Signal of Harm From Impella Has an Easy Answer

John Mandrola, MD


November 19, 2019

Two late-breaking studies presented at the American Heart Association (AHA) Scientific Sessions 2019 suggest a signal of serious harm from the mechanical cardiac support device Impella.

The studies, from two separate non–industry-supported research teams, used different observational data sources to come to the same conclusion: Impella (Abiomed) was associated with significantly worse outcomes, including death, bleeding, stroke, and acute kidney injury (AKI).

To be clear, these are observational studies that cannot prove that Impella caused these worse outcomes. Observational studies always come with the possibility of selection bias from confounding factors not seen on data spreadsheets.

That does not mean we should ignore this provocative evidence. Let's first consider the areas of agreement on Impella:

  • Its use is increasing. Both studies show growing use of Impella in the United States.

  • Its high price increases the cost of care. When I recently visited a National Health Service hospital in Scotland, physicians there told me that cost was the main reason they don't use Impella.

  • It is an invasive device, inserted through a large-bore arterial access in patients who will be on anti-thrombotic drugs.

  • Few data from randomized controlled trials (RCTs) support Impella for either of its two main uses: cardiogenic shock and protection during high-risk percutaneous coronary intervention (PCI).

Sanket S. Dhruva, MD, MHS, from University of California, San Francisco (UCSF), led off his presentation at AHA with the fact that only two RCTs—published in 2008 and 2017—have compared intra-aortic balloon pump (IABP) and Impella in patients with cardiogenic shock. These RCTs enrolled a total of 74 patients and found no difference in 30-day mortality rate but higher bleeding rates with Impella. Null results of these trials are especially problematic because a meta-analysis of IABP vs medical therapy found no benefit for IABP.

The PROTECT II randomized trial of Impella vs IABP for high-risk PCI was stopped for futility and did not show a statistical difference in the primary endpoint of major adverse cardiovascular events at 30 days.

The Two AHA Studies

Dhruva and a team of researchers at Yale and UCSF searched a linked cohort of the National Cardiovascular Data Registry  CathPCI Registry (>1500 US hospitals) and Chest Pain-MI Registry (>1000 US hospitals) to identify patients with cardiogenic shock due to acute myocardial infarction.

They used propensity matching (of 75 variables) to get two groups of just under 1700 patients who received Impella or IABP, and then compared outcomes. The rate of death in the Impella arm was 45% vs 34% for IABP. The rates of major bleeding were 31% vs 16%, respectively.

Amit Amin MD, MSc, from Washington University in St. Louis, presented the second study, which was simultaneously published in Circulation . This  research team used the Premier Healthcare Database to analyze about 48,000 patients undergoing PCI with mechanical support at 432 hospitals from 2004 through 2016.

Their approach was an ecological one, comparing costs, utilization, and outcomes (in-hospital mortality, bleeding, AKI, and stroke) in the "pre-Impella" era (2004 to 2007) vs the "Impella" era (2008 to 2016).

They saw sharp rises in Impella use from 2011 onward but also a greater than fivefold variation in device use across hospitals. This variation allowed the authors to compare outcomes in hospitals with high vs low Impella use. Hospital, death, bleeding, AKI, and stroke were all significantly higher in hospitals that used the device most.

The authors also found more adverse outcomes in the Impella era compared with the earlier time period: Death was 24% higher, bleeding 10% higher, and stroke 34% higher. All these reached statistical significance.


The fact that two different groups used different techniques and different databases and showed similar patterns of harm deserves attention.

Obviously, the main issue is whether Impella caused the harm or whether sicker patients got the device. One way to sort this out is to consider plausibility.

I asked senior author of the Yale/UCSF study, Nihar Desai, MD, from Yale University in New Haven, Connecticut, if there was a plausible explanation. Yes, he said, and pointed to the 15% higher absolute rate of bleeding in the Impella arm of their study. Major bleeding in patients with acute myocardial infarction is associated with a higher risk for death.

Dhruva also pointed me to the degree of risk seen. All observational studies can have selection bias, but he believes that a greater than 10% absolute increase in risk is more than one would see with confounding.

I think the Premier Healthcare Database study makes it harder to posit selection bias. The before-and-after nature of this analysis and the fact that hospitals with high Impella use had worse outcomes make it less likely that a systematic bias affects these results.


Putting this together, we have an approved device that is costly and invasive and is being increasingly used—in the absence of high-level evidence that it is better than a device that was no better than medical therapy.

These two observational studies do not prove Impella is harming people, but, taken together, they do show a strong signal of concern. Despite good intentions, and the fact that Impella produces more hemodynamic support than IABP, the device may be harming more people than it helps.

I am not saying it is, but medicine is replete with examples of therapies that seemed effective until they were tested in a proper trial. The time to study such therapies is before they become accepted practice. Imagine if years ago payers or regulators had said Impella could be used, but only as part of a trial.

If the device is as beneficial as proponents say, this trial would not require many patients, nor would it need long follow-up since the types of patients who get Impella have high event rates.

At a recent meeting in my hospital, a colleague who implants a lot of Impella devices said there will not be a trial because there is not equipoise. These two studies, plus the history of medicine, strongly oppose that notion.
Now—not later—is the time to do a proper trial of this device. In the meantime, it seems reasonable to consider a more conservative approach to managing these patients.

John Mandrola 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 @drjohnm. For more from | Medscape Cardiology, join us on Twitter, Facebook, and Instagram .


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.