A conversation with Dr. Greg Albers, professor of neurology at Stanford and the principal investigator for DEFUSE 3.
Interviewed by Dr. Kaustubh Limaye, assistant professor of neurology in the division of cerebrovascular diseases at the University of Iowa, at the International Stroke Conference 2018 following the presentation of the final results of DEFUSE 3 and a simultaneous publication in the New England Journal of Medicine.
Dr. Limaye: Dr. Albers, first accept my hearty congratulations on the phenomenal success of the DEFUSE 3 trial.
Dr. Albers: Thank you so much.
Dr. Limaye: Just like everybody else, I was patiently waiting to hear what the results were going to be. All of us are delighted looking at the strong treatment effect that DEFUSE 3 showed in this extended time window. Thanks again for taking time out from your busy schedule. I’m sure this conference was extremely busy for you.
Dr. Albers: It’s been a very exciting week. We’ve been anticipating this for some time, and it’s wonderful to see this come to fruition.
Dr. Limaye: We recognize that obtaining funding and organizing a clinical trial is difficult. Would you tell us a little bit more about how much time it did take, from conceptualization of the study to bringing the study to fruition?
Dr. Albers: It actually took about 20 years. The conceptualization occurred in the mid–1990s, when we were fortunate enough to have a faculty member at Stanford, Mike Moseley, identify that diffusion imaging could see a stroke in real time, because before that, practicing as a neurologist, it was frustrating to go to the ER and see somebody with a major clinical deficit, and a normal CT scan, so you didn’t really know for certain that it was a stroke. You certainly didn’t know how much tissue was irreversibly injured, even the location. You made your best estimate from what you knew as a clinical neurologist, but oftentimes, you could turn out to be wrong.
When diffusion imaging came along, it was just a complete game-changer for us at Stanford. We started to do diffusion imaging on every single stroke patient, and we would see patients who would come in very early after stroke onset, with an MCA syndrome, and have the entire hemisphere diffusion restricted with massive injury. That’s what we had expected from how we were taught about stroke in medical school, that it progresses very quickly, but as we started scanning more patients, we found out that there were a large number of patients who arrived at Stanford after being transported long distances, and arrived at 12, 14, 16 hours, and looked identical, same clinical scenario, same high NIH stroke scale, yet a very small diffusion lesion. When we took a picture a couple days later, it blossomed into the full MCA distribution stroke.
At that point, we knew that there was tremendous variability in stroke evolution, and we wanted to try to identify that stroke evolution variability with an NIH grant that was called the DEFUSE grant, which now we call DEFUSE 1. In that study we used intravenous tPA at twice the treatment window. It was three hours back then. We doubled it to six, and what we found is that if you had a favorable profile on MRI, with a small DWI core and a large area of perfusion abnormality, this predicted a very favorable clinical and radiographic response to reperfusion, even at a late time window.
We developed the RAPID software based on those observations because we needed to have a way to identify the irreversibly injured tissue, as well as tissue that was hypoperfused, but likely to go on to infarct. Once we had that software, we were fortunate that we had a new and exciting way to treat stroke, which was the thrombectomy procedure. In DEFUSE 2, we went out to 12 hours, using MR only, and could see that the same favorable profile that we call “target mismatch” responded beautifully to endovascular reperfusion, all the way out to 12 hours.
Realizing that most hospitals can’t do urgent MR, we had to modify the software to handle CT technology, so the CT perfusion version of RAPID was born, and we tested it in three studies, CRISP, EXTEND-IA and SWIFT PRIME. We were quite thrilled to see that those studies had spectacular good outcome rates in patients with both early and late window thrombectomy. SWIFT PRIME, EXTEND-IA were early window; CRISP was late window. In all of the studies, it appeared that time didn’t seem to matter very much. If you had a favorable imaging profile, it essentially removed time from the equation, in that we could get really good outcomes at later time windows, that appeared to match the early time window outcomes, if you had a small core, and a large amount of salvageable tissue.
