The shiny new toy in stroke treatment is endovascular therapy. There have now been 12 randomized controlled trials (RCTs) on endovascular stroke therapy (EST), with eight of the last nine showing positive results – stunningly positive. This flood of positive trials has led to new guidelines from the American Heart Association (AHA) and American Stroke Association (ASA)that extend the treatment window potentially as far out as 24 hours after last known well, and has spawned a movement to completely overhaul how we deliver care for patients with acute ischemic stroke (AIS). With all of the enthusiasm for EST, it is important to review the evolution of this new approach, to review and critique the evidence, and to evaluate what this means in clinical practice.
The initial positive evidence that led a lot of people to think that endovascular therapies would be the next great advance in stroke therapy came from the PROACT II trial in 1999. PROACT II used intra-arterial (IA) prourokinase, and although the study demonstrated benefit, the U.S. Food and Drug Administration didn’t approve it for this indication, citing the need for a confirmatory trial. Though never truly replicating PROACT II, researchers continued to investigate endovascular treatment options, and began to look at mechanical thrombectomy in addition to IA lytics. The trials showed that these devices were effective at opening vessels but didn’t show clinical benefit (MERCI in 2004 was the most well-known of these). Newer, supposedly better devices were developed and studied, with variable results. Further studies were initiated at this time, but took what seemed like an eternity to come to publication.
After three negative trials were published in the February 7, 2013 issue of NEJM (IMS-III, MR RESCUE and SYNTHESIS), the tide turned in 2015.
MR CLEAN was a big deal. Published on New Year’s Day after being presented at the International Stroke Conference (ISC) the previous November, it was the first major RCT to show significant improvement in clinical outcomes with endovascular therapy. So, let’s take a closer look at this one…
MR CLEAN was an open-label multi-center RCT of 500 patients from 16 centers in the Netherlands. They tried to address all of the issues raised with the 3 previous negative RCTs. They refined their inclusion criteria – to be included, patients had to have a large anterior circulation stroke with large vessel occlusion of the distal internal carotid (ICA), anterior cerebral artery (ACA) or middle cerebral artery (MCA) demonstrated by computerized tomographic angiography (CTA) or magnetic resonance angiography (MRA) (different from the negative 2013 RCTs) and be able to have intervention within 6 hours of symptom onset.
The results were impressive. The primary outcome of modified Rankin scale (mRS)of 0-2 at 90 days favored the intervention group, 32.6% vs 19%, number needed to treat (NNT) 7. There was no statistically significant difference in mortality or symptomatic intracranial hemorrhage (sICH), but 5.6% of the intervention group developed a new stroke in a different distribution (presumed due to embolization during intervention) vs. 0.4% in the control group, number needed to harm (NNH) 19.
Multiple other RCTs were underway when MR CLEAN’s results were released. All of them stopped enrolling patients, did an interim analysis, and sprinted to the finish line to get their results published.
Four of these trials were published in NEJM over the next six months:
- EXTEND-IA had only 70 patients enrolled at the time of the interim analysis and extended the treatment window slightly (intervention had to be started within 6 hours and completed within 8). The treatment benefit was huge – the NNT for reperfusion and, more importantly, for early neurologic improvement (reduction of at least 8 points on the National Institutes of Health Stroke Scale (NIHSS) or an mRS of 0-1) was about 2, and for a 90-day mRS improvement based on an ordinal analysis (more on this later) was 5.
- ESCAPE extended the treatment window even further, out to 12 hours. They used the ASPECT score on non-contrast CT (rather than CTA/MRA). The NNT for an mRS of 2 or less at 90dayswas 4 – again, a huge benefit.
- SWIFT PRIME required no large ischemic core on perfusion imaging. They also changed their imaging criteria half way through, a huge methodologic no-no which curiously did not generate much concern among those reviewing the study. Their primary outcome (functional independence at 90 days) had a NNT of 4 which, again, is extremely impressive.
- REVASCAT looked at patients within an 8-hour window who had no large infarct, and like the rest, had stunningly positive results.
- THERAPY was published in September 2016, despite not showing a statistically significant benefit (like all of the post-MR CLEAN studies, it was stopped early due to publication of other trials).What is most interesting about THERAPY is what they cited as the reason for stopping the trial. Less than a year after publication of the first trial showing clinical benefit for EST, the authors stopped their trial because they concluded that giving IV tPA alone would be “unethical.”
