Research Translation Podcast

The AHA’s cholesterol guideline cites a large review of trials from 2012 as the principal evidence for treating cholesterol in healthy people.

The following table is from that review, webfigure 9, in the supplement. It’s okay for your eyes to glaze over, I’ll summarize the important parts.

People in statin trials were assessed for their 5-year risk of having a heart problem. The table, which shows deaths in statin and placebo groups, divides people by this risk. The first two rows are the lowest risk, at <5% and 5-10%.

Note the ‘box-and-whisker’ in these first two rows. It crosses the solid vertical line. Meaning, no detectable difference in deaths between statin and placebo.

The last column shows relative risks and confidence intervals that include 1. Confirmed: Statins and placebo were the same.

We can therefore make the following statement, and it is undisputed: For people at <10% risk, there was no survival benefit.

For higher risk people mortality benefits accrued over time to 2% or more. At the highest risk levels, therefore, about 1 in 50 people who took statins lived longer.

Now check yourself.

Plug in your numbers for ‘ASCVD’ risk on the AHA calculator website. If you are less than 20% (in 10 years) then the first two rows are the most relevant to you.

Case closed.

Most people who take statins are in the first two rows. Rows 3 to 5 are mostly for people with known vascular disease, in which case reducing cholesterol is reasonable and uncontroversial.

For everyone else, however, statins did not prolong life. Yet that is precisely the benefit people are led to believe they are getting from their daily statin drugs.

Incidentally, the authors of the 2012 review claimed that the benefits of statins for the lowest risk groups “greatly exceed any known hazards.”

Huh. What benefits, and what hazards?

Certainly not mortality, which they refuse to acknowledge, saying: “There were too few deaths in the lower risk participants to allow reliable direct assessment of the effects of statin therapy.”

Did they see their own data?

The numbers (53K low risk people, 1,800+ deaths) are unprecedented and allowed them to potentially detect a mortality benefit as small as 1 in every 250 people. They didn’t.

So if not mortality, what benefits DO they mean?

The 2012 review headlined a custom-made composite, “major vascular events.” This was a combination of strokes, nonfatal MIs, ‘coronary deaths’, and revascularization procedures. A combo for which the authors report a 1.1% overall benefit in low risk people. Or, as they say it, 11 people per 1,000.

Let us break that benefit into its parts. (Then we’ll talk hazards). Using the trial figures, the 11 events break down roughly as follows.

About 30-35% of it is ‘revascularizations’, 30-35% is nonfatal MI, 15% is coronary deaths, and 20% is nonfatal stroke.

On revascularizations, we now know procedures like stents for stable disease don’t save lives or prevent heart attacks or strokes. Cardiologists did not know this when designing the statin trials, so we can politely recognize the error and remove it from the calculation.

On ‘nonfatal MI’, see part 2. In short, nonfatal MI in trials is not a condition that represents, or is even associated with, death or disability. It is a false surrogate, a historical bait-and-switch that used to be critically important (when it was defined differently), and now mostly constitutes a lab test derangement.

The next part, at about 15%, is ‘coronary deaths’. But deaths are not prevented in this risk group—see table above. ‘Coronary deaths’ is therefore a red herring. Death re-labeled, then placed under a shell, then moved. It is a ghost in the data that disappears when using the patient-centered version—all-cause mortality.

Finally, let us accept nonfatal stroke as a benefit, at about 2-3 strokes per 1000, or about 1 in every 300 people.

Now let’s compare this to statin harms. And for readers who are sad because we removed nonfatal MI, let’s toss them back in, at 3 per 1000.

It turns out the CTT group, responsible for the most pro-statin paper in history, also had the data on statin-induced diabetes. In 2024, twelve years after their review that put statins in the mouths of millions, they showed it to us. Whereupon we learned that the lowest dose of statins cause diabetes in 6 per 1000 people. At higher doses it’s 60 (or 1 in every 17 people taking a statin).

Even at the lowest statin dose this harm eclipses the stroke and MI benefit.