That set the stage for DEFUSE 3. DEFUSE 3 was a 6- to 16-hour trial where we used the same target mismatch profile to select the patients. We couldn’t be more thrilled with the results, particularly looking at the endovascular arm. We saw that the clinical response to thrombectomy at six hours was identical to 16, truly removing time from the equation. Interestingly, in the medical arm, time did seem to make an impact. The medical arm group did very poorly, and particularly towards the end of the time window. Medical arm patients had absolutely dismal good outcome rates, approaching zero, if you look at the 12- to 16-hour population. What that meant is that the response to therapy was larger towards the end of our window, so the ultimate paradox: We have a late-window trial showing bigger treatment effects than early window trials, in particular, the biggest treatment effects towards the far end of the time window in DEFUSE 3!
Part of the explanation for this is the lack of the use of an IV thrombolytic in DEFUSE 3. A nice example of the benefits of an IV thrombolytic we can look back to is SWIFT PRIME and EXTEND-IA, which were taking the same type of favorable target mismatch patients, and we can look at both the clinical outcomes in the tPA alone arm, versus the tPA plus the stent-retriever arm. We can see, when we compare that to studies that were less selective on imaging-based selection criteria, like MR CLEAN or THRACE, that there’s a huge benefit in the tPA alone arm in SWIFT PRIME and EXTEND-IA. If you have a favorable imaging profile, there’s about a 40% response to tPA; 40% of these patients, treated only with IV tPA, have complete reperfusion, and their clinical outcomes are just as good as in the endovascular patients who reperfuse. However, in the endovascular arm, reperfusion occurred 80% of the time, so that’s why SWIFT PRIME and EXTEND-IA showed such large benefits in the endovascular arms.
This shows us that if we take imaging-selected patients, and we treat them with IV tPA, that we anticipate 35% to 40% good outcome rates. If you add the stent retriever, you bump it up to 65% to 70%. Take these same patients in the late window, and you can see that the medical arm is not doing nearly as well, because they’re missing the thrombolytic, and now that we have Bruce Campbell’s exciting TNK versus tPA data, you can imagine what might have happened to the medical arm of DEFUSE 3, if they had received a TNK bolus. We suspect that it would push those 15% good outcome rates up to 30-35%, a huge benefit for those patients. This is where we’d like to go in the next trial, and in addition, we would like to see TNK given to the late window endovascular patients as well. We’re looking at 45% to 50% good outcomes in the endovascular arm of DEFUSE 3 and DAWN, but in SWIFT PRIME and EXTEND IA, good outcomes were 65% to 70%. I suspect a large part of the difference was the lack of an IV thrombolytic in the late window trials.
This opens up, we think, a paradigm shift for how stroke trials can be thought of in the future. We need to get away from the arbitrary time windows and focus on finding patients with salvageable tissue and test appropriate therapies. This strategy should allow us to treat many more patients successfully in the future.
Dr. Limaye: Amazing indeed! What were the important differences between DAWN and DEFUSE 3?
Dr. Albers: A very good question, and we look very closely at that, and report exactly what the differences are in our New England Journal of Medicine paper. Primarily, it’s two categories. One is the mild NIH stroke patients, because they were using a clinical/core mismatch, they needed to have severe clinical deficits to be confident there was going to be very likely a large volume of tissue that was hypo-perfused.
Because we used the perfusion/core mismatch in DEFUSE 3, we were able to take patients with lower NIH stroke scales, so patients with scores of 6 or higher were included. In particular, the difference is that patients with NIHSS scores of 6-9 were excluded from DAWN, and we had a fair number of those in DEFUSE 3. When we look at the response of those patients, particularly if you look at our primary endpoint, the ordinal analysis of the full Rankin scale, those patients had statistically significant benefits all the way down to an NIHSS score of 6. We’re very excited that this expands the benefits of late window therapy to a lower NIH stroke scale group.
