Jul 12 2024 This Week in Cardiology

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, download the Medscape app or subscribe 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 July 12, 2024, John Mandrola, MD, comments on the following news and features stories.

Vascular Closure Devices

Venous vascular closure system vs. figure-of-eight suture following atrial fibrillation ablation: the STYLE-AF Study

I reported on an observational study of venous closure devices after AF ablation. A listener sent me notice that there was an actual RCT out of the Univ of Lubeck.

It’s called STYLE AF, published in Europace with first author, Roland Tilz, MD. They compared venous closure systems to figure-of-eight suture and manual compression. 125 patients with AF were randomized. They found that the device arm had shorter time to ambulation (109 min vs 269 min), hemostasis, and discharge. They also found a trend towards fewer minor vascular access complications.

I like trials such as this. Yes, it’s small. Yes, the primary outcome, time to ambulation, is a surrogate, it’s unblinded and funded by industry. But at least treatment is randomized. We need more of this.

While the authors rightly conclude that closure systems improved surrogate outcomes, I would note that less than 10% of patients received protamine. Not reversing protamine would surely favor a closure device. We reverse heparin in all cases. There was also no difference in same day discharge which was less than 15%. That’s also different from many labs in the US which send patients home on the same-day.

The authors don’t emphasize the massive difference in costs. So I stand by my original take that spending many hundreds of extra dollars for a few minutes shorter time to ambulation hardly seems worth it. The figure of eight stich with a stopcock instead of a knot remains one of the most elegant techniques in all of cardiology.

GLP-1s Unknowns

Since the placebo-controlled SELECT trial found that semaglutide reduced CV outcomes in patients with established ASCVD and obesity, GLP-1s can now be called disease-modifying agents in patients with heart disease. That is big.

Also big is the vast number of patients who will be eligible for these drugs. One other observation: the drugs seem to require long-term usage. Because patients who stop the drug gain weight back. Plus, why would you stop a disease-modifying drug?

One important issue, not just for individual patients, but also for public health, is the possibility of long-term adverse effects. RCTs are done for a handful of years; but if millions, perhaps billions, of patients take these drugs for decades, even a low-incidence adverse effect could be important.

Professor Rod Hayward wrote in Circ Outcomes that even statin drugs have unknown unknowns when it comes to decades of use. The same hold true for GLP-1s.

If we are to detect low-incidence AE, we will need observational data. Two studies published recently purport to detect signals from GLP-1s. Both studies fail miserably at their goal. And why and how they fail are instructive for those interested in critical appraisal.

JAMA-Ophthalmology published the first study—a matched cohort study of two groups of patients who presented to a neuro-ophthalmology clinic at Harvard Medical school. The authors concluded that there was a possible association between semaglutide use and rare but serious cause of blindness called non-arteritic anterior ischemic optic neuropathy. Before this week, I had never heard of NAION. In fact, it is pathetic how little understanding I have of the human eye. Even though I see eye doctors often—for my far-sightedness.

NAION, I have learned from Google and Youtube, is the second most common form of optic neuropathy and significant cause of blindness in adults.

Nearly 250 news outlets covered the paper, with headlines such as Weight Loss drug linked to rare cause of blindness. The Altmetric score as of yesterday was 2300.

Two doctors, each of whom I respect the greatly, sent me the paper—but for opposite reasons. One smart doc sent me the paper because it was methodologically so poor. The other doc sent me the paper because the findings were interesting.

Here is what the mostly eye-surgeon authors did: They write that the study included more than 16,000 patients referred to their specialty clinic. But that is misleading because the actual study included 710 patients with diabetes and 980 patients with obesity.

Of the 710 with T2D, nearly a third had semaglutide exposure and the rest were treated with non-GLP-1s. Similar for the 980 with obesity: about a third had semaglutide exposure and two-thirds did not.

They then did matching of the two uneven groups –based on a few co-variates.

Now they looked back at how many in each group had NAION.

