COMMENTARY

CardioMEMS Remote HF Monitoring: Resist the Spin and FOMO

John M. Mandrola, MD

Disclosures

March 18, 2022

What happened with the CardioMEMS HF System (Abbott) is not normal. This is a story about dubious trial analysis, spin, lax regulation, and the growth of low-value care.

The paperclip-sized implantable device records pressures from the pulmonary artery of patients with heart failure and then wirelessly transmits these numbers to clinicians. It's nifty, and it makes sense that such monitoring ought to help patients.

Except we have been here before with the Swan-Ganz pulmonary artery catheter, which also provided clinicians with hemodynamic data on hospitalized patients with heart failure but did not improve mortality or hospitalizations in the ESCAPE randomized controlled trial.

But the CardioMEMS pulmonary artery monitor is new and wireless.

The US Food and Drug Administration (FDA) approved this device in 2014 for a highly select group of patients. This was based on the CHAMPION trial, which randomly assigned patients with New York Heart Association (NYHA) class III heart failure and a recent hospitalization for heart failure to management with the device or standard care. Patients in the device arm had significantly fewer hospitalizations for heart failure.

Yet device uptake has been low, perhaps because of the selective CHAMPION inclusion criteria.

Recently, the FDA approved an expanded indication for the device based on the results of the GUIDE-HF trial, which enrolled a much broader group of patients with heart failure. The expanded indication means an additional 1.2 million U.S. patients will be eligible for CardioMEMS.

Let me first present the trial and then its contentious analysis.

The GUIDE-HF Study

In GUIDE-HF, about 1000 patients with NYHA class II to IV symptoms and either a recent heart failure admission or elevated natriuretic peptides were randomly assigned to management guided by the device or standard care.

Enrolled patients were about 70 years old, and their mean left ventricular ejection fraction was around 40%. The primary endpoint had three components: death, urgent heart failure visits, and heart failure hospitalizations.

There were 253 primary outcome events in the treatment arm (0.56 events per patient-year) and 289 primary outcome events in the control arm (0.64 events per patient-year). The hazard ratio was 0.88 (95% CI, 0.75 - 1.05). That 12% relative risk reduction did not reach statistical significance (P = .16).

This should have led to the conclusion that device-based management did not significantly improve important outcomes—much like the results of the ESCAPE Swan-Ganz catheter trial two decades before.

But that is not what happened. Instead, the investigators performed a "pre–COVID-19 impact analysis," which yielded a 19% lower rate of composite primary endpoints that barely met statistical significance (P = .049).

Although all patients were enrolled before the pandemic, about a third of the follow-up of GUIDE-HF occurred during the pandemic. The findings from the pre–COVID-19 post hoc analysis led the makers of the device to declare that the results of GUIDE-HF support the expanded indication.

Dubious COVID-19 Impact Analysis

The first problem with the pre–COVID-19 analysis is the fact that you can't just make up new ways to analyze the data after the trial.

You have to prespecify both the endpoints and analytic method. The investigators could not have predicted the pandemic, so it's not surprising that the pre–COVID-19 analysis was not prespecified in any draft of the study protocol.

The authors write that they worked with the FDA in doing this analysis and that the data were still blinded. But neither of these points make it a scientifically sound analysis.

The interesting thing, though, is that when you dig into this analysis, it actually shows that the device does not likely improve outcomes. When you analyze the components of the primary endpoint in the pre-COVID analysis, you find strong evidence that performance bias, not device efficacy, drove the benefit.

Performance Bias

Performance bias is not necessarily nefarious; it occurs when there are systematic differences between groups in the care that is provided. Unblinded trials are especially susceptible to this bias.

For instance, I was part of the His-Synch trial comparing His-bundle pacing to biventricular pacing. I am also a proponent of His-bundle pacing. When I operated on patients randomly assigned to His-bundle pacing, I tried hard to make the technically demanding His-bundle lead work.

