Medicare Data on AMICS Too Muddied for Firm Conclusions on Impella

For patients who have acute MI with cardiogenic shock (AMICS), Medicare claims data aren’t detailed enough to show whether use of the Impella percutaneous microaxial left ventricular assist device (LVAD) actually leads to higher 30-day mortality compared other options like medical therapy and intra-aortic balloon pump (IABP) support, according to a new study.

That concern stems from a handful of observational analyses looking at the Impella device (Abiomed) in the AMICS setting—many attempted to overcome potential confounding by propensity matching, but ultimately spurred debate over whether the increased mortality seen among Impella-treated patients was merely an association or a sign of a true problem. As such the current study, published in JAMA Cardiology, is the latest effort to gain this much-needed knowledge while the field awaits results from randomized controlled trials, a complex undertaking in AMICS.

What researchers found is that differences in patient and institutional characteristics, coupled with changes over time and lack of detail on patients’ illness severity, together make it impossible to draw firm conclusions.

According to Robert W. Yeh, MD, MSc (Beth Israel Deaconess Medical Center, Boston, MA), senior author of the new paper, the study was initiated by the US Food and Drug Administration. The investigators set out to ask whether it’s possible, even when using the best statistical approaches, to overcome the limitations of observational data.

Going in, Yeh said, it was clear that subtle factors are driving clinicians to choose a certain strategy when treating the AMICS patient in front of them. “[It’s] a setup for confounding, because we tend to choose the more powerful devices like an Impella or [extracorporeal membrane oxygenation] in patients who are doing worse. . . . We put in a more-powerful device because I think that we're worried that we can't get away with certainly nothing and a balloon pump may not be enough support,” he explained.

Although in some ways the study results could be considered “negative,” in that they don’t provide clear answers on how Impella stacks up against the other treatment options, “we wanted in part to publish this paper as an example of how observational research ought to be conducted, written up, and discussed. . . . It was not a fishing expedition,” Yeh said.

Manesh R. Patel, MD (Duke Clinical Research Institute, Durham, NC), who co-wrote an editorial for the paper with David J. Cohen, MD, MSc (St. Francis Hospital, Roslyn, NY), praised the investigators not only for doing multiple high-level statistical methods but also for being transparent.

Instead of simply saying there’s an association, they went further by asking if there was some way to discern a causal relationship. This “is a really critical question as people use more and more observational data,” he pointed out to TCTMD. For very sick AMICS patients being treated for different indications across different centers, the data are “unfortunately [too] horribly confounded for us to be able to get the answer,” said Patel.

We wanted in part to publish this paper as an example of how observational research ought to be conducted, written up, and discussed. . . . It was not a fishing expedition. Robert W. Yeh

It’s somewhat reassuring that the new analyses were unable to show neither harm nor efficacy, he noted, but for clinicians, “having used the device and knowing that the device can work in some patients, the question [still] is, can the science identify the right patients at the right time?”

To generate that evidence, as the editorial notes, will require applying new strategies to help research reach the randomized-trial stage more rapidly.

“Unfortunately or fortunately, the device is being used as it is, and I don't think [the current paper will] change how it's being used,” said Patel. “What I think we all wish is that we would have had a better evidence base before the broad use [of Impella] happened, rightly so because patients are sick and clinicians are faced with managing these sick patients. . . . I fully understand why clinicians are making the decisions they are, and those decisions cannot be captured in these observational data, as we've seen.”

Medicare Claims Data

For the study, led by Zaid I. Almarzooq, MBBCh, MPH (Beth Israel Deaconess Medical Center), the researchers analyzed data from Medicare fee-for-service claims for 23,478 patients (mean age 73.9 years; 60.8% male) admitted with AMICS undergoing PCI between October 2015 and December 2019. Among them, 17.3% received an Impella percutaneous microaxial LVAD on the day of PCI, while 29.7% received an IABP and 53.0% got no mechanical circulatory support. (Due to various reasons, another 33,505 AMICS patients hospitalized during the same time period were excluded.)