- THRACE is interesting. They too showed benefit, but not as profound of a benefit as the previous four trials. The NNT of 9 is still pretty darn good. But it is a higher NNT than the other post-MR CLEAN trials. And when the results were examined with an ordinal analysis, the difference was no longer statistically significant. An ordinal analysis is a statistical method used to detect changes in a graded measurement (in this case, the mRS, which is scored from 0-6) that are missed by a dichotomous measurement (i.e., 0-2 vs. 3 or greater), and to adjust for the differences in magnitude between various grades (i.e., applying a statistical correction if the difference between, say, an mRS of 2-3 is greater than the difference between a score of 3-4). Because it is able to identify patients who have improvements on the scale but don’t meet the definition of a good outcome (mRS ≤2), it has become the statistical crown jewel of the pro-stroke intervention crowd, making the results look more positive than if they were analyzed with the traditional dichotomous outcome alone. So the second largest trial in terms of patients enrolled showed much less profound benefit than the smaller trials that were stopped earlier, and that smaller difference disappeared when the data were examined with an ordinal analysis. I don’t want to give a critique of the ordinal analysis here (Rory Spiegel has already done that with his usual eloquence), but remember, when the ordinal analysis was needed to show benefit, the pro-intervention crowd trumpeted it as the greatest statistical advancement since the slide rule. When the same ordinal analysis contradicted the dichotomous outcome (+/- mRS ≤2) and didn’t show benefit in THRACE, it was only mentioned in passing in the article (one sentence).
What’s even more interesting is that the first five trials (MR CLEAN and the four very positive trials that rapidly followed) generated a ton of fanfare and were published in NEJM, while these two which showed smaller benefit if any at all were still in major journals (Stroke and Lancet Neurology), but not in the journal that has been the hub for all of the EST research, and they barely generated a whisper.
In November 2017, the next landmark trial in the EST literature was published. The DAWN trial looked exclusively at patients in an extended window, not within 6 hours, but at least 6 hours out (6-24 hours, specifically). So they were able to include the “wake up” strokes, the patients who were fine when they went to bed, and woke up with a stroke. Their results were fantastic. So fantastic, in fact, that they too stopped early after an interim analysis of the first 206 patients. For their co-primary endpoint of mRS 0-2 at 90d, the NNT was 3. Three!!! For a group of patients that we’ve always had nothing to offer except for PT/OT/ST and aspirin…
After the eye-opening results of the DAWN trial were released, the EST community breathlessly awaited completion of the next RCT in the pipeline. DEFUSE3, like DAWN, looked at patients exclusively in the extended window (6-16 hours). An interim analysis was done after the DAWN results were reported prior to publication, and the trial was (surprise!) stopped early. We waited for months to see the results, and finally, in a dramatic turn of events masterfully orchestrated at the International Stroke Conference (ISC) in Los Angeles in January of 2018, two important things happened simultaneously.
The DEFUSE3 results were published online and presented at the conference (again, wildly positive), and the AHA/ASA announced the online publication that day of their new guidelines. They included a Level Ia recommendation for EST for patients who meet the MR CLEAN criteria and can be treated within six hours of last known normal, as well as a Level Ia recommendation for EST in the 6-16 hour window and a Level IIa recommendation for patients meeting the DAWN or DEFUSE-3 criteria (two positive RCTs are required for a Level I recommendation, and since only DAWN went out to 24 hours, the recommendation for that time window was Level II).
Here is a summary of all of the studies to date on EST, in a well-designed chart that has appeared on this site before. One thing about this chart that is very different from the similar chart on the RCTs for IV tPA is the amount of green in it – the number of positive trials. One thing that EST has that IV tPA does not have is results that have proven to be reproducible.
Let’s pump the brakes for a minute…there are some concerns with all of this, and we need to mention the flaws and biases in these studies.
So,let’s start with the statistical and methodological issues with these studies.
- All of the trials underway when MR CLEAN’s results were released were stopped early. When a trial is stopped early and it is positive, the benefit tends to be magnified. Considering that MR CLEAN was completed and THRACE was one of the closest to completion when it was stopped, and that those two studies had the largest enrollments of all of the trials, and that their results were the least positive of all the trials…it seems like a reasonable conclusion that the NNTs of 3-5 in the smaller trials that were stopped early would likely have been higher if the trials were seen through to completion. Then, when you take a bunch of positive trials stopped early and combine them into a meta-analysis (like the BMJ, Lancet and JAMA did recently), that amplifies the effect even further.
- All were either partially or fully supported by the manufacturers of the devices and/or drugs being used. While this obviously does not negate the results, it is a cause for skepticism.
- Ordinal analyses, with all of their inherent flaws, were used in most of these studies.
- All of these studies compared EST to tPA alone – none compared to placebo. In the world of the investigators, where the utility of IV tPA for stroke is a settled issue, this makes perfect sense. In the world of emergency medicine (and in the world of methodologists whose serious questions about that data have never been answered), where it is far from settled, it is a different story. When there is no settled gold standard, the intervention needs to be tested against placebo.
