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 April 7, 2023, John Mandrola, MD, comments on the following news and features stories.
First, I want to thank the University of Wisconsin cardiology program for hosting me for cardiovascular (CV) grand rounds. I appreciate the opportunity to make the case for a medically conservative approach to evidence appraisal. Ninety people in the audience was great, as that’s 90 chances to spread the word about a skeptical but non-cynical approach to using evidence.
Next week I head to the Portuguese Cardiac Society meeting. Gosh I am excited to attend this fantastic meeting again.
Also, I want to say thanks to the listeners of this podcast. TWIC has surpassed 600 ratings on Apple Podcasts. That is amazing. If you haven’t done so take the time to rate and review this podcast on whatever app you use.
Artificial Intelligence
The journal Diabetes Care published a study looking at how well Chat GPT did answering questions about diabetes self-management. These were common questions in four domains of diabetes self-care.
The results are amazing. Artificial intelligence (AI) did really well; not perfect but really well. I lead with this topic not because of how large language models (LLM) will affect diabetes, but because I believe we are on the brink of something similar to the Internet in 1995.
When I trained in medicine and electrophysiology (EP) in the 1990s, there was no Google, no apps, no smartphones. I could not have even imagined a world with these things. When I travel and talk to young people, I often get blank stares when I remind them that doctors my age practiced when there was no Internet.
Now I think we are at a similar place. LLM shock me. If you haven’t explored it, you should. You will be shocked too.
I was writing an article about heart failure with preserved ejection fraction (HFpEF). I wanted to know some basic epidemiology. ChatGPT cranked out a few paragraphs that could have been used in the column. The stats were correct. I searched Google and it led me to large documents that had the info, but I had to search the documents. This was just the enhanced search function — like 1/100th of ChatGPT’s capabilities.
The title of the article on Medscape is “Can ChatGPT Replace Diabetes Educators?” I don’t know that answer, but this is the right question. To wit, what will ChatGPT replace? The answer is a lot.
The authors of the paper emphasize the mistakes the LLM made. My counters to its imperfection are, just wait, GPT4 is already better than GPT3; And what human doesn’t make mistake; and what patient or user of information doesn’t grade received advice for accuracy?
If your job is something rote or robotic, I would be worried. If your job involves human-to-human interaction or creativity, LLM may make you more, not less, valuable.
I hope LLM will make writing great again. The readability of Chat GPT is already better than that in most medical journals. I like to write in a different way, with my voice. Perhaps those of us who write like that will stand out as different or better than LLM.
I also think LLM will make public speaking more valuable. When learners can simply ask a LLM to explain a complex topic at a college or high school level, what hope does a poor lecturer have?
Public speaking that offers a human view, a creative view, will become much more valuable. Speakers who read PowerPoint text slides should be worried.
Another hope I have is the power of LLM to help with documentation at the bedside. LLM may be able to summarize big electronic health record charts, write notes, and allow more time for human-to-human care.
This is the first kilometer of a marathon. I will surely speak more on this. I also want to say that I follow the rationalist community, and there are some — Eliezer Yudkowsky and Robin Hanson — who express serious worry about the existential risks of AI
Hanson has also written about the possibility of a "hard takeoff" scenario, in which AI rapidly surpasses human intelligence and becomes uncontainable. He argues that this scenario is a real possibility and that we should be taking steps to prepare for it.
Yudkowsky believes that the risks of AI are significant because once an AI system becomes sufficiently advanced, it may be able to recursively self-improve and rapidly become much more intelligent than humans. This could lead to a situation where the AI system's goals are no longer aligned with human goals, and it could take actions that are harmful to humanity as a whole.
By the way, these last two paragraphs were written by ChatGPT! So, learn to write like you.
An Unrecognized Beneficial Therapy for HFpEF?
I discovered something about treating HFpEF this week. I had no idea, until publication of a well-written American Heart Association/American College of Cardiology (AHA/ACC) review on exercise training for HFpEF (lead author, Dr Vandana Sachdev) that there were so many randomized controlled trials (RCTs) of supervised exercise training for patients with HFpEF.
