Chapters Transcript Video Real-World Interventional Outcomes for Cardiogenic Shock Complicating Acute Myocardial Infarction Dr. John Brush describes the Sentara Cardiogenic Shock Project and related trials with outcomes and statistical analysis. Good morning. We'll go ahead and get started. I'm John Brush. Um Many of you know, me because I've practiced interventional cardiology in this area for 32 years and more recently become Chief Research Officer for Centa Health Care. So it's my pleasure to give this talk this morning. Thanks for inviting me. I wanted to talk about um acute my cardial infarction complicated by cardiogenic shock. I have no disclosures or any type of relationship within industry related to this talk. Um This talk is summarized in a paper that we've submitted now, uh real world interventional outcomes for cardiogenic shock complicating acute myocardial infarction. And you can see uh my co-author Ann Harper worked with me um on the analysis of this Luke Cohen. Um two students, uh Zachary Booker and uh Kayla Da Dan de Moody and also Dak Cal Asia. So I wanted to um call out uh my collaborators in putting this paper together. So um this started out uh years ago um where I uh and this was when I was still practicing interventional cardiology. And the hypothesis was that uh devices for cardiogenic shock don't show effectiveness because at both ends of the risks, risk spectrum, the devices are futile at the, at the high end uh of risk people, you know, no matter what you do or, or uh you know, may, may not make it through. And at the low end, uh really nothing is necessary because you know, it may be cardiogenic shock related to uh temporary right ventricular dysfunction or, or, and so, predicting the midrange risk may identify patients who would actually benefit from interventional procedures. So I um I submitted an Irbi I, it was approved by the IRB actually five years ago and I did some initial work with Enterprise Analytics and, and had C uh an organization at EV MS that did statistical analysis at the time. Enterprise Analytics uh was not quite geared up for this type of work. A nurse at Enterprise Analytics helped me and, and someone would help me pull the data. Um and they, they delivered deidentified data to me. And, and so what I noticed is I noticed that um we had difficulty matching our N CD R data with Epic and we had lots of missing data and lots of uh missing matches. Well, enterprise analytics has has improved. Um uh Angela TPP was hired and she's brought on new people and one of those people is Anne Harper and many of, you know, Anne Harper, she was added to our study team. And so, um after sort of, you know, thinking that this project wasn't gonna be possible. We were able to revive this uh this project last year, Anne Harper and I met every Monday for months uh working on this and working on uh the data extraction and making sure the data uh match properly and, and working on on the analytics. And so I wanted to uh just to say that I really appreciate Anne's hard work and persistence and and and collaboration. So just some background cardiogenic shock is a complication of acute myocardial infarction that occurs in about 10% of semi patients. The in hospital mor mortality for uh ami with cardiogenic shock is about 40%. Um And randomized control trials are hard to do in this setting. We're focused on door to balloon time, getting people uh to the Cath lab as quickly as possible because we know that uh that, that uh has an impact on mortality. Um And so, uh you know, in that rush setting, it's, it's very hard to uh do informed consent, but also it's difficult to get uh informed consent in the setting of taking care of extremely sick people. Many of these people have had cardiac arrest, they may be unconscious or they may be in uh severe distress, respiratory distress related to pulmonary edema. And so it's very, very difficult to do randomized controlled trials in this setting. So there have been a few randomized controlled trials and, and everyone uh knows about the shock trial which was led by Judith Hockman as published in New England Journal in 1999. And it showed that uh doing a primary angioplasty um in patients who come in with cardiogenic shock improved the six month and one year mortality. This was with plain old balloon angioplasty. Back in 1999. More recently, there was the Aib P shock two trial. Uh This was published in the New England Journal in 2012 and it showed that balloon pump counterpulsation had uh no effect, either positive or negative effect in 600 patients with acute myocardial infarction and cardiogenic shock. There have been a uh two undersized randomized controlled trials that compared balloon pump and impala and basically found no difference between those two interventions. There's was the um ECLS shock trial um that was published in the New England Journal um in, in 2003 and it showed no effect of eco in 420 patients with uh ami and cardiogenic shock. And then just uh just last March was the publication of the danger shock trial was presented at the AC C and that with simultaneous