Another Stroke Study; Time is Brain, or is it?
As the spouse of a neuro-interventionalist I pay close attention to most of the stroke data because it impacts my day-to-day life. For those unfamiliar with stroke, this is an event in which a blood clot forms in one of the blood vessels that supplies the brain. This stops blood flow to a specific region depriving the brain of precious oxygen and nutrients. Within a few hours that region of the brain is dead. As a result the patient loses function to the part of the body that region of the brain controlled. This could be speech, swallow, vision, balance, arm, or leg strength, it all depends on which region of the brain is affected.
Over the past decade a new technology has been developing that allows specialists to extract these clots if patients do not respond, or were ineligible for, medicines known as thrombolytics (clot busters).
Managing an acute stroke is a lot like trying to diffuse a bomb before the timer reaches zero. The initial treatment window from symptom onset was 4 hours. As research progressed it became 8 hours, then 12 hours. It further advanced to wake up strokes that had favorable perfusion studies, etc.
A perfusion study is a specialized CT scan that determines if an area of the brain is receiving blood flow, and to what degree. Patients that had poor blood flow to the affected area (a large core infarct) were considered poor candidates for clot extraction and deferred for medical management.
A couple of weeks ago I was scrolling through twitter when I came across this study: https://www.nejm.org/doi/full/10.1056/NEJMoa2214403
The publication demonstrated that patients with > 80% core infarct may be candidates for thrombectomy, regardless of time of symptom onset.
I had to do a deep dive to see if this could be true. My entire medicine training, when it comes to stroke, time is brain. Time is all that matter. Recognize early, work-up quickly, and intervene as soon as possible. Now time is irrelevant. What gives?
This is the SELECT 2 trial. SELECT 2 is an open label randomized international study initiated by the neurology department at the Cleveland Clinic. Patients presenting within 24 hours of symptom onset with a large clot in either their carotid artery or first level of the middle cerebral artery were eligible for the study
These patients had to have a large core stroke volume of at least 50mL of at-risk brain with an ASPECT score of 3-5. The ASPECT score ranges from 0-10 with lower scores suggesting a larger stroke. They were randomized 1:1 to thrombectomy (experimental group) or medical therapy (control). The primary endpoint was Modified Rankin Score (MRS) at 90 days. A table of the MRS is below:
The trial was stopped early for benefit after 352 patients were enrolled. A total of 20% (36) of patients in the thrombectomy group vs. 7% (12) in the medical care group were functionally independent at 90 days. There was no difference in mortality.
The Mann-Whitney test for superiority based on MRS score did not achieve superiority however. The median MRS score was 4 with a confidence interval of (3-6) in the thrombectomy group vs. 5 (4-6) in the control with a probability of superiority of 60%.
The researchers “reported” a single symptomatic intracranial hemorrhage in the thrombectomy group. Further complications of thrombectomy were 10 vessel dissections, 7 cerebral vessel perforations, 11 vessel vasospasms. Some of these may have occurred in the same patients. Many of these likely resulted in subsequent procedures.
In this study the authors demonstrated a potential benefit of thrombectomy in late presenting strokes with large core infarcts. This is contrary to the current practice in which patients with large infarcts are delegated to medical management. The benefit was modest. The median outcome of the treatment arm in this study left many with significant disability (MRS 4). The absolute risk difference for functional independence appears large (20% vs. 7%). However, the total number of individuals functionally independent was small. Also, the inclusion criteria were narrow. They sampled healthy patients that had strong functional independence prior to their stroke.
I assume most practitioners are already giving some of these patients a chance in the real world. When a young healthy person comes in with a devastating stroke we tend to error on the aggressive side given what they stand to lose. Also, in the setting of a clinical trial it is likely that many of these patients were given more time to recover to an MRS score of 4/5 than what occurs in the real world. Outside of a clinical trial, we typically do not keep patients in the 80’s on ventilators after a devastating stroke that in the best-case scenario will place them in a nursing home for the remainder of their lives.
