I’ve written a few shameless promotional pieces for Primary Care/General Internal Medicine. Here is another one. My daughter recently told me that I write a lot ot listicles, so here is another one: seven things I love about being a general internist

Getting paid to answer people’s questions and work toward making them better

This realization inspired this post. I have mornings and afternoons when I don’t have patients scheduled. These hours are supposed to be filled with writing, lecture preparation, administrative responsibilities, and tasks left over from clinic sessions. In truth, what often occupies this time is answering calls, texts, emails, and MyChart messages from people who need help managing their health. I find it amazing that I get paid to basically be a 311 line for my patients. I have trouble imagining a more satisfying and rewarding way to spend a day.

Puzzling through medical mysteries

I’d be lying if I said that the majority, or even a large minority, of my clinical time is spent solving Holmesian (Housian?) level mysteries. Much of what I do is routine, completing diagnostic or therapeutic tasks that I’ve done a thousand times before: evaluating dysphagia, treating a UTI, adjusting insulin dosing, and the like.

But, a few times a week, people come in with symptoms whose etiology is not obvious. I love the puzzle of trying to make a diagnosis parsimoniously, causing the patient the least trouble and spending the fewest healthcare dollars.

Trying, and failing, to master medicine

I’ve written this many times in many different ways. There may be a finite number of diseases, but with over eight billion people, the possible presentations are essentially infinite.

Every day, I get to read articles about diseases and treatments, not just because they are interesting or so I’ll have more to say at a cocktail party, but to actually help people. I’ll see a patient with an interesting problem, and I’ll have to dive into the medical literature, broadly defined, to solve it.

Although I deeply love this challenge, when I occasionally, invariably, get things wrong, the last thing I am thinking is, “Isn’t it great that medicine is so hard!”

Building and maintaining relationships

I’ve been in the same practice in the neighborhood where I live and where my kids went to school for 29 years. It has been an absolute pleasure to watch my patients change, and our relationships develop. I have patients I started seeing when they were in their teens who are now in their 40s. My feeling for some of these people could best be described as paternal.

It has been delightful watching a web of patients’ families and friends become my patients. It once made me uncomfortable, but I now enjoy socializing with my patients at work, at school functions, around the neighborhood, and even at my own colonoscopy.

I feel ridiculously fortunate that my work has introduced me to people I would never have otherwise met. I am sure that these relationships have made me a better person.

Working with colleagues

My clinical colleagues are a total joy. These are smart, accomplished people from whom I constantly learn. These are people who derive the same pleasure from caring for people as I do. There is never a week that I don’t walk away from a conversation with a colleague impressed.

I get that there are a whole lot of problems with social media. I won’t even list a handful of them as examples for fear of excluding important ones. Acknowledging that, through Twitter, X, and now Substack, my circle of colleagues has grown exponentially. It is a pleasure to interact with people from, quite literally, around the world, about patient care issues, academic pursuits, and even homemade ice cream sandwiches. These relationships have led me to visits near and far. I thought I’d come back from these trips with another line on my CV, a bit of ego boost, and — occasionally — a small honorarium. In fact, I come back enriched with new ideas, new enthusiasm for my work, new colleagues, and, sometimes, new friends.

Caring for everything (with backup)

As a resident, I’d complain when surgery declined to admit a patient with cholecystitis or a partial SBO. I’d whine when neurology passed on a patient with a completed stroke. I’d seethe when anyone pregnant ended up on my service. Although none of these patients required surgical, neurologic, or obstetrical care, they were not well enough to go home. So, they ended up on medicine. Medicine can never refuse a patient.

Then I grew up. I realized that the beauty of the field in general, and of my specialty in particular, is that we care for anything and everything. I’ll happily manage your heart failure, your thyroiditis, and your pneumonia. I’ll also think hard about when my patients would do better (or worse) seeing a subspecialist.

Being a good general internist means knowing what every specialist can and will offer. It means knowing enough about everything that you know what you don’t know enough.

The money is good

Talking about money is generally considered unseemly. But we hear about people choosing ortho, or ophtho, or derm for the money, so it’s probably worth acknowledging. I am more than satisfied with the money I make. In fact, I can’t believe I am paid so well to do something I so fully love. I certainly don’t make master-of-the-universe money. (Nobody in medicine does, and I chuckle at those who think they do.) I don’t deserve those kinds of riches. I’ve taken no risks; I just chose a path, worked hard, and got lucky.

Here is some recent data about physician income.

Even for a single-earner family, this puts a PCP in the top 10% of yearly incomes for Americans. Given that 20% of doctors are married to other doctors, those people are comfortably in the top 1%.

Ok, there are caveats. Most doctors don’t start earning real money until their early 30’s. That’s 10 years of income to make up if you compare yourself to a peer who got a well-paying job right out of college. It is also true that many students graduate with quite a bit of debt. (Though it is also true that a large proportion of medical students come from families of means who pay their way. Take a look at this terrific study by previous Sensible Medicine contributor @armanshahriar.)

Considering how good the job is, to be compensated handsomely is pretty great. You’ll never get rich as a primary care general internist, but if you live within your means, you’ll be fine.

Share

• Side Effect Patterns in a Crossover Trial of Statin, Placebo, and No Treatment

Cost-Effectiveness of Fecal Immunochemical Testing Alone vs Co-Testing With Helicobacter pylori Stool Antigen

Share

Starting next Tuesday, we will begin by re-posting Season 1 ICYMI. We will post these weekly, and make these episodes available to everyone. No more paywall! If you enjoy the episodes, send the videos to your students, residents, fellows, friends and colleagues.

Then, 10 weeks from now, we will start posting Season 2 episodes. We will post these bi-weekly for our subscribers.

How Not to Get Fooled is motivated by our view that, in 2026, it is easy to be mislead by medical science, and many forces are out there trying to mislead you. The sponsors of specific clinical trials, retrospective observational studies, and cost-effectiveness analyses stand to make billions of dollars if they can persuade regulators and doctors to adopt their products. Clinical societies and expert physicians— so called “Key Opinion Leaders”— often lack expertise in critically evaluating these research efforts. At times, there may be financial incentives to look the other way.

Even “non-conflicted” researchers may have conflicts. In biomedicine, there is a huge incentive to claim discovery- That you have found something new and actionable that no one else has. Yet, time and again, these efforts do not always replicate. Are there ways to tell which science will be enduring?

That’s why randomization matters. Not always, but sometimes (we explain when). Control arms matter. Pre-specifying your primary endpoint matters. Sticking to your statistical analysis plan matters. Endpoints matter. Power matters. Hierarchical design matters. Drug run in periods— or double run ins— can matter. Representation matters. Pragmatism matters. Concealment matters. Blinding matters (well, again, sometimes).

We believe this 19 part lecture series captures the best of each of us, but brings something more. Each lecture is typically half didactic and half discussion. Even though we are 3 like-minded medical conservatives, you will see considerable disagreement between us. Even when we agree, we phrase things differently. There is a synergy when we are all in the room together. That’s why we travelled to make this possible.

We are delighted to share this lecture series with you, dear reader. You will find that I have already unlocked the first episode, but starting next week, we will post them directly to you— and they should also appear on the Sensible Medicine podcast feed. Subscribe on Itunes, if you haven’t already.

Finally, let us not forget that despite the fact our lecture series is about arming yourself with the mental tools to avoid being fooled— many times in medicine, we have witnessed real advances. There has been considerable progress made in all specialties in the 20 years I have been in medicine, the 40 years John has been, and the 275 years since Adam blood-let George Washington. Progress is wonderful, but all that glitters is not gold.

Our lecture series aims to help you tease apart the real nuggets from pyrite.

PS: Why am I introducing the series? Only because I drew the short straw and have to edit the videos! If there are errors, these are mine alone.

Beta-blockers in HFrEF is complicated

Today I want to revisit one of my biggest frustrations in cardiology: being forced to use specific beta-blockers in patients with heart failure (due to low EF) and atrial fibrillation (AF) when a) there may be better alternatives and b) there may be no benefit to these drugs in this scenario.

What do you mean beta-blockers may not have benefit in HFrEF? Mandrola, there are 10 trials that show that the drugs carvedilol, metoprolol XL and bisoprolol reduce mortality vs placebo in patients with HFrEF. CIBIS, COPERNICUS and MERIT HF were all landmark trials establishing use of beta-blockers in HFrEF.

My problem stems from the need to treat patients with AF. Does the presence of AF alter the benefit-harm ratio of these drugs? For instance, sometimes I would rather use digoxin to treat AF. But this brings an alert on the computer; and calls from my colleagues that I am not practicing evidence-based medicine.

Yet isn’t it possible that the rhythm may alter the treatment effect of beta-blocker? We give this the awful name treatment effect heterogeneity.

Eleven years ago, a British team led by Professor Dipak Kotecha combined 11 trials (having individual patient data) to assess the benefits of beta-blockers in patients with heart failure who had SR and AF. The Altmetric score of this vitally important meta-analysis is a measly 131. People did not like the results.