- Intrinsic bias: many of these researchers had been involved in these projects for years or even decades, and have tremendous career capital invested in the effectiveness of this therapy. An illustration of this is the site selection process for the REVASCAT trial. For a site to be included in REVASCAT, there had to be at least 60 of these procedures done in the previous 12 months – and this was at a time when there was no evidence of clinical benefit for the procedure. These people were believers, or else they wouldn’t have been performing the procedure based on faith rather than evidence.
- All of the studies were non-blinded, which leads us to another problem…
- The mRS has been shown to have poor inter-rater reliability. Rory Speigel has written extensively on this as well, and a systematic review of 10 trials involving the mRS showed significant interobserver variability. That being said, it may be the best we have. But there is a real problem with using subjective measures with poor inter-rater reliability in non-blinded studies. When the people assigning the mRS values are not blinded and know whether or not the patient got the “special treatment,” the likelihood of that knowledge influencing their rating, particularly for a score with such high interobserver variability, is higher than anyone should be comfortable with.
So there is a plethora of statistical and methodological issues for academics to argue over. But there are bigger issues than these. Let’s assume that the data are correct – that the methodological issues above aren’t enough to discard the studies. Now what? Who would benefit from this therapy? The initial negative trials casted a wide net, and failed. MR CLEAN and the subsequent positive trials all looked at a very selected population – anterior circulation strokes involving an identifiable clot in a proximal large vessel (distal ICA, ACA, MCA) with a small infarct core and a large salvageable penumbra. But how many patients having stroke meet this criteria?
Look at the number of patients per center per year that were enrolled in each of these trials. Add all of these up and do a crude calculation, and you get 4.6 patients per center per year – about one every 2-3 months.
Let’s look at this another way. We are talking about roughly 1 in 770 stroke patients that would benefit from this therapy. A little over 0.1%. Where does that number come from? Most of these trials didn’t tell us how many patients they screened to get the patients they registered. Only one did – EXTEND-IA. In that trial, they screened almost 7800 patients, and registered 70. That’s 0.9%, or about 1/110, that were eligiblefor the treatment. The NNT in MR CLEAN, the largest and most methodologically sound of these trials, was 7. If you have to screen 110 patients to find one eligible, and need to treat 7 to benefit one, then about one out of every 770 patients having a stroke would benefit from this therapy.
In order to find that one patient, we’d have to screen them all. We would need to revamp our entire EMS system to divert every stroke patient to a center that is capable of rapidly performing and interpreting CTA or MRA. That’s 770 CTAs, at well over $1000 a pop, plus the radiation and contrast dye, for every one who benefits. Then the patients who were eligible would have to be sent to one of the few centers that have the devices and the staff available to do the procedure, many by helicopter. Are you doing the math on the societal costs of this? The societal and personal costs of stroke are massive as well, but the cost of revamping our entire system for this ultra-select group of patients, for a therapy with somewhat questionable evidence, is almost certain to blow the societal benefits out of the water. And that is what is being lobbied for as we speak.
It doesn’t stop there, either. Imagine you’re a hospital administrator, or you’re the interventional radiologist. You’ve spent all of this money on these devices, all this time and energy recruiting and training the staff and interventionalists to be on call 24/7 – and you’re only going to do it once every couple of months? Seriously?
So what are you going to do? You’re probably going to start stretching the indications, just like what is being done now for IV tPA for stroke. You just know that they’ll be doing these procedures on patients outside of the study protocols. That may not always be horrible. But remember, when we tried this therapy on a population that wasn’t highly selected, back in 2013, the results were completely negative. So if indication creep occurs, as it almost certainly will, given the factors noted above, we will be performing this procedure on the same patients we were performing it on in IMS-III, MR RESCUE and SYNTHESIS. And those patients did not benefit.
Clinical Bottom Line
- 8 RCTs have demonstrated the benefit of EST (NNT: 2-12) for a highly selected group of patients: large anterior strokes with a demonstrable clot in a large vessel (distal ICA, ACA, MCA) with a small ischemic core and large salvageable penumbra
- There are issues with the data, but unlike IV tPA for stroke, the results for EST have been consistently replicated
- EST benefits only about 1 of every 770 patients with acute ischemic stroke
- The extended treatment windows resulting from the DAWN and DEFUSE-3 trials may bring that number down
- Areas for future study:
- Attempting to identify more patients who benefit
- EST alone for patients in the window for IV tPA (unlikely in the current climate)
- Confirming benefit out to 24 hours (i.e., replicating the DAWN trial)
Endovascular stroke therapy will be one of the most hotly debated topics for the next several years. Despite the flaws in the evidence, the reproducibility and wildly positive results of the studies provide some optimism that, for the tiny subset of patients who meet the inclusion criteria, significant benefit is possible. But the societal costs and potential for indication creep are major causes for concern, and will undoubtedly lead to battle lines being drawn and fiercely defended, just as has been the case for IV thrombolysis for stroke for the past two decades.