Moreover, the one randomized trial of statin muscle effects found that 5% experienced statin-induced chronic muscle pains. That is 48 per 1000. Add this to the 6 diabetes cases and we have now eclipsed the stroke and nonfatal MI benefit by ten times over.

One might say that the hazards of statins greatly exceed any known benefits.

There is an old adage in medicine: Doctors are trained to seek pathology and diagnose disease. But patients just want to feel better and live longer.

In cholesterol management, that tension is the problem. We began by trying to prevent death. We ended by counting events most people would never have known occurred.

The result is an approach to cholesterol that serves the system—but not the patient.

That is the Big Mistake.

Next week, in part 4 we’ll talk about how this all happened, and what was lost in translation.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

Author note: What follows is part 2 of a series intended to summarize and explain just how, historically, we went off the rails with cholesterol. Check out part 1 here.

In the early 1990s, Eugene Braunwald and the cardiology community had a problem.

In the throes of a major heart attack new clot-buster drugs and angioplasty procedures were working—saving lives. Cardiologists could see it even months later, in the clinic.

But they couldn’t see it in the data.

Published trials were short, lasting weeks or months. But reperfusion—re-opening occluded arteries with drugs or procedures—didn’t just prevent instant deaths. It also halted and reversed massive heart attacks, rescuing enough heart muscle to prevent heart failure and deaths that result months or even years later. The benefit was real, but delayed. And therefore invisible in trials.

Braunwald and his colleagues therefore fashioned a solution.

They noted: Not all myocardial infarctions, i.e. heart attacks, are equal. While small MIs often pass without danger, big ones that damage lots of heart muscle can cripple a heart, leading to failure, decline, and death. In these cases even supposedly ‘nonfatal’ ones are a harbinger—a surrogate—for delayed disability and death.

In a landmark paper they called on researchers to therefore count not just deaths, but also these large ‘nonfatal MIs’.

The paper mapped risk carefully, detailing a hierarchy: Larger infarcts predicted death, while small ones rarely did.

The argument was elegant, mathematical, and tightly anchored to both context (heart attacks in progress) and definition (big ones). And it worked. Reperfusion was shown to be a game-changer.

Then, ‘nonfatal MI’ was hijacked.

Braunwald’s careful creation was relocated and redefined. Transposed into trials of prevention (rather than heart attacks) and expanded to include any attack (not just big ones), ‘nonfatal MI’ became something entirely different.

In prevention studies of healthy people nonfatal heart attacks are infrequent. They’re also heterogeneous, and—crucially—not dangerous. Why? Because the trials run for years. Big, dangerous heart attacks that lead to medium-term heart failure and death are easily captured.

In this context ‘nonfatal MI’ therefore no longer means large heart attacks with important and often fatal effects, it means overwhelmingly small ones, where even medium-term death is a virtual impossibility. Unlike nonfatal strokes, which sometimes leave people permanently disabled and dependent, small heart attacks do not.

Thus ‘nonfatal MI’, a term that for Braunwald meant a harbinger of death, became the opposite: a mark of guaranteed survival.

And we knew this early.

Fibrate drug trials of lowering cholesterol in the 1980s found that despite preventing nonfatal MIs, more people taking the drugs died. This debunked ‘nonfatal MI’ as a surrogate for death in prevention trials. Citing the frightening incongruity, a BMJ meta-analysis called for a moratorium on all cholesterol-lowering drugs.

And yet, nonfatal MI did not disappear—it grew. Soon it was the backbone of all cholesterol lowering trials, and the darling of industry researchers.

Why? Because rather than indicating mortal risk, nonfatal MI now distracted from it, functioning as a decoy that pulled the reader’s eye from outcomes people care about: disability, heart failure, dependence, death.

Take WOSCOPS, the trial that first ‘proved’ lowering cholesterol in primary prevention was worthy. In WOSCOPS there was no reduction in death. Or deaths from heart disease. Or strokes.

The only difference between groups was in ‘nonfatal MI’.