Table 1 of their document lists the 10 RCTs of supervised exercise training.
Most looked at the primary endpoint of peak VO2 and most were positive.
Secondary endpoints of quality of life (QOL) also improved.
These trials differ from drug trials in that drug trials record clinical endpoints such as CV death or hospitalization for HF (HHF).
My friend Adrian Elliot, an academic exercise physiologist in Adelaide, points out that it may be easier to get positive results in trials looking at VO2 max vs clinical endpoints.
That may be true, but the counter is that patients value functional status and QOL, perhaps even more than a HHF.
Notable also is that these were supervised exercise programs. The document also listed a handful of home-based exercise studies, but these were quite small and reliable conclusions could not be made.
Here’s the thing that this podcast has said many times: HFpEF is far more heterogenous and complex than HF caused by a reduced EF.
In HFpEF, symptoms can surely stem from diastolic dysfunction in the heart, but also there are pulmonary, vascular, and skeletal muscle components.
Drugs hit mostly the cardiac component; exercise hits everything – lungs, blood vessels, muscles, and surely the mind and soul.
Yet, use of exercise in our current healthcare delivery model has headwinds. The document and news coverage on Medscape delves into these. Things like supervised exercise is different from cardiac rehab, which is a more comprehensive program of health education, and the fact that there isn’t a reimbursable model for supervised exercise training in HFpEF.
I hate that so much of therapeutics comes down to profit. Yet that is the reality. Exercise has few champions. Consider the difference between exercise and drugs/devices.
Drugs have the backing of industry, and Industry backs professional societies and key opinion leaders, so drugs get promoted.
Drug and device therapy becomes a therapeutic fashion – like left atrial appendage occlusion or ablation for atrial fibrillation or coronary CT scans. All of these interventions have champions because being a champion for these brings money and influence.
You see an older patient with a stiff heart and a couple of admissions for HF and everyone thinks drugs. Ok, but no one sees the fact that this patient is totally and utterly deconditioned. Her muscles and lungs are toast.
So, few of us even think of structured exercise. It’s not the fashion because it has no champions. Because there is no benefit in being said champion. But that doesn’t make it any less worthy.
In EP, device companies could not make money from conduction system pacing. They had little interest in funding an intervention that decreases cardiac resynchronization therapy units. They funded no trials. But now, because doctors pushed it, conduction system pacing is hot and becoming established, and industry had to follow our lead, and get in the game.
For my HF colleagues out there, exercise therapy in HFpEF needs an organic conduction system pacing -like enthusiasm. Industry and professional societies are unlikely to help. It’s going to have to come from the working stiff doctors as conduction system pacing did.
Statins and Muscles
Oh my, the Journal of the ACC has published a very interesting study on statins and exercise from Dutch authors in Nijmegen. You will like this.
Some background:
Statins reduce cardiac outcomes by 25%. As I have said many times, these are heart attack reducing drugs.
But statins can cause statin-associated muscle symptoms (SAMS). We can argue about the cause, drug effect or nocebo but there is no denying that people on statins have SAMS.
This can affect physical activity. People with muscle symptoms may not exercise or they may stop taking statins. Neither are ideal, because exercise and statins improve health.
Previous studies have suggested that vigorous or eccentric exercise may increase muscle damage markers, such as creatine kinase, more in statin users compared with statin nonusers.
There are less data on moderate level exercise. There are also questions regarding mechanism of SAMS that involve coenzyme Q10 (CoQ10) and mitochondrial dysfunction.
Now to the study. In Nijmegen, Netherlands, there is an event called the Four Days Marches. People walk 30, 40, or 50 km per day at a self-selected pace. Pause there. 50 km = 30 miles. Have you ever walked that much? I have walked 20 miles and it hurts.
For this study, investigators recruited three groups:asymptomatic statin users; statin users who had SAMS; and controls.
They made baseline measurements and then studied as a primary endpoint muscle damage markers, such as CK, myoglobin, LDH, troponin, and BNP.
Secondary measures were muscle pain, strength, fatigue, and CoQ10 levels.