What they found was that the incidence of NAION in patients with T2D or obesity was 4-6-fold higher for semaglutide vs non-semaglutide patients.

They displayed the KM curves with a y-axis cropped from 90%-100%.

Comments:

I worry about the long-term effects of GLP-1s. If this were a true association, it would be very worrisome. But my gosh, this is not the way to find out.

First of all, there is a huge collider bias –as these were patients referred not just to an eye surgeon, but a specialty neuro-ophthalmology clinic at Harvard. If you are referred to such a clinic you are special.

Brian Locke, MD an intensivist at U of Utah noted on Twitter that

You could also see an apparent increase in hazard of NAION if GLP-1s reduce the risk of other causes of ophth referral. both groups had to have some reason for referral in order to be included, so ‘event free’, is actually another referral trigger.

If patients receiving GLP-1s were referred *faster* after starting their new med than others (plausible, as they managed to get on a GLP.), that would also lead to the finding since they analyzed using hazard ratio (less exposure time bfore Dx). And Locke notes that

It doesn't even have to be a causal reduction... if patients on other meds are more likely to be referred for other types of issues that aren't balanced by their matching - that could explain it. Esp when access to GLP-1s is limited... those likely have very diff ocular issues.

My friends Bogdan Enache, MD an electrophysiologist doc in Romania, and Venk Murthy, MD a professor of cardiology at UM, noted the problem with time zero and immortal time bias. This occurs because the authors calculated person time from the first prescription of semaglutide vs non-GLP-1 diabetic or weight loss medicine. If something else caused the eye issue, then people on the newer drug will have a higher rate on the person-time scale.

Other problems – the effect sizes are huge. But they are calculated not from a general population but from a highly selected group of patients. The authors even calculated incidences of this condition by dividing the rate of NAION by the number of patients in the sample at the clinic. This is beyond comprehension—since the overall incidence of NAION is super rare, in the range of 1 in 10,000.

Another issue: they don’t control for diabetes severity. Semaglutide is used in more advanced cases of diabetes. That’s important because T2D and obesity are risk factors for NAION.

Another – they propensity matched, which is really a weak way to approximate comparable groups. Even if you believe there is a real association, which I doubt, there is a really good chance that these are not truly matched patients.

Finally, there are wide confidence intervals and an early huge signal that then plateaus. To me this suggests a different patient population.

Now for a meta-comment: That this paper gets published in JAMA, then JAMA promotes it, then it gets big media is deeply problematic. I think it is 99% due to incentives. Incentive to do such a flawed study, incentive for the journal to publish it b/c of the attention, and incentive to cover it the news. I don’t have the solution, but one start could be some journal editor just says no. We aren’t publishing this.

If you wanted to know if the drugs caused this rare cause of blindness, you would need a far more rigorous study. Perhaps the Danish registry could allow for detecting associations in large populations? But surely not a small series from one special referral center.

GLP-1 drugs and Cancer.

Glucagon-Like Peptide 1 Receptor Agonists and 13 Obesity-Associated Cancers in Patients With Type 2 Diabetes

JAMA-Network Open has published an equally dubious study looking at the question of GLP-1s and cancer.

The thinking goes that obesity likely causes cancer. Or, at least obesity strongly associates with cancer. GLP1 drugs reduce obesity, so the question becomes: do GLP1 drugs reduce cancer?

Again, such a GLP1 and cancer signal is unlikely to be found in RCTs, because of the short duration of follow up.

Authors from Case Western in Ohio used a large EHR database of patients over 64 healthcare organizations.

They then search the database of 16 million patients to find 1.6 million with T2D who had no previous dx of cancer and were prescribed GLP1 drugs, insulin, or metformin between 2005-2018.

The primary outcome measure was the first diagnosis of 13 different obesity-related cancers. They then compared cancer rates in those receiving GLP-1s vs insulin and in those receiving GLP-1s vs metformin.

