Jan 27, 2023 This Week in Cardiology Podcast

John M. Mandrola, MD


January 27, 2023

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 January 27, 2023, John Mandrola, MD comments on the following news and features stories.

Exercise and AF

The Adelaide group of electrophysiologists, led by Prash Sanders, is at it again. This month, the increasingly prominent journal, JACC: Clinical Electrophysiology, published results of the ACTIVE AF randomized controlled trial (RCT). ACTIVE AF screened patients with symptomatic atrial fibrillation (AF) who were referred to three tertiary care hospitals to participate in a pragmatic trial of an exercise intervention.

In all, 369 patients were screened and 120 were randomly assigned, so right off the bat, you know this is a select group. Kudos to the authors for including the screened vs randomized numbers. By the way, this is an important thing to look at when starting to read a trial, because it bears on how representative the study patients were to normal patients.

First author Dr. Adrian Elliott is an exercise physiologist and the exercise intervention in ACTIVE-AF was carefully tailored to the patients. Some were older and retired adults; others were younger and still working or with family commitments. So, they needed different strategies.

  • All patients in the exercise arm had a baseline VO2 max test, which guided the parameters of the exercise program, which was supervised initially in clinic and then at home. Exercise included high-intensity aerobic intervals.

  • For the first 3 months, patients underwent weekly supervised exercise visits, followed by supervised visits every 2 weeks. They also had an exercise prescription for home-based activity with goals of getting to 210 minutes/week.

  • The control arm got standard advice on exercise with goals to reach 150 minutes/week of exercise. They had a repeat visit at 3 months to reinforce the exercise advice. Remember this is a medical group known for their risk factor modification.

  • Patients were on average 65 years old; 75% were male; baseline body mass index (BMI) was 30-31.

  • The co-primary endpoint of AF recurrence (off antiarrhythmic drugs and without ablation) and symptom severity score.

  • By 12 months, freedom from AF was achieved in 40% of patients in the exercise group and 20% of 60 patients in the control group (hazard ratio [HR]: 0.50; 95% confidence interval [CI]: 0.33 to 0.78). 

  • There was also a statistically significant improvement in AF symptom score, although this barely met significance.

VO2 max — a measure of exercise performance — improved in the exercise group but interestingly, weight, BMI, and blood pressure (BP) did not. That’s sort of important because a) it re-enforces the notion that weight loss takes more than exercise; and b) the reduction in AF recurrence occurred with exercise alone, not risk factor modification or medication.

Also notable, the same number of patients in both arms ultimately had pulmonary vein isolation (PVI), which speaks to the modest effect size. I would add however, that Supplemental Figure 1 found a strong trend that overall AF-free survival with antiarrhythmic drugs or ablation was better in the exercise arm. So, it seemed that fitness adds to other rhythm control therapies.

Comments. I have to disclose that I am friends with Adrian and Prash and others on this team. Second, I strongly believe that AF ablation in our fee-for-service model is wildly overused. I don’t want anyone to think #TWICPodcast hides its biases.

We should set out first that randomization in a trial is a much better way to study the effects of an intervention like exercise than is an observational design. That’s because the observational design will have severe problems with selection bias. Healthier patients are far more apt to exercise and gain fitness, and it is likely those factors — not just the exercise — may lead to lower rates of AF.

Bottom line on the results:

  • This is a positive trial.

  • It showed statistically significant and clinically important reductions in the primary endpoint.

  • Add to that, the fact that exercise has essentially no adverse effects and has many other health benefits.

  • Thus, this is an extremely valuable addition to the literature.

Another clarifying fact. The issue of AF being increased in endurance athletes does NOT pertain here. That association occurs in folks who insist on years and years of intense endurance exercise. These are people who could be said to overdose on endurance sport. The goal weekly dose of 210 minutes per week of exercise in ACTIVE AF amounts to an easy day of training for many of the people who get AF from too much exercise.

But there were limitations of this trial:

  • First, the external validity. The authors clearly state that they excluded 250 of the 370 patients they screened. Many refused. Some had orthopedic issues that precluded exercise. So, as the authors write, this was a select group of patients. Structured exercise is not going to work for all patients.

  • Another external validity concern: This was a single-group study, a group that is intensely interested in using risk factor modification to improve AF outcomes. And Adrian Elliott is a talented and motivated leader. I’d listen to him if he said do this amount of exercise. In other words, this group is full up of champions for exercise and fitness.

  • That’s an important factor for translating this evidence because whether this scales to places with less enthusiasm for fitness, like 95% of us in the United States, is in question.

Now to the internal validity questions:

  • The study was obviously unblinded so I don’t think we can put much weight on any quality of life (QOL) questionnaire. In fact, I am not sure QOL should be used in any trial that is not blinded. The arm with the more care will tend to feel better. Caring signals should never be discounted.

  • Another issue — and again, the authors highlighted this — was the lack of continuous monitoring. When you measure AF recurrences as an endpoint, you like to have continuous loop monitors. But I totally understand the limits of funding and the tension between doing an imperfect trial and not doing any trial.