publication in the New England Journal last March. So there have been several outcomes uh studies over the over the years. Um looking at cardiogenic shock, there was a match registry where um uh investigators um took uh took uh prospectively collected data on uh impella and compared that to match patients from the IABP shock two trial and they found that there was no effect of impala, but there was uh no positive effect on, on, on mortality of the impella. But impella was associated with some serious complications. There was a National N CD R analysis and they used probabilistic matching so that they could match data sets and then they use propensity matching to compare impala with comparison groups. And they found that impella actually had a higher death and bleeding rates. There have been three administrative claim studies uh again using propensity score matching. Uh and each of those uh three studies also showed that Impala had higher death rates and higher complication rates uh compared to matching controls. Mhm. So, um so now along comes the danger shock trial, danger shock trial began recruiting patients 11 years ago. So it took 11 years to recruit patients into this trial. It was expanded from Denmark to Germany. That's the name Dan J. Uh And, and then uh also expanded to England. They screened 1211 cardiogenic shock patients to enroll 360 patients. So 70% of the patients were excluded and they actually completed the study with 355 patients. So, independent confirmation of randomized controlled trials is always good. Typically for AAA level 11 a recommendation that's typically two randomized controlled trials. But I think we can probably agree that this randomized controlled trial will never be repeated. Um It had very uh very tight ex inclusion criteria. Uh Participants had to have a systolic blood pressure of less than 100 or they had to be on pressers. Every one of them had an echocardiogram before they were randomized and they all had to have an injection fraction of less than 45% and no evidence of right ventricular failure. All of them had an arterial lactate and, and uh they required to have a lactate level of greater than or equal to 2.5 millimoles per liter. They excluded patients out of hospital cardiac arrest if they had persistent um loss of consciousness. They also excluded patients with severe peripheral arterial disease and frailty. Although um I have to say that uh they, they weren't very explicit on how they define uh severe peripheral arterial disease or frailty randomized was stratified by timing of the randomization compared to the revascularization. And by uh whether the patients had anterior M I or not. So, 57% of those patients got impala uh placed before they actually had revascularization and the remainder had revascularization first. So here are the results at the top is um the Kappa Meyer curve showing uh results over, over time in patients with standard care who had a higher percentage of, of, of, of death versus patients with um with impella with microvascular uh um um left ventricular device. And you, as you can see, um the results were uh were barely significant with a P value of 0.04. And um down below, there was uh a secondary composite endpoint uh is listed uh which was escalation of therapy, trans, uh either cardiac transplantation, escalation of therapy or death. And you can see in that situation, they, for some reason, didn't put a P value on this one when uh uh but it, it looks like the um confidence intervals for the HAZ hazard ratio doesn't cross one, but they didn't put the, the PP value. And my calculation would be that the P value would be slightly greater than 0.05. People have done something called the fragility index calculation. And they've uh determined that two less deaths in one group or two more in the other group would have eliminated the significance. So we've got a borderline significant study that took 11 years, which was highly exclusive for their inclusion and exclusion criteria. Um And, and the study will probably never be validated with another randomized controlled trial. So let's look at the R dari shock trials. If you looked at uh the country where they were randomized, uh the patients that were randomized in Denmark were uh highly significant but the patients who were randomized in Germany and the UK were not. And if you looked at other factors, um and you just scan down here to see which uh factors were uh were strongly positive, uh It clearly men. Uh whereas women, there was no no difference at all. Um and uh multi vessel disease, whereas in single vessel disease, no difference at all. Um And uh and so, and, and also uh uh oh here uh pa patients with uh very low blood pressure. Uh So, so men with very low blood pressure, with multi vessel disease had the most dramatic effect. Whereas uh women uh had no effect. Uh patients with single vessel disease had no effect. And so um really um raises questions about how you want to uh utilize this therapy. It had a lot of crossovers. There, there were nine crossovers in the impala group and three in the control group. And if