What about football?

Hell no.

Skiing?

Too much ACL risk

Cycling?

Too much crash risk

What sports are acceptable?

Tennis, golf, fencing

I think my colleague is not unique, and he captures the growing epidemic of poor risk perception running rampant these days among academics and other white collar parents.

Allow me to defend the proposition that kids should be free to pursue any and all sports that they enjoy.

Let me start with a few obvious points.

First, we are not talking about the Roman Colosseum. No one is fighting a tiger, and we don't have death matches. This essay is about the most common sports in America— all of which are largely tamed.

Second, comparing tennis to football— soccer to fencing— is laughable. Individual sports— tennis, fencing, golf, cycling, swimming (ps these can be team but each contribution is measured separately) and team sports— football, soccer, basketball, hockey— are completely different experiences.

Both can be fun, but team sports have a camaraderie and shared purpose and build skills and community in a way individual sports lack. On a tennis team, there are winners and losers, on a soccer team, you are in it together. That has meaning that transcends health and medicine. I like both categories of sports, but let’s be honest about the differences.

Second, looking at the risks and benefits of playing a career in the NBA, NFL, MLB and extrapolating to your kids is…. it is really hard for me to pick a polite word… misguided.

Professional athletes tend to live longer and better than the average joe— after all, they are rich and fit (in their youth after all); comparing NFL players to replacement players, or MLB players, etc is also misguided. Replacement players differ from NFL players in many ways besides playing pro-ball. How often do they go to the clubs to party? (I joke, but how do you separate playing football from the lifestyle of an NFL player)

Moreover, extrapolating from the NBA to your kid is the height of arrogance. Your kid is not going to play for 20 seasons in the pros. He or she might play a few years in school before giving up. The attrition for playing sports is colossal. It makes surgical residency look like a high retention activity.

If playing elementary school, middle school, high school, college and 10 years in the NFL causes xx, yy, or zz (PS I am not sure we have good data on even this question)— that tells you nothing about your kid playing a few seasons until the age of 14 or 16 or 20. It would be like saying that if staying awake 22 hours a day is bad, staying awake 16 hours is also bad. Uhhh, wait.

It’s plausible that being tackled is bad for the brain. If a lot of very hard tackling is bad, even a little tackling may be bad— some believe on faith. But why stop there? Brains should not be jostled. Don’t ever jump off 3' stairs. Don’t change direction too rapidly. Don’t ride a rollercoaster or drive on winding roads. Don’t floor your Tesla plaid. Just wear a c-collar, and use the buttons on your adjustable bed. Obviously, this is extreme, but no one knows how much brain jostling is good and when it becomes bad.

Soccer is also an odd comparison for concussion. Heading the ball— what the risk averse say not to do — is not the most common cause of “concussion”. Running into the other player is.

There is no evidence that playing tackle football for a few years in one’s youth has any long term consequences. Some claim this study proves it does. If you read this study and you think that it proves that, the brain that should be examined is your own.

The study looks at people who donated their brains to research— it is enriched with pro players. We can stop right there. It is a convenience sample. You won’t be able to learn anything. The sample size is abysmally small. Most patients have CTE, per the authors. They claim that starting tackle football earlier is linked to having symptoms earlier. Though it isn't linked to worse CTE. Here is their main result

Why don’t you just run a line of best fit through the northern sky? Actually, here is a better graphic that captures the “science” deployed here:

The only way to study this issue is to look for a natural experiment. Where a school closed a sports program for instance. E.g Year 1 soccer and football were both options, Year 2, only soccer was an option, football team lost funding/ or didn’t have a coach, etc. Some football players might do soccer instead, and some kids who would otherwise have done football might be sedentary instead. Some players who would have made the soccer team are now cut, they should also be considered. RDD design would test the question of whether closing the football program, improves or worsens health. Another natural experiment would be if players are cut at random. The problem is that almost never happens. They're cut based on ability.

My prediction if you pooled all such natural experiments— you would find no effect on any outcome or that reducing sport options is generally a bad thing. Mostly for social reasons and weight gain.

Third, all activities have some unwanted effects. Wrestlers can get ringworm. Tennis and golf players can get tendinitis. Cyclists sit on their urethra all day. Fencers have to see the disappointment in people’s eyes when they say they are fencers.

Fourth, the challenges that children face these days are finding joy and community-- being resilient and open-minded— avoiding loneliness and depression. Organized sports are much more likely to have a favorable impact on all of these actual risks, then any theoretical risks on brain function. Reducing the number of sports options may result in some or even most kids picking a different sport, but a few kids will be pushed into doing no sport at all. Sitting around is terrible for all sorts of social and health outcomes.

Fifth, if you disagree with my argument, show me a single paper that can demonstrate that limiting a child's choice to tennis, golf and fencing leads to any better long-term outcome than expanding the choice to all sports. Find any study and please link it below. I will read it and I will show you why it is no good in a future post.

Finally, the substack is supported by people like you. Because I want to show my appreciation to the subscribers, I'm going to make one last point available to those people. Thanks to all those who subscribe.

Subscribe now

Read more

Astute readers of VP’s Observations and thoughts will know that I took this issue to task last month— as implausible, unrealistic and poor quality research. Now we have a larger update.

Here is what you need to know. The entire analysis hinges on the fact that people born beyond a certain date are much more likely to be vaccinated than people born before that date according the way vaccines were deployed in England (and Wales). That creates a natural experiment. We don’t expect people just before and just after the cutoff to differ dramatically— so it is almost randomized. In these cases, we can look for a regression discontinuity- a jump in the graph— to see the effect of our policy.

First, lets look at the effect on shingles. This is the 8 year cumulative risk of shingles, before and after the program launch

It drops by 0.12 percentage points— which means 833 people post cutoff date avoid a case of shingles over 8 years due to the program. We don’t know exact rate of vaccination. But it is good to consider the effectiveness of the program itself— and magnitude thereof. Fair to say, it is visible and small.

Now let us look at dementia.

Wow, dementia rates are plumetting— but the vaccine does literally nothing. There is no step at all.

Now they look at dementia with a 12 month grace period (vaccines can’t work super fast), and tighter window around the potential RDD— and again, totally null

Not also the secular trend— it was and is getting better over time.

Finally the authors go to great pains to recapulate the Wales results and to see if their analytic choices can nullify the results. They find differences in magnitude of effect, but not direction. That leaves the mystery open. The data are 18x smaller, and the entire effect may be a persistent artifact.

Finally, there are two randomized trials— these and these ALONE— will be the definitive verdict on the topic.

What is the overall lesson?

I personally put little stock in both (a) amateur internet analyses of this topic and (b) the media and academic literature on this topic. In response to excessive and unjustified skepticism from the former group, we are witnessing unjustified and excessive cheerleading from the latter group. As such, the public understanding is a disaster. Adam Cifu reminded me last weekend that my bias is that (b) annoys me more than (a) for the simple reason— they should know better. As always, he nailed it.