Across decades of trials that are now embedded in guidelines, the touted benefits are simply never driven by death, disability, heart failure, or stroke. They are driven by a proxy for guaranteed survival.

Remove ‘nonfatal MI’ from the measurement, and any advantage disappears.

Today the bait-and-switch is even more extreme than in WOSCOPS. Troponin blood tests are a newer gold standard for diagnosing heart attacks, precisely because they capture increasingly smaller—ultimately microscopic—events. These are now labeled heart attacks, but would literally never have registered in earlier trials. Braunwald would have called them ‘angina’, a far cry from the massive, disabling events he fought to include as an endpoint in trials.

Recognizing that nonfatal MI was failing as a surrogate for mortality, in 2021 researchers studied it. Analyzing decades of data in 144 studies, they found zero connection between nonfatal MI and death—the outcome it is supposed to herald.

Below are the key scatter plots from the paper, preceded by an example of what they should look like if nonfatal MI is a surrogate—as Braunwald intended. While none of the graphs even hints at a meaningful association, note Figure ‘2C’, where the analysis includes studies focused on the association between nonfatal MI and death over time.

That is not a surrogate. It is a decoy.

Would we all like to avoid even tiny heart attacks? Of course. But by center-staging events that are often microscopic, and basically preclude serious consequences, we risk medicalizing the equivalent of cardiac wrinkles.

Skin ages. It wrinkles, thins, and loses elasticity. So do coronary vessels. They stiffen, crack, and change. But if these changes never affect life span or function, should they be treated as equal in studies to a paralyzing stroke? Or crippling heart failure? Or death?

Right now, they are.

Which has led generations of well-intended doctors to conflate the findings reported in cholesterol trials with life-altering or life-saving effects.

Which, by definition, they are not.

And that is the fatal mistake of nonfatal MI—the engine that manufactures cholesterol-lowering ‘benefit’.

Next week, in part 3, we’ll look at cholesterol-lowering trial results through the prism of part 1 (LDL as a non-risk factor) and part 2 (nonfatal MI as the driver of benefit).

Get full access to Research Translation at researchtranslation.substack.com/subscribe

The AHA’s new cholesterol guideline is asking professionals to start earlier, screen more, and treat cholesterol more aggressively, decades before any heart or vascular disease could plausibly occur. What follows, therefore, is the first of a series in which I will attempt to show how doctors (and most others) have misunderstood the science on cholesterol, from the beginning. I will be citing and showing original data from seminal studies, to help clarify what went wrong in translation, and how we can use the same data, productively, today.

----------

“Right there,” the man winced as I pressed on McBurney’s point. It hurt him to palpate his abdomen on the right side, two thirds of the way between his belly button and his pelvic bone.

I ordered a CAT scan and blood work, then presented the case to my attending. “Appendicitis” I said, with a flourish. “I called surgery, they’re on their way.”

Later, I found the attending scrutinizing the CAT scan results. He turned from the computer, and looked me in the eye. “This gentleman has no appendix. Had it out years ago. Other thoughts?”

In the early days of modern medicine, before guidelines and consensus statements, there was a simple rule:

**BEFORE YOU TREAT A RISK FACTOR, YOU MUST PROVE IT IS ONE**

That rule, a first principle in epidemiology, built everything. It built Framingham. It built the case for controlling blood pressure. It built the case against smoking. And it rests on a straightforward sequence:

First, find a characteristic like high blood pressure, or obesity, or smoking. Then look for strong, consistent associations with bad outcomes like heart attacks, strokes, and death.

These associations, however, cannot be tenuous. They must be mathematically robust and durable—across populations, and over long stretches of time.

If you find that ironclad association, then (and only then) you may ask whether changing the characteristic can prevent bad outcomes. But the association comes first.

Smoking is a risk factor that is robust and durable. It is associated, everywhere and always, with heart attacks, strokes, and death. High blood pressure is the same.

These relationships don’t flicker. They don’t disappear when you follow people longer. They don’t depend on how you slice the data.