you looked at um at uh as treated, if you looked at the, as treated uh results, there was no significant difference between impala and um and standard therapy. By the way, standard therapy did not include balloon pump, counterpulsation in Europe. They take the the balloon pump uh trial very seriously and they don't use balloon pump. Whereas I think in America and certainly here, I think balloon pump still uh is being utilized. So first do no harm. Uh This, this graph was in the supplemental material. It actually wasn't in the New England journal uh publication, but it was in the supplemental material. But it's uh really important because there were a lot of complications. Uh This composite safety endpoint which was severe bleeding, limb ischemia, hemolysis, device failure or worsening A I was significantly higher in patients with the impala versus the control. And you can see in 24% of patients they had this complication, moderate or severe bleeding, and 22% of patients with impala uh renal replacement therapy. And this is a real mystery to me because renal replacement therapy, dialysis was extremely, uh common in both the, the impala group and the control group. I, I can only guess that, uh, that in Europe, there may be, you know, very aggressive uh use of dialysis in patients in cardiogenic shock, uh to uh try to alleviate uh worsening renal function. Um um also uh and also sepsis with positive uh blood cultures uh were um more likely in patients with receiving impeller treatment versus the control group. The number needed to harm was six. The number needed to treat was eight. Typically we like to see trials where the number needed to harm is about four times at least four times higher than the number needed to treat. The number needed to harm. How many patients do you need to treat before you're gonna harm someone? Only six. How many patients are you gonna treat before you uh have mortality benefit? Eight. So there's, and, and so, and, and, and these patients harm. You're, you're talking about patients who are sick to begin with. So somebody who has a vascular uh complication uh or sepsis superimposed on a very, very severe uh problem in the first place. With a 40% baseline mortality that could make the difference. So, um uh and, and so, but somehow or another, they managed to pick the patients over 11 years where the harm was outweighed by the benefit. Whereas in every single observational study that's ever been done, it was the other way around and there are no other randomized control trials for, for validation. So questions remain the trial. Uh uh Authors of the study said that the trial supports judicious use of impella in the setting of an ami complicated by cardiogenic shock. And I think we can all agree that case selection is the key. Um They did uh echocardiograms on everyone. And so they chose people with ejection fraction of less than 45% and severe arterial lactate elevation. Oh and, and, and uh by the way, no RV involvement, I think that's really important, severe uh arterial lactate elevation in the map in the sixties. Uh Men only maybe, I don't know, um avoid patients with RV, failure, avoid patients with P AD. That's certainly a good idea, but it's certainly not settled yet. And despite the fact that I don't think it's really settled. Uh a similar trial that was underway in America, the recover for trial was canceled. So, um it's very disappointing it, they felt felt that there, there wasn't true equipoise. Um And, and they couldn't uh they couldn't ethically randomize people into the uh cover for trial because of the uh danger trial results. So, confirmatory uh randomized controlled trials are gonna be very unlikely. So it's in that setting that we can report out our study. Our study is a single center study. Um looking at ami and cardiogenic shock and conti and continuously uh treated patients at Centa because we're a single center, we didn't have to do probabilistic matching. We could do direct matching of patients because we, we, we've got their, their medical record number, we can directly match uh patients. And so we can match match data from multiple sources, which is what we did. The primary data source was the N CD RC PC I version 4.4 and five data over a period of six years. But we were able to match that data from vital signs and some labs that we were able to get from expert. So it was, it was information that was available to the operator at the time of the procedure that what we, we were trying to uh use uh uh data that was available. Um And, and mimicking what the operator would see at the time of the procedure. Also labs and Ivy drips, whether they were pres and iron troops, we were able to get that from epic and we were able to also get additional information from the master death index and epic regarding mortality. And uh because we're a single center and had the ability to match, we had a 99% match rate after uh uh we revitalize the study with Ann Harper. So the patient population was patients from eight hospitals to take care of sties um for uh using uh primary PC I for, for sty care. It was over