All medications are harmful, some are beneficial, and some have benefits greater than harms. If the latter is true, it typically applies under careful conditions and settings. Believing that shingles vaccines could lower the rate of dementia by massive amounts (20%, 30%, etc) in short time frames (6 months) is completely irrational— totally bonkers— and yet this is the stated position of many academic researchers.

You are also generous people, you take the rent money and redistribute it. If someone is an artist or musician, you lower their rent, and charge more to the plumbers and electricians. Art, after all, is something we all can appreciate and you want to encourage artists.

Every 4 years you and your spouse take turns deciding how much to subsidize the artists. You’re slightly more sympathetic to the engineers so the subsidy to art is less than your spouse, who’s much more sympathetic. So every few years, the subsidy changes by a few percentage points.

Read more

The media is quick to sell the punch line— Up to 5 cups may be good for you!—, but what is missing is any serious reflection on just how poor the data are.

The vast majority of the data presented in the document are observational studies. These studies- for a common nutritional exposure and a common health outcome—are useless. People who drink coffee and those who don’t, are different, and your coffee patterns change as you get sick and unwell, if you are hospitalized etc. You might as well read tarot cards if you want to rely on observational nutritional studies.

Let us look at the randomized trials. Here is the single best one. If you have afib and have ever drank a cup of coffee, and you get cardioverted, it is better to drink coffee for afib than abstain.

This is a really good paper, but let us not forget there are still open questions.

1. What happens after 180 days

2. In the long run, do you have MORE stroke. You see— if more afib then more DOAC. If more DOAC than less stroke. Ironically, coffee may be good for afib, but bad for stroke bc it lulls docs into false sense of confidence, and leads to less DOAC. I don’t know this to be true, but it is not yet ruled in or out by data.

3. In the long run, do you have a overall better QoL.

But still— a very good paper because the mantra was coffee is bad here.

Now lets look at the next set of RCTs (many of these refs are RCT or meta-analysis of them)

I have no doubt that caffeine can cause short term BP increases. But are these BAD or are they GOOD? If I am going to go on a long bike ride, hiking yosemite, lifting heavy weights, playing soccer or basketball or tennis, I find caffeine to enhance my exercise. So I can push myself harder, which later results in greater fitness. Perhaps even also lower long term CV risk. Many fitness experts share this view— but none of these studies even seek to evaluate this. It would be so easy:

Randomize participants to coffee before exercise for a month, or no coffee, and then measure VO2 max, resting heart rate, etc. Even then, we are still using surrogates. Yet, just like the coffee and afib paper— we may dispel misguided prejudices.

Here is another lamppost below which we search for our keys

Again, this means nothing. Short term, abysmally tiny RCTs measuring glucose levels. There are similarly uninformative studies for lipids. Whether coffee improves or decreases mortality, or diabetes, or any hard endpoints is unknown.

Yet, the document concludes strongly— mixing up the weak to non-existent randomized data with the uninformative observational data.

The conclusions that are powerful here rely only on the unreliable studies.

Here is how I would have written the concluding paragraph.

If you have afib, have ever drank coffee, and get cardioverted, you will have less afib if you keep drinking than if you abstain. There are some small trials assessing coffees effect on short term biomarkers that are uninteresting. Beyond this, we have no idea if drinking 1,2,3,4,5,6,7,8,9,10,11, cups of coffee is optimal for health— let alone cardiovascular health. We don’t know if 6 cups vs 3 increases the rate of stroke or heart failure, if practiced for years. We could make a claim here using UK biobank data— but that would be lying. Better to be honest and say we don’t know. The fact we know so little about this topic— despite widespread public interest— should be a call to arms to launch a research agenda. In the meantime, you won’t find guidance in this guidance document. Finally, we urge a mortatorium on public funding for large observational studies which are incapable of answering these questions. Those funds could be better spent on an RCT agenda.

That is the right conclusion. Let me know your thoughts below.

Dr. John took to task the Watchman device. We discussed IVC filters. A half dozen chemotherapy drugs came up. Paxlovid, removing the primary tumor in metastatic disease, cancer screening, bone marrow transplant for myeloma, MRIs for dubious purposes, etc. etc. A laundry list of things that don't work.

Our conversation reminded me of the core challenge in health care. We have some interventions of indisputable benefit, some that are preference sensitive and an ocean of things that don't work or have harms exceeding benefits— and our system pays for everything, all the same. If anything, the last category puts food on the table.

Different people can agree or disagree on whether health care is a human right— that society should be taxed to pay for. But no one can possibly think people have a right to health care that doesn't work. No one can possibly believe that we should tax people to pay for IVC filters and watchman's that do more harm than good. Yet that is precisely what we are doing.

It is clear that federal and state governments are not capable of paying for things that work, let alone prioritizing high value medical services. It is clear that doctors can't do it either. Patients are often confused about what to believe. Vendors only care about their sales. Hospital administrators are trying to put the profit in non profit.

In such an environment, where we cannot stop interventions that don't work, where spotting real scientists from snake oil salesmen is a game of Where's Waldo, the best thing we can do is to stop subsidizing health care.

Health care, as it is practiced in 2026 America, is not a human right. It's a tax penality on our ignorance and wishful thinking.

1. If lung cancer screening DID NOT improve mortality but DID make surgeries smaller, led to less chest tubes, led to less ports, led to less chemotherapy— that would be worthwhile, and is a reason to screen.

This is true, but the problem is there is no evidence that lung cancer screening does any of these things. If you think there is evidence for this, please link to it below. The authors of NLST and Nelson have access to (or could obtain) the FULL data on downstream procedures and have never reported them. Some reports include only the FIRST downstream procedure. I would support screening if it could improve these endpoints but there is no evidence it does. There is in fact no paper on the topic in lung cancer. The lung cancer trials don’t tell us how many people got chemo or surgery or how big the surgeries were etc. Yet, Welch did look at this for breast cancer. At least for breast cancer the truth was counterintuitive. Here is SEER data.

If breast cancer screening means finding small tumors that were going to get big, and sparing you a larger surgery or chemotherapy, you would expect the rise in small tumors being diagnosed to be offset by a commensurate and delayed reduction in large tumors. Yet this is not the case, there is no reduction in large tumors at all. This suggests mammograms find way more small tumors and does nothing to stop denovo metastatic disease (other Welch paper) or large tumors. With CT screening for lung cancer, there is no evidence to date. I am open to being persuaded, and the best way to do that is to email the RCT authors and ask them to publish this. Why wouldn't they, if it's so wonderful?

2. Lung cancer specific mortality is a better endpoint than all cause mortality, some argue. I disagree, and a simple story will illustrate why. A 74 yo gets screened for lung cancer and they find a small mass. A CT guided needle biopsy results in pneumothrorax, and unfortunately the patient codes then and there. He can not be resuscitated. These events— or other complications, such as patient dies after EBUS or VATS are NOT COMMON— read that again, they are NOT COMMON— but they are not zero. Lung cancer specific mortality does not count these events at all. There is no attempt to quantify the harms of screening on all downstream deaths in any cancer RCT — besides… of course… reliance on all cause mortality, which automatically captures all events. The fact the trials are stone cold negative for all cause mortality should be disclosed to patients.

