Mar 4, 2022 This Week in Cardiology Podcast

John M. Mandrola, MD


March 04, 2022

Please note that the text below is not a full transcript and has not been copyedited. For more insight and commentary on these stories, subscribe to the This Week in Cardiology podcast on Apple Podcasts, Spotify, or your preferred podcast provider. This podcast is intended for healthcare professionals only.

In This Week’s Podcast

For the week ending March 4, 2022, John Mandrola, MD comments on the following news and features stories.

Update to CardioMEMS Discussion

Last week I discussed the GUIDE-HF trial and how it led to CardioMEM’s FDA approval. I strongly opposed it. Today I want to give a brief update on something that I did not make clear in that discussion. If you haven’t listened to last week, it would be useful to go back and listen to the first 10 minutes or so. I was unclear on the post-hoc, pre-COVID analysis, which the authors purported to show benefit for the CardioMEMs group, actually and inadvertently, but which strongly suggests performance bias not actual efficacy as the explanation for the finding. Let me explain. GUIDE-HF compared heart failure (HF) management with the pulmonary artery monitor called CardioMEMs, to standard care. The main analysis did not show a statistically significant reduction in the primary endpoint, but the pre-COVID analysis did meet statistical significance.

COVID-19 affected the follow-up of the trial; 28% of the follow-up occurred after an emergency declaration of the pandemic in March 2020. The other 72% of the follow-up occurred before the pandemic. To understand my argument here, I have to tell you the three components of the endpoint.

  • The first component of the endpoint was mortality—zero bias.

  • The second component was urgent HF visits, which the authors describe as unscheduled and unplanned. This endpoint seems relatively low in bias.

  • The third component was HF hospitalization (HHF), which was determined by a clinician who was aware of the treatment assignment.

    • This is highly susceptible to bias. This is not necessarily nefarious, it is just human clinicians acting like humans.

Ok. Now to Table 2 in the Lancet paper. When you look at the pre-pandemic events, there were no differences in mortality or urgent unplanned HF visits. The clinician determined HHF drove the benefit. HHFs were lower in the CardioMEMs arm. Was this a biased endpoint? Were clinicians more apt to hospitalize a control patient and less apt to hospitalize a device patient? Or was the CardioMEMs brilliant? To answer this question, you have to look at what happened during the pandemic. The authors don’t tell us event rates during pandemic follow-up, but you can simply subtract pre-pandemic events from total. If you do this, you find that there are now no differences in any endpoints. Specifically, the lower rate of HHF disappears.

Is it clicking yet? Consider that during the pandemic there was a high bar to admit patients to the hospital. You had to be sick. The pandemic essentially rendered HHF a far less biased endpoint. And, if HHF was not a biased endpoint, for example, like urgent HF visits, and CardioMEMs actually reduced HHF, you’d expect the same findings pre-and post-pandemic. But that is not what happened. Instead, during the pandemic the device failed to reduce HHF. Which is especially weird because you’d expect the wirelessly transmitted data, which reduces in-person interactions, to be more beneficial. What shocks me is that FDA reviewers did not see this.

This device is likely to take off. It has a real chance of becoming the therapeutic fashion. I have a column coming on these issues. I often say on this podcast that I try to remain skeptical not cynical. But stuff like this makes it harder.

ARBs and Cancer

This is real hot potato! There has been conflicting information about ARB use and cancer risk. Eleven years ago, a group from Turkey published a meta-analysis suggesting an increased risk of cancer with ARB use. Subsequently, FDA reviewed the data and found no increased risk. Yet senior FDA reviewer Thomas Marciniak, who is now retired, conducted his own individual patient-level meta-analysis and found an increased risk of lung cancer. Marciniak also noted problems with the original FDA review, including, for instance, not counting lung carcinoma as lung cancer. Numerous other meta-analyses have been published in the intervening decade and the results have been mixed. Recently, Dr. IIKe Sipahi has published another meta-analysis of ARB trials and cancer risk. PLOS-One published the paper. Notably, it is a single-author paper. The unique aspect of this most recent meta-analysis was that Dr. Sipahi looked specifically at cumulative exposure and cancer risk. If ARBs cause cancer, there should be a dose-response signal. Namely, more drug exposure (higher dose and longer duration) should increase the risk.

One way to look at this question is to use what’s called a meta-regression, where you plot the trials by cumulative exposure on the x-axis and log-risk ratio on the y-axis. The trials with the greater exposure had the higher rates of cancer.

  • This was a trial-level analysis including 15 trials taken from the ARB trialists’ collaboration.

  • There were about 75,000 patients and more than 170,000 person-years of exposure to high-dose ARB.

  • There was a highly significant correlation between the degree of cumulative exposure to ARBs and risk of all cancers combined and also lung cancer.

In trials where the cumulative exposure was greater than 3 years of exposure to daily high dose ARBs, there was a statistically significant increase in risk of all cancers combined (P = 0.006).There was a statistically significant increase in risk of lung cancers in trials where the cumulative exposure was greater than 2.5 years (P = 0.03).In trials with lower cumulative exposure to ARBs, there was no increased risk of all cancers combined or lung cancer. Cumulative exposure-risk relationship with ARBs was independent of background angiotensin-converting enzyme (ACE) inhibitor treatment or the type of control (i.e. placebo or non-placebo control).