a six year period from the beginning of 2017 to the end of 2022. If duplicate, if, if a patient had duplicate admissions for Stey during that period of time, we took the last hospitalization uh with an emergent PC I to avoid immoral time bias. If, if we'd taken the first, we, you know that by definition, they would have looked um and also the index event was the first procedure during that hospitalization because that's a hot, that's the procedure where the critical decision making was performed. Some patients were brought back to the Cath lab for, you know, a additional uh procedures of other arteries or, or maybe even an additional procedure of the that artery. But we uh the index event was the first PC I procedure during the hospitalization. Uh The data that we used um was the coded data according to the C A PC I definitions, we did not go back and try to re judicate or, or change any of the data, whatever data was was in the C A PC I registry. That's the data that we use. And the covariates that we use. The predictive variables that we use were data elements that would have been available to the care team at the time of the procedure. The outcome variables were death at 30 days. That's similar to the Aib P shock two trial or 100 and 80 days, which is similar to the danger trial. So we looked at both 30 day and 100 and 80 day mortality using, primarily using the master death index. Secondary outcomes were stroke, major bleeding, vascular site, uh access site injury, sepsis and new dialysis. I'll just say one thing that um one thing that we discovered is that when patients are transferred from one hospital to another hospital, our, our coders stop collecting in C uh in CD R data. And so that's a limitation of our study. It's not a limitation for death because we were able to match that with the master death index no matter where he went, no matter where you died. We, we, we were able to find that. But for some of these secondary outcomes, you'll see that um that there, there are limitations on whether we truly captured all of these secondary outcomes or not. So we did two statis statistical analyses we analyzed and compared at the patient level and at the hospital level. And we use Multivariate logistic regression model to uh to make the comparisons. So we call hospitals, malva, hospitals versus non malva hospitals, microvascular uh left ventricular device. Um And so ends up, there are three hospitals where uh those hospitals are prone to use Malva uh or Andela. Uh and there are um uh five hospitals in our system that also treat stem that are, that are not uh prone to use um impella. So, uh and, and the where the patients happen to go into the hospital is a relatively random thing, right? You know, they don't say, oh, I'm having a sty and I think I'm going into cardiogenic shock, maybe I'm gonna go to Norfolk General because they use impella. That's just not the way it works. People are randomly allocated based basically based on what is the nearest hospital. Some of them were transferred from one hospital to another hospital. And we, we uh well, you'll see we controlled for that. But there uh but there was sort of a natural experiment here where patients were allocated in almost random way to uh hospitals that are uh impeller prone versus non impeller prone. And so we compared the two treatments that are still being used in our system, intra aortic balloon pump and uh micro uh axial flow with impala. Now, there these two devices have very different effects and these are pre pressure volume curves uh with uh balloon pump versus micro actual flow pumps. And this is from the group uh in, in uh um in, in Denmark uh who are involved in the danger trial. So you can see there's a very slight decrease in, in uh after load, a decrease in preload. Uh the stroke volume stays pretty much the same. One thing that's not shown is that balloon pump presumably improves coronary blood flow because it inflates during diastole when coronary blood flow occurs. And so that's not shown here. Micro A uh flow pump has a very dramatic effect. Decreases blood pressure, markedly decreases stroke volume and decreases uh filling pressures, gives a steady pressure rather than a pulsing pressure. Um And there are people, I, I, I'm not an interventional cardiologist anymore. I, I don't uh put in impelled devices, but there are people that do that tell me that they are absolutely convinced that they have put impelled devices in patients and that is absolutely saved their life. And I don't, I don't, I don't question that for a minute. I believe that I believe that that's true. Impella roughly doubles your uh cardiac output. And uh and so uh there are clearly uh people that you can look at anecdotally and say the impella, the uh uh save that person's life. And so um I don't doubt that, but I, I do think that there are also people with a severe peripheral vascular disease where impella um may have caused more harm than good. So, um so these are the results um over a six year period of time, we had 3812 patients who uh were cared for at cento facilities with a merchant