The other large group that did not qualify for DAWN was because their ischemic core size was too large. In DAWN, you had to look at the age of the patient and their NIH stroke scale and their core size to determine eligibility. For example, for those older than 80 years in DAWN, their ischemic core had to be less than 20 ml, where we would allow those patients to have a core up to 70 ml. Similarly, depending on age and NIH stroke scale, there was a group that was excluded for core > 30 or > 50 ml. Many of our DEFUSE 3 patients had core sizes that were too large to meet these thresholds. When we look at these larger core patients, we can see that the good outcome rate in both the endovascular and the medical control arm in DEFUSE 3 declines with larger core sizes, but the separation between the two, the treatment effect, looks to be very stable. Therefore, we feel that we have now expanded the treatment benefit to a larger core population. The only other group that differed was based on the modified Rankin score (mRS) pre-stroke. Most of the thrombectomy trials required a baseline mRS of 0-1 to qualify. We decided to push that up to 0-2. That’s because our primary outcome was an ordinal shift across all levels of disability.
What we found is that there was a really incredible shift on our primary analysis. The odds ratio of 2.8 is the largest odds ratio reported for any thrombectomy trial, whether it’s early or late window, and when we corrected for imbalances in the stratification factors, the Or increased to 3.4, which is really phenomenal. The reason for that is because we have these more severe, larger core patients, who are not getting to a zero, one, or two, but they are shifting out of the five and six range. The mRS 5 scores are so devastating to have people who are completely dependent on others for all activities of life, confined to a nursing home, bed-bound. We’d wanted to be able to shift away from death and the most severe disability. We dropped that mRS 5-6 outcome from 42% to 22%, a 47% relative risk reduction, which was highly statistically significant, and that’s unprecedented in prior studies. Of all the results from the study, it’s the one that I’m most proud of.
Dr. Limaye: Great.
Dr. Albers: Overall, about 40% of our patients would not have met that DAWN criteria, and our projection was that probably the treatment effect in this subgroup wouldn’t be quite as robust as what we would see in the DAWN-eligible patients, because the DAWN criteria were carefully selected to really find ideal patients: very small core, large NIH stroke scale. This was expected to be the “cream of the crop” for reperfusion response.
We were extremely pleased to see that there was no difference in the odds ratio for the primary outcome between DAWN eligible and ineligible patients in DEFSUE 3 and both groups had a statistically significant benefit. Our primary outcome was as good, if not slightly better, for the patients in the non-DAWN eligible category. Obviously, we have tremendous reassurance that the patients who meet the DAWN criteria benefit as we now have two studies that demonstrate a large treatment effect. But we are also confident that this benefit can be extended to the non-DAWN patient.
What about DAWN eligible patients who did not qualify for DEFUSE 3? The key patients there are the 16- to 24-hour patients. We were very interested in those patients, but we polled our StrokeNet investigators before we started, and said, “What time window are you most comfortable with?” They were more comfortable going at 16 than 24. We have up to 16-hour data, but as I mentioned earlier, the treatment effect was not waning. If anything, it was expanding towards the end of our window. Therefore, we feel very confident that we can use the DEFUSE 3 criteria beyond 16, and we probably will be able to feel even more confident when we perform an analysis of the DAWN patients in that window, and we apply the DEFUSE 3 criteria.
One of the things many people don’t realize about DAWN was even though it was a clinical/core mismatch trial, they have perfusion data, because the majority of their patients were enrolled based on CT perfusion, using the same RAPID software that we used in DEFUSE 3. When you have the RAPID output, you get the ischemic core, which comes out in pink, and you also get the green hypoperfusion volume that represents tissue likely to go on to infarct.
Among DAWN patients, virtually all have perfusion/core mismatch, and that’s exactly what the clinical/ core mismatch criteria were meant to do. In the MRI group in DAWN, there was no requirement to do a perfusion scan (which requires a gadolinium bolus), but many of the sites did that anyway so the data are available.
What this means is that we can use the DEFUSE 3 criteria within the DAWN study, just as we were able to apply DAWN criteria to a subset of our DEFUSE 3 patients, and we’ve submitted an abstract on this topic to the European Stroke Organization meeting.
Dr. Limaye: How generalizable are these results to a larger population base than academic centers as compared to community centers? How do you think it is going to impact the management of acute ischemic stroke patients with large vessel occlusion in the U.S.?