Compared with insulin, GLP-1-treated patients had a significantly lower risk of 10 of the 13 cancers. Some of lowest HR were for gall bladder cancer, meningioma, and pancreatic cancer. The incidence of gall bladder cancer was <0.02 vs 0.04. Fewer than 10 cases per 50,000 patients treated with GLP-1s.

Compared with metformin, GLP-1s were not associated with a lower risks of most cancers, except for a 54% higher rate of kidney cancer. The incidences were 0.78 vs 0.55 respectively.

The KM figures for two positive associations, that is, colorectal and liver cancer for GLP-1s vs insulin show the major flaw of this analysis: immediate separation of the curves. In other words – the authors published the fatal flaw in one of the main figures.

Because, there is no way GLP-1 drugs have an immediate effect on cancer. It’s an open journal. Go look at Figure 3.

Again, my friends, we need to have information on these drugs, but such EHR mining is not the way forward. I am not a trained epidemiologist, but when I see curves separate basically at time zero, I know that the patients are different.

Meta-comment again: what the heck is happening in the JAMA family of journals. Perhaps there just needs to be fewer publications. I want to propose to you that flawed data is worse than no data.

The ECG

I am emotional about ECGs. Reading ECGs is one of the reasons I learned to love cardiology. Charles Fisch, MD, at Indiana was a mentor. So, it pains me to say something less than positive about a study about ECGs.

JAMA-IM has published an observational study looking at the impact of ECG screening in working age people in outpatient clinics.

You might wonder why such a study is needed b/c oodles of studies have found associations between ECG abnormalities and future CV outcomes. The authors write that few have looked at general populations and few have provided a gradient of risk based on how many or how severe the ECG abnormalities are.

The authors gradient of abnormalities were minor and major. They provide a list in the supplement. A premature atrial contraction (PAC) or nonspecific -ST-T change are minor, for instance. A left bundle branch block (LBBB), or AF are major. The list I have to say is somewhat arbitrary.

The authors from Harvard and Kyoto University used a large Japanese insurance database covering many millions of people.

In Japan, the authors write Every year, insurers are required to provide their employees with opportunities for health screenings with full subsidization. Employees who undergo screenings receive their results directly from the screening facilities and must report them to their insurers. Those diagnosed with major ECG abnormalities are advised by their insurers to seek further examinations, which are covered by their health insurance.

This was a big data study. More than 3.6 million people in this database. Mean age 47. Primary outcome of interest was overall death and MI, stroke, or HF.

Findings	5.5 year fu	Hazard for composite vs Normal ECG
17% had 1 minor abnormality		1.19
4% had ≥2 abnormalities		1.37
1.5% had 1 major abnormality		1.96

To give an idea on absolute risks, if you had a normal ECG, the incidence rate of the composite outcome (death, MI, stroke, HF) was 93 per 10,000. For one minor abnormality it was 129 per 10,000 and for a major abnormality it was 266 per 10,000.

I know, that is a ridiculous amount of numbers. So the delta in risk between normal and 1 major abnormality is about 1% higher.

The authors conclude:

The findings of this study suggest that the potential role of routine ECG screening for early prevention of CVD events, along with the optimal follow-up strategy, should be examined in future studies.

Comments:

I don’t agree with the authors conclusion. These data do not suggest any role for ECG screening for early prevention of CVD. All these data say is that there is a very small increase in risk of a CV outcome for patients with varying degrees of ECG abnormalities.

You could do a similar correlation for many things—age for instance. Getting older is a risk for a future health event. Even if the numbers were real, how does a working age person with non-specific ST changes use the fact that his risk for a CV outcome is 5 in 1000 higher than if had a normal ECG? All that this can do is scare someone.

Screening always creates tension in my noggin. Here is why: the vast majority of people getting an ECG will have no effect (normal), a large group will be referred for testing. Many of which will lead to over-diagnosis, over-treatment, and even harm. I could tell you stories about how shadows seen on an Echo have led to serious harm. And don’t even get me started on how many ECGs would lead to coronary artery calcium and the downstream shenanigans from that.