PC I for stemming and of those 505 consecutive patients with cardiogenic shock thir roughly 13.2% which is a around what we would predict uh had cardiogenic shock. According to the definition of the N CD RC PC I registry, the average age was uh 66 years. Um 100 and 63 or 32.32% died during hospitalization. That's uh a little bit better than what we would expect about 40%. So we're a little bit better than what we would expect. Um 38% died within 3030 days. 41% died within 100 and 80 days. Um And so of those patients, 100 and 60 patients were treated with uh intra e ball and counterpulsation and 73% were treated with impella. So if you look at the baseline characteristics of patients who were treated with balloon pump versus impala, they are actually remarkably similar. Our plan was to do propensity score matching. We really didn't have to because they were already pretty well matched. Except for this uh in FARC location in FARC location, uh was a significantly uh anterior and FARC location was significantly higher in patients who received impala compared to intra aortic balloon pump. Um and inferior M I was more common in patients who received intraoral balloon pump. So we know that this is a difference between the two groups and this is something that we're gonna need to control for. So if you look at outcomes and this is just uh un uh unadjusted outcomes uh discharge status according to the N CD R registry was significantly higher in the impala group compared to the Bloom pump group. But this may be uh affected by the fact that patients were transferred out and we didn't have fully capture everyone. Uh And, and here, here is the transfer rate. Uh 24% of patients uh with impella and 31% of 32% of patients with balloon pump were transferred out. And you can imagine what happened here. You know, they got a balloon pump uh if they're treated with uh for a QM I and got a balloon pump at Lee Hospital or immediately transferred to Norfolk General because we don't take care of patients with a balloon pump at Lee Hospital or Martha Jefferson or, or uh uh wherever. And so, uh so this is a problem and we, we got around this problem with uh by looking at the master death index, which is our is uh to, to look at 30 day and 100 and 80 day mortality, which is our primary outcome. But for some of these secondary outcomes, um we, we've just have limitations and, and so you can see that the bleeding tended to be higher in the impeller group. Um And uh uh compared to um compared to bing pump. So this is a univa analysis uh for patients who died versus patients who lived. And uh so you can see that these are all co uh uh candidate covariates for a multi variable regression analysis. So mean age um creatinine uh prep procedure, creatinine whether they had a pre procedure, cardiac arrest or not was a strong predictor. Um where uh in FARC location inferior M I actually uh was uh was a predictor for um uh being alive whereas uh anterior lateral M I uh uh wasn't independent or wasn't uh predicted here, lactate level and lactate levels were uh lactate levels that we have are all extremely high because lactate levels were selectively drawn. Uh And they weren't uh uniformly drawn on everybody. So the, the lactate level, the mean lactate level that we have in this population is extremely high, but it was significantly higher in patients who died, patients with severely low. Uh ph um was higher, mean heart rate was higher. Um And also uh human dynamic support with impala um was more likely in patients who died compared to alive. And so is that due to uh its uh correlation with some of these other variables, we have to see with multi variable regression analysis. This is a forest plot that shows the result of multi variable logistic regression uh looking at 30 day mortality. And you can see that the uh the variables that were independently uh predictive uh according to this model where heart rate very tightly uh uh distributed uh ph of greater than 7.2 baseline B un. Use of impala age, high lactate and cardiac arrest. So those are uh independently predictive of 30 day mortality. Uh In FARC location didn't make the cut for uh A as an independent variable. In this analysis, looking at 100 and 80 day mortality. It was the those same variables plus narc location, anterior and lateral. M I was a predictor for 100 and 80 day mortality. So now let's look at the hospital comparison. And for a hospital comparison, we wanted to exclude the people who are deliberately uh transferred to a hospital so that we could create, we could set ourselves up for that natural experiment. We wanted to look, we wanted to find the people who just naturally and randomly were allocated to one type of hospital versus another. And so uh looking at the Malva hospitals versus the non ma non malva hospitals. So we're stric for transfers in uh there were eight transfers to, to non mal vet hospitals, but there were 38 transfers to the malve hospitals. The malve hospitals were, were our tertiary care hospitals. Uh Norfolk General uh And uh primarily, and so um we wanted to restrict those. So uh