Dr. Albers: I think it’s going to have a huge impact! I think stepping back, in addition to the tremendous benefit to patients, we have a huge benefit into how we now look at how strokes evolve. In the journal Stroke, simultaneously with the DEFUSE 3 publication, we wrote a paper called “Late Window Paradox,” which tries to explain what’s happening here. What we can see, and is outlined in that paper, is about 50% of the large vessel occlusion patients evolve their stroke at a rate that nobody would have expected, which is at a very, very slow rate for many hours, and then at some point, the collaterals give out, and the stroke takes off.
During the time you’re in that slow progression phase, you have a huge treatment opportunity. Once the collaterals give out and the infarct grows, your treatment opportunity disappears. The other side of the coin is these early window patients. There’s about 20% of these patients who have terrible collaterals. Their stroke grows very, very rapidly, so even though they come in in the golden hour, and they get treated right away, basically what you’re doing is reperfusing a completed infarct. This is part of the reason that you see this huge “time is brain” effect when you don’t separate out these different populations, because it’s that malignant growth population where not only does every minute matter, but every second matters for those patients.
If you can treat those patients ultra-early, you can have a benefit. As you get to the later time window for those patients, not only do they not have a benefit, but we’re quite convinced that we harm some of these patients with reperfusion, and we’ve seen this over the years at Stanford, because we perform advanced imaging in all our patients. There’s a subset of these fast-growing patients who have such horrible collaterals that they essentially have no blood volume in the area of the infarct. We can see this on the RAPID maps we call the image columns. They give you a color-coded map of the cerebral blood volume. We’ll see these occasional patients who come in early, but the cerebral blood volume is essentially zero in a large region of the infarct.
This is akin to a cardiac arrest of that region of brain, and we know that when you have zero blood flow to the brain, that six minutes may be your window of opportunity, not three hours, or six hours, or sixteen hours! What we’ve seen, and we’ve reported this data with our colleague from Australia, Bruce Campbell, is that when you reperfuse an area that has no blood volume, that you are at risk for two very bad outcomes. One is obviously brain hemorrhage, because you’ve got damage to the blood-brain barrier, but the other, which may even be more common, is massive edema. You have tremendous leakage of the blood-brain barrier, you restore blood flow to this area very damaged blood-brain barrier, and you can get massive edema, and these patients are going to need hemicraniectomy, or they wind up often having withdrawal of care, because they get massive shift, edema. These are patients that we don’t do any favor for when we reperfuse them. That’s why in a study like DEFUSE 3 or DAWN, where you’ve eliminated that patient population, you have the opportunity to get huge treatment results.
Dr. Limaye: What changes and systems of care do you think we’ll have to implement to use the protocols for DAWN and DEFUSE 3?
Dr. Albers: This is a game-changer. This is a complete landscape change. Akin to what happened three years ago when we had the five New England Journal trials showing that thrombectomy works. It worked for people who are selected with more advanced imaging, but it also worked in people who were selected just with a plain CT and a CTA. What was the landscape change there? It was that primary stroke centers, people who are receiving stroke patients within six hours, had to add CTA. If they didn’t add CTA, they were not meeting the standard of care, because if you had a patient with an M1 occlusion present to your hospital within six hours, and you didn’t clarify if that patient was a thrombectomy candidate, you were doing a huge disservice to that patient.
So now in primary stroke centers, everybody has to do CTA. If they don’t, and they get a sub-six-hour window patient, they need to ship that patient out immediately. They cannot accept those patients, and really, they should be bypassed if they can’t perform a CTA. What’s going to happen in 2018 for six to 24 hours? There is now a level 1A guideline, that if these patients meet DAWN or DEFUSE 3 criteria, they need to be treated! It steps things up quite a bit.
You now have a choice of CT perfusion or an MRI scan. Most primary centers are not going to be able to do an urgent MR. If they can, that’s terrific. We love MR at Stanford. We developed the RAPID software based on MR, but we translated it to CT perfusion because we knew that CT is how stroke patients are evaluated, not only in the U.S., but throughout the world. Any scanner that can do a CTA can do a CT perfusion. What we’re pushing the manufacturers to do is to do both studies, CTA and CTP, with a single contrast bolus, and many scanners can do it.