The authors benefited from the fact that 17% or 3066 patients in these trials had AF while 76% (13,946 pts) had sinus rhythm. The first finding is that patients with AF had a higher death rate than those in SR (21% vs 16%). They were sicker.

Here are the graphs for mortality (SR on the left; AF on the right):

Patients in SR enjoyed a 27% highly significant reduction in death vs placebo. Patients in AF had no benefit from beta-blockers. It turns out that there is a statistical test for an interaction. As I understand it, if there was no interaction, how surprising would these differences be. The p for interaction in this case was extremely low at 0.002.

Here are the graphs for CV hospitalizations

The pattern is similar. Nearly all the benefit is in patients in SR.

This graph shows that there was not one trial driving the result. The positive effect and size of effect of beta-blockers was similar in all the trials while the lack of effect was also similar. (A is SR and B is AF).

The authors must have had trouble understanding this observation because they also looked at subgroups of patients with AF who may have benefited from beta-blockers. None were found. Here is the graph:

No subgroup of patients with AF seemed to benefit from beta-blockers.

The authors make a clear conclusion from this paper.

The results of our analysis suggest that the substantial benefit identified in patients with sinus rhythm should not be extrapolated to patients with atrial fibrillation.

Comments

I re-up this paper (my post in 2023) because it’s so provocative.

First, for statistical purists it raises the tension between using only the main effect in a trial vs that in subgroups. Trials are powered for the entire group not subgroups. The purist interpretation of a trial, therefore, should always be the main effect.

I understand this concept, but I also understand that it depends on whether you like the overall effect or the subgroup finding.

Take the DANISH trial, for instance. In this RCT of the ICD vs no-ICD in patients with non-ischemic cardiomyopathy, the main results were totally null. Look at those curves. Nada. No more ICDs for this huge group of patients, right?

Electrophysiologists don’t love that result. So most of my colleagues point to this subgroup analysis based on age.

See…the ICD works in young people. And I admit there is a biological plausibility to the fact that young people may get more ICD benefit because there is less competing risks for death, and perhaps fewer complications.

I show you this example because if you turn down the beta-blocker analysis based on it being a subgroup analysis, well, then, you are going to have to stop putting in ICDs in patients with heart failure not due to ischemic causes.

The second tension of this beta-blocker analysis is that it sure seems real. A) the authors had individual patient data, B) there were more than 3000 patients in the AF arm, C) the p for interaction was highly significant, and D) you can tell a biological story that patients with AF may be too sick to benefit from the negative inotrope.

Yet neither the American nor European HF treatment guidelines give much weight to this subgroup analysis. Quality measures in the hospitals surely do not. I can’t help believing this has to do with not liking the finding.

The third tension that has come to light is that we now have two trials that (taken together at least) found benefit from another drug useful in AF—digoxin.

The DIGIT HF trial found benefit from digitoxin vs placebo. And the Dutch DECISION trial found a 19% nonstatisically significant reduction in the composite endpoint of CV death and HF events.

I wonder what would happen if I tried to treat a patient with AF and HFrEF with digoxin instead of carvedilol. It would align with evidence but would surely be shunned by those measuring how well I practiced evidence-based medicine.

Let me know what you think.

Subscribe now

The phenomenon Foy believes is happening— an efficacy-effectiveness gap— runs rampant in cancer medicine. Some new, $200,000 a year therapy eeks out a 1.5 month PFS or OS benefit in a randomized study. Oncology KOLs and the fellows that aspire to be them jump out of their chairs applauding. What few discuss is that the patients in these randomized studies are different than the average patients in our clinic. Patients on trial are younger, often as much as a decade. They have better performance status, which means they're more active around their house. They tend to have fewer comorbidities. Their liver function and kidney function are generally preserved. And they've waited to be enrolled on study without the disease exploding— they may have more indolent biology. Even in this setting, there are toxicities, but we're always reassured that they are manageable. That's a nice word to say when you're not the one suffering.

Some look at these studies and argue the median is not the message. Although the median person has a one-month improvement in survival, look at the tail, there are some people living 18 months or more. Problem I have with this argument is: look at the other tail. The group of people not assigned the magic new drug, assigned the older medicine, sometimes assigned placebo also has a tail. Is your tail that much higher? Are you sure? How many people are still at risk? Is this a statistically robust conclusion?

And there's the other argument. It's entirely possible that somebody does have seven or 8 months of life added by taking the drug, but if the median is 1, wouldn't that mean there's somebody whose life is shortened?

If you survey gamblers coming out of a Vegas casino, you may find that on average, they lose a little bit of money. If you find someone who scored big, then probably is someone who lost big. Of course this is the case with cancer drugs, we have all taken care of people who've experienced a grade 5 adverse event—which is death. The challenge with medicine is that it is difficult to identify the individual counterfactual.

Inevitably, with time, cancer therapies become used in populations not originally studied. Older people, people with comorbidities, people with second cancers, people with diminished kidney function. Almost always the benefits are eroded and not enhanced. The toxicities become more formidable. People experiencing them do not always use the word manageable, instead they replace that word with an expletive.

The heart failure community was quick to criticize Andrew Foy. They continue to pound on the drum of guideline directed therapy— the bizarre idea that it is in the best interest of someone with newly diagnosed heart failure to start on all of the medicines simultaneously, rather than sequentially.

Often in real clinical practice, this poses pragmatic challenges. Some people are overwhelmed by the number of tablets. Others struggle with diminished blood pressure, and there is a limit on how much these drugs can be pushed. A patient ends up on a homeopathic dose of four medicines rather than a disease-modifying dose of one or two.

Remember, the randomized studies supporting approval often do not support diminished doses, and don’t support starting all at the same time.

One argument made in response to Foy is that randomized controlled trials generate on average treatment effects, but we can use covariates to create individualized treatment effects. You can predict someone's relative and absolute risk reductions, which do not reflect the average relative and absolute risk reductions.

We should remember this is a hypothesis in need of empirical verification. The way you would verify it is to run a large randomized study, generate these subgroup specific effect sizes, and then pick a few subgroups and run the randomized study again to see how closely you predict the actual effect. If you don't do this, you are a tremendous risk of overfitting.

You could also split your data set.

Or you could generate estimates for the people on the edges of the study and then test those estimates formally in a pragmatic randomized study of people ineligible for your trial. That would be my preferred option.

But anyway you slice it. I have not seen any of these external verifications and as such it's just a convenient story without scientific backing.

Is the guideline directed therapy paradigm good? Well that depends on for whom. For the manufacturers of the therapies, it is a windfall. How do you sell a drug that is an add-on drug with additive side effects in a marketplace that has many effective, generic options. Creating a mantra that starting them all together is preferable to starting with the tried and true drugs and expanding, if and only if they are tolerated, is a great marketing strategy. But the burden of proof is on he who lays charges. Run a study where starting all of them is compared against starting on an ace or an ARB and cranking that up and then moving from there, logically and sequentially. That has never been done. Instead some cling to observational studies that compare people who can tolerate multiple drugs against those who can tolerate only one or two. But you don't need to be a rocket scientist to know these are different people.

It's true that most medical care is not supported by randomized evidence. It's also true that you don't need randomized evidence to change practice, or approve new therapies and there are many examples of such products. But there is also a juggernaut of people encouraging the use of medical practices that are incredibly costly, of dubious efficacy, and that generate massive profits. If patients had skin in the game and had to use their own money, we would have a slew of third parties evaluating this evidence, skeptically and making prudent recommendations. But because there is little skin in the game, and the money is siphoned off public funds, the only answer is to give more medicine, earlier, in combination, and we have a legion of KOLs happy to endorse that view— evidence not required.

I should ignore the longevity gurus. I should tell myself:

There have always been snake oil salesmen, and this new well-dressed, photogenic, muscled internet type deserves no more attention than the bearded man selling wares from a covered wagon. Sure, they’ve gone from potions and salves to supplements, AI chatbots, and MRIs, but really, they are no different. They will make money off a few people, but they are unimportant.

As much as I try, though, they really get under my skin. In an effort to eradicate these modern chiggers, 5 reasons I can’t stand the longevity gurus.1

1. They are wrong

The longevity gurus have nothing to offer. They pitch early detection, but our most successful and widely adopted early detection strategies, breast cancer and colon cancer screening, have never been shown to improve overall mortality. Our most successful cancer screening, cervical cancer screening, will soon be unnecessary because of hard-won breakthroughs in prevention (the HPV vaccine). New technologies have either failed (cell-free DNA) or are untested and likely to fail (whole-body MRI).

The treatments they foresee remain hypotheses, conjectures, or fairy tales. Whether it is metformin, rapamycin, GLP-1s, nicotinamide, or some made-up adaptogen, we have found no substance that makes an average healthy human live longer. If you haven’t read it yet, the TLDR for Ending Medical Reversal2 might be:

A good story, an animal model, a mention in ancient scripture or indigenous mythology, observational data, or even private equity funding does not tell you that something will work. The only thing that tells you something will work is a well-designed clinical trial.