  • I have seen some raise the issue of performance bias, meaning that patients in the exercise arm benefited from more interactions with caregivers. The authors address this concern by saying, well, the bias of increased interactions in the exercise arm may have helped reduce AF, but it could have been countered by the increased chance that caregivers would pick up AF.

  • I’d argue it a different way. I’d say performance bias was the point here. You are studying an intervention of more involvement with caregiver’s vs less. I don’t see performance bias in this study as a problem. It’s not like EAST AF.

In EAST-AF, the question was: does maintenance of sinus rhythm improve outcomes over standard of care, mostly rate control. EAST AF found that early rhythm control (ERC) led to a reduction in hard major adverse cardiac event (MACE) outcomes — cardiovascular (CV) death, stroke, heart failure (HF), acute coronary syndrome (ACS). Well, in that trial, patients in the ERC arm had many more interactions with healthcare providers, and it could have been those interactions that led to better outcomes in things like HF, ACS, and stroke. I say that because the actual difference in AF in the two arms was minimal. If you want to show that ERC is better, the control arm should have the same amount of care. And you have to show a larger difference in AF in one arm. Otherwise, there are other factors besides AF that may be reducing non-AF related outcomes.

Future directions: This exercise/fitness signal ought to be studied further. Why wouldn’t this scale to cardiac rehab clinics? The infrastructure is already there. Such clinics are well-suited for this. I’d love for the government to fund a multicenter trial randomly assigning patients with AF to cardiac rehab-led exercise vs standard of care.

Take-home for my and your clinic today:

  • Exercise is way under-used. I routinely tell patients that 15-30 minutes of exercise that leads to sweating is like a heart pill. It should be taken every day.

  • Now I can tell patients with AF that there is a chance that exercise may help reduce AF. Along with weight loss, sleep apnea treatment, and alcohol reduction, there is a chance their AF may be vanquished without obliterating atrial myocardium.

  • And the best argument to incorporate these findings: Even if this isn’t replicated in a big trial — what are the downsides? Even if AF remains, there are huge benefits from gaining fitness, especially for 65-year-old patients who have a mean BMI of 30.

GLP-1 Agonists and Retinopathy

The next topic involves a drug-complication that I did not know about. The new wonder-class of drugs called glucagon-like peptide 1 (GLP-1) agonists, specifically semaglutide, have been shown to have a possible harm in the worsening of diabetic retinopathy.

I highlight this meta-analysis of GLP-1 agonists, published in the Diabetes and Metabolic Syndrome journal and covered by journalist Marlene Busko, for two reasons.

  • GLP-1s are a hot new drug class, for diabetics and patients with obesity, and we best know about safety signals, especially since cardiologists treat a lot of patients with diabetes and diabetic retinopathy.

  • And it’s always good to remember that drugs may block one or two channels (or in the case of amiodarone, many), but their effects are not limited by this one action.

Please do read Marlene Busko’s summary; it is good. The meta-analysis combined six of the GLP-1 CV outcome trials – the ones that evaluated diabetic retinopathy.

While it is well-established that GLP-1 agonists reduce MACE in patients with diabetes, this meta-analysis found that semaglutide increased the rate of diabetic retinopathy (HR 1.73), with a CI going from 10% higher to 2.7 times higher.

The authors also did a meta-regression, which is a type of statistical analysis used to study the relationship between different variables in a large group of studies. Its purpose is to try to understand how different factors might be affecting the results of different studies. For example, a meta-regression might be used to look at how different study designs, sample sizes, or other factors may influence the results of studies on a particular topic. Meta-regression basically uses the study results rather than the patients as measures to plot. The goal is to identify patterns or trends in the data that can help researchers understand the underlying causes of the results they are seeing.

The meta-regression of these studies found the following signals:

  • The A1C reduction correlated with both MACE reductions and increases in diabetic retinopathy.

  • The change in diabetic retinopathy was predominantly found with subcutaneous semaglutide given for > 1 year (relative rate [rr] = 1.559,1.068,2.276, P = 0.022) and with reductions in A1C > 1.0% (rr = 1.59,1.092,2.316, P = 0.016).

The authors wrote that, when evaluating semaglutide alone, the effect on worsening retinopathy was not associated with drug exposure alone or with oral semaglutide. Rather, there was a correlation of worsening retinopathy associated with longer trial duration and greater decreases in A1C.

However, the reassuring part of this analysis was that the absolute reductions in cardiac outcomes was much larger than the increases in retinopathy. Previous studies seemed to suggest that the proportion of retinopathy complications was greater among patients with a history of retinopathy at baseline (8.5% vs 6.2%). 


  • Specific message — don’t shoot from the hip. If you have an overweight diabetic patient that you are thinking about prescribing semaglutide to because you read an article in the New England Journal of Medicine, think about diabetic retinopathy. If the patient has it, think about chatting with the eye doctor or deferring to diabetes doctors.

  • Larger message — even though a drug blocks one pathway, its effects can be complicated. Always remember that most trials are 2 to 4 years in duration, and most regulatory trials are designed for efficacy, not safety.