when we did that, we had, we had roughly an equal number of patients in uh in those two uh two groups. And interestingly enough, um there was no difference in, in FARC location when we uh when we looked at, um uh when we looked at this analysis, um and so uh this sort of random allocation if you will, because of this natural experiment appears to be valid. Uh Actually uh where um where, you know, patients with anterior M I didn't have a choice. They went to the nearest hospital ends up there, obviously, where there was a difference in uh the treatment, somehow 14 per 14 patients were treated at Non Malva Hospital. So there's a crossover effect if you will and that crossover effect, if anything would have softened our difference. Um But uh but, but we, we do need to recognize that there is a a crossover effect uh there. So here are the outcomes by hospital, hospital type, a discharge status. This is from N CD R data. Um And so uh there's clearly a difference here. Now malve hospitals, there were very, very few of those were transferred out. Um You know, 16 of those were transferred out. Um It may have been somebody from Rockingham Memorial that was transferred to U VA I I don't know how that uh you know, not many people are transferred out of Norfolk General. Um But um and and many more people were transferred out um in the non Malva group, but looking at uh 30 day and 100 and 80 day mortality, there was a trend towards a worse 30 day mortality in the Malva hospitals and there was a significant increase in 100 and 80 day mortality. This is the mortality. This is the outcome that was evaluated by the danger shock trial. Uh And so our, our results are different than the danger shock trial, looking at some of these other outcomes. And again, these are limited because of our, our our limitation and follow up. But there was clearly more bleeding abnormalities uh in patients with mal back. Now, this is this is a solid number because the transfer out is relatively low. But this may be under counting here. Also a requirement for dialysis uh much lower in our patients than in the danger shock trial. So from this, we can conclude that intraortic balloon pump versus malva versus impala uh were well matched at the patient level except for in my location. So using multi variable logistic regression impala was an independent predictor of mortality with an odds ratio of 1.9 for 30 day mortality and 2.0 for 180 day mortality. Other predictors of mortality were low ph high lactate level heart rate, cardiac arrest, baseline B UN age, um uh age and, and for 100 and 80 day mortality, uh anterior or lateral M I. Other outcome measures were affected by transfers out but bleeding tended to be higher in the impala patients. Understandable. Um In the comparison of the MAL VD versus non Malva hospitals, they were geographically dispersed and isolated uh relatively isolated. And so patients were taken to the nearest hospital which set up a sort of a natural experiment after excluding the patients who were deliberately transferred into those hospitals. The pa the ho those hospitals, the patients who were treated at those hospitals were extremely well matched. And in that analysis, the ma that hospitals had a higher 180 day mortality and bleeding rates and trends for a higher 30 day mortality rate, access site injury and new requirement for dialysis. So uh clearly um device choice was uh based on whether you're a prone to put in uh impala, whether you're skilled at putting impala, whether your hospital has the capability of putting impala, so that clearly affected cho V choice. It appears that in FARC location also affect that. And so in FARC location also affected mortality. So in FARC location is a confounder. And so the question is, is whether we adequately controlled for confounding and if we did, then we can uh infer causation between the device choice and mortality. But that's the critical question. Did we adequately control for the confounders? Clearly, we need to address the bleeding and site uh access site complications in patients who are treated with impella. We need to be extremely selective regarding impella use to replicate the gen the dangerous shock trials. They were extremely selective. We need to be uh extremely selective. I don't I think that the reason for our different the difference between our study and their uh danger shock trial is we weren't quite as selective. We, we may have put it in an inferior in fars that may have had a small uh small LV. And, and I had uh uh right ventricular involvement and a small in LV with a, with a device uh sitting in it may have been actually harmful and clearly uh the lead complications. More research is needed to confirm the danger shock uh trial and better to find better case selection. And so uh you're not gonna get it from randomized controlled trials. But uh observational studies like this uh will help and product development is needed. Uh you know, studies like this will spur the company to uh develop the product further making smaller devices that may be less likely to cause vascular complications. So, if