2. They are inexperienced

Wait, that’s not right; they are woefully inexperienced. I know that the gurus often argue that their outsider status is exactly what enables them to be disrupters (see #3, below). But really, only a tech bro, or residency dropout, or doctor with a practice of 100 healthy, wealthy, and privileged patients can possibly think that a non-specific longevity tonic is possible.

What do I see in my patients in their 9^th or 10^th decade? I see people who have lived well and taken care of themselves, who are failing. When we fix something, we are mostly preventing further decline rather than restoring them to good health. And, each fix usually comes with a complication or another diagnosis. The idea of a single fix is laughable.

3. They are overconfident

I love disruption. I am happy that Uber disrupted the taxi industry. I am happy that Airbnb has given us more options when we travel. I love FedEx. But most disruption falls short. How many of us utilize the Wi-Fi capabilities of our refrigerators?

The history of people searching for the fountain of youth should make us all just smile and walk on by when we hear people pitching longevity. The actual work to improve longevity is long and hard. The true longevity gurus are not those with pretty websites with a “buy now” button but the people who have developed the science and deployment of interventions to prevent and effectively treat diseases that once caused premature death; premature deaths which occurred from infancy through the 70s. The strides these people have made are amazing, and I am optimistic that the trajectory here might continue to improve in the coming years.

If I am honest, I get a little insulted by these people. It is kind of like how I can criticize the US, but if a foreigner does, I get all defensive. I spend a lot of time on Sensible Medicine criticizing medicine, but when someone outside the field tells us that he has discovered the fountain of youth, which medicine has failed to discover, well, then…

4. They are childish

It is completely normal for children to fear death. Most kids experience a period between about five and eight years old when they worry a little, or a whole lot, about death. I remember lying in bed pondering — actually getting completely freaked by — the idea of no longer existing.

But, you know what happens to most people? They grow up. They realize that we all die. They realize that part of the wonder of our time on this planet is its transience. It is naïve to believe that we can meaningfully alter our biology.

5. They are selfish

The state of the planet should be telling us that we need fewer people, not more. Sure, there are many complicated solutions to the world’s problems, but a simple one would be for fewer of us. I want to stick around for as long as possible, but I also would like to leave a magical, livable world to my children. We, and our predecessors, clearly have not thought enough about this.

Conclusion

Medicine has done amazing things and I expect we will continue to make slow steady progress at making peoples lives longer and better. So, maybe this is too harsh. Maybe today’s longevity gurus are not uninformed, inexperienced, overconfident, selfish children. Maybe they are just avaricious charlatans. Part of me hopes this is the case. It is true that their sales pitch is often good, starting with common-sense advice (eat well, get a good night’s sleep, exercise, have friends) and using terms like wellness and healthspan as an entrée to their kookier and more expensive ideas.

But even as a proud capitalist, I have trouble embracing their profit-driven approach because their products can cause real harm. Yes, “a fool and his money are soon parted,” but the longevity gurus give people false hope, and their screening advice drives healthy people into the public system through false positives and overdiagnosis, thus driving up costs for those not tempted by their falsehoods. At the hand of the longevity gurus, a fool risks even the wisest of us being parted from our money.

OK. I feel better. Next week, I will tell you why you won’t have a carbon-neutral flying car in 5 years.

Share

I had just finished writing this essay when I got the July 6/13 issue of the New Yorker

and read this wonderful article: Something Is Very Wrong with Modern Longevity Science. The author, Dhruv Khullar, points out flaws in much of the observational and animal data that support the favorite recommendations of longevity gurus. He takes appropriate aim at the Blue Zone ridiculousness and does a nice job of extending the lessons to current fads. He writes:

One hardly needs a list of blue zones to think of commonsense ways to improve life span. We can invest in biomedical research and enact laws and regulations to reduce pollution, design walkable neighborhoods, establish safe public spaces, create affordable housing, and curb smoking. Such strategies might not trend on TikTok, but they would be more effective than the latest juice cleanse. Unfortunately, the pursuit of longevity—like so much else in American life—is increasingly an individualistic endeavor. We adopt new health fads as though they are forms of personal expression, experimenting with cold plunges and red-light therapy, wearable devices and peptide injections. Though it’s nice to think that striving to be healthy is cool or popular—it wasn’t always—the quest for extending life can become monomaniacal, and we easily forget that one’s life span is not, by and large, a life-style choice. Instead, longevity has always been a communal affair.

The article is a good read, and a bit less rant-y than mine.

This is a reminder that in our zeal to help patients we can do more harm than good by uncritically applying clinical trial results to daily patient care.

The Randomized Aldactone Evaluation Study (RALES) compared spironolactone to placebo in 1,663 patients with heart failure with a left ventricular ejection fraction of </= 35% and a New York Heart Association (NYHA) class of III or IV. Patients were followed for an average of 2 years. Spironolactone reduced death by a whopping 11% (number needed to treat 9) and hospitalizations for cardiac causes by 8%, and there was no statistically significant difference in cases of serious hyperkalemia.

In the words of my two and a half year old son, you could drive a big old truck through the separation in the survival curves.

On a per-patient basis, RALES showed the largest absolute difference in survival for any large-scale clinical trial testing drug therapy in patients with chronic systolic heart failure.

RALES was published in September of 1999. Several years following its publication, Juurlink et al. conducted a population-based time-series analysis to examine trends in the rate of spironolactone prescriptions and the rate of hospitalizations for hyperkalemia before and after the publication of RALES.

The authors linked prescription-claims data and hospital-admission records for more than 1.3 million adults in Ontario, Canada, for the period from 1994 through 2001.

Among patients recently hospitalized for heart failure, the spironolactone prescription rate was 34 per 1,000 patients in 1994 and increased immediately after the publication of RALES, to 149 per 1,000 patients by late 2001.

Over that time, hospitalizations for hyperkalemia rose from 2.4 to 11 per 1,000 patients and the associated mortality rose from 0.3 to 2.0 per 1,000 patients. The publication of RALES was NOT associated with significant decreases in the rates of readmission for heart failure or death.

What gives? How could increased use of a drug that performed so well in a clinical trial not translate to better results in the real world?

To understand this, you must understand the concepts of internal validity and external validity and appreciate the difference between the two.

Internal validity is the extent to which a clinical trial accurately proves that the tested treatment directly caused the observed results. It confirms that the results are due to the intervention itself and not to bias, confounding variables, or measurement errors.

Like most drug trials published in high-impact journals, RALES has high internal validity. I’m confident that if spironolactone is used in patients, like those included in RALES, and the same procedures are applied (dose initiation, up-titration, clinical and laboratory monitoring, etc.), that similar results would be achieved. They wouldn’t necessarily be observable on an individual basis, but they would be detectable at a population level.

But internal validity is not nearly as important as external validity when it comes to clinical translation. External validity is the extent to which the study’s results can be generalized to the broader, “real-world” population of patients with the clinical condition of interest. It determines whether the treatment’s success will hold true when applied to everyday medical practice involving the “target population.”

Who was the target population in RALES? Fundamentally speaking, it was patients with systolic heart failure. If you drill down a little further, the average age was 65. In the Ontario population studied by Juurlink et al., the average age was 78 and 10% resided in a long-term care facility. This is a massive difference.

Remember this: patients in clinical trials tend to represent the most robust, ambulatory, educated, and highly motivated of those in the broader target population. This is true whether explicit inclusion/exclusion criteria are placed on these conditions or not.

Then there is the matter of trial procedures. Monitoring conditions in clinical trials is difficult to replicate in the real world.

In RALES, follow-up evaluations and laboratory measurements were conducted every 4 weeks for the first 12 weeks, then every 3 months for up to 1 year and every 6 months thereafter. Additional clinical laboratory tests were also performed at weeks 1 and 5. That’s a lot of monitoring—and it’s hard to do this in regular clinical practice.

Clear differences in age structure between RALES and the Ontario population studied by Juurlink, which will necessarily encompass significant differences in illness severity, frailty, multimorbidity, polypharmacy, etc., and differences in procedures likely drove the unexpected findings by Juurlink. It is a classic cautionary tale but one that is not appreciated nearly enough these days.

In my own teaching, I have developed a set of rules for applying clinical trial results to patient care:

1. When you know your patient would not be included in an RCT, proceed cautiously (if at all)

   1. Advanced age, frailty, multimorbidity, or meets explicit exclusion criteria for trial

2. When you are unsure if your patient would be included, proceed cautiously (consider reducing the treatment effect by half and doubling the risk of harms and think, “Does this still seem like an attractive option?”)

3. When you are confident your patient would be included in an RCT, attempt to replicate RCT procedures (i.e., dose initiation, up-titration schedule, safety monitoring, etc.)