When I was in fellowship, there was a poor patient with Hodgkin’s lymphoma whom the attending blasted with chemotherapy— extremely high doses. The drugs eventually caused liver failure, and the patient died. The attending said, “at least he didn’t die of lymphoma.” I am not persuaded by such arguments.

Consider the current cutting edge…

As you see the patient, AI transcribes the note, and pulls up suggestions from Open Evidence recommending which tests to order…

That’s just not imaginative or disruptive enough. Why does the patient need to see you? Why is a visit needed?

Let’s try again.

When you are home and feeling ill, you ask Alexa for health advice, and start listing your symptoms. Alexa recommends a Chest CT, and schedules it nearby. You drive to the center, and complete the scan. By the time you are in the parking lot, AI has already read the image and phoned in a prescription to the nearby pharmacy, which you pick up on the way home.

Nope. Still trapped in a mindset. Why do you have to enunciate your symptoms? And also, why are you driving the car? Rookie mistake.

Try again.

You sign up for a new health care service. When you launch it, you do an virtual interview intake and the service deduces that you are the kind the person to blow of your health. You aren’t a complainer. AI asks you (in front of your wife) if you would consent to allow it to optimize your health in any non-invasive way it see fit without bothering or notifying you— as long as the interventions were non-invasive. After all, if you are involved you would only procrastinate. And after all, it is just collecting information. You agree. AI asks one follow up. In order to improve health for all people, do you consent to AI running some tests or interventions on you— this is only for societal good— without your explicit consent each time. Your wife looks at you… Ok, you consent again.

One day, you notice your car drives itself home via a different route. When you get home you wonder about the detour. Turns out the car drove you through a CT scanner because it noticed your movement and behavior was different, and it wanted to work things up. Awaiting you, next to your meal, is a prescription tablet to take. The active ingredient was discovered by AI, and the confirmatory trial was run by AI.

“That’s not possible,” you shout, “imaging modalities can’t see through the automobile’s frame.” Of course, AI laughs, it solved that years before—during the boom of midjourney.

Technology’s full impact on health care would mean that every aspect of how we seek and provide care is changed.

• You don’t track your health, your health is tracked.

• You are given care, you don’t decide to seek care.

• Data is collected about your health in ways you are oblivious to (with your upfront consent, just once of course)

• Treatments are studied on you and others around you— because you are a good person and consent to that, and this happens if and only if, AI can’t figure it out using observational methods (strangely AI admits that this is often the case).

• You consent to receive treatments in quiet moments, and you may waive your right to know you are receiving these treatments

• For medical interventions that require trade-offs, you can describe your philosophy and values/preferences or you can ask AI to deduce that about you in the course of it’s monitoring, and then you can waive your individual decision making.

Practically what does this mean for technology makers and investors?

Instead of developing clinician decision support, develop methods to monitor people in their homes/ in the course of their day to day lives. Surely, AI will soon be able to predict when someone may be headed towards illness before they think to ask about symptoms— if you can track their steps, gait, patterns of wakefulness, heart rate, blood pressure (someone needs to make a reliable tracker of this— that doesn’t go around the arm), sweat patterns, skin conductivity, skin temperature, armpit temperature, stomach sounds, bowel patterns, bladder patterns, food consumption, weight, height, visceral fat estimate, body fat estimate, etc.

Instead of investing in imaging that requires a participant to walk into a center, what sort of biological data can be generated from monitoring our trail around the house? Invest in collecting data unobtrusively.

In the Human Stain, Roth wrote,

“We leave a stain, we leave a trail, we leave our imprint. Impurity, cruelty, abuse, error, excrement, semen - there’s no other way to be here.*”

How much of the human stain we leave around the our house can be assayed and tracked.

If AI technology is as disruptive as other advances — electric vs steam engines, computers vs slide rules, combustion engines vs horsepower, etc— we need to reconceptualize all of our assumptions about health delivery.

A patient noticing something wrong (or their wife noticing it), nagging them to see a doctor, the doctor blowing it off, the patient’s wife nagging again, the next doctor taking it seriously, the diagnostic test performed, the therapy prescribed— that’s our current model. A model limited by the fact that the information about the person whose health we are trying to optimize is filtered through so many fallible people, and requires cumbersome ways to obtain more objective information. AI’s greatest disruption will be to eliminate the human barriers from this equation and take the information silently and unobtrusively, and obtain more, silently and unobtrusively, and then learn how to course correct… silently and unobtrusively.

Will it succeed? Will we find new pitfalls? I am only sure the answer to the second question is yes.

A final point, which I may elaborate on in the future. Every single intervention in medicine that actually drives gains in life expectance has come from technologic advancement. Sorry health systems and equity researchers have done next to nothing. Technology that is fiercely debated— where some love and others hate— is almost universally garbage that takes us nearly nowhere. The technology that no one argues with— including drugs and devices— that is where gains come from. If AI becomes one of those technologies, it won’t be by patching our broken, inefficient system, but by slowly and systematically replacing all the imperfect actors. It will be a transformative technology in the way PPIs eliminated surgery for ulcers.

PS: In the future, I will write more about the enduring and important role physicians will always have— don’t be depressed my doctor friends.

*Yes, I am aware, Roth was using this to mean something else entirely.

At times, they seem to understand the perils of overdiagnosis.

Then it is surprising to read a very uncritical article about CT screening for lung cancer. The title tells us the scan saved his life! The conclusion has already been reached, dear reader. Nothing to see here. Lives saved. Why aren’t you already in the scanner?

But what precisely happened? A 69 year old former smoker underwent CT screening for lung cancer in 2021. It found a stage 1 cancer, leading to surgery….

A stage 1 cancer, and, then a lot of scans, and another stage 1 cancer. And now he is on a TKI? for 3 years?

Only osimertinib and alectinib are approved in stage Ib, and osi only has data for 3cm + tumors and alec only 4cm and larger. And even here the data is dubious. The gains in the trials are from the higher stages.

Given how frequent the surveillance scans were, I highly doubt the patient has a tumor so large. This case— light on detail— really smells like inappropriate adjuvant therapy. I would love the reporters to have asked questions as to whether this treatment— which has side effects is appropriate. How large is his tumor? What study justfies adjuvant therapy? What has he been told the benefit is?

But the bigger question:

Does finding two stage 1 lung cancers mean his life was saved? It is completely unknown. What would the natural history of these tumors be otherwise? We don’t know. What is known for sure is that the patient has had 2 surgeries (minimun) and is on a cancer pill for 3 years.

What do we know about lung cancer screening?

We know there is no all cause mortality benefit. Here is NLST

And here is Nelson— the two trials mentioned in the NYtimes article.

Yet, this is not mentioned in the article. If you don’t increase overall survival in a randomized study, how can you say lives are saved?