you look at trial co uh comparisons, um uh the danger shock excluded 70% of their patients. Uh IBP uh was less than that uh dropped from the study. This is an interesting thing dropped from the study. One patient from the treatment group and four patients from the control troop were kind of mysteriously dropped from the danger shock. Trial terms of crossovers. Nine patients crossed over from uh uh in the treatment group and three from the control group and in the control group, they did not treat them with balloon pump counterpulsation. If they needed escalation of therapy, they went to ECMO these people got ECMO. And so 30 37 patients got escalation of therapy to ECMO that may have harmed patients. The, the really interesting thing in the danger shock trial is the complication rate in their control patients, which is much uh higher you'll see than our com our our patients. So this is a busy uh uh busy uh slide. But this is uh Sana patients who got neither patients who got balloon pump patients who got impala, the danger of shocked patients who got impala versus standard care. And I be shocked two child who got uh balloon pump or got control. You can see that uh we had a much higher percentage of women as compared to the danger shock trial. As, as did the uh balloon pump tr uh trial, uh danger shock had absolutely no blocks. Uh We had, uh, we had um roughly 20 to 20% African Americans, 10 to 20% African Americans and hypertension was substantially higher in our population. 74% versus uh roughly 50%. So you can imagine an African American patient with long standing hypertension has got severe concentric left in high pery. Uh that, that patient uh may have a different echocardiogram on admission when cardiogenic shock than, than, than some of the dangerous shock patients. Diabetes was significantly higher in our patients compared to the danger shock trial. A QM I history of QM I was higher cardiac arrest. We did not exclude patients in our population due to cardiac arrest. Uh whereas uh roughly 20% of the patients in the danger shock trial were excluded. So how relatable is a danger shock trial to our population? Other trial comparisons, uh all cause 30 day mortality. Uh patients who were treated with uh with impella here were roughly comparable to the patients uh in uh or, or maybe slightly higher than the patients who were treated with uh impella in the danger shock trial. That may be uh because of the inclusion of patients with peripheral vascular disease. More women patients with inferior M I with RV, uh dysfunction rather than LV, dysfunction. Um The 180 day mortality was pretty similar um uh in hospital stroke uh was quite similar uh peripheral complications, uh similar severe bleeding. Um you know, we, we had maybe slightly more again because we were a little bit less selective um act actually, compared to the imp patients who got impella. We were, we were right in the range actually. So I I take that back, we were, we're right in the range. So we did not have more bleeding complications. Uh compared to impel. Um new dialysis was dramatically different and, and that I can't explain because the the dialysis rate uh in Europe was was um strangely elevated. So uh conclusions uh use uh my conclusion is not, not throw out impala, impala is a useful thing. Uh It, it works, it doubles cardiac output but use it selectively should it only be used in, in men or maybe men is just a marker for larger patients, certainly larger patients with larger arteries. Uh patients with anterior M I or, and uh and, and not with an interior M I, you have to make that decision, a snap decision and I'm not sure that we're gonna have the, the ability here in the middle of the night at all eight hospitals to do an echocardiogram before you decide, make your decision, you're gonna have to make your decision based on clinical factors and some of the clinical factors that we use uh in our study uh may be uh may be helpful. So inferior M I where you're suspecting RV involvement may not be such a good idea. Um Certainly low ph is, is something and high lactate level lac doing lactates on everybody might be something. Um And, and strongly evaluating cardiac arrest. We do that anyhow. Uh we don't uh you know, we, we're a little bit uh more conservative about jumping in and somebody who's had a prolonged cardiac arrest and has uh evidence of severe neurologic dysfunction. So, conclusions, right? Uh randomized control trials are not always the definitive answer. I think that this randomized controlled trial raises serious questions um in randomized control. And it's not because they didn't try that, you know, kudos to the to the investigators, they stuck with it for 11 years, but randomized control trials are hard to perform in severely ill patients. It's hard to generalize from trials that uh that may not represent our patients. No black patients, uh lower patients with hyper uh with hypertension or diabetes. Uh More uh fewer women. Um And so, and so because of that observational research has a role, it enables us to look at our, our, our local uh uh data and, and