   1. If you cannot, proceed cautiously

   2. When a run-in phase is employed, start your patient on the medicine like they are in a run-in phase, and if they do poorly, stop

4. Pay attention to entry conditions in clinical trials.

5. Appreciate that certain patients are not included in RCTs because investigators know they WILL NOT be able to show benefit – so why would you try it on them?

   ---

Editor’s note: This may be the most important post of the year. I could not agree more strongly with Andrew. For many reasons, not least quality measures, clinical translation of trials is massively misunderstood in the real-world. We thank Andrew for this incredibly important piece. JMM

Subscribe now

Three takes on the issue.

Read more

Synthesizing basic medical science, empirical evidence, and the complexities of an individual’s physiology, values, and goals grants medicine its status as both a science and an art, practiced in real time and bespoke for each patient. Yet, despite incredible strides in medical science, medicine remains rife with ambiguities. While fundamental facts are typically stable, science changes. For example, when I was a medical student, I was told that the old way to treat atrial fibrillation had been to control the abnormal rhythm: get everyone back into sinus. But the AFFIRM trial in 2002 challenged this: it turned out that just controlling the rate worked about as well—and avoided the sometimes nasty side-effects of antiarrhythmics. But then, in real time, I watched practice change. Now it appears that controlling the rhythm early with an ablation is often superior.

The scientific paradigm shift is the process through which theories are built, tested, and replaced. Ideally, the good is replaced by the better. But too often, an established practice is found to be pointless or harmful and undergoes reversal. The ongoing proliferation of medical research threatens the levees we’ve built around what we think we know. Making matters worse, a significant portion of scientific research isn’t replicable. Whether due to poor design, incomplete data, or faulty analysis, many manuscripts are relegated to the far corners of PubMed, collecting dust rather than citations. And even high-quality research is not without its controversies: for a paradigmatic example, look no further than the PARADIGM trial.

In the face of this data tsunami, there has been a proliferation of clinical practice guidelines. But there’s something lurking behind the crisp, multicolor tables and lists of citations. The danger lies precisely with their greatest asset: simplification. The full complexity of a clinical question is reduced by a guideline to a recommendation. This is a godsend, but it carries with it a risk: confidence.

Clinical guidelines provide an easy out but disincentivize both research into the topic and scrutiny of how the guidelines themselves were created. By collapsing the jumble of research into an easily interpreted format, the underlying uncertainty and inherent contingency of that research are lost. Decisions of the guideline committee are likewise obscured by the final product. It is through this laundering of uncertainty that guidelines risk disrupting the judicious, patient-centered practice of medicine. In the balance of medical decision-making, strong medical justifications often overshadow unique patient factors. Artificially augmenting the confidence in a recommendation risks alienating care from the individual patient in the pursuit of benefit to the average patient.

Like clinical guidelines, UpToDate provides fast answers to complex questions. By providing articles detailing the diagnosis and treatment of numerous ailments, it has all but supplanted textbooks and the search of the primary literature. What is often forgotten is that UpToDate articles are written by individuals. They are useful not because they are the unerring sum of medical knowledge, but because they are convenient and well presented.

Large language model-based systems have recently joined clinical practice guidelines and UpToDate. Whether driven by unrealistic efficiency expectations in our increasingly corporatized medical systems or by individual physician choice, LLM use has grown rapidly. OpenEvidence, in particular, is popular thanks to its ability to answer complex questions with citations—often to clinical guidelines. At their best, these tools can provide a jumping-off point for deeper reading. But there is an incentive to read the provided answer and stop. Contributing to this bias towards cognitive closure is LLMs’ proclivity for zealous confidence. It remains unclear whether this central tendency toward agreement and obsequiousness can be reversed by software engineers (let alone tolerated by market forces).

There are other concerns with LLMs. Consider the possibility of a literature search revealing a morass of contradictory evidence, while the language model frames a version that appears authoritative and complete. It is not that the information is wrong, but how our view of it is shifted.

Clinical guidelines, UpToDate, and AI are powerful resources. They can distill and present complex medical knowledge into a usable form. It is impossible for physicians to have the knowledge of a medical researcher in every domain in which they practice. Even the best textbook will lag behind the most recent papers; even the most recent paper will lag behind the preprint. My concern is that an over-reliance on these consolidated sources risks trading the bitter medicine of ambiguous truth for the hollow satisfaction of false confidence. I fear that where once there may have been a question, only certainty will remain.

Combatting the trend towards algorithmic medicine should start with medical education. Evaluating scientific literature is a skill, and like any skill, it is learned through practice. Curricula should focus on equipping student doctors with the tools they need to reason clinically: scientific knowledge and a critical eye for research. Basic medical science provides the foundation, while deft literature appraisal allows for lifelong knowledge-building.

Unfortunately, it is becoming increasingly popular for medical schools to shorten their course of study. This is usually justified with reference to the often slow-paced final 6-8 months of undergraduate medical education. And yet many new curricula include cuts to pre-clinical coursework. While more medical knowledge is generated each year, medical students are being given less time to learn it. Even if unnecessary material is removed, introducing a curriculum on evidence appraisal is likely to be net neutral. The study of medical science and the evaluation of medical evidence must be given adequate time, rather than being pushed into evening lectures during busy clinical rotations.

For residents and practicing physicians, solutions are less clear. I suspect that concerns of litigation steer many physicians towards a greater reliance on guidelines. It’s hard to argue a physician acted inappropriately if they followed the recommendations. And even if the guidelines later turned out to be wrong, who could blame them?

What separates modern medicine from superstition is the recognition of what we do not know. Admitting uncertainty builds trust and lends weight to our claims of knowledge. We owe it to our patients and to ourselves to grapple with ambiguity and reject the false comfort of confidence. The alternative is that we become mere confidence men.

Share

Jordan Schneider, MD, is a resident in Internal Medicine at NYU in New York. The perspectives written above are solely his own and do not reflect the views, policies, or positions of the NYU Grossman School of Medicine, NYU Langone Health, NYC Health+Hospitals, the Veterans Administration, or any affiliated organization.

Photo Credit: Annie Spratt

There is a reflex in emergency medicine and cardiology that goes something like this: patient arrives in atrial fibrillation, the monitor confirms it, and a cardioversion, an electric shock applied externally to the heart, makes it disappear. Problem solved. Everyone goes home happy.

The reflex is intuitive. It is also, it turns out, frequently unnecessary.

At Cardiology Trials Substack, we recently reviewed RACE 7 ACWAS - an important trial that has not received the attention it deserves.

RACE 7 ACWAS, published in the New England Journal of Medicine in 2019, randomized 437 patients presenting to Dutch hospitals with stable, symptomatic, recent-onset AF of less than 36 hours duration to either immediate cardioversion or a wait-and-see approach - rate control, discharge, and cardioversion only if the arrhythmia had not resolved within 48 hours.

The primary endpoint was simple: presence of sinus rhythm at four weeks.

The result: 91% of the delayed group were in sinus rhythm at four weeks versus 94% in the early cardioversion group. A difference of −2.9 percentage points that was within the non-inferiority margin the authors specified. Restoring sinus rhythm quickly does not buy durable rhythm stability. It changes the route, not the destination.

But the primary endpoint is not the main story, in the wait-and-see group, 69% of patients converted spontaneously within 48 hours on rate-control medication alone. Only 28% required a cardioversion.

Meanwhile, in the early cardioversion group, 16% converted on their own before the cardioversion could be performed. And 20% of otherwise eligible patients never made it into the trial at all because they had already returned to sinus rhythm in the waiting room. Read those numbers again. Most patients with recent-onset AF will convert without any intervention beyond rate control.

Cardioversion is not a minor procedural inconvenience. It requires sedation, monitoring, physician time, and carries a small but real risk of complications. It is not nothing.

What the trial cannot tell us: RACE 7 ACWAS was not powered for hard outcomes such as stroke, death, or hospitalization. The small event numbers in both arms are reassuring but not conclusive. The population is also worth noting. Of 3,706 screened patients, 2,581 (70% ) were ineligible, most often because the AF had been going on longer than 36 hours or there was a history of persistent atrial fibrillation. RACE 7 trial results apply to the stable, uncomplicated, early-presenting patient. Not the unstable patient, not the patient with atrial fibrillation of unknown duration, not the patient in whom rate control fails.

Why this still matters in 2026: The trial was published in 2019. The findings are arguably more relevant now than they were then.

Wearables and digital devices are detecting AF in patients who may be asymptomatic or minimally symptomatic, prompting emergency visits for arrhythmias that might have gone unnoticed, and untreated, without consequence a decade ago. The pressure to do something when a monitor shows AF is considerable. RACE 7 ACWAS is a useful antidote to that pressure.

For many patients with new AF, the wait-and-see approach is not a passive strategy. It is an active decision to avoid a procedure most patients don’t need, backed by a randomized trial.

Similar Story with Premature Ventricular Complexes (PVC)

The same principle extends beyond atrial fibrillation. In patients with frequent premature ventricular contractions, a study found that 44 of 100 patients with a median PVC burden of 18.4% experienced spontaneous reduction to below 1% at a median of 15.4 months - without any suppressive therapy.