Then again, it would be of value to discover asymptomatic LBBB in a 50 year old. That person might have an ejection fraction of 35% and not know it, and benefit from medical therapy. I’ve seen that too. A good exam might sort out reversed splitting of s2 but few clinicians have this skill anymore.

Consider AF.

We now know that two robust screening trials (LOOP and STROKESTOP) fail to find substantial benefits of screening for AF. I know, I know, STROKESTOP was slightly positive but the difference was tiny.

If you can’t show benefit from screening for a major abnormality in high-risk people, then it’s extremely unlikely that screening would be positive in an all-comers working-age population.

Even though I love ECGs, and I think it’s one of the most under-rated tests in all of medicine, I don’t love this study. The ECG is wonderful for sorting out diagnoses of people who have complaints. People who are asking for our help.

ECGs for screening will be unlikely to improve health. Consider for example, the time spent explaining a PAC could be used to explain the benefits of exercising every day.

TEER for Secondary MR

One of the great mysteries in cardiology right now is the story of transcatheter edge-to-edge repair (TEER) of secondary mitral regurgitation (MR). AKA mitra-clip.

As an EP, I am an outside observer. But the non-industry supported MITRA-FR trial, found absolutely no benefit from the clip. The industry-sponsored COAPT trial, on the other hand, found that the clip was amazing. The effect size was akin to clean drinking water and sanitation on tuberculosis a century ago.

I know, Milton Packer, MD explained it all on proportional MR. And that’s possible, but I come from Indiana Univ, where Dr Feigenbaum banged into our heads how hard it was to grade MR. (or aortic insufficiency for that matter).

Neutral Martians look at table 1 of baseline characteristics in this trial and have trouble seeing enough difference to explain the widely disparate results.

The ESC meeting in London is coming soon will feature a third trial of TEER for secondary MR.

RESHAPE HF 2 trialists have published the baseline characteristics. The trial compares mitraclip to medical therapy

It may not—exactly—be a tie-breaker because RESHAPE will have far fewer patients with severe MR (less than half of the patients). RESHAPE will also have a slightly less effective regurgitant orifice area. Basically, less sick patients with FMR.

There was also a change in primary endpoint. The original design had a composite of total HF hospitalizations and CVD during 2 years of follow up. That was similar to COAPT and Mitra-FR (Mitra-FR was 1year).

An interim review however found that all-cause mortality rates were trending lower than expected. (recall less sick patients). When this happens, trialists have options: extend the follow up, recruit more patients, or change the endpoint.

The authors decided to randomize more patients and add endpoints. More patients doesn’t bother me. But the change in endpoint surely did. The two extra endpoints included total HF hospitalizations as its own endpoint and change in KCCQ score from baseline.

They will then use something called a Hochberg procedure analysis which sounds like a modified bonferri procedure. I found this copy and paste from a statistics site>

The calculation of adjusted p-values in the Hochberg procedure involves comparing each individual p-value to a critical value or threshold. The critical value is determined based on the desired false discovery rate (FDR) control.

I always love learning new statistic approaches, but I am really worried about this change. Both the HF hospitalization and KCCQ endpoints are problematic. HHF because it requires a decision from a clinician, who may be aware of a patients treatment arm. It’s not nefarious but if a doctor knows a patient has a clip, they may be more likely to send the patient home on diuretics rather than admit. And you can’t blind caregivers because clips stick out like a sore thumb on echo.

The KCCQ endpoint is way worse, because you can’t measure quality of life indices without a proper sham control. In these trials, one group gets a major intervention, and the other group does not. It boggles my mind that medical scientists repeatedly try to convince about quality of life endpoints without proper controls.

We shall see at ESC, but I would encourage all listeners to look closely at the hard outcomes in RESHAPE HF

Comments

Commenting is limited to medical professionals. To comment please Log-in.

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.