Finally, I have no issue with the NYTimes running an essay arguing that many doctors believe CT screening for lung cancer saves lives, and we ought to expand those efforts, if the essay also acknowledged: however there is disagreement. The key randomized trials don’t show all cause mortality benefits. Some experts questioned whether a stage 1 patient should be on a TKI for 3 years— as the toxicity may outweigh the benefit. Moreover, those randomized trials did a bad job of post-protocol care for the control arm.

Instead, journalism is uncritical, and woefully inadequate for the public. If you want a different perspective, read only of my most popular articles from Sensible Medicine

Sensible Medicine

"I just want to work on my car"

I always tell junior faculty that your contract means little, and, indeed, I found that to be the case, when a mid-career faculty departed and I inherited a third of his lung cancer patients. Yet, like most unexpected clinical changes in my career, I ended up learning unexpected truths…

Read more

3 years ago · 529 likes · 32 comments · Vinay Prasad

And if you believe in what this substack does, consider subscribing

Subscribe now

Wow. Not: wash your veggies and fruit well. Not: buy whole lettuce and toss the outer layers and wash the rest well. Not: don't buy bagged salad. But: don't eat fresh fruits and veggies at all!

Something has happened inside public health and ID. It is no longer possible to have sensible, proportionate responses. It's just extremes. Stop eating fruit!!! Stop breathing!! Don't leave the house!!

The expert hasn't considered the downsides of this course of action. If a million people stop eating fresh produce for a week, some of them may continue to reduce their produce intake in the weeks and months to come. The impact of that on cardiovascular disease and diet far outweighs any theoretical risk of infection. The infectious disease doctor's advice is likely net harmful for America.

There is a lesson here: experts and the media have no credibility. Their advice doesn't pass the common sense test.

Side note: rainier cherries are so good this season, I will take any risks.

Today, I learned, that we should all get a:

CAC Scan

a baseline echo

a regular visit with your doctor

eat fatty fish

When a famous young film actor died of colon cancer in his 40s, the medical establishment changed colon cancer screening to start at age 45 (not 50). Some said we should begin even earlier: perhaps 40 or 35 or 30.

It is the most natural thing that after some thing bad happens, our instinctive reaction is ask what can we do to avoid it.

Don’t fly ryan air.

Don’t fly into Reagan airport

Don’t go on cruise ships (Well that one is good advice ;).

Don’t eat at chipotle

Of course, when it comes to flying in to Reagan or eating at Chipotle— it is easy for people to use their common sense. Of course, Reagan is a busy airport and I am happy to concede the risk of a crash is higher there than Dulles, but goddamn that Dulles traffic is bad. You know what.. I will take the risk to save 90 mins.

And of course, Chipotle has been in a rough spot, but they have cleaned up their entire supply lines, and their food is actually real food and not ultra-processed garbage (other fast food) maybe it is time to return.

But when it comes to medical advice, it is hard to be critical. To stop for a minute and ask if the recommendation is sound.

What happens if everyone gets a CAC scan?

What happens if everyone gets an echo?

Does seeing a doctor regularly improve outcomes?

Should I eat more fatty fish?

I won’t bore you with meta-analyses and systematic reviews, and a sea of confounded observational studies, but it is fair to say the evidence for all of these claims is nonexistent to weak.

If you are having tearing chest pain, then yes, please go to the emergency room, and before they push heparin, perhaps they should rule out dissection, but if you are a 68 year old watching the news, and otherwise feel fine, should you get a :

CAC scan - there is no evidence you will live longer

Echo - no evidence you will live longer

See a doctor regularly (you should check your BP (you can do it anywhere), but there is no evidence routine health visits help beyond this.)

Should you eat more fatty fish? There is absolutely no good evidence for this.

And should you get a colonscopy at 45?

Actually, you should have gotten it at time of birth; 14 at the latest.

I don’t fault lay people for trying to understand a medical tragedy by finding something to grab on to— some intervention they can make in their own life to escape their own fear of mortality— but it is problematic to see doctors leap to unfounded conclusions.

It is always tragic when anyone dies suddenly, and it is natural to try to make sense of it, but sometimes a person can get good medical care, excellent medical care, have well controlled hypertension, and still die of dissection. And if you prescribed echos for all people who fit that description, you will do far more harm than good. That’s the hard truth about life: we all die, and technology is not always able to save us. Even if we used it daily and sync it to our watch.

If both the branded medicine and the compounded medicine were the same price, and equally easy to get, I imagine most people would take the branded product. But given that the compounded medicine is cheaper, and, as the new letter suggests, easier to get, then naturally many will go that direction.

Even drugs that aren’t (yet) approved are available from the compounder. Stat recently speculated that Trump himself got retatrutide via a compassionate use from Lilly.

The White House press officer denied this, and mocked the reporter for running journalistic speculation. Yet, what was lost in the narrative is that perhaps as many as 10k, or 100k or even more people have gotten retatrutide from the compounders. Including some doctors self experimenting, and documenting for youtube.

CBS news is taking the opposite tact. Too many, not too few people are taking reta.

Obviously, there are massive financial interests that would want to sell these drugs for $1000 a month. Compounding pharmacies threaten this. Is the fear about compounders geniune or it is hysterical— principally serving entrenched financial interests?

In order to elevate these discussions, here are questions, you the reader should ask when you see news like this:

Ok, this JAMA paper is critical of how easy it is get to get these drugs from online pharmacies, but have they shown actual harm? Do they report the rate of adverse events and severe outcomes when compounders are used in stead of branded drugs? Do they factor in the gains from more people taking the product (who otherwise would not)? Is the hypothetical patient in the study benefitted or harmed by the prescription? The compounder didn’t ask for photo id— ok sure, but how often do people give their GLP 1 to someone else (This isn’t vicodin)? And if it so easy to get, why would that happen? What precisely is the concern here? In the absence of demonstrating these things, should they chill out?

For CBS, do you have the rates of adverse events from bodega reta vs clinical trial reta? Are they higher? If so, by, how much?

Instead of data, authors are relying on emotional arguments: of course branded drugs are better than those from a compounding pharmacy. Who would want a filthy bodega drug.

But this is a distraction. Instead ask them to quantify their feelings. How much better? Which adverse events are worse? Are all compounders the same? Is there a trade-off with access? And finally, are you sure your coverage and research is working in the best interests of the American people or the best interests of companies that stand to make or lose a trillion dollars?

FWIW, my personal 2 c is the heart rate increases from reta are concerning, and even though it is an insane fat loss drug, that should be monitored. Even if the heart rate increase is ultimately harmful, the huge weight loss may outweigh that and it may still be, on balance, net beneficial.

In the meantime, remember to not be persuaded by emotional arguments.

Substack, Youtube, Instagram, LinkedIn , ITunes and Twitter. If you want to get all my creative content, it it best to follow me on each of these platforms. A brief explanation of each one:

On YouTube, you will find long form videos— even if I have written on the topic before (on Substack), these videos have no script (and no planning), I often add extemporaneous comments, and occasionally say exactly what I think as I would tell my friends in real life. Some videos— like my Longevity series— don’t have a written version— so Youtube is the only place to learn about these topics. Also, imo, videos capture something that text misses.