And this is not unique to cardiology. Most people do not present to a physician when they develop flu-like symptoms, abdominal pain, or diarrhea, and for many, waiting a few days at home is entirely appropriate.

Good medicine is not defined by how much is done, but by the wisdom to know when action is warranted and when restraint is the better prescription. The two are equally important.

Subscribe now

Over my now long career, I have observed behaviors among my patients that are not associated with better blood sugar levels or blood pressure, or even with weight loss or quality of life; they are associated with changes in overall mortality — the only endpoint Drs. Prasad and Mandrola think is important. Changing these behaviors would therefore save lives.

I know the traditional approach would be to test my so-called hypotheses in an observational trial, using statistical analyses to control for confounding. Then, if the associations were supported, I would design a robust, multicenter, double-blind, placebo-controlled RCT to demonstrate the effectiveness of the proposals I outline below. But, come on, I am a doctor, a professor of medicine, with two Ivy League schools on my CV.2 How could I possibly be wrong?

Here are four simple interventions that will reduce mortality among our patients. These need to be instituted immediately. They include changes to our practice, our medical centers, and our medical record system.

1. Disconnect all landlines

Patients who have landlines listed in their contact information have higher mortality rates than those who do not. This association even has a dose-response curve; those whose landline is their preferred line, or their only number, have the highest mortality. Disconnecting all of Ma Bell’s phones will save lives.

2. Make valet parking at medical centers free for all

Some of my patients walk to visits with me, some ride public transportation, some ride Pace Paratransit. Others drive themselves; they park on the street, in the medical center garage, or they use the valet parking service. When you exclude people who walk to their visits, those who use the valet service have a statistically and clinically significant lower mortality rate than street and garage parkers.

If we enable everyone to use valet parking, the mortality benefits will be immediate.

3. Add preferred pharmacies to patients’ medical records

Most of my patients have one or two pharmacies listed in the electronic health system. They usually list the medical center pharmacy and one close to home. Depending on their insurance, some list a local pharmacy and a mail-order one. A few patients have long lists of pharmacies, with locations such as Martha’s Vineyard, Lake Tahoe, Santa Fe, and Scottsdale. Patients with this diverse array of pharmacies live decades longer than those who only use Chicago pharmacies. Add a smattering of pharmacies located in American Blue Zones, and we will bend the mortality curve.3

4. Never remove “general healthcare maintenance” from the problem list

When I precept the residents, I am a stickler for cleaning up medication lists and problem lists. “Make them clean and accurate so they help you care for your patients better and more efficiently,” I say again and again. I have the residents remove duplicate diagnoses and ones that are out of date.

One thing that I caution them never to remove is the “Screening” or “General Healthcare Maintenance” problems under which people track due dates for things like colonoscopies, mammograms, and bone density studies. My experience has been that whenever this problem is deleted, patient demise soon follows.

I also oppose deprescribing statins or any drug on the Beers’ list. These actions always lead to patients exiting this mortal coil.

Share

Rest assured, I will continue to observe associations between my patients' behaviors and their outcomes, and I expect to routinely update this list of recommendations. Recently, I have noticed that patients who pay for the Galleri® Multi-Cancer Test or full-body MRIs have decreased mortality. I might start recommending these, but because these people usually have multiple pharmacies listed in their medical record (see #3), I am not sure if these are independent, causative factors. Yet!

I hope this article can help.

To keep it interactive— lets start with a series of polls. In each scenario vote for who needs the disease expert.

From Psychiatry:

Patient 1 - a 7 year old kid says they feel anxious, and you are deciding whether or not they would benefit from starting an SSRI

Patient 2 - a 47 year old has stable bipolar disorder on Lithium for 15 years, but increasing urination

Patient 1 - A 57 year old had a stent placed two weeks ago for an occluded right coronary artery, and now presents with worsening shortness of breath— worse than before the procedure and it started suddenly.

Patient 2 - A 27 year old has LDL of 105 and normal weight, blood pressure and no other risk factors and is considering getting a coronary CT, lp(a) or starting a statin.

Patient 1 - A 57 year old has pancreas cancer and is on FOLFIRINOX. Feels weak after 3 months, and is wondering if he should go on hospice.

Patient 2 - A 57 year old woman (no symptoms) wondering if she should get a mammogram

Patient 1 - 42 year old wondering if he should get a fall covid booster?

Patient 2 - 48 year old with HIV on Biktarvy, with diarrhea that is not getting better, wondering if he needs more loperamide or should switch therapies?

It should be clear but deciding if a 7 year old should start on SSRIs, if a 27 year old needs a statin for cv risk, if a 57 year old should be screened for breast cancer, and if a 42 year old needs a annual respiratory virus shot— are data expert questions and not disease expert questions. What is the data that these interventions work? and are they worth it?

In contrast, how to manage lithium toxicity, the complications of a recent stent, whether a cancer therapy should be halted, or how to manage the side effects of HIV therapy —benefit from someone in the field.

The first category of questions have to do with whether, on average, for people in these situations, there is net benefit from a course of action. You can appeal to data, and if there is not great data, then one must appeal to one’s sense of risk and benefit. Do you think it is better or not to take a statin for 50 years without clear proof of benefit? Do you think a mammogram is worth it, given lack of all cause mortality benefit, and real risk of overdiagnosis, but possibility of reduction in cause specific mortality? No disease expert has any advantage over a thoughtful reader and thinker. In fact, they may be at disadvantage.

The latter category of questions are those that require modifying a planned course of action. There is, often, no gold standard evidence on when and how to modify, and seeking the guidance of someone who has rules of thumb may be of value. Even here, patient preference has a role, and even here, I would be careful not to assume the disease specific expert is correct. At the same time, it is nearly impossible to formulate an opinion in these situations based solely on reading. A data expert has no advantage.

A critic will say I am wrong. A child psychiatrist has a wealth of experience giving sertraline to 7 year olds, and knows when to give it and when not to give it. Furthermore, they know when to recommend CBT. What do you know?— they will shout. A data-expert can read this study, but has no idea how much anxiety is normal and when treatment is warranted.

Only the psychiatrist knows best, they will say. Oh really? 8 billion people are alive today, and some estimate that 120 billion people have ever lived. Of these 1.5% were and are 7 year olds— say a billion 7 year olds have ever existed. Researchers say 10% of kids may have anxiety. That might have been higher in times of war or famine. Let’s say maybe there have been 100 million 7 year olds with anxiety throughout human history. I believe that is probably a conservative estimate. What percentage have been seen by psychiatrists? 0.000000…1%.

And what do psychiatrists have to offer? In a short term study (inadequately short), there are tiny effect sizes from some course(s) of action, and a whopping placebo effect. What could a disease expert know about whether or not to give a child an SSRI based on their clinical experience when the effect sizes are microscopic, and no (good) effort has been made to see what are the long term consequences of starting a child on such a medication, and potentially taking it for 40 years (more on that below). Instead, in such a wicked learning environment the doctor may have learned superstitious behaviors. Akin to Skinner’s experiment giving rewards at random. A wicked learning environment is a situation where you cannot get ahead by experience— this is often the case in medicine when effect sizes are only visible from space.

Instead when it comes to a 7 year old that may or not be anxious, the dominant factor here is whether the parents believe this anxiety is something within normal limits, or something they will grow out of, or whether they believe 7 year olds should take psychiatric meds (I suspect most don’t agree with this). For 99.999999999999% of parents in human existence the solution was clear.

You may wonder why I have picked this extreme example. Why a 7 year old? I didn’t pick it! A JAMA expert did. Explained here.

Vinay Prasad's Observations and Thoughts

Should you take an SSRI for 40+ years?

Listen now

22 days ago · 79 likes · 10 comments · Vinay Prasad

A cardiologist knows more about lp(a) than you do— the critic will argue. That is true, but none of that has any bearing on whether or not a healthy 27 year old should take a statin for 50+ years. Someone who works in a movie theater knows more about where a soda was spilled last week—but that doesn’t help you decide if you want to the produce the Odyssey film.

I often see doctors on social media post videos saying “as a surgical oncologist, I know you should get a colonoscopy at 45…” That is as logical as saying “as a whitewater rafting guide, you should definitely drink a peaty scotch.” “as an interventional cardiologist, the right age to start zetia is in your twenties.” is akin to “as a circus performer, I always recommend 22 inch tires on your Tesla.”

Life is short, art is long, opportunity fleeting, experiment treacherous, judgment difficult. Knowing when you need a disease and data expert is knowing when you need experience and when you need experiment— judging which is which is difficult. On SM, we consider these matters carefully.

John Mandrola: I am torn on the PolyPill concept

Is Universal Testing for Lp(a) Warranted? I Have Questions.