But with cholesterol, they don't just flicker. They are almost entirely absent.

Fun fact: In the original Framingham study—the origin story for cardiovascular risk—LDL cholesterol was not a risk factor for death.

Below is a graph from the final paper. The solid line represents overall mortality. It crosses 1.0, and never strays far above or below. Mathematically, in other words, there is no relationship between mortality and LDL.

No important study has ever found the same for high blood pressure or smoking.

But it’s not just death. LDL, the ‘bad cholesterol’, has no relationship to any of the outcomes that matter most: Stroke? No association. Heart failure? None. Mortality? None. Until, of course, LDL levels are extraordinarily high, over 200—i.e. grossly abnormal. This is far beyond thresholds targeted by modern guidelines.

Above is a figure from a 2021 study of LDL cholesterol in US adults, plotted against deaths. Note that the horizontal black line on the vertical y-axis is expected death rate. The solid red line is actual death rate. Until 200 mg/dL the red line does not begin to separate from expected mortality. This, by the way, is LDL alone, not total cholesterol. Meanwhile, the AHA hits the panic button at 160 or above, labeling this its very highest risk category—well before LDL is even a risk factor for death.

All of this helps explain why lowering LDL in primary prevention has produced, if anything, hilariously small ‘benefits’. Reductions in stroke are at best fatuous. There is no effect on mortality. And even nonfatal heart attack, the prize pig of the fair, is barely touched. (Stay tuned next week, for more on this one).

Contrast this with smoking or blood pressure. When smoking or blood pressure is reduced, there are clear reductions in stroke, MI, and death. That is because the associations are, and always have been, robust and consistent. But cholesterol never had that foundation.

Which makes the current moment bizarre.

Instead of reconsidering their premise when history’s biggest and best study of risk factors showed last year that ‘high LDL’, as currently defined, has no association with heart disease or death, the AHA is doubling down.

That is because they have reversed the scientific method.

They are not starting with a strong association and asking whether modifying it helps. Instead, they are starting with an assumption—that LDL is the central cause of heart disease—then searching for data to make the assumption look true.

This is why the gargantuan efforts and enormous costs that have been poured into obsessing over, and diligently lowering, cholesterol to prevent heart disease have yielded nothing.

I once diagnosed a man with appendicitis before I knew whether he had an appendix. Bad call.

First things first: Association is the first principle in establishing a risk factor. Everything else follows.

Cholesterol doesn’t have it.

Next week I’ll write about heart attacks, specifically nonfatal ones. You’ll see why.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

In the early 19th century, the French physician Pierre-Charles-Alexandre Louis did something medicine had rarely done before: he counted.

At the time, bloodletting was standard. It was elegant, logical, and universally accepted. If an imbalance of humors caused disease, removing blood should fix it. The theory made sense. The practice was everywhere.

Louis compared outcomes among patients with pneumonia who were bled early and aggressively to those who were not, and found that bloodletting didn’t help.

If anything, it made things worse.

That single act—placing belief next to data—began to unravel one of medicine’s most entrenched practices, forcing doctors to confront a dangerous possibility: a treatment can be coherent, widely accepted, and completely wrong.

Now, imagine if the medical establishment had ignored that finding, and doubled down—expanding bloodletting to younger and healthier people.

That is where we are with cholesterol.

In March of last year, the New England Journal of Medicine published a paper that should have ignited a revolution. Spanning millions of participants, hundreds of cohorts, and decades of follow-up, it is the largest and most comprehensive study ever performed on cardiovascular risk factors.

If the Lipid Hypothesis—the idea that elevated cholesterol is the primary cause of heart disease—were true, this study should have cemented it.

But it didn’t.

Every other classic risk factor behaved as expected, showing a robust statistical association with heart disease and death: Smoking. High blood pressure. Diabetes. Obesity.

But not high cholesterol.

Hyperlipidemia, as defined by the AHA, was the only ‘risk factor’ whose point estimates drifted to the wrong side of the ledger. Numerically, it was associated with longer life and less heart disease.