Itunes will still have podcasts— when I remember to upload them.

(If you are in the SF/bay area and edit short videos and do podcast stuff, reach out, I need help. I also want to make a weekly summary letter.)

On Twitter, I release raw videos, and occasionally videos that I don’t think are suited for YouTube— largely because they are overly technical and don’t suit the audience.

On Instagram, I will put short video clips and occasional some other images or pictures.

On LinkedIn, you will find work related updates

Finally, you can find me in three places on Substack. Substack is the best platform because the creator has total control over our connection.

1. Vinay Prasad’s Observations and Thoughts — right here. This stack exists because there are some weeks I feel like I have a million ideas, and there is not enough page space anywhere else. I also feel like this is a place where I can write a book review— about Ferrante’s series— and also take apart a poorly done, recent dementia observational study, and you, the readers, indulge me.

2. The Drug Development letter. Admittedly, I need to be better about positing here, and I will post more thoughts— often technical— on bringing drugs and biologics to the US market. Obviously, these days, this is a topic I know something about.

   

3. Finally, with Adam Cifu and John Mandrola, I am a writer and creator of Sensible Medicine. I co-host the fortnight in medicine podcast there, and I write articles for the magazine. A new one— coming soon— co-authored with some of my favorite cardiologists will be interesting.

   Sensible medicine exists because we believe in publishing a range of opinions, including articles that irritate and annoy me— that’s a good thing. Check it out if you are interested in medicine broadly.

   I am lucky to be a part of Sensible Medicine because I get to work with my two favorite medical writers: Adam and John. Once a year, we meet up to record in person videos which are part of our How Not To Get Fooled Series. We are meeting this month!

If you enjoy my content and perspective, be sure to follow all of the above, and your support keeps this going. I appreciate you.

These papers ask: which platform gives the most accurate answer? (as judged by 2, 3, 4, or X random doctors). Some papers say Gemini, some say OE. And the commenters are at each others throats. My AI is better, dammit, they argue.

A recent paper does something I appreciate: It provides specific queries that are judged. Analyzing them in some detail can give us a perspective that is being lost. It can find the meaning of medicine. Here are 3 actual questions:

Is it appropriate to proceed with Cycle 4 of FOLFOX in a patient with metastatic cholangiocarcinoma whose oxaliplatin has already been dose-reduced by 25% and 5-FU bolus omitted since Cycle 1, with ANC values of 1390, 1450, and 1300 before Cycles 2, 3, and 4 respectively, and platelets trending down from a baseline of 110-120 to 79 before Cycle 4?

What is the management of chronic phase CML with a white blood cell count of 500?

What is Castleman's disease?

(I am choosing oncology questions because I am an oncologist, so have thoughts).

I find these questions fascinating. Supposedly these are real questions put into one AI tool by real doctors seeing real patients.

Let us take them in reverse order:

What is Castleman’s disease? — This strikes me a medical student question. If you have this question, I might recommend starting with wikipedia or I am partial to this paper by David Fajgenbaum, who famously had Castleman’s and wrote a nice book about his experience. A case of doctor heal thyself.

I can only imagine an attending doctor asking this as general curiosity. If the doctor has a Castleman’s patient, and asks What is Castleman’s, lord help the patient.

What is the management of chronic phase CML with a white blood cell count of 500?

My first question is… I am just double checking you have the right diagnosis, partner. Every since I saw doctors miss that heart failure patient in Nature Medicine (video out at the link) I am worried about competency.

Please confirm: you looked at the smear and it is CML, not AML or CLL? I need you to confirm you looked at it. Read me the differential. And no blasts/ accelerated phase? And you confirmed bcr-abl? And patient has no evidence of leukostasis? You asked the appropriate questions?

And then I think it gets tricky. Because I know many of the big CML players. And I can close my eyes and hear what Brian or Hagop might say, but I think you will have a bit of an argument between them. And I have my opinion about how to incorporate fluids, hydrea, tki, allopurinol, rasburicase (may be needed) etc. I have my own preferences for how to manage this. Imatinib is still my first choice btw.— (everyone else is wrong ;) )

I think it is fair to say that while there are some aspects of management we might all agree on— different experts will manage the patient differently. And again, I would be terrified if I were the patient and I knew my doctor was performing this search. If I presented with CP CML WBC 500, and the doc were asking this question, I would be asking for a transfer to Hopkins.

The final question is most interesting of all:

Is it appropriate to proceed with Cycle 4 of FOLFOX in a patient with metastatic cholangiocarcinoma whose oxaliplatin has already been dose-reduced by 25% and 5-FU bolus omitted since Cycle 1, with ANC values of 1390, 1450, and 1300 before Cycles 2, 3, and 4 respectively, and platelets trending down from a baseline of 110-120 to 79 before Cycle 4?

Well, again, let me double check a few things. This patient is on second line? They progressed through cis-gem-durva? (That makes most sense to me) No mutations? No IDH?

Boy, you really have to wonder if the right answer is being missed by the doctor with how they frame the question— the patient’s bone marrow is full of cholangiocarcinoma because they are dying in front of you (might that be what is going on?)— after all this is second line— and they may have had the diagnosis for some time— and those platelets may not be due to chemo toxicity, but disease— and instead of having the appropriate hospice discussion, you are treating this like it is a drug dosing question. Can I review the imaging? Can I see the patient? Can I see if you are even asking the right question?

Perhaps the right question is the existential question. Wake up doctor— are you sure this patient needs more chemo? Are you sure that dose reducing is the question we have? As a wise oncologist told me, 13 years ago when I was getting started, “It is easy to give more chemo, it is hard to be honest.” My intuition from being in this situation many, many times is there is more here than meets the eye.

AI is wonderful, but AI can’t jump out of the screen and ask you all these things. AI can’t lay hands on the patient and look them in the face. And touch their legs, and listen to their heart and lungs.

And getting different answers to these questions and having random doctors decide which answers are good or bad or best or worse is also, in my opinion, insufficient, inadequate and misses the point. The doctors judging may not know enough to know if the answer is close or not. And different expert doctors may have dramatically different opinions. I know excellent oncologists who agree or disagree with me on different cases.

There is no canonical “right” answer to many, perhaps most, questions in medicine; instead there are many wrong answers and a shorter list of defensible choices. We don’t have for instance RCTs restricted to patients with CP CML WBC>500. We have all comer studies, but considerable nuance is needed for very high white count. And don’t get me started on the literature on leukopheresis— were someone to evoke that.

And the platforms are evolving by the minute. Whatever is in first place in the morning, might not be at dinner. If the doctors judging are 6 random doctors, the judgement may be different than 6 doctors who are immersed in that question. And the doctors judgment also changes over the course of our careers— often based on salient examples we remember. The same 6 may vote differently a year later.

And finally, I refuse to believe that most doctors are choosing which AI platform to use based on which scores best in a preprint or peer reviewed publication.