Guest Post Written By:
- Furlan A et. al. Intra-arterial prourokinase for acute ischemic stroke. The PROACT II study: a randomized controlled trial. Prolyse in Acute Cerebral Thromboembolism. JAMA 1999. PMID: 10591382.
- Gobin YP et. al. MERCI 1: a phase 1 study of Mechanical Embolus Removal in Cerebral Ischemia. Stroke 2004. PMID: 15514171.
- Quinn TJ et. al. Reliability of the modified Rankin Scale: a systematic review. Stroke 2009. PMID: 19679846
- Kidwell CS et. al. A trial of imaging selection and endovascular treatment for ischemic stroke. N Engl J Med 2013. PMID: 23394476.
- Ciccone A et. al. Endovascular treatment for acute ischemic stroke. N Engl J Med 2013. PMID: 23387822.
- Broderick JP et. al. Endovascular therapy afterintravenous t-PA versus t-PA alone for stroke.N Engl J Med 2013. PMID: 23390923.
- Rory Spiegel (EM Nerd) at emcrit.org: The Adventure of the Cardboard Box.
- Berkhemer OA et. al. A randomized trial of intraarterial treatment for acute ischemic stroke. N Engl J Med 2015. PMID: 25517348.
- Goyal M et. al. Randomized assessment of rapid endovascular treatment of ischemic stroke. N Engl J Med 2015. PMID: 25671798.
- Campbell BC et. al. Endovascular therapy for ischemic stroke with perfusion-imaging selection. N Engl J Med 2015. PMID: 25671797.
- Saver JL et. al. Stent-retriever thrombectomy after intravenous t-PA vs. t-PA alone in stroke. N Engl J Med 2015. PMID: 25882376.
- Jovin TG et. al. Thrombectomy within 8 hours after symptom onset in ischemic stroke. N Engl J Med 2015.PMID: 25882510.
- Rory Spiegel (EM Nerd) at emcrit.org: The Adventure of the Cardboard Box Revisited.
- Bracard S et. al. Mechanical Thrombectomy After Intravenous Alteplase Versus Alteplase Alone After Stroke (THRACE): A Randomised Controlled Trial. Lancet Neurol 2016. PMID: 27567239.
- Mocco J et. al. Aspiration Thrombectomy After Intravenous Alteplase Versus Intravenous Alteplase Alone. Stroke 2016. PMID: 27486173.
- Badhiwala JH et. al. Endovascular Thrombectomyfor Acute Ischemic Stroke: A Meta-analysis. JAMA 2015. PMID: 26529161.
- Rodriguez FB et. al. Endovascular treatment versus medical care alone for ischaemic stroke: systematic review and meta-analysis. BMJ 2016. PMID: 27091337.
- Goyal M et. al. Endovascular thrombectomyafterlarge-vesselischaemicstroke: a meta-analysisof individualpatientdatafrom fiverandomisedtrials. Lancet 2016. PMID: 26898852.
- Nogueira AP et. al. Thrombectomy 6 to 24 Hours After Stroke With a Mismatch Between Deficit and Infarct. N Engl J Med. PMID: 29129157.
- Albers GW et. al. Thrombectomyfor Strokeat 6 to 16 Hours with Selection by Perfusion Imaging. N Engl J Med 2018. PMID: 29364767.
- Powers WJ et. al. 2018 Guidelines for the Early Management of Patients With Acute Ischemic Stroke: A Guideline for Healthcare Professionals From the American Heart Association/American Stroke Association. Stroke 2018. PMID: 29367334.
- Salim Rezaie at REBEL EM: Endovascular Therapy for Acute Ischemic Stroke.
For more on this Topic, Check Out:
- Rory Spiegel (EM Nerd) at emcrit.org: The Case of the Erratic Pendulum.
- Anton Helman at EM Cases: Journal Jam 10 Part 2: Endovascular Therapy for Stroke
- Rory Spiegel (EM Nerd) at emcrit.org: The Case of Corporeal Clock.
- Justin Morganstern at First 10 EM: Endovascular Therapy for Stroke – The Evidence.
- Ken Milne at the SGEM: SGEM#137: A Foggy Day – Endovascular Treatment for Acute Ischemic Stroke.
Post Peer Reviewed By: Jenny Beck-Esmay, MD (Twitter: @jbeckesmay) and Salim R. Rezaie, MD (Twitter: @srrezaie)