I recently received a distressed email from an internist colleague and friend who was upset by the results of a blood test (lightly edited to anonymize):

“I slept poorly last night. [My PCP] wrote me a note saying that I had to tell all my first-degree relatives to get checked for Lp(a) and that I should have coronary calcium scoring. It weighs heavily on me that I could have passed this on to my kids. Because there is nothing to do about it, I feel like it’s my death sentence.”

I wrote back to better understand the cause of such intense feelings. Turns out that my colleague had expressed concern to her doctor about her family history. Her father had an aortic valve replacement for calcific aortic stenosis in his early 70s and recently died from prosthetic valve degeneration. She also had 2 other relatives who had clinical atherosclerosis and ischemic strokes in their later years. She wanted to better understand her cardiovascular risk and how she could mitigate it. So, her doctor ordered an Lp(a) test. It came back “High” (scary highlight original):

I had been hearing about Lp(a) as a risk factor for atherosclerotic cardiovascular disease (ASCVD) and calcific aortic stenosis (AS) for several years. This association had long been known and more recently clarified by several large database studies showing a correlation with cardiovascular risk. The mechanism by which Lp(a) causes increased risk remains incompletely understood. The talks usually ended with excited anticipation of trials of new pharmaceutical products which lower Lp(a) - none of which have yet been completed or reported.

So why is Lp(a) measurement being promoted now when there is no specific treatment for it? I turned to the recently published 2026 ACC-AHA Dyslipidemia Guidelines for some answers:

A cardiologist colleague also provided this nice infographic:

But this information just led to more questions and concerns:

1. What is the benefit of a test without a clear plan of action based on the test result? Such testing may just cause non-productive anxiety and distress. Labeling patients with a new disease (? Lp(a)-itis) has psychological consequences, especially if there is no specific treatment for it. The guidelines suggest using Lp(a) as a “risk enhancer” to select patients for more intensive lipid-lowering with current drugs but provide no evidence that this application of the test improves patients’ health or outcomes net of the cost and harms. Interestingly, the most commonly recommended and prescribed lipid-lowering drugs, statins, have little effect on Lp(a) levels.

2. The universal testing recommendation is made outside of any Bayesian framework. A “40% increased risk” can sound scary to someone worried about their cardiovascular health. But if that person’s baseline 10-year absolute clinical risk of an ASCVD event is only 2% (the case with my colleague), then a “40% increased risk” translates to an absolute risk increase of less than 0.1% per year – clinically meaningless. Pre-test probability matters. Absolute risk matters. This is clinical reasoning 101. For most patients with low baseline risk, reinforcing Life’s Essential Eight lifestyle prevention may be more healthful and empowering than pointless lab testing.

3. The universal recommendation for measuring Lp(a) creates an impression (a false impression, I hope) that the medical profession has been enlisted to prepare a market for new pharmaceutical products which have yet to show any net clinical benefits for actual patients. If the trials of these products fail, what happens to the class I testing recommendation? Will it need to be revised? This would raise questions about why it was made in the first place and undermine public confidence in other guideline recommendations.

4. There have been many markers of risk for all kinds of diseases, conditions, and outcomes that have emerged from database research and mechanistic studies. Only a small percentage make their way into clinical practice. In my field of EP, they have included microvolt T-wave alternans, baroreflex sensitivity, and heart rate variability. Most fail when put to the ultimate challenge: Does the test improve healthspan or lifespan net of costs and adverse effects? There are many tests which predict sudden death risk, but few that have met the latter challenge of clinical utility.

In our medical training, we are programmed to rely upon testing. Medical students and interns learn the ritual of “morning labs” on every hospitalized patient. Repeated checks of potassium and magnesium levels on every cardiac patient getting diuretics (remember those nursing calls at 2AM for K of 3.4?). Lots of X-rays, CTs, and MRIs. When in doubt (and medical trainees have lots of doubts), do more tests.

Academic specialists should also be sensitive to the burdens placed upon primary care clinicians. When academic specialists make a new universal screening recommendation, someone else (usually) has to fulfill it. This is the classic unfunded mandate. Primary care specialists are already inundated with hundreds of screening and other preventive measures they are encouraged to perform and interpret for large numbers of patients, in addition to managing their acute and chronic medical problems. Patients are also bringing them results from genetic tests, wearable device data, and whole-body scans the primary care doctor never asked for. Little wonder that there is so much burnout in the field. Academic specialists should therefore ensure that new broad screening recommendations imposed upon primary care clinicians and the public are fully warranted and evidence-based.

So, what am I going to do with Lp(a)? Based on my experience and review of the data so far, I will not be rushing to test all my patients for Lp(a). I will look forward to critically examining the clinical trial results of Lp(a) lowering drugs. Does this make me a bad doctor, or a sensible one? I will let the reader decide. What do you think?

Share

Dr. Joseph Marine is an arrhythmia cardiologist who practices in Baltimore, Maryland.

Implantable cardiac defibrillators (ICDs) can save lives if a person goes into a shockable rhythm (VT or VF), which would otherwise progress to impaired perfusion and death (termed sudden cardiac death, SCD), but instead, the ICD shocks them back to sinus rhythm.

Some models can also terminate VT by administering rapid ventricular pacing to abort the problematic rhythm — so-called anti-tachycardic pacing (ATP).

Currently, most ICDs are placed when a patient’s left ventricular ejection fraction is less than 30-35% (after 3 months of optimal medical therapy). Other less common indications include hypertrophic cardiomyopathy and inherited arrhythmia syndromes such as long QT.

Key Problem: Despite three decades of placing ICDs, many people still die of SCD without getting an ICD, and most people who get an ICD die without it ever doing anything — it never goes off.

There are then 2 important research questions: How can we identify groups who may benefit from ICDs (who don’t currently receive them), and how can we identify people who currently receive ICDs but don’t need them?

In a new Nature paper, five authors claim to have identified a new ECG risk marker that may do the former.

At the outset, let us be clear: If you claim to have found a new group that may benefit from ICD placement, the definitive way to prove it is to randomize participants to ICD placement or no intervention and show an all-cause mortality benefit.

The authors have not yet done that, though they say they will in an ongoing study. Importantly, the control arm of such a study should not be banned from receiving an ICD, but should be permitted to do so the moment they have an indication (low EF, inducible VT, etc).

In the absence of an RCT, one can try to construct a chain of causality. The authors attempt to do this; here, we outline 5 considerations that warrant further exploration.

1. “Sudden death” occurs for many reasons (massive pulmonary embolism, aortic rupture, brain bleeding, etc). You need to identify people with sudden cardiac death due to an arrhythmia that the ICD can stop. There are only two: rapid VT or VF and asystole (bradycardia). The former is far more common. The authors do this by relying on death certificate data. But being coded as a sudden cardiac death on a death certificate does not mean it was a sudden cardiac death that could have been stopped with an ICD. For instance, the most common cause of VF is ST elevation MI. Here the actual cause of death would is coronary occlusion—the ICD may shock the VF but if the coronary artery is not opened, death occurs. The authors anticipate this objection and performed a complex analysis linking their prediction method to either sudden death or just VT/VF without SCD in a Swedish dataset. It was a nice attempt but even they admit serious limitations to their methods.

2. You have to be careful to train the AI tool on naive EKGs and not on the byproducts of medical interventions aimed at someone whom the doctor is worried about the risk of sudden cardiac death. The Nature paper does not have pure, naive EKGs. As an example, one of the early attempts at AI famously trained the algorithm to recognize abnormal blood smears. Slides from around the office were used. The algorithm learned that dust on the coverslip was a predictor of normal blood smears, as abnormal slides received more views from residents. Similarly, you do not want to train the algorithm for SCD on EKG markers that occur if you already have an ICD in place.

   In the paper, 7% of patients who died of sudden cardiac death (by death certificate) died with an ICD in place. ICDs were present in 5% of people at the time of baseline EKG. It is possible that AI is learning to recognize some artifact of ICD placement — pacing, for instance, or perhaps creation of the ICD pocket alters the EKG in a way we do not fully appreciate — by making providers place the leads in slightly different spots. Pacing is definitely visible on the EKG with nearly every beat. It is not clear that the Nature paper controls for pacing and post-pacing ECG changes — they are not mentioned.

3. You need to know that these rhythms would have been shockable (ATP, a shock, or shock series will restore sinus rhythm). Not all SCD is shockable. Not all VT and VF can be successfully shocked. Cardiac amyloidosis can result in cardiomyopathy, with risk of SCD, but it is also difficult to cardiovert. Dead meat, don’t beat, cardiologists say. The authors of the Nature paper may well identify a group of people prone to VT, but that is not sufficient. They have to have an arrhythmia that can be successfully cardioverted. To overcome this objection, the authors rely on an observational study showing that those at risk of SCD who received ICDs had lower mortality than a theoretically modeled control group — but, like all observational studies, this is at risk of serious residual confounding and guarantee time bias.

4. The people you identify as needing an ICD cannot have a high competing cause of death. Patients need to die of cardiovertible SCD, but not for lots of other reasons. For example, some ICD studies have shown that ICDs don’t improve survival in folks over the age of 75 with low EF — or in patients on dialysis, partly because competing risk is so high.

