Impella vs IABP: AHA Analyses Signal Urgent Need for Randomized Control Trial

John M. Mandrola, MD; Sanket S. Dhruva, MD, MHS; Nihar R. Desai, MD, MPH


December 12, 2019

This discussion was recorded November 17, 2019, at the American Heart Association Scientific Sessions; the transcript has been edited for clarity.

John M. Mandrola, MD: Hi, everyone. This is John Mandrola from | Medscape Cardiology. I'm here at the American Heart Association meeting in Philadelphia, and I'm very pleased to have Sanket Dhruva from the University of California, San Francisco, and Nihar Desai from Yale University. We're going to talk about a study they presented today as a late-breaking clinical trial on the Impella device. Tell us what your study found.

Sanket S. Dhruva, MD, MHS: Thanks very much for having us. We studied outcomes in patients who had an acute myocardial infarction (AMI) with cardiogenic shock (CS) who received Impella or an intra-aortic balloon pump (IABP). We found that patients who were treated with Impella had an almost 11% higher risk of dying in the hospital compared with those treated with IABP. We found also that patients treated with Impella had about a 15% higher risk of bleeding. Now, I just want to emphasize that these are absolute risk differences. So the absolute risk of death and bleeding was significantly higher among patients treated with Impella compared with IABP.

Mandrola: That's provocative. Nihar, can you tell us why you studied this and how you went about it?

Nihar R. Desai, MD, MPH: We were interested in asking this question because it was clinically relevant. About 10% of patients that have an AMI experience shock. We know that there is substantial morbidity and mortality for these patients, and yet there's a relative absence of high-quality evidence. So we brought a team together in collaboration with the US Food and Drug Administration (FDA) and the American College of Cardiology, and the National Cardiovascular Data Registry (NCDR) platform, to use that really rich data to answer what we thought was a clinically important question.

Mandrola: I was struck by the absence of data. How could there be an absence of data for an approved device like this?

Dhruva: It's an interesting question. Impella came to the market in 2008—a little over a decade ago—through what's called a 510(k) clearance pathway. That is, the device was shown to be substantially equivalent to devices that had previously been on the market. Since then, different iterations of the device have been approved through the FDA premarket approval pathway, which requires clinical data. But when we looked at the evidence for use of Impella in patients with AMI and CS, we found only two randomized clinical trials and they enrolled a total of 74 patients. Now, this device is being used in thousands of patients in this clinical scenario, so we thought there was a relative absence of robust evidence.

Mandrola: Just quickly, what did those randomized controlled trials conclude, those 74 patients?

Dhruva: Overall, they found no significant difference in mortality among patients who were treated with Impella and IABP. But they did find higher rates of bleeding among patients who were treated with Impella.

Mandrola: Tell us how you went about this.

Desai: What we tried to do was bring the rich clinical data that's captured in the registries in an observational way, using the most robust methods we could. We did a propensity-matched analysis of nearly 1700 matched pairs, so almost 3400 patients in total. We compared those patients that received an Impella device with those that received IABP.

Mandrola: It's not randomized. How do these robust methods try to get close to randomization?

Desai: This was not a randomized trial, so there's always a concern about confounding or differences in clinical case mix across these groups. The best tool that we have available to us in that setting is to try to match patients as well as we can. We used 75 clinical variables, including sociodemographics, prior clinical history, and laboratory data, and because we were leveraging the NCDR data, we had coronary anatomy, infarct location, and other anatomical features that we thought were important to match on.

Mandrola: Sanket, you made the comment about absolute risk differences in the topline results. When we look at cardiovascular studies, there were absolute risk differences of 10%-15%. These were huge.

Dhruva: We think that these are important findings because these are absolute risk differences. And because of that, we think that there's really a need for additional data to help us understand how these devices can and should be used in AMI with CS. We performed as robust an analysis as we could, and as Nihar mentioned, there could be limitations to this observational analysis, but we used as robust propensity score matching as we could. We also performed instrumental variable analysis. And we think that the results are as strong as they could be in the absence of a high-quality randomized trial that's adequately powered to endpoints.