local adaptation and implementation of therapies observing uh uh uh patients uh also uh can be hypo hypothesis generating. Uh uh they certainly have a role for demonstrating a situation where a randomized control trial might be useful. Observational studies are not randomized but they can be controlled. So, more on observational research, there are people out there that say we should only believe experiments, any evidence other than randomized controlled trials is flawed. Well, I I say tell that to Copernicus and Galileo, we would still think that uh the sun revolved around the earth. Um If we didn't use observational studies, some questions just can't be uh easily randomized. Um And so, uh and, and observations can be controlled. So every action we observe our actions and based on those observations, we feed back uh to our actions. Observations are critical for feedback. And they're behavioral economists, uh economists who um are Nobel Prize winning uh behavioral econo economist who say that feedback is critically important for uh for calibrating our intuition. That's, that's required to make judgments um in um in, in medicine but beyond medicine. So observation, I I wouldn't be so quick to toss out observational studies. I, I'm firmly convinced that they have a role and you can use. People are, are, are, are, are really working on how to, to derive causal inference from uh from observational studies. There's a guy Judea Pearl who wrote the book of Why, which is an interesting book. But he's been, he is a data scientist. He's at UCL A, he's working on ways that uh you can start to develop um um causal inference. And one of the things is to develop these directed a cyclic uh graphs that that can can tell the story. Judea Pearl says, tell the story because the story behind the data is as important as the data itself. And if you know the story behind the data, you can start to control for confounding, you can start to figure out colliders and confounders and you can start to control your data in a way that you can start to create causal inference. Natural experiments are also something that's uh important Bobby Y at uh at Harvard and, and the B I, they've created the Smith Center for uh Outcomes Research and cardiology. Um And he wrote this article that was published just last year, bringing the credibility revolution to observational research and cardiology said it it's, it's possible to perform observational studies with methodological strategies that are better equipped to address threats to the validity of causal inference. And he said he also said it may help to share knowledge and methods from adjacent fields such as economics. In 2020 the Nobel Prize in economics was given to three people who uh developed the use of natural experiments to understand causal effects and policies in economics. You can't do, you can't do uh you can't do experiments to look at a lot of economic issues, but you can look for natural experiments. And these guys won the Nobel Prize for, for, for developing that. Here's an example, correlation between states that did Medicaid expansion versus uh other states and looking at all cause mortality and, and they're sure enough, there's a, there is a difference. So that's another example of a natural experiment where you can make causal inferences from a natural experiment and other other uh people are writing about using big data to emulate target trials when randomized controlled trials are not available. So um I think we need to learn more about that, learn more about how to do that. I'm trying to uh learn more about that myself. And um I think it's a a fruitful area of research. So um one of the things I think I thought one of the reasons why there are two reasons why it was important, I think to, to present this number one is to present the data. Uh I think there is something to be learned from our own experience related to acute myocardial infarction, complicated by cardiogenic shock. Many of the things that uh that we presented that I presented, I think you already probably knew or or maybe, but this I think is is confirming and, and it strengthens um your your maybe your assumptions and, and and your judgments. The other reason why I wanted to present it is to present. Um This is an example, we have built the capacity I think to do to, to take data from our registry, merge it with other data sources. Uh Pulling data from clarity and merging that data to create a data set where we can do sophisticated outcomes research here. And so, uh that's uh the main reason why I wanted to present that to show this and, and to encourage people who want to do this type of research, we're building the capacity in conjunction with enterprise analytics here at SA to do this type of research. And so that's another reason why I wanted to present this. So I, I thank you for your attention and I'll, I'll stop there and open it up for questions. Thank you. Published November 14, 2024 Created by Related Presenters John Brush, M.D., F.A.C.C. Sentara Cardiology Specialists -Interventional Cardiology View full profile
BROADCASTMED