We’re working on automating the CTA processing, just as we’ve automated the CTP, so that once that scan is done, the information goes to the software, and then it’s not only sent to PACS, but it’s sent through secure email to whoever you want to see that information. If you’re a primary stroke center, and you don’t do thrombectomy, and you have a relationship with a comprehensive stroke center, those images can immediately go to the stroke team at that comprehensive center. This way the comprehensive center knows at the same moment what you do, that there is a patient in the late window who has met the DEFUSE 3 or DAWN criteria, and they can start to get ready to treat.
We’ve already been instituting this protocol at Stanford. We have some of our referring hospitals that have the RAPID software. I get the map on my iPhone the same time that they do, and when I’m on call, I know this is a patient for whom I’m about to get a phone call. We’re calling the transfer center, we’re sending the helicopter, because this is a patient that we want to get to Stanford right away. It’s a huge advantage for us, because Stanford Hospital is very full, and we don’t want to transfer every patient who comes in to our referring hospitals in a six- to 24-hour window. We’d be completely overwhelmed. It’s only about 50% of these patients, with an M1 or an ICA occlusion in that time window, who meet the DEFUSE 3 criteria.
DEFUSE 3 criteria is more generous than DAWN, so we use DEFUSE 3 criteria. If you think of a patient with an M1 or an ICA going into your referral primary center, about half of those patients are going to be appropriate for transfer, and about half are not. The half who are not belong at home, where their primary doctor is, where their family is. They shouldn’t be filling the beds of a comprehensive center. The half that need the thrombectomy should be transported immediately to the thrombectomy center, where we can really help them and have these huge benefits, and then as soon as they’re stable, they need to head home.
Dr. Limaye: As DEFUSE 3 used the newly created NIH StrokeNet platform, how was your experience working with this new built clinical trial platform?
Dr. Albers: It’s pretty exciting. We had to be patient because we were ready to launch DEFUSE 3 for quite a while. We had to wait for the network to be formed. The network will take you a few additional steps to get your grant funded. There are more committees; there are more approvals. There’s more oversight, so it delays your grant being funded, but once you’re funded, there’s some pretty huge advantages. They’ve got this network of the top hospitals in the country, running on idle, waiting for your protocol.
Once that protocol gets approved, then the contracting goes more quickly, there’s a central IRB that can expedite some of the IRB work. The local IRBs, they still want to take a look and provide some input. They’re not completely out of the loop. You do have to communicate with the local as well as the central IRB, but then once you launch, activating the centers can occur very quickly. We activated our centers pretty fast. We had 40 approved centers. We were able to activate 38 over a year. By the time we got to about 20 of those centers activated, we were enrolling at 20 patients per month, which is a faster rate than any of the prior thrombectomy trials.
As we headed up towards our 40 sites, we were clearly going to be enrolling at least 25 to 30 patients per month, which is just phenomenal, to think that you could enroll 300 patients in a year. You can answer critical questions very, very fast. Before we were fully up to speed, the DAWN trial news broke, and we had to stop. The advantage to the StrokeNet is that this group was able to enroll thrombectomy patients at a very rapid rate. You have access to very experienced centers who are not only experienced with thrombectomy, they’re experienced doing clinical trials.
We had a tremendous number of patients from very high ranking academic centers, but StrokeNet means that you’re not only working with academic centers, which are typically the regional coordinating centers, but they’re finding the best primary stroke centers in their region, and some of the primary stroke centers in DEFUSE 3 had exceptional enrollment.
As opposed to some trials, where they get the vast majority of their enrollment from two, three, four centers who are truly expert centers and probably have very low complication rates and very fast treatment times, DEFUSE 3 had enrollment from 38 different centers, so the results are clearly applicable across the country. They’re applicable to primary centers who do thrombectomy. They’re applicable to the comprehensive centers.
Dr. Limaye: Anything else that you would like to add that we did not discuss?