All muscle injury markers were comparable at baseline and increased following exercise, with no differences in the magnitude of exercise-induced elevations among groups.
Muscle pain scores were higher at baseline in symptomatic statin users (P < 0.001) and increased similarly in all groups following exercise (P < 0.001).
CoQ10 levels also did not differ in the groups.
Conclusions:
Statin use and the presence of statin-associated muscle symptoms does not exacerbate exercise-induced muscle injury after moderate exercise. Muscle injury markers were not related to leukocyte CoQ10 levels.
This is nice data. The authors state clearly that this moderate level exercise may not translate to more vigorous types, but I think they undersell the degree of exercise this is. It’s a lot, in my opinion. It’s not intense but it is still hard to walk that much.
Nonetheless, it contrasts with previous studies of more intense exercise in that walking for many hours does not cause any demonstrable negative effects on muscles. Of course, another limitation is that these are Dutch people who were willing and able to walk 20 to 30 miles per day. I think we can use this paper to add to our education of patients with statin-related muscle complaints.
First we can discuss SAMSON data showing that statins exert their negative effects mostly through a nocebo effect.
Then we can add this paper, which shows that if there are mild to moderate SAMS, exercising at moderate levels, even for hours on end, will not harm the muscles.
And by doing so, we can encourage our patients to stay on a drug that reduces outcomes and more importantly to continue exercising.
To be sure, I want patients to have all available statin information but, in the end, it is up to them to decide if the risk reduction is enough to take a pill every day.
Statin Use After CV Events
Secondary prevention of another cardiac event in patients who have had one or more events is one of cardiology’s most important jobs. It’s unlike primary prevention in that these patients have proved that they have atherosclerotic disease. While secondary prevention is still prevention, we are treating people who have disease and have asked for our help.
Statin therapy is a key part of secondary prevention. But how should we use these drugs? Should we simply prescribe high-intensity statins, as was done in RCTs? Or should we start with more moderate intensity statins and treat to low density lipoprotein (LDL) targets?
If you get low enough LDLs with lower intensity drugs, you might reduce side effects and improve adherence. On the other hand, treat-to-target involves more work (I’ll come to that).
It turns out that these strategies have not been compared in RCTs. That is, until the recently published LODESTAR trial from South Korean investigators, first author Sung-Jin Hong. This is also a fun trial to discuss.
4400 Patients with coronary artery disease.
One group is treated to a target LDL of 50 to 70 mg/dL. The other group received high-intensity statin therapy (rosuvastatin, 20 mg, or atorvastatin, 40 mg). Three-year follow-up with composite major adverse cardiac events (MACE) as the primary outcome.
Investigators chose a noninferiority (NI) design, with the high-intensity arm as the standard and the treat-to-target as the experimental arm (that is weird, I will also return to that).
If it is an NI design, you have to have an NI margin, wherein you say the experimental therapy can be this much worse and is still ok (or noninferior). To young listeners, the choice of NI margin is a matter of great debate and it is content dependent.
Here the investigators expected a 12% event rate for the composite of death, MI, stroke, or coronary revascularization. They set out an NI margin in absolute risk increase of 3%. So, if the upper bound of the risk of the risk difference was less than 3%, treat to target would be called NI to the simple just-use-high-intensity-statin strategy.
Before I tell you the results, it’s important to say this was an open-label strategy trial. Treat-to-target patients required lots of changes in dosing, mostly to increase statins to get below the LDL goal of 70 mg/dL, but sometimes backing off when levels went below 50 mg/dL.
Another caveat. Investigators discouraged the use of non-statin drugs such as ezetimibe. This choice illustrates the tension between having a “cleaner” trial — here, isolating the effects of statin use — and simulating practice (e.g pragmatic). If you simulated practice and encouraged non-statin drugs, and there was an imbalance in the two groups, it would be harder to pin causality of any differences on either of the statin strategies.
The results:
The mean LDL at 3 years in the treat-to-target group was 69 mg/dL vs 68 mg/dL in the high-intensity arm.
Only 54% of treat-to-target patients received high-intensity statins compared with 92% of those in the high-intensity statin arm.