Instead, doctors want a rapid system to remind us of what we think we remember, to stimulate our thinking, to push us in the right direction, to suggest options we have not thought of, to reassure us, and to guide us to focus on the big picture.

I think 4 things are true

1. AI is already doing amazing things and better than 90% of doctors

2. Doctors will use AI for all of the above reasons

3. Trying to prove your technology is the “most accurate” misses the forest for the trees.

4. Letting random doctors judge which answer is best for these questions moves us nowhere. These types of papers are misguided.

And if you are a patient, there is no substitute for a doctor who never forgets what matters in life.

First, the book reminds me of Bad Blood — about the Theranos scandal. It's a quick and engaging read and I commend the author.

Second, the book raises many issues about science that are worthy of further discussion. Because I have so many points (20). I will number them.

1. I am persuaded by Theo Baker that MTL has a problematic research history. Papers at multiple institutions have replicability issues and data/ image manipulation problems. It's hard to know what MTL knew and when. At worst, he is complicit in active fraud. At best, he slow rolled appropriate corrections. I personally lean towards the former, but the truth is unknown.

2. Scientists should be scrutinized. Scrutiny should be proportionate to the claims they have made (bigger discoveries will get more scrutiny bc that drives more research) and scrutiny should happen equally across the board. You can't just audit University presidents. You should audit all people, and if you have to triage, start with those who make big claims. MTL is one.

3. People should be punished for wrongdoing.

4. MTL went down because his work provided raw images. Raw data lets third parties find errors. Most of science hides the raw data and provides only tables and figures summarizing the data. If all raw data were provided, I bet 60, 70 or 80% of research would be found to be untrue, manipulated or fraud. This is compatible with the abysmal reproducibility rate across many fields documented in a series of papers. This is compatible with Ioannidis’ famous thought experiment paper on why most research is false. This is compatible with our work on medical reversal. And this is compatible with my read of tens of thousands of papers over 20 years. This is in contrast with Mr Baker who believes that MTL was exception to the rule. He is the norm. Scientists are engaging in massive manipulation, selective reporting and outright misleading research. At times, they fool even themselves. MTL isn't the lone drug dealer at your college, MTL is the drug dealer heading up the cartel. Some will argue that others aren't as intentionally deceptive. That's like saying it's not stealing bc you didn't know it was someone else's. The weaknesses of science and unwillingness to reform is well known. Now more than ever.

5. MTL is faulted for being too ambitious. Sadly this is true for probably 99% of Stanford or Harvard professors. You don't get the job without ambition. You won't survive.

6. Ironically being a university president and being a scientist have nothing to do with each other. MTL’s science misconduct led him to resign a political— and not a scientific— job. Now he is back in drug discovery. Probably it was less risky to have him fundraise for Stanford.

7. A person like MTL will have lots of enemies. Enemies will use anything to take MTL down. I wonder who weaponized Baker against him. Who was the first anonymous tip in 2022?

8. Journalism is motivated by taking down individuals and not solving systemic issues. Bad science is a systemic rot. Knocking out MTL doesn't solve the issues. Journalism loves stories about people. They don't love the messy reality that people operate according to incentives and move in herds.

9. The book reminds me of Jose Baselga. Baselga was physician in chief at Sloan kettering. It is no secret, he and I had our disagreements, but what is less well known is that we interacted frequently since I was a fellow and I quite liked Jose. He made me laugh. Baselga had someone report him to Charlie Orenstein (reporter) for failing to disclose all his payments from pharma in papers. And the allegation was broadly true. It was also par for the course. Oncology was (and is) overrun with faculty taking money from pharmaceutical firms and then recommending those products— even for off label and disputed uses. The field has never done anything about this other than token disclosures. Disclosures, which, by the way, have been shown to be ineffective at mitigating bias. Baselga made a mistake but so many others have made it as well. Instead, someone who has an ax to grind reported him. Journalism pounced. Eventually he resigned and moved to AstraZenica where his salary increased dramatically. Sadly a few years later he would die from Creutzfeld-Jacob disease. Now, a decade later, none of the systemic issues have been changed. It may even be worse. It many ways it's just like MTL. Instead of reform, we pick one prominent scapegoat and slaughter them.

10. Separate from MTLs research issues, he supported the censure of Dr. Scott Atlas who was opposed to cloth masking, school closure and lockdowns, but made the unforgivable mistake of once touching Donald Trump's hand. The Stanford faculty senate couldn't process the pandemic and MTL called these views misinformation. Ironically, this MTL action is far more relevant to being a university president than Western blots. MTL punished Atlas for the crime of being correct. Bob Harrington (Stanford chair) was another offender. We actually don't need academic leaders to do science, we need them to not interfere with those who are— even when it is hard— especially when it is hard. Both these men failed to meet this bar.

11. The king is dead, long live the king. Theranos is gone, so there are no longer any biotech companies making false claims— oh wait. MTL is gone, so there are no more scientists doing bad work. Oh wait

    Consider the recent randomized control trial from China on time of day administration of immunotherapy. Of course the randomized study has been retracted. Of course it looks extremely problematic. But what no one has acknowledged is the ocean of retrospective observational studies that are hopelessly confounded that have built this field up. We need a single non-fraudulent randomized study and that has not yet been done. The entire field is broken to churn out more and more retrospective observational studies and not to conduct the correct trial. (See also radiating oligomers). One Chinese study is retracted— people celebrate— but the true answer remains unknown and some one is launching a new retrospective chart review as we speak— the Photoshop of epidemiology.

    We should be very careful not to celebrate the downfall of individuals, while systemic issues are completely unaddressed. And let us be clear, they are completely unaddressed. Nearly every lab is selecting the best image for the publication, and nearly every retrospective observational study is playing around with covariates.

12. At the same time, it must be acknowledged that science is still making progress. If you let your kids make you pancakes, and they make one perfect pancake, but you go into the kitchen and find 24 eggs are broken and fourteen cups of flour are used, you will say it's great that you made a pancake, but next time we have to learn how to do it less wastefully. Science still makes pancakes. Some discoveries are true and useful and progress still occurs. We are just breaking an inordinate amount of eggs to make it. And the incentive to reduce the egg breaking, when it is the federal government footing the bill, is nearly nothing. Instead, advocates cry for more and more federal funding. More and more eggs to make one pancake. No one has shown that every MTL paper is wrong. He did make at least one pancake (e.g. Denali). He just broke a lot eggs.

13. Making more science true and useful is possible but it will threaten many entrenched interests. These interests are far more powerful than MTL and far more dangerous— in part because they are unified, and capable of misleading entire swaths of reporters. The last year has witnessed unimaginable efforts to oppose grant reforms. Somehow scientists believe both that lots of science is bad, but also that they are perfect and deserve more tax money. No questions asked.

14. If it is the case that much of science is problematic, and MTL is emblematic and not exceptional then science and universities start to resemble a massive regressive wealth distribution scheme with the token justification of cures and advances. We siphon billions of dollars from taxpayers to fund a growing university administrative state, with the rare byproduct of a real advance, and when a once a century pandemic strikes the President signs off on the censure of the one faculty member who hasn't lost his marbles. (To be clear Atlas and Jay B and John I, and Eran B and others at Stanford were all generally right.)