The absence of blinding in these studies could heavily influence post stroke care, as suggested by the 19 patients discharged to hospice in the medical group vs. 11 in the thrombectomy group, and 72 patients discharged to inpatient rehab vs. 65 in the medical care group. Had those numbers been similar your MRS difference likely vanishes. However, important outcomes like functional independence and independent ambulation remained significant at 90 days, which is a strong signal that some patients are going to do well.
After a study like this there is usually a rush to update guidelines and expand indications for a procedure. The evidence based medicine website UpToDate already has a 1B recommendation attached to this paper. The highest recommendation level is 1A.
On the surface this is a positive study that should help people. After my initial read, I became quite depressed. I have quite a bias against positive stroke studies because I have a spouse that does this procedure and I like having her around. She and her partners already spend most of their days doing thrombectomies and coiling aneurysms, there aren’t many physicians in training ready to take on expanded indications, so where will the average hospital system find the manpower to meet this new demand? Where is the capacity going to come from? When will the robots be ready? We’re waiting…
So, this is my attempt at cross-examination (knuckles cracking). There are several major issues I have with this study. First, the perfect patient. Second, their reported low complication rate. Third, they stopped the trial early for “benefit.” Fourth, the use of the Mann-Whitney test. Finally, what does this mean for the interaction between communities and healthcare systems.
First, the perfect patients. Patients had to have a pre-stroke MRS score of 0-1, which would exclude many patients presenting with a stroke. This skewed their population young with few medical illnesses. In fact, they didn’t even report previous illnesses on their demographics table. They report median time from symptom onset. This indicates their time from symptom onset was not evenly distributed and skewed in some manner. A significant number of the patients enrolled presented within 6 hours of symptom onset. Only 1/3 presented after 12 hours. The trial was designed with this selection bias in mind.
Second, the low complication rate. The authors reported only 1 symptomatic post procedure cerebral hemorrhage. This would be a complication rate of 0.5%, which would be significantly below the known hemorrhage rate of roughly 4-11% depending on which study you choose to reference. There were 10 vessel dissections, 7 perforations, 11 vasospasm cases. They reported neurologic worsening in 25% of the patients in the first 24 hours post procedure as well. Yet somehow, these patients did better?
Third, they stopped the trial early. Internal review boards like to have interim analyses performed to ensure safety during the study. This trial was stopped early for benefit. Typically, when this is done the review board believes the group randomized to the control is missing out on a potential lifesaving intervention. This may be true in this case. However, when you have small event rates there is a risk the study was stopped at peak efficacy. What I mean is, unbeknownst to the researchers, they happened to peek at the data at the optimal time, and it wasn’t going to get any better. For example, let’s say they enroll 50 more patients in each arm. Of that 50, 5 are functionally independent in the medical arm, and only 1 ends up independent in the thrombectomy arm. This would not be unreasonable considering the 12% functional independence across the study. In this scenario your margin shrinks from 20 vs. 7%, to 16 vs. 8% with only 6 more events. Unlikely: sure. Impossible: No
Finally, the Mann-Whitney test. I’m sure most readers don’t pay much attention to the symbols superscripted above the data, but occasionally, they are important, and this is a good example. This test was required for the MRS data. What it indicates is that the data is not evenly distributed. No bell-curve. Therefore, the tails of the data are drastically skewed. My guess in this case is to the disability end. The Mann-Whitney indicates the probability that there is a difference between the two groups. In this case, it estimated with 60% probability of superiority. Kind of reminds me of Paul Rudd in Anchorman, “60% of the time it works every time.”
Despite my arguments one cannot deny the fact that this is a positive study, and that each one of my arguments has an agreeable rebuttal. There’s a strong chance that my initial heart sunk feeling will turn into a reality. Every governing body that has anything to do with stroke will likely pounce all over this.
This outcome is the absolute worst. I say that with full awareness of my explicit bias. It’s not good. But it’s good enough to change practice. These results do not compare to the original thrombectomy trials. MR. CLEAN demonstrated a 47% independence at 90 days and a full 2-point shift in MRS score that was statistically significant. ESCAPE demonstrated improved mortality with a 53% functional independence. EXTEND-IA showed a 70% functional independence, and this is after thrombolytics. SELECT 2 is 20%, with no significant improvement in overall MRS. It’s so lukewarm, it shouldn’t have happened. Let me explain. This is long, but it’s my blog, so I get to write what I want.