These were not exactly small absolute risks: 120 patients needed to be treated with the maximal daily dose of an ARB for 4.7 years for one excess cancer diagnosis. And In 2011, it was calculated that about 200 million individuals are treated with an ARB globally.

Comments. This is a super-tough call. Sue Hughes has amazing news coverage. She has a detailed account of the findings and a great discussion with experts on both sides. I’ve exchanged emails with Drs. Franz Messerli and Thomas Marciniak.

On the one hand, the most recent meta-analysis and meta-regression is worrisome. It comports with Marciniak’s meta-analysis in 2013. The dose response curve is compelling, because if you posit this is a spurious finding, say post-hoc cherry-picking, then how do you explain the dose response curve. Another question is plausibility. Is it plausible that ARBs cause cancer or lung cancer, specifically?

In his discussion, Sipahi spends many words speculating about the recent recalls of ARBs. as these were related to finding low levels of possible carcinogens in the compounds. I am not as swayed by plausibility because if cancer risk is truly increased, the cause does not matter. I also find compelling the fact that Marciniak’s and Sipahi’s findings line up well. I am no statistician, but both look to be well-conducted analyses with pre-specified methods.

On the other hand, we also have to be fair about the limitations of this analysis. It used trial-level data, so it’s hard to know things like smoking status. (Though because these were randomized trials, you’d expect randomization to mostly balance co-variates.)

The most compelling argument against this safety concern is the one made by Franz Messerli and Andrew Althouse, namely the matter of competing risks of death: since ARBs, especially when used at high doses — presumably what is needed to treat people with substantial heart disease or hypertension, reduce cardiovascular (CV) events, and prolong life — you will see more cancer.

Competing risks is such an important concept. That is, we all have an expiration date. We are mortal. So, if you treat older people and reduce their death rate from heart disease, these patients can live long enough to get cancer. I see this in my implantable cardioverter-defibrillator (ICD) clinic. ICDs and modern medical therapy dramatically reduce death from heart disease. Most of my patients with HF and ICDs live long enough to get other diseases—like cancer. They would not be at risk for cancer if they have ventricular fibrillation or a fatal stroke at an earlier age. Althouse – who is fantastic to follow on Twitter — points out that to sort out competing risks of death you need individual patient-level data. which Sipahi did not have.

“When there are some patients dying during the study, the only way to tell whether the intervention actually increased the risk of other health-related complications is to have an analysis that properly accounts for each patient's time-at-risk of the outcome.”

Finally, there is a resolution to this question: Sipahi wants the US FDA to re-study this issue. He points out that they have patient-level data. Why not open the data up and let neutral scientists analyze it and tell us their findings?

Troponins After Cardiac Surgery

The diagnosis of myocardial infarction (MI) entails three things: chest pain, electrocardiographic (ECG) changes, and enzyme rises. Cardiac surgery messes with each of these three components. After heart surgery, all patients have chest pain and enzyme elevations, and nearly all have ECG changes. And as troponins get more sensitive, the question becomes, what threshold of troponin (or myocardial injury) is high enough to predict poor outcomes such as death at 30 days.

The New England Journal of Medicine (NEJM) has published an interesting observational study from the group at McMasters looking at post-cardiac-surgery high-sensitive (Hs) troponin values and outcomes. This was an international prospective cohort study including many different types of heart surgery.

Hs-troponin-I levels were measured 3 to 12 hours post op, then at days 1, 2, and 3. They then did regressions looking at the relationship between peak troponin and death at 30 days. They included nearly 14,000 patients of whom 296 or 2.1% died within a month.

The key findings:

  • Among patients who had coronary artery bypass graft (CABG) or aortic valve replacement (AVR), the day-1 threshold troponin level that was associated with an adjusted hazard ratio of more than 1.00 for death within 30 days was 5670 ng/liter, a level 218 times the upper reference limit.

  • Among patients who underwent other cardiac surgery, the corresponding threshold troponin level was 12,981 ng/liter, a level 499 times the upper reference limit.

To place this in context, the troponin thresholds currently recommended in consensus statements (> 10, ≥ 35, and ≥ 70 times the upper reference limit) were exceeded in 97.5%, 89.4%, and 74.7% of patients, respectively, within the first day after surgery.

Comments. I don’t see this as surprising. Ablation causes troponin release. Cardioversion causes troponin release. Obviously, cardiac surgery will cause troponin release. But every time one says “troponin release,” one should also think myocardial injury.

The issue this study gets to is that as troponin assays become more sensitive, we will have to adapt upper limits of normal before making intervention decisions—say, post-op angiography or stress testing or medical therapies. This paper does not tell us where to put these dichotomous limits, but it does tell us the current thresholds are way too low.