   The coefficients in the paper show that the authors are finding a new group at much higher risk of competing death than heart failure with reduced ejection fraction. While low EF pts are roughly 3X at risk of all death vs SCD, the new algorithm is roughly 5x the risk of all-cause death. This is a huge problem because competing risk will kill — no pun intended — the potential for your intervention to save lives with the ICD.

Competing risk issues are raised further by looking at a Figure on page 5 of the supplement:

The strikingly high total mortality in the “ECG high-risk group” (25% at 3 months) suggests that the high-risk group contained many subjects who had their ECG taken in the context of a serious/life-threatening illness or imminent death. Curves that separate instantly in retrospective studies often indicate that you are looking at two entirely different groups. No one needs an AI model to tell you that a critically ill patient is more likely to die in the next year than a healthy clinic patient getting a checkup. And no cardiologist would be keen on placing a ten-thousand-dollar device in a person with a high competing risk of death. Clinical context matters.

5. People identified as dying of SCD with the new method can’t later develop a standard indication for an ICD before SCD. In other words, if they were going to die on Friday from VT, and AI found they needed a device on Monday, but the doctor was going to find they needed a device on Tuesday, you didn’t help them. They were still going to get the device before Friday. Just like the control arm of the validation study is not — no ICD, not ever — but ICD when you otherwise develop an indication. It is not clear whether the Nature paper considers this. The algorithm is only of value if it finds a person who will have shockable SCD before they have another reason to have an ICD placed.

Ultimately, the concerns in our analysis are neither damning nor exculpatory. The only way to settle the debate is a prospective randomized trial. Take patients who do not have an indication for ICD, screen them with the new AI tool, and randomize them to ICD now vs ICD, if and only if it is needed in the course of usual care. The endpoint must be overall survival. If we were betting, the high risk of competing death, the possibility that some artifact is being learned, and the chance that patients at high risk may later qualify for an ICD are all reasons to be skeptical that this approach will change practice.

We agree with the authors who write, “many promising predictors in the past have failed to identify patients who benefit from defibrillators.” Failed predictors include microvolt T-wave alternans, baroreflex sensitivity, heart rate variability, and heart rate turbulence. Each of these had a reasonable physiological basis and abundant preclinical evidence suggesting that they would be clinically useful risk predictors, but, at some point in development, failed to show any value in predicting clinical benefit from an ICD.

Modern medicine has made the ICD selection method even harder: it is now clear that sudden death mortality in common conditions like heart failure is declining. A recent paper from Switzerland has also confirmed substantial decreases in the incidence of appropriate ICD therapies over the past two decades. A declining arrhythmia incidence reduces the proportion of people who can benefit from an ICD. In fact, a new trial in Europe called PROFID is retesting the positive results of the 25-year-old MADIT II trial.

There are a number of other questions that should be addressed before even considering exposing patients to the risk of such an ICD trial:

1. Is the new ECG biomarker stable over time in a given patient?

2. Is it reproducible from ECG to ECG from one day to the next?

3. Is it affected by metabolic, acute, illness, position, or other changes?

4. Does it predict risk independently of other known variables such as other cardiac diagnoses, age, heart failure status, other comorbidities, etc?

5. Does it predict risk in stable ambulatory patients and in subjects with no other clinical indication for an ECG?

6. Is the model accurate in people with bundle branch block, paced QRS or widened QRS for other reasons?

7. Is the model affected by lead placement or ECG system used to make the recording?

Two final points:

1. We note that a Washington Post reporter has already posted that the paper is “amazing.” While we would hope a reporter at a major newspaper would have enough in-house expertise to sort out limitations of a medical paper, his enthusiasm is not that unusual—even amongst doctors.

2. While we are eager for the results of a randomized trial, we note that many machine learning or AI-suggested algorithms still lack such verification, despite many years having elapsed since the claim of “discovery.”

In short, the new proposed ECG biomarker approach generates a lot of buzz, but there are open conceptual issues with the paper. Newspapers and physicians accepting the results of science that they either did not or cannot vet is problematic.

One vexing problem with ICDs is that the vast majority of patients implanted with an ICD never use it. They incur all the harms (inappropriate shocks and infection), but none of the benefits. We wonder if an ECG model could predict patients unlikely to benefit from the ICD. Such an algorithm could be used to design an RCT at the time of ICD generator change—the most risky time for patients with the device.

Subscribe now

I met Mrs. S, a 99-year-old woman with a history of an irregular heartbeat, under dire circumstances. She was an Indian indentured immigrant who had spent most of her life toiling in the cane fields. Despite her age and frailty, there was a steely determination in her eyes.

She had developed an arterial clot in her leg but flatly refused hospitalization or surgery. “I’m too old for that,” she said firmly, her voice a mix of resignation and resolve. My initial reaction was one of frustration. I saw a clear medical solution to her problem, but she was unwilling to take it.

Her home was next to my office. Over the next few days, I focused on managing her pain and tried to convince her to be admitted. Despite my best efforts, she remained resolute. Her leg, deprived of proper blood flow, began to turn black and necrotic.

Her living conditions were humble, a stark contrast to the sterile, high-tech environment of my office. As I entered, the familiar scent of mangoes filled the air, mingling oddly with the sharp odour of gangrene.

Mrs. S lay on a simple bed, her leg now a mottled, lifeless mass. Despite her condition, she greeted me with a weak smile. “Doctor, you came,” she whispered, her voice barely audible over the hum of a nearby fan. I sat by her side. We talked. Not about her medical condition, but about her life, her struggles, and her memories. She told me stories of her youth and the back-breaking work in the cane fields. Her husband was an excellent man but died at the age of 50 from a heart attack. Her only daughter was battling cancer at hospital.

As I was about to leave, she reached under her pillow and pulled out a small, worn purse. “I want to pay you,” she insisted, her hands trembling as she counted out a few crumpled bills. I protested, telling her it was unnecessary, but she was adamant. “You mad or what? You are my doctor!” It was a matter of pride and dignity for her.

As I stood to leave, she held my hand and whispered, “We finally have a good doctor here.” Those words hit me harder than any criticism or praise I had received during my training. In that moment, I understood the true essence of being a doctor. It wasn’t just about diagnosing and treating -- it was about listening, understanding, and connecting with patients.

Before I left, a neighbour handed me a basket of mangoes. “This is from her tree. She asked me to pick these for you, doctor,” she said, her eyes twinkling with gratitude.

Mrs. S passed away a few days later. Her death left a void, but her words and the lessons she imparted stayed with me. She taught me that medical knowledge, while important, is just one part of the equation. Compassion, empathy, and respect for a patient’s wishes are equally vital.

The village came together for her funeral, a solemn yet poignant ceremony held by the river. I attended, standing among her family and friends as they paid their respects. The air was thick with the scent of incense and the sound of prayers. Her body, wrapped in a simple white cloth, was placed on a traditional funeral pyre. As the flames began to consume the wood, casting flickering shadows, I felt a sense of closure. This was not just a farewell to a patient but to a teacher who had imparted invaluable lessons on the human spirit and the true art of healing.

Share

Dr. Joel David Teelucksingh is a Consultant in Internal Medicine, Endocrinology and Diabetes. He is a Lecturer in Adult Medicine at the University of the West Indies

When I was writing my section of our opinion post on whether you can trust your doctor, I reread the oath and, not surprisingly, came away feeling that Dr. Lasagna’s 60-year-old oath needed an update. Of course, much of what it means to be a doctor does not change. Even the original Greek oath has sections that we would all feel are important:

Into whatsoever houses I enter, I will enter to help the sick, and I will abstain from all intentional wrong-doing and harm…

And whatsoever I shall see or hear in the course of my profession, as well as outside my profession in my intercourse with men, if it be what should not be published abroad, I will never divulge, holding such things to be holy secrets.

But so much of the field has changed. In 1964, we could not have imagined the treatments we employ, the information patients have access to, and the conflicts of interest in medicine today.

Here is a Hippocratic Oath for 2026. It is not a final version. I am hoping we can workshop it as Andrew Weir did with The Martian. Please suggest additions, subtractions, and rewrites. I’ll incorporate the suggestions and post something we can hope schools will adopt.

I swear to fulfill, to the best of my ability and judgment, this covenant:

I will apply the hard-won scientific gains of medicine for the benefit of the sick, avoiding those twin traps of overtreatment and therapeutic nihilism. I will always consider the ultimate consequences of my actions and think in advance about how test results will change my management.

I will remember that there is art to medicine as well as science, and that warmth, sympathy, and empathy may outperform the surgeon’s knife or the chemist’s drug. In addition to my extensive and hard-won medical expertise, common sense should never be discounted.

I will never be ashamed to say, “I know not”, nor will I fail to call in colleagues when the skills of another are needed. I will work with affiliated healthcare professionals, as they often have expertise that doctors lack.