In my eyes, the paper raises a lot of questions. First, how did they get such a good result. If imaging suggests the infarct is complete…how did the brain regain function? If time is brain, how was time discounted in this trial? Is a perfusion image not giving us the information we originally thought or need? What characteristics did the 20% that improved of have that could have predicted success? Finally, and most important, what happened to all the MRS 3,4, and 5’s that didn’t die? You know, the other 80%.
A patient of mine passed away a few months ago. He was in his late 70’s, A bit of a loner. He had one son that would check on him from time to time. He was divorced and never remarried. He had significant health issues and really did not put much effort into improving his health. One morning the son couldn’t get a hold of him, so he called 911. The paramedics found him lying on his living room floor, still alive. He was rushed to the emergency room where he was found to have a large stroke. The imaging showed a near completed stroke, but there were no catastrophic changes on CT. The son didn’t know what his wishes were, so it was decided to “do everything.” I was surprised that the neuro-interventionalist took him for thrombectomy. The procedure went well but he never really recovered. He ended up unable to breath on his own with a permanent breathing and feeding tube. He was discharged to a nursing home. After a few weeks he was readmitted to the hospital with pneumonia and died.
I knew this patient. I’m sure he would’ve just rather died in that ER. He just didn’t have it written down on paper and hadn’t told anyone. This scenario is NOT an outlier. This is every day, at every hospital system, in America.
For all the usefulness MRS scores brings to the world of research, there is an equal amount of shortcoming in reality. Disability is a wide spectrum. Being unable to speak, feed, and walk is still alive, but may not be a life worth living. Having to move into your child’s home, spend thousands of dollars on adaptive equipment, home care, needing to be cleaned after every bowel movement by someone else because you can’t use the right side of your body, is disabled (MRS 3). It’s a reality that many of our elderly face after a stroke. Most people don’t plan for this because they fail to imagine it could happen to them. Stroke doesn't kill patients, it just severely maims them.
According to this study, for every 1 in 5 patients that potentially walk out of the hospital, there are 4 in 5 that are left to live in a nuclear wasteland. Families are left to pick up the pieces, often with limited resources. To further scrutinize the data, the control group was enrolled in hospice at a similar frequency as the success of the experimental group. This suggests an either explicit or implicit bias against patients that undergo thrombectomy that do not have a good outcomes.
The other issue to consider in circumstances with great disparity between success and failure is optionality. Optionality may be a good thing when you’re purchasing an automobile. However, in medicine it can be a detriment. This study will no doubt expand the use of thrombectomy, which will ultimately bring fewer ideal candidates for intervention into the fold. This will compress the margin of success leaving more disabled individuals that probably wouldn’t have made the same decision in hindsight. Clinical trials present a simplified picture of the world, when implemented in practice, nature reminds us it's not so simple.
Now onto surrogate decision makers. Often times when patients have massive strokes they can't talk, or understand language. That part of the brain is severely injured. The default setting for surrogate decision makers in these intense moments will be to call the “hail mary.” Go for it. "Do everything." Most decision makers will choose to gamble the 1 in 5 odds, rather than accept a certain loss, even if the loss is quicker and less painful. For most, based on this data, pain and suffering will be prolonged unnecessarily.
Unfortunately, one thing that can’t be done is to put pandora back in its box. As I said earlier, clinical trials present a simplified view of the world. We take this lens and change practice with it. Rarely do we look through another lens to see if the change makes sense. On the front lines we often discover the world is more complicated than we thought. The memory of our wins keeps us pushing, but the weight of the losses gets heavier and heavier for everyone to bear.
End of Rant
Copied from UpToDate. A comment after providing their recommendation. Helps prove my case.
"Note that all three trials enrolled patients who had very severe stroke deficits at baseline and enrolled very few octogenarians or excluded them entirely. Despite the benefit of MT, the majority of these patients were left with substantial disability; in the MT arms in SELECT2 and ANGEL-ASPECT at 90 days, the median mRS score was 4. Nevertheless, MT should now be considered the standard for patients with very disabling deficits, even if they have large ischemic core."