I may be old fashioned, but I still see tons of value in correlating troponins with other findings. Yes, the ECG is hard to use post-op; there is usually a degree of pericarditis, but we can still look for focal findings. Sometimes a careful echocardiogram can help sort out a new wall motion abnormality. We should not be nihilistic about using clinical criteria for diagnosing true post-op MI.

The other thing to say about elevations in troponin levels is that there are tons of other causal factors. I can imagine a patient with chronic kidney disease (CKD), chronic obstructive pulmonary disease (COPD), or severe left ventricular dysfunction who is operated on acutely will have high troponins. But is myocardial injury the only causal factor in this person’s 30-day death? The authors adjust for clinical scores, but is that enough to sort through the effects of every clinical factor? I don’t know; I am asking. Another concern I have is the 30-day mortality. Yes, this makes the study easier to do, but what about 1- or 2-year mortality. Might there be a lower threshold for post-op troponin levels that predicts higher hazards at 1 to 2 years? Finally, this has implications for the interpretation of the EXCEL Trial that compared percutaneous coronary intervention (PCI) vs CABG in patients with left main disease. EXCEL used a biomarker-heavy definition of MI. There were higher rates of MI after CABG but later there were more Mis in the PCI arm. The criticism is that if one did not include these post-CABG MIs (due to troponin rises) after CABG, the rate of MIs would have been higher in the PCI arm, and since this was a component of the composite primary outcome, CABG would have looked even better than it did.

FDA Approves Empagliflozin

The US FDA has approved an expanded HF indication for the SGLT2 inhibitor empagliflozin that now includes HF with mid-range or preserved left-ventricular ejection fraction (HFpEF)). This is an expanded indication; empagliflozin was approved for use in patients with HF with reduced EF (HFrEF) last August. Now we have two expensive drugs approved for HFpEF (empagliflozin and sacubitril/valsartan) that have been approved with dubious trials.

First empagliflozin: EMPEROR-Preserved; 6000 patients with HFpEF, empagliflozin vs placebo.

  • The primary endpoint (PEP) was reduced by 21% but the PEP of CV death and HHF was driven only by HHF.

  • No significant difference in CV death, all-cause death, composite renal outcome.

  • HHF accounted for a tiny fraction (well less than 20%) of total hospitalizations. In fact, total hospitalizations did not differ.

I use SGLT2 inhibitors and believe they are beneficial drugs for certain indications, but there use in patients with HF not due to a reduced EF is incredibly low value care.

Now sacubitril/valsartan in HFpeEF: PARAGON-HF; 4800 patients with HFpEF, sacubitril/valsartan vs valsartan alone.

  • The PEP was reduced by only 13% and this did not reach significance, though the P-value was 0.06.

  • The PEP of CV death and HHF was driven by HHF not CV death.

  • No significant reductions in CV death, all-cause death.

  • Yet the FDA approves the drug based on subgroup analyses which suggested women and those with a low-normal EF benefited more from the drug.

PARAGON authors don’t tell us total hospitalizations so we can’t assess how important a reduction in HHF hospitalizations really are, but I suspect it’s the same story as EMPEROR-Preserved.

Low-Value Care

I wonder if you noticed a theme here: CardioMEMS, empagliflozin, sacubitril/valsartan in HFpEF. Yes, each of these are examples of new innovations that add little to progress.

As we wrote in the medical conservative essay, to determine genuine progress, medical conservatives endorse evidence-based medicine and critical appraisal, because many developments promoted as medical advances offer, at best, marginal benefits.

That brings me to the final topic: The American Heart Association (AHA) has added another scientific statement about strategies to reduce low-value CV care. First author Vinay Kini. They write that there is a critical need to reduce this kind of care. They target examples like serial troponin testing in low-risk patients, or routine stress testing after revascularization, coronary artery calcium (CAC) scores in patients with known coronary artery disease, or dual chamber ICDs in patients without a pacing indication.

One of the ways they suggest to reduce low-value care is to promote shared decisions with patients and nudges for doctors.

I have three comments about this paper:

  • First, I think it’s great that our leadership and professional societies recognize low-value care. Even if no one reads the details, it’s great for clinicians to see leaders talking about value.

  • Second, yes, we all agree that the examples—the choosing wisely type things—are indeed super low-value care. I would argue, however, that these are not the main drivers of waste in our system. The main drivers of waste are the everyday, middle of the curve things, that have become therapeutic fashion.

    • In my field, AF ablation done before risk factor management is a norm and it’s a huge waste.

    • Left atrial appendage closure has become a norm, and it too is low in value.

    • High-cost drugs like empaglifozin and sacubitril/valsartan in HFpEF will soon be a norm and it is waste. The failure to critically appraise normal stuff leads to far more waste than outliers like serial troponins.

  • Third, I strongly push back when people suggest shared decision making (SDM) be used as a means to reduce low-value care. That is not what SDM is for. While it may reduce low-value care, SDM is simply the right thing to do. It’s moral and ethical on its own grounds. It should not be used to nudge patients. It should be used to inform patients and increase their participation in medical decision making.

Reducing low-value care is on us, the clinicians. And the first step is not to be bamboozled by dubious evidence.


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.