Dr. Albers: I’m incredibly excited that we’ve gotten to this point. There was a tremendous amount of skepticism over the years, because we had this concept of everyone has rapid infarct growth ingrained, and I’m not exactly sure why this got ingrained. It was certainly a wonderful mantra for getting patients motivated to call 911, getting doctors to rush around. It’s obviously the right thing to do for patients, to move as quickly as possible. But the downside of that is that it gave the impression that the wake-up stroke or the 12-hour stroke had missed the boat, and that I don’t know how many times in my career I’ve had to say to the patient, “I’m so sorry, but you didn’t come in soon enough for tPA,” or, “I’m sorry, you didn’t come in soon enough for that clot-retrieval procedure you read about in the newspaper. I can’t offer that. It might be dangerous. We don’t have any evidence. It’s not in the guidelines.”
That approach restricted our ability to treat a huge number of patients over the years who could have benefited. It’s a very, very happy time for us to see that this idea that stroke patients should be treated as individuals, rather than based on arbitrary time windows, has now panned out so beautifully with the new data.
Dr. Limaye: Do you think in the future that the time windows will be completely obsolete in treating acute ischemic strokes?
Dr. Albers: I don’t think they will completely vanish. I think we certainly need to expand out the thrombolysis time window. We could see patients from DEFUSE 1, who had beautiful results when we treated them with IV TPA at six hours. I’m very confident that if we can safely reperfuse these large perfusion deficits at 24 hours with a catheter, that we can also safely reperfuse a lot of these patients with thrombolytics in late window. It makes me excited about the M2 the M3 occlusions, the patients who may not be going to be great candidates for thrombectomy, because the vessel is really too small to get the stent retriever into; these patients are going to have an opportunity to benefit from an intravenous thrombolytic agent, even if they present late.
There’s going to be some limit, right? We know that we can go out to 16 hours and still have about half of our M1s and ICAs having salvageable tissue, but clearly, if we go out beyond 24 or 36 hours, this window typically closes. I think that’s partly what we’re seeing in our medical arm of DEFUSE 3. The spontaneous recanalization in that medical arm was 18%. It was 18% throughout the time window, but what you can see is that those patients who are having that spontaneous recanalization late in the time window, they were not having a benefit from it, and that’s probably because the collaterals were running on empty there. If you take the picture and the core is small at 16 hours, and then three hours later, you get your spontaneous reperfusion, there’s a chance that when you’re that far out in the window, that’s around the time you’re going to see collateral failure.
We see a form of this not infrequently in our clinical practice. There will be patients who present with an MCA occlusion. They have a big deficit, but by the time they get to the ER, their NIH score has dropped dramatically. They may be a two or a three. They’ve got an M1 occlusion. They’ve got a large penumbral region, but they look great clinically, because of those collaterals. The endovascular therapists are quite hesitant. They don’t want to take somebody with an NIH of one or two to their cath lab, so we put the patients in the ICU. We watch them carefully. We make sure their blood pressure’s good, they have good fluid status. Then we go to bed.
Then the pager goes off at 2 a.m.; the patient has now dramatically changed. They have their original deficit back, their aphasic, and the hemiplegia is back. We say, “We need to get this patient to the cath lab right away.” That means calling everybody in who are also in bed, driving into the cath lab, firing it up. The phone call to reperfusion time at 2 a.m. is not 90 minutes. It is not what we could have achieved if we had treated that patient earlier, but it’s an example of what can happen when the collaterals fail.
I think that this is really a boon for imaging. It tells us that we need to know more about the collaterals. We need to be able to predict who is the patient where the collaterals will last? We do have the 10% to 15% of DEFUSE 3 patients in the medical arm who did great. What’s happening with those? Those appear to be the patients who have such terrific collaterals that they don’t fail. Here’s a real challenge for us moving forward: How do you identify the good collateral patient, where they’re good enough that they’re going to make it, they’re going to get discharged home without a deficit, versus the one who you’re going to get the 2 a.m. phone call?
Dr. Limaye: Thank you, Dr. Albers, for taking time out from your very busy schedule to talk to us. It was amazing to talk to you and learn about DEFUSE 3 and its impact on systems of care for acute ischemic stroke for large vessel occlusions in the U.S. Thank you!
Dr. Albers: My pleasure.