  • When we use drugs for longer periods than trials, there are known unknowns, and unknown unknowns. These unknowns may be relevant given recent news regarding the drug’s effect in young people with obesity.

ChatGPT Can Pass USMLE Exam Questions

I will take some of the next from a preprint published on an artificial intelligence (AI) tool’s ability to answer medical test questions. The preprint is already accepted for publication in PLOS-Digital Health. But first, I need to set out that everyone needs to know about ChatGPT. It is a large language model (LLM) developed by OpenAI. LLM are different than previous deep learning models, which simply recognize patterns.

Get this. It’s ridiculous:

  • LLM are trained to predict the likelihood of a given sequence of words based on the context of the words that come before it.

  • If trained on large enough amounts of text data, they can generate novel sequences of words never observed previously by the model, and these words, typed out before your eyes, are plausible.

  • Anecdotal use indicates that ChatGPT exhibits evidence of deductive reasoning and chain of thought, as well as long-term dependency skills.

Here is what one of the smartest persons on the Internet says; George Mason University’s Economics’ professor Tyler Cowen wrote:

 “In only a few years, these models have gone from being curiosities to being integral to the work routines of many people I know. This semester I’ll be teaching my students how to write a paper using LLMs.

“At some point, GPT may substitute directly for some of my writings... I expect I will invest more in personal talks, face to face, and also “charisma.”  Why not?

“Well-known, established writers will be able to “ride it out” for long enough, if they so choose.  There are enough other older people who still care what they think, as named individuals, and that will not change until an entire generational turnover has taken place.

“I expect the entire calculus here is very different for someone who is twenty years old.

“Today, those who learn how to use GPT and related products will be significantly more productive.  They will lead integrated small teams to produce the next influential “big thing” in learning and also in media. Most current contributors will miss that train almost entirely, just as so many people missed the importance of the internet for learning and also for media....”

In the preprint description, a group of people from Ansible Health, a company that manages homebound patients with chronic lung disease, tested the AI bot’s capabilities in taking the USMLE exam. In fact, the bot wrote the abstract and results section of the actual preprint, and contributed large parts of the intro and methods section, as multiple authors of a paper would do!

For each step, the researchers prompted the chatbot in three ways.

  • First, it was given a theoretical patient's signs and symptoms and asked to pontificate on what might be the underlying cause or diagnosis.

  • After ChatGPT was refreshed to eliminate potential bias from any retained information from the previous exercise, it was given the questions from the exam and asked to pick an answer.

  • After again refreshing ChatGPT, the researchers asked it to "please explain why the correct answers are correct and why the incorrect answers are incorrect."

The answers were reviewed and scored by three board-certified, licensed physicians. The bot did well, getting answers in the 50% to 60% range overall. Of course, this was only a fraction of the test. The team could not ask questions that used images or sounds.

As for my comments, I don’t know exactly what to say, except I have the same feeling now about these AI models, as I did when sitting in the VA clinic in 1990, playing around with this new thing called the Internet. Clearly, medical licensing exams will need to change. Memorizing questions that an AI bot can answer will need to cede to questions that only humans can do.

HF Prognosis and Implications for New Therapies

The European Heart Journal published a nice paper on HF prognosis. The results explain a lot of what I talk about on #TWICPodcast.

This was an observational study from a Danish registry, with first author Jonas Bruhn.

  • The researchers included just over 100,000 adults who were free of cancer and had an HF diagnosis between 1997 and the end of 2016.

  • The purpose of the paper was to explore competing causes of death in patients with HF. This is important because improved HF survival may increase the risk of cancer because of a competing risk.

Datasets like the national registry in Denmark allow for looking at temporal trends like this. Also, this is appropriate use of observational data, unlike the many studies that try to simulate randomization from nonrandom comparisons. Think users and nonusers of hormone replacement therapy to prevent CV disease in post-menopausal women.

The authors report two main findings.

  • One was that the incidence of cancer 5 years after a HF diagnosis over the 20-year period remained stable.

  • The other was that the 5-year cumulative incidence of survival for patients with HF increased with advancing calendar years, going from 55.9% (1997–2001) to 74.3% (2012–2016).

Both of these results surprised me. I would have thought cancer incidence would have increased. But the improved survival after a HF diagnosis is a super-important observation because it helps explain a lot of nonsignificant findings in recent years.

  • DANISH, the trial of implantable cardioverter-defibrillator therapy in nonischemic cardiomyopathy was nonsignificant. Many of the trials of HF with preserved ejection fraction (HFpEF) were non-significant, and even the ones that were significant, were driven by HF hospitalizations not CV death or all-cause mortality.

  • Recall that the early trials in HF, ACE inhibitors, beta-blockers, and spironolactone, all had mortality as an endpoint. But newer HF trials are, mostly, positive only because they reduce some non-fatal surrogate outcome.

  • In sum, the good news is that we do a lot better treating patients with HF. The bad news is that, given fixed human lifespans, it’s much harder for new therapies to make much of a difference.

  • The new frontiers of HF may lie in better palliative care. Mortality may no longer be the best endpoint of HF trials. And adopting therapies that improve the quality of life for patients with HF may be a huge positive.


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.