Mandrola: How likely do you think it is that Impella patients did worse because they were just sicker patients? For instance, the doctor sensed something that wasn't on a spreadsheet that you can’t measure and chose to use Impella in patients that were inherently sicker than those with an IABP?

Dhruva: That's certainly an important question, which we tried to address as best as possible. We did have some vital sign data, and we had coronary anatomy data, so lots of rich clinical details. But we weren't there at that moment when that clinical decision was made to put in the Impella or the IABP. We were missing some labs that sometimes can be helpful, such as a lactic acid (lactate) level. But we think that across the lab values that we did have, it was quite robust. And as I mentioned, we also conducted an instrumental variable analysis where the results were consistent.

The findings of a 9%-10% absolute increase in mortality in patients receiving Impella, even if there is some potential unmeasured confounding, is quite a dramatic increase. Perhaps the most important endpoint for our patients is, can we keep them alive?

Mandrola: One of the things that we do in critical appraisal and trying to bring this evidence to the bedside is look at other studies. There was another presentation on Impella during the same session that involved an observational dataset. I know it's not your study, but it showed similar signals.

Desai: In the same session following Dr Dhruva’s presentation was a presentation by Amit Amin, MD, from Washington University School of Medicine in St. Louis, using the large PREMIER observational dataset. Over 45,000 patients were included in that analysis, again comparing Impella versus balloon pump. Essentially, the findings were in concordance with our analysis—two independent study teams using different datasets, coming to relatively consistent results about Impella versus balloon pump.

Mandrola: You guys are pretty expert in the regulatory framework, and I'm just a clinician. Could you speak to how much observational or randomized controlled data FDA needs before a device gets on the market or taken off the market?

Dhruva: That's a good question, John. I'm not able to speak to the regulatory decision-making. The FDA, as the agency in charge of regulating medical devices in our country, has tons of expertise and is able to make those decisions. What I can say is that there have been studies of Impella and these met the bar for FDA approval at this point in time. But overall there has been limited randomized data, and in some cases these have been single-arm studies. There's been a relative absence of high-quality comparative data that's focused on the clinical outcomes that matter most to us for our patients, particularly being able to keep them alive. These patients are super-sick. We were trying to fill that gap.

Mandrola: Nihar, one of the things that could help give us confidence in this finding is if there was plausibility. Are there any plausible reasons why Impella would do worse than IABP?

Desai: I think there are at least two important mechanisms for us to think about. One is the substantial increase in bleeding. There's a body of evidence to suggest that those patients that have a major bleeding complication are much more likely to have adverse events. Whether that's because of a withdrawal of antithrombotic therapy or transfusions or a combination of those things, they have higher event rates. In addition to that, again, I think it's important for us to remember that this analysis was looking at the device performance in the real world in unselected settings. How the device performs in the real world might be different from how it would perform within the confines of a clinical trial.

Mandrola: This seems to me a strong enough signal to say that we should do a randomized controlled trial, but I know there are people in my laboratory and elsewhere who feel that there's not equipoise here. I think this shows that there's at least equipoise. What do you think?

Desai: I agree. I think it's for the clinical community to really come together and say we need more high-quality evidence in this area. These patients are being seen at facilities literally every day with very little guidance or evidence on what the optimal management strategies are. In addition to the randomized evidence that I think we do need, I think it's important that we also have real-world evidence, because the way a device like this performs is likely to be quite different from how it performs in carefully selected centers with carefully selected investigators.

Mandrola: I want to thank you guys for joining me today and for doing such important work.

Desai and Dhruva: Thanks for having us, John.

John Mandrola, MD, practices cardiac electrophysiology in Louisville, Kentucky, and is a writer and podcaster for Medscape. He espouses a conservative approach to medical practice. He participates in clinical research and writes often about the state of medical evidence.

Follow John Mandrola on Twitter

Follow | Medscape Cardiology on Twitter

Follow Medscape on Facebook, Twitter, Instagram, and YouTube


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.