The Impella-supported patients tended to be younger and were more likely to be male, present with NSTEMI, and have histories of AMI and heart failure. Notably, the Medicare data did not include information on the severity of cardiogenic shock, vasopressor dosing, or hemodynamic parameters.

Jointly with the FDA, Almarzooq and colleagues planned various statistical approaches in an attempt to see whether any could actually pin down how Impella performed in comparison to alternative treatments.

With the inverse probability of treatment weighting analysis, patients who received Impella had higher 30-day mortality (risk difference 14.9%; 95% CI 12.9%-17.0%) compared with those who had alternative treatments. The same pattern was seen with a grace-period approach looking at Impella use within 48 hours of PCI (risk difference 18.4%; 95% CI 12.1%-24.7%). But the Impella-treated patients also appeared to have more-severe illness—for instance, they were more apt to be intubated, receive vasopressors, and have right-sided heart catheterization—thus raising the possibility of unknown confounders.

With the instrumental variable analysis, 30-day mortality was also higher with Impella (risk difference 13.5%; 95% CI 3.9%-23.2%), but the confidence interval was wider than with the other approaches and patient and hospital characteristics differed in a way that again suggested possible confounding.

With the instrumented difference-in-differences analysis, the association between the percutaneous microaxial LVAD and mortality was imprecise, producing “very wide CIs compatible with both benefit or harm,” the researchers say.

They conclude: “Our findings suggest that commonly used observational data sets cannot support a causal interpretation of the estimates produced by different analyses used for the evaluation of percutaneous mechanical support devices in cardiogenic shock.”

What Can Be Done Differently

Ajay Kirtane, MD, SM (NewYork-Presbyterian/Columbia University Medical Center, New York, NY), associate editor for JAMA Cardiology, said this study “stresses the need for randomized data for percutaneous MCS. . . . At this juncture, it’s hard to get around the fact that well-done observational analyses certainly don’t show benefit (in fact they hint the other way) and there are no adequately sized randomized trials showing benefit either.”

There’s no high-quality, compelling evidence of benefit, yet physicians are still being taught in AMICS that support should be started upfront and early, said Kirtane. “I will freely admit that I myself gave lectures stating that position several years ago, and even said I would have difficulty randomizing due to lack of equipoise. But I have changed my mind. As confounded as many of these analyses are, if there was an incredible benefit, we should be able to discern it somehow,” he stressed to TCTMD in an email.

“I daresay it is next to impossible to find a physician who has used percutaneous MCS who hasn’t seen complications occur. I also know that when higher-risk devices are placed in centers with less experience, the risk/benefit ratio changes,” Kirtane added. “The reason I went into cardiology was because of the rich tradition of producing high-quality data through the conduct of RCTs. The absence of these in this field right now is simply striking.”

In their editorial, Cohen and Patel describe the new study’s findings as “sobering.”

“it is important to consider how we arrived at this situation, where we are attempting to use administrative claims data and complex statistical methods to determine efficacy of a device with broad clinical adoption,” they say.

The FDA’s premarket approval of Impella was based on a combination of hemodynamic factors, surrogate endpoints, and a single randomized trial in high-risk PCI that excluded patients with recent MI. Medicare coverage and favorable reimbursement soon followed, “allowing for rapid expansion of off-label use by clinicians in an effort to support patients who were often in extremis,” they write. “These factors, coupled with pathophysiologic reasoning and the substantial challenges of performing clinical trials in patients with cardiogenic shock, led to loss of equipoise on the part of clinicians and led the cardiology community to the current dilemma.”

To avoid this situation in the future, there needs to be disease-based standardization of terminology and data collection as well as the infrastructure for long-term follow-up of real-world device use, Cohen and Patel suggest. Payers also should be able to incentivize manufacturers, through conditional coverage, to perform clinical trials that generate evidence even after device approval. 

In terms of Impella, only time will tell what comes from the ongoing DanGer trial and another planned trial, RECOVER 4, which is projected to be complete in late 2027.