That finding, the clearest demonstration that hyperlipidemia as defined by the AHA is not, and never was, an important risk factor for heart disease, should have triggered a reckoning in medicine.

Instead, it was ignored. The new AHA guideline—134 pages devoted to identifying and treating cardiovascular risk—does not mention, cite, or even acknowledge the most important study ever published on cardiovascular risk.

But the study’s findings didn’t come out of nowhere.

In 2011, a JAMA study of more than 500,000 patients with a first-ever heart attack found that just 28% had hyperlipidemia. In other words, nearly three quarters of people having their first heart attack did not have the condition the AHA says causes first heart attacks.

Think about that.

It gets worse: Among healthy adults, roughly half meet the AHA’s definition of hyperlipidemia. Yet the 2011 study found that among people having a heart attack, hyperlipidemia is present in barely a quarter.

That made high cholesterol unique among the study’s risk factors, because it was more common among people without heart problems than people having heart attacks.

Is that a risk factor? Um, no.

A risk factor is supposed to be more common among those headed toward the disease than among those who are not. Hyperlipidemia failed that test years ago.

Meanwhile, lipid-lowering therapy has expanded relentlessly for decades. Statins became ubiquitous. Newer drugs drive LDL levels to previously unthinkable lows.

If cholesterol were the central driver of heart disease, we should therefore be witnessing the disappearance of heart disease.

We are not.

Heart disease mortality was falling precipitously well before the advent of statins, a decline that slowed once the drugs became ubiquitous. And now, with cholesterol levels lower than ever, the decline has disappeared—the curve has flattened.

The AHA’s latest guideline, however, doubles down—recommending cholesterol screening and treatment not just for people like those in trials of the drugs, who already had heart disease or were at ultra-high risk. But for teenagers and children—heart healthy youths, decades from any possibility of heart disease.

The AHA then parades their self-delusion, classifying this as a ‘IA’ recommendation, a designation to indicate their strongest possible recommendation (‘I’) , ostensibly based on multiple randomized trials (‘A’). But no trial has ever shown a benefit for testing, much less treating, cholesterol in children.

Thus the AHA brazenly disregards even its own classification system, in order to promote the treatment of lipids.

Breathtaking.

They could have gone in a different direction. Acknowledging the failure of the Lipid Hypothesis would not require conceding that cholesterol is irrelevant. Extreme abnormalities matter. Treating lipids in high-risk or secondary prevention populations can still be reasonable and helpful.

But a risk factor for heart disease that fails to be a risk factor in the best study ever done, and that’s more common in healthy people than those having heart attacks—cannot credibly be called the central cause of heart disease.

Yet that is exactly the AHA’s theory, and the silent infrastructure for the guideline.

The AHA has become something different from what it claims to be: not a scientific body responding to evidence, but a signaling institution—translating a dumbed-down, cholesterol-centric theory into screening programs, drug pathways, and public messaging.

And once a theory becomes infrastructure—once it supports careers, industries, and the identity of an institution—it stops being something you test.

It becomes something you protect. At all costs.

The sad fact is that medicine didn’t abandon bloodletting by arguing about it. They abandoned it when continuing to defend it became embarrassing.

Cholesterol has quietly reached that point. But the AHA, with guideline committees living in a bubble that protects them from scientific fact, doesn’t know it.

Or worse—they do, and they’ve decided it’s still better to keep bleeding us all.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

Earlier this week media reported that Robert F. Kennedy Jr., the Secretary of Health and Human Services, had surgery to repair a torn rotator cuff.

That’s interesting. Not just because I wrote about rotator cuff surgery last week.

Kennedy has spent years arguing that medical treatments should be recommended only after they’re proven to work in rigorous trials.

Which raises the inevitable question: What do rigorous trials show about rotator cuff surgery?

RT is 100% reader-supported. To help it continue, please consider becoming a paid subscriber.

At half a million performed each year in the U.S., rotator cuff surgery is among the most common orthopedic procedures. Which is curious, because the handful