we still have about 30% of our patients showing adverse T-2 changes early in follow-up. And I was very curious what that would look like on TRAM. It's difficult ask because, honestly, calibration of TRAM is to vascular flow, which is really high in an AVM. So that area should become silent.
And I've only one case, and it's a disclaimer, do not try this at home. I'm not endorsing it. I did it because I was curious, and I know the Brainlab guys are back and laughing. They already know I'm a renegade. But here is your flow voids. Interestingly, at this point, the T-2 is not showing any flare change, but look at the changes on TRAM. Now, this, I don't understand, and I'll have to work with the group in Israel to understand what I'm really seeing here. But this is very interesting to me. It has to be related to the pathology because it's following the white matter track, going right up against the ventricle. But this is something just as a curiosity factor. So stay tuned, there's more to come, and we're gonna keep doing this.
The take-home part of this is we've integrated Elements into a non-LINAC-based radiosurgery platform to make up for what we can't do with some of the software that we already have. It enhances our capabilities. It has really opened a window into what we can do with follow-up, and we'll have many more cases because we have now incorporated it into our flow system. And once it's in the workflow, it's gonna be done on a routine basis. And I still think, like you have the GBM study, we need some histology studies to prove if mets are doing exactly what GBMs do. But we'll get there. So, hopefully, by the next ASTRO, or maybe the one after that, we'll have some more robust data to show you. But this is our early experience, and I just thought that it would be interesting for the audience to hear. Thank you.