I recognize that successful treatment often requires temporary harm. I will never cause harm without first having honest discussions with my patients.

I will remember that I always have more to offer a person in poor health than one who is healthy.

The medical decisions I make will be guided by my expertise and my patients’ values. I will rigorously appraise the information that supports my recommendations, whether it comes from articles, guidelines, or decision support tools. I will take responsibility for the consequences of my decisions.

I will respect the privacy of my patients. Their problems are disclosed to me only so that I may care for them.

I will remember that I do not treat a fever chart, a test result, or a cancerous growth, but a sick human being, whose illness may affect the person’s family and finances. My responsibility includes treating the person as a whole. I will endeavor to treat all patients equally.

I shall never deprive my patients of hope, nor will I dispense false hope.

I accept that being a doctor requires sacrifice, including the effort and time to learn and stay abreast of current practice and the need to be available to my patients.

I will always act to preserve the finest traditions of my calling, and may I long experience the joy of healing those who seek my help.

Share

Adam Cifu

Created in 1955 by Andrew Garfield, the impact factor (IF) is one measure – albeit an important one – of the quality of a journal. It is a relatively simple calculation, with numerous nuances. Calculated by Clarivate – there is a numerator – all citations to every article published in a journal over a 2-year period – divided by what Clarivate considers “citable” articles. Citable articles are research reports and long reviews. Brief article types, under 2-3 pages, with a limited number of references, are generally not considered citable. Specifically, editorials, research letters, perspectives, viewpoints, and other brief articles do not count in the denominator, but the citations they accumulate do count in the numerator.

Released every June, the IFs are already somewhat out of date. The newest IFs, released in June, are the 2025 IFs. They reflect data from 2023 and 2024. Recent impact factors offer insights into many of the nuances of this measure.

The 2025 IF of NEJM was 85 (51,607/611), whereas the IF for Lancet was higher at 109 (51,655/474). But as you can see from the denominator, Lancet has been very careful to “control” its denominator. It publishes far fewer “citable articles” than does NEJM. This has been true for many years. But that is not the entire story. Of the top 10 articles in Lancet, 5 were Global Burden of Disease papers, and 4 were worldwide estimates of various diseases. These 9 articles accounted for 24% (12,117/51,655) of the citations! Of the top 10 cited papers in NEJM, 8 were RCTs, and the other 2 were translational science papers, accounting for 10% (5386/51607) of citations. This highlights important differences in the journals. NEJM publishes what many would consider traditional research, RCTs, and other scientific reports, whereas Lancet publishes more epidemiological studies. From my standpoint, RCTs change practice, whereas epidemiological studies are more likely to influence health policy.

A similar difference is seen in the IF of the leading general cardiology journals, EHJ, Circulation, JAAC, and JAMA Cardiology. The 2025 IFs were 45 (22986/507), 41 (21872/529), 22 (13,095/588), and 15 (3470/228), respectfully. But again, a more detailed examination highlights differences in content. Of the 10 most cited papers in EHJ, 6 were guidelines, reflecting how often the European Heart Society produces them. Circulation published 3 guidelines and JAAC 2. The top 10 cited articles in EHJ accounted for 28% (6434/22986) of the citations! Although it should be noted that 2 of the top 10 papers in Circulation were reports from the American Heart Association. JACC published 3 consensus pathways from the ACC. JAMA Cardiology had no guidelines or society reports in its top 10 cited articles. Are guidelines, consensus statements, and reports from professional societies research? These publications certainly impact clinical care, but many would not consider them “traditional” research.

The major fully open-access journals offer more nuances. BMC (formerly Biomed Central), which began publishing in 2000, has an IF of 8.7 (9287/1062), and PLOS One, which began publishing in 2006, has an IF of 2.8 (88,549/31,104). These are among the oldest fully OA journals and represent this “new” publishing model. PLOS Medicine, which began publishing in 2004, has an IF of 9.9 (2838/287), and is meant to be more selective than PLOS One. BMJ Open debuted in 2011 and has an IF of 2.5 (19,449/7909), while JAMA Network Open, which debuted in 2018, has an IF of 11.7 (39,580/3391). These titles are representative of the thousands of OA journals that come in all shapes and sizes. In general, as fully OA journals publish more manuscripts, which drives their revenue, it is difficult to maintain a high IF – the exception has been JAMA Network Open, which, despite a relatively large denominator, has a healthy IF.

There are several other nuances that bear discussion.

• Should a journal release an important article in November, when it will count for fewer months in the IF than if it were released in January?

• Are journals adding more short article types – hoping to add citations from articles that don’t count in the denominator?

• Is there article creep, short article types that are longer with more references, but still not “citable?”

• Should journals add “big” epidemiology studies, knowing that they are well cited?

• Do journals focus on their denominator, “citable articles,” and at launch of a new journal keep it very limited, or if an established journal, over time, shrink the denominator, hoping to eliminate those articles that are not well cited?

Comparing journals from different fields is risky. Cardiology and oncology journals have higher IFs than journals in other fields simply because there are more journals in those fields – hence, more citations. Should editors of general medical journals give preference to those subject areas, knowing that those articles are likely to accumulate more citations?

The IF is not perfect – it can be manipulated – by “controlling” the denominator, by publishing certain types of articles that are known to be well cited, or by publishing more articles that might be cited, opinion pieces, or research letters, but do not count in the denominator. The IF does not reflect the quality of an individual article or the research work of an author. Yet, it has stood the test of time – and all the other journal metrics – CiteScore, Scimago Lab – have their own challenges. For example, both Lancet (92.4) and NEJM (74.8) have very high CiteScores, which is the number of citations accumulated over 4 years divided by all published reports. JAMA does not (26). This is in part because the CiteScore counts all authored reports in a journal, and JAMA has a very robust, popular authored news section, which adversely affects its CiteScore.

When selecting a journal to publish in authors should consider not only the IF, but whether their article is in scope for the journal (hint, read the journal before submitting to it!), the services provided (think turnaround time and quality of the editing), its audience, including print circulation and reach in social media, and increasingly whether the journal offers open access, which is often mandated by funders. In addition, cost is likely to become an increasingly important part of the discussion. With open-access fees exceeding $10,000 US for some journals, these will simply be out of reach for many investigators.

Share

Howard Bauchner, MD, is a Professor of Pediatrics and Public Health at the Boston University Chobanian & Avedisian School of Medicine. He is also a Visiting Scholar at the National University of Singapore. Dr. Bauchner is also the former Editor in Chief of JAMA and the JAMA Network and former Editor in Chief of Archives of Disease in Childhood.

Read more

Using multiple cameras and digital 3-d reconstruction, offsides can be called when as little as 1 mm of a player has advanced past the defender. Offsides has long been a rule in soccer, but these days the enforcement is rigid.

Some soccer fans don’t like the new 3-d enforcement. They argue that this technology has changed the spirit of the game. Specifically, it has removed the “attacking nature” of soccer. Because players can be disqualified for the smallest offsides position, they are more tame, reluctant to leap at a scoring opportunity.

Offsides is a simple, unchanging rule, but it is different when it is performed in real time with flags, when it is reviewed on videotape, and when it is determined by automated digital reconstruction systems. It raises the question of whether or not all offsides is the same? Can a difference in degree become a difference in kind?

Of course the analogy could not be more apt for medicine. We all know what a stroke is, a cerebral perfusion defect and symptoms. But what is a stroke in the era of extremely sensitive MRI machines? Sometimes patients are flagged for having a stroke who have no neurologic deficits.

We all know what a blood clot in the leg is, but is it different if you present a painful calf versus if you're asymptomatic and it is found on screening ultrasound?

Is it still the same afib if you find it with your apple watch? In one trial, 14 day screening found more afib, lead to more anticoagulation, but there was no difference in stroke. Apple watch trials are ongoing.

And with cancer, is all metastatic disease the same? We have widespread stage migration.

Even the endpoint in our clinical studies is not exactly what we see in the clinic. Every doctor knows what shingles is. It's bad. The endpoint of the randomized controlled trial of Shingrix is not exactly shingles. It is detection of the virus after this happens:

Is there a difference between shingles when the patient, untrained in Shingles, notices something is wrong and seeks the care of the doctor versus shingles when the patient is educated to be vigilant for shingles, keeps a diary about shingles, and contacts the doctor for anything that may or may not be shingles with a hair trigger?

Is there a difference between being offsides when it's visible to the sidelines referee and the flag is in the air, and semi-automatic offsides detection technology that finds your elbow is past the defender?

Soccer purists say it changes the attacking spirit, and similarly, using diagnostic categories that bend with technology may alter the therapeutic benefit of interventions— finding disease that doesn’t need treatment— or creating a false perception of the benefit of a therapy— exaggerating the value of a preventive by averting pauci-symptomatic disease that would not have been diagnosed without a course in detection.

The idea that technology can change a rule, not by definition, but by the margin of detection, is a lesson both soccer fans and doctors should appreciate.