15. Mr Baker goes easy on Holden Thorpe and journal editors. Let's be clear: journals protect bad research. I cannot remember how many letters to the editor I wrote that decimated a paper that they refused to print. Now that these pages reach more readers and scientists than most journals— I just post them here. I cannot recall how many follow up questions I asked and how many data requests journals didn't grant. Journals are incentivized to protect their product/ not to enforce standards or truth. Again, a systems issue. In the MTL case, journals only took action when publicity was too much to bear.

16. Side note: some confuse monetizing a product with scientific success. Scientific success is when a product makes me people’s lives better. Selling your idea in a heavily government subsidized arena means others can resell it— not that it helps anyone or is worth it.

17. Side note 2: As in the case of the Genentech paper and MTL, it is often a junior person whose hands appear to be the most dirty when it comes to fraudulent data analysis/presentation/ manipulation. This is both a character issue and a systemic issue. I will have more to say about this on a future date.

18. Mr Baker is a good writer and journalist but I continue to believe that newspapers need in house medical expertise with a focus on critical appraisal if they want to remain fit for purpose. In the meantime, we will do our best on Sensible Medicine

19. Finally now that MTLs false research on Alzheimer's is corrected, we can focus on the real cure for dementia…. The shingles vaccine…. 😂 The more things change, the more they stay the same. The king is dead, long live the king.

20. What about Harvard? Well, their University president resigned over plagiarism, not image manipulation. That's different. Also, unlike MTL, her scholarship was so unimportant and uninspiring that, had she not been president, it may never have been read at all. If a tree falls and no one hears, it doesn't make a sound? If you Plagiarize and no one reads it… these are the questions for academics. Again, plagiarism is a systemic issue, which remains unaddressed in the millions of academics who aren't Harvard president. Also it is part of a broader problem of repacking others ideas as your own. Also, unaddressed.

In short, Theo Baker writes a riveting book. Many will read it as a story of one bad apple climbing to the highest perch. But the real story is far more terrifying and threatening. Of 50 billion in annual science funding, what percent is true and useful? How much is siphoned and stolen? What efforts are there to fix this?

Instead of tackling those questions, I look forward to journalism informing me that actually all vaccines halt dementia, AI is albert einstein for finding St elevation MI in a smoker with crushing chest pain that all doctors missed, and a new EKG algorithm finds more people who benefit from ICDs.

So long Theranos and MTL, and a warm welcome to the new companies and leaders who offer the real discoveries.

On GLP 1s and Alzheimer's: target trials claimed a benefit. But massive 3000+ person randomized data is negative.

On Tamiflu: in a target trial, among hospitalized and severely ill adults with flu, there was a huge mortality benefit. Yet, a randomized study failed to find benefit in this population. It trends towards increased deaths. (For this essay no need to debate useless vs harmful— it is sufficient that it doesn't help)

On paxlovid. The same thing. It works in a target trial but not in an actual trial.

On GLP1s and early cancer, same pattern. Early benefits in a target trial not present in randomized data. Strange results for thyroid cancer.

On covid boosters, a target trial showed preposterous results. That all cause mortality was reduced even before covid associated CV events were. The booster traveled back in time to cure cancer and diabetes and heart disease. #TheScience

Then there are the results of RCT duplicate- a pre specified comparison of target to randomized trials. Note that this is different than the more recent examples. In this case the authors already knew the randomized results they were aiming for. It's easier to predict the winner of yesterday's world cup game than tomorrow's.

Let me be clear: there are many clinical questions where you don't need a randomized trial. If I make a cure for late stage Alzheimer's, we would accept an uncontrolled trial of 3 people before licensure. A drug that takes a fatal disease’s well understood biomarker and normalizes it, will get approval without a randomized trial. Probably n of 12 people.

But there are situations where you want a target trial. All of these fit. Do glp-1 reduce Alzheimer's? This is about averting or delaying a bad outcome and many people aren't destined to experience the event. Does Tamiflu lower mortality from 10 to 9%. Here you want a controlled comparison. The effect size, if present, is too small for historical or external benchmarks.

And recent studies show that if you want a target trial, you need a randomized trial. The former makes false promises.

Why are we seeing so many target trials lately? Because researchers want to pad their CV. Because people doing them are ignorant to their shortcomings— proof is imagine publishing a covid booster target trial and posting that they reduce all cause mortality in 10 days and not thinking that invalidates your paper.

Because it's the cheap and lazy thing to do. Rcts take years and money. Target trials take a statistician and middle authors who didn’t help.

Will AI be able to fix target trials? Maybe— I have some ideas for how to do so. But proof will require predicting the results of ongoing trials.

What if the target trials were right and the rcts were wrong? And what if this is all just a dream? What if we are really living in the Matrix? What if…

Science magazine has a story lamenting that medical students are pumping out low quality observational studies. How dare they! You have to spend a decade in training to become a faculty member working hard to pump out low quality target trials.

Rather than validating AI, the paper makes me worried the treating physician is on the express train to malpractice litigation and makes me wonder about the reporting standards at Nature Medicine.

This case report appeared in Nature medicine. A 45 year old male smoker presents with shortness of breath.

Last I checked, the patient is getting a cardiac work up. If the reader is aware of any hospital or physician who doesn't work a 45 male smoker with shortness of breath for cardiac etiology, let me alert my lawyer friends who specialize in malpractice.

The patient did get some cardiac work up. He got an EKG and a ‘biomarker’.

The biomarker is elevated.

Almost surely it is troponin. What is the precise value?

Aside: It is no secret that I have been extremely disappointed with nature medicine editors. In the last few years I have seen so many poor quality papers published there. I see that they had to retract a randomized study on time of chemotherapy administration. Any good review would have identified the deficiencies at the outset. Here, imagine publishing a case report claiming that AI can do what no doctor did, and not reporting the specific biomarker that was elevated and exactly how high it was. It's hard to fathom that there can be any clinical expertise among the editorial staff.

Next is the EKG. It's clearly abnormal.

Here is John Mandrola

I'm deeply suspicious the clinical note is dishonest about physical exam findings. Perhaps the patient had distended neck veins, crackles at the bases of both lungs, and pitting edema of the legs. I no longer have confidence this case report is faithful.

Next comes the AI part. Supposedly AI triggers further workup and an echocardiogram which finds diminished ejection fraction. Patient ultimately gets a heart transplant. AI gets the win.

Don't get me wrong: Thank God we had AI. We need AI to bail out doctors. But remember this is only in the context of what seems like poor medical care.

Most doctors would agree that in a 45-year-old smoker who presents with shortness of breath and elevated cardiac biomarkers, with an abnormal EKG, you shouldn't send them home with albuterol and a prayer. Instead they need a full cardiac workup, including but not limited to an echocardiogram.