The primary end point occurred in 8.1% in the treat-to-target group and 8.7% in the high-intensity statin group (absolute difference, – 0.6 percentage points; upper boundary of the one-sided 97.5% confidence interval [CI], 1.1 percentage points), which was well less than the NI margin of 3%.
Treat-to-target is NI.
Conclusion:
Among patients with coronary artery disease, a treat-to-target LDL-C strategy of 50 to 70 mg/dL as the goal was noninferior to a high-intensity statin therapy for the 3-year composite of death, myocardial infarction, stroke, or coronary revascularization. These findings provide additional evidence supporting the suitability of a treat-to-target strategy that may allow a tailored approach with consideration for individual variability in drug response to statin therapy.
Comments. The first thing to say is that Steve Stiles is an excellent journalist. Go read his coverage on the theHeart.org | Medscape Cardiology.
The next thing to say, and I don’t want to sound like a typical panelist at a late-breaker, but, really, I laud the South Korean trialists for studying this question empirically. This is a hard trial to pull off, and since I believe we need more empirical study of our interventions, I really do love that they did this trial.
The third thing to say is that evidence can go out of date quickly. I have a number of slides showing this. Consider implantable cardioverter-defibrillator (ICD) use in nonischemic cardiomyopathy. The SCD-HeFT trial in 2004 was positive. The DANISH trial in 2016 was null. Same therapy, the ICD, but different time.
Well, LODESTAR began in 2015. There are now four non-statin drug classes that lower LDL and three of them reduce outcomes: ezetimibe, PCSK9-I, and bempedoic acid. The twice-yearly small-interfering RNA, inclisiran also markedly reduces LDL.
The reality now is that many high-risk patients with established disease could be on two agents. A clinician could easily drive an LDL cholesterol level much lower than 50 to 70 mg/dL with a combination of the two drugs. These 8% event rates could be lowered even further.
Taking this trial as it is, not how we want it to be, I’d say the two strategies achieve the same LDL at 3 years and the results are similar. That’s not too surprising.
It’s hard to know how to apply these results in 2023.
I like the treat-to-target strategy because it’s always good to use less drugs. But the high-intensity arm has one great advantage — simplicity. It decreases the work of being a patient. To me, this outweighs the treat-to-target for most but not all patients.
Having looked at the evidence and seen patients for decades, I am not that convinced that the side effects of moderate vs high-intensity statins differ that much on average.
I’d say the next trial would want to look at high-intensity statins vs a lower LDL target that could be achieved with a combination of drugs. But that won’t be an easier trial to interpret either.
Let’s say the LDL-is-everything proponents are right and such a low LDL target produces a statistically significant absolute risk reduction of only 1%. In that case, we will be faced with making cost-efficacy decisions. In other words, is it worth it?
Or perhaps lower LDL below 50 to 60 mg/dL won’t make much difference? Perhaps there is, in 2023, a lower bound of LDL benefit.
For the in-the-weeds people, LODESTAR is yet another NI trial that overestimates event rates. They predicted 12% and got 8%. Why is that important?
It is important because they chose the NI margin as a risk difference. They chose 3% absolute risk difference based on the 12% expected event rate. When the actual events are 8%, well, 3% makes it much easier to reach NI; 3% is a much bigger fraction of 8 than it is of 12.
If you knew it was 8% then you’d choose a lower risk difference as a NI margin.
I don’t know why all NI trials don’t use both relative and absolute risk margins. I calculated relative risk here with 95% CI and I think treat-to-target would have still made NI, as the upper bound was 1.13-ish.
But always be aware of expected vs observed event rates in NI trials. You might wonder why most NI trials expect more events than they observe. It has to with the constraints of funding trials. If you expect 8% event rates, you need more patients or longer follow-up. That means more costs. Trials don’t pay for themselves, so cost constraints are always factor in trial design. It’s why I wish for more efficient trial formats.
© 2023 WebMD, LLC
Any views expressed above are the author's own and do not necessarily reflect the views of WebMD or Medscape.
Cite this: Apr 07, 2023 This Week in Cardiology Podcast - Medscape - Apr 07, 2023.
Comments