In Table 2 of the Lancet GUIDE-HF paper, the authors report the total number of events for each component of the endpoint for both the primary and pre–COVID-19 analysis. You can then subtract the numbers in the pre-COVID period from the total to get the number of events that occurred during the pandemic.

Before the pandemic, the driver of the significant reduction in the CardioMEMS arm was fewer heart failure hospitalizations (185 vs 225). There were no meaningful differences in urgent heart failure visits or deaths.

Now consider the potential biases in these endpoints. Mortality has no bias. Urgent heart failure visits were unscheduled and unplanned, which seem less susceptible to bias. But the hospitalization of a patient with heart failure requires a decision by a treating physician—who was aware of the treatment arms.

During the pandemic phase, there remained no differences in mortality or urgent heart failure visits. But the difference in heart failure hospitalizations disappeared. It was now higher in the CardioMEMS arm (61 vs 49).

Next consider that the pandemic increased the threshold to hospitalize patients: that is, it reduced the bias in this endpoint.

If the decision to admit a patient for heart failure were not susceptible to bias, there should be no difference during the pandemic. The fact that the only driver of the significant results in the pre–COVID-19 analysis was lower rates of hospitalizations for heart failure strongly suggests performance bias, not device efficacy, as the cause.

Strengthening this argument is the wise comment from heart failure specialist Larry Allen, MD, who said, "One might hypothesize, that in the setting of limited in-person visits with loss of physical examination, perhaps CardioMEMS would be more—not less—helpful during the pandemic. And yet the opposite was seen."

How Important Are Heart Failure Hospitalizations?

On the website clinicaltrials.gov, GUIDE-HF authors list total hospitalizations as an outcome they will measure. But nowhere in the paper or supplement do the authors report this outcome. Maybe it is forthcoming.

The point is that a consumer of medical evidence cannot know whether the reductions of heart failure hospitalizations in GUIDE-HF reduced total hospitalizations. What if heart failure hospitalizations represent only a small fraction of total hospitalizations—as was the case in the EMPEROR-Preserved trial of empagliflozin in patients with heart failure and a preserved ejection fraction?

Implications

I have colleagues who argue that the FDA expanded indication means nothing; clinicians can look at the dubious analysis of GUIDE-HF and decide not to use this device. This is technically true, but it is not what happens in the competitive U.S. healthcare system.

One day after the FDA expanded the indication, a colleague told me that an Abbott company representative had sponsored a lunch for clinicians. He said one of the main messages to the attendees was that two competing hospitals were already using the CardioMEMS device.

This is how medical marketing works. Industry understands our fear of missing out (FOMO). If big-name hospitals and doctors are using (and promoting) a therapy, other hospitals and clinicians will feel pressure to use it too.

This is what happened with left atrial appendage closure devices and ventricular assist devices, such as Impella (Abiomed).

Once a device (or drug) becomes the therapeutic fashion, unconvincing regulatory trials are gradually forgotten. Those who resist are labeled Luddites and contrarians and lose status in the hierarchy of influence within the hospital system and professional societies.

Opportunity Costs and Critical Appraisal

Proponents of CardioMEMS may argue that the device should be widely available because there are select patients whom it helps. That might be true, but when the cost is high and the device is invasive, the onus of proof is on the proponents to show us exactly which group of patients benefit.

Another take-home involves opportunity costs. A CardioMEMS device will cost hospitals approximately $20,000. If expanded to the 1 million extra patients, the cost to the healthcare system would extend into the tens of billions. Imagine the good that these billions could do if directed at improving basic care for uninsured or underinsured people.

The final point centers on our core task of helping patients. To know which interventions (really) work, we have to do our own critical appraisal.

The CardioMEMS story teaches us that we cannot rely on the neutrality of journal editors, researchers, regulatory bodies, or even colleagues. Critical appraisal may add to the work of being a clinician, but it has never been more vital.

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

Follow theheart.org | Medscape Cardiology on Twitter

Follow Medscape on Facebook, Twitter, Instagram, and YouTube

Comments

3090D553-9492-4563-8681-AD288FA52ACE
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.

processing....

Recommendations