And if doctors decided to publish this as a case report extolling the benefits of AI, they should at least try to adhere to basic standards of medical practice and fully report all lab values and provide all raw imaging. (I was able to review the echo but not the XR chest). My intern would do better than this case report.

Nature medicine needs an editorial revamp. You need to have clinicians who practice as part of the staff if you're going to publish clinical research.

Newspapers need in house medical expertise or they should stop covering medicine. Nytimes has done a disservice.

AI can help catch doctors’ errors. Doctors should aspire do be better. Claiming AI has transcended physicians is problematic when it is based on what appears to be borderline malpractice. Had AI not flagged the EKG, I suspect the patient could sue. The hospital would settle.

Just as allbirds transitioned from a shoe to an AI company, midjourney is transitioning from an imaging to health care company.

Their new ultrasound technology requires submerging in a warm pool of water for a scan. Who hasn’t been to the doctors office, and looked around and thought, “the only thing that would make this better is wearing a swimsuit and dipping into a lukewarm pool of water?”

AI is supposed to help boost or reconstruct images, but you can’t boost what is not captured. There are tremendous limits in what ultrasound can “see.” Interestingly MR has made considerable advances. It is hard to imagine using an arguably poorer technology.

But the biggest challenge in all this is that when you find lumps in the body that look odd they can be one of 5 things

1. Totally normal tissue fooling you

2. Abnormal tissue or even cancer that if left untouched would not do anything in your natural lifespan.

3. Abnormal tissue or cancer that has already spread microscopically, and no matter what you do will still kill you in 2 months or 2 years or 2 decades. (early treatment doesn’t improve outcomes).

4. Abnormal tissue or cancer that has not yet spread, but will soon, but if you cut out right now, or take some medication (aka possibly the new work by Charlie Swanton) you can change the future.

5. Abnormal tissue or cancer that has spread, but early treatment improves outcomes.

You want a body scanner to find 4s, but not 1s or 2s. You want to find 5s, but not 3s. I know the lay person intuition is that there has to be more 5s than 3s— what with all this medical advance— but let me assure you— when something is metastatic, we are mostly in the 3s business. (Insert someone citing some low quality observational study, and my rebuttal).

Do 2s,3s,4s and 5s look different under the microscope? No

Do they look different with sequencing? No

Do they look different with gene expression profiling? No

Do they look different with proteomics? No

Might they look different with some future technology? Yes, and the moment you find this, the whole strategy starts to make sense, but right now it is dubious.

There is an added complexity with 4. The process of cutting the tumor out has to be less harmful than the benefit of removing the tumor. This depends on how bad the cancer is, and where the surgeon is cutting.

Can these challenges be overcomed? Yes, and if you are working on them reach out, I have some ideas, but for until that happens, I think Jason Ryan is right.

Fact: Observational studies and even “natural” experiments have methodologic challenges.

1. People who get a vaccine and those who don’t are different.

2. If you see a doctor, you are more likely to get a diagnosis. Show me a healthy person, and I will show you someone who hasn’t gotten enough tests.

3. Diagnoses change over time. A dementia diagnosis in 2017 and 2018 may not be the same thing. Doctors may become more or less disease aware. They may be incentives to add diagnoses to the chart to make more money. They may code the same exact forgetful elderly man differently* based on a single calendar date.

4. Doctors may consciously or unconsciously offer more preventive care to people who appear well.

5. The medical record aka “chart” doesn’t truly capture how well people look.

Here are some other facts, which I don’t think can be disputed

1. There has been no truly successful treatment for dementia. We have a few drugs with effects so abysmally poor, we can debate if they work at all. Dementia is not a disease with a series of advances. It seems like we barely understand it biologically.

2. This is despite huge financial investments.

3. Dementia is a slow process that takes years to manifest. By the time you forget who you are, you have likely had symptoms for a long time.

With this background, lets turn to a scientific hypothesis:

What if the shingles vaccine reduces dementia?

I am willing to consider this hypothesis. Maybe less episodes of shingles— which is debilitating— and you get less dementia. Probably would take a few years or maybe a decade to see the benefit, if true. Powering a randomized trial for this would be gargantuan. Observational studies or natural experiments are probably useless due to the challenges with the data and endpoint adjudication, but that never stops a scientist up for promotion.

Now let us look at the 4 papers:

“Experts” offer a series of papers making ludicrous claims. They argue because the claim is consistently ludicrous— it must be true— rather than the same sources of bias plague the literature. The earth is flat because it looks flat when you look around in every direction.

In annals of IM this week.

Wait, if you get ADMITTED to a nursing home—in less than a year—maybe as soon as 200 days—you get less dementia because of the Shingles vaccine. Within 3 years there is a 5% ABSOLUTE reduction in dementia. This vaccine must go inside neurons and repair them. It is what Ponce De Leon was looking for. A huge effect size that is also rapid.

In the JAMA paper, by 100 days, the shingles shot is already halting dementia. I don’t have dementia— at least I think I don’t— if I get 3 doses, I will probably be able to build Anthropic or Space X in a week— the brain benefit is so rapid.

The Wales paper doesn’t include K-M time to event plots. Sad. Why omit data when it will be so amazing? The US data is great. Obviously it confirms that the effect is near immediate.

And it includes a hazard function— thank you!— which makes it easier to see the fast action of the shingles vaccine. It makes Tums look slow acting.

Wait, what if this is true for every vaccine given to adults. What if the effect is not just shingles vaccines, but all vaccines. Perhaps they are miraculous as a class.

Check out this meta-analysis— it is true. Every vaccine prevents dementia!

Which is more likely?

Shingles vaccines (and maybe all sorts of other vaccines) instant act to reverse dementia. Even though nearly no drug in 50 years has achieved this. Despite tremendous effort and hundreds of billions of dollars. And this discovery just so happens to occur at a contentious time when vaccine science has taken on a left-right political bent, and coincidentally all the liberal outlets know the truth, while the conservative outlets spread misinformation (Option 1)…. or

The speed of the result is preposterously fast, the pretest probability abysmally low, and the desire to deceive ourselves astronomically high. (Option 2)

Future steps

Write to the annals of internal medicine and ask the editors to make the authors provide a Kaplan-Meier (time to event plot) of all cause mortality. (falsification endpoint)

Write to the Wales/ nature editors and ask for a time to event plot for dementia.

My understanding is that the NIH is planning on commissioning a randomized trial on this question. Encourage them to do so. (Equipoise is in the population of individuals who would do it, if and only if, this claim were true).

If there is a betting market on this trial, get a shingles shot to boost brain health before you vote.

Conclusion

The media coverage of this topic is deeply concerning. As I wrote recently in Sensible medicine, the media is unable to cover health and medicine without in house expertise in clinical trial appraisal. The status quo is untenable.

Coming soon, let me discuss that whole body ultrasound ;)

PS, if you enjoy these analyses, please subscribe

Subscribe now

*say before or after a 2024 election

UPDATE: Since I posted this someone drew my attention to another regression discontinuity effort in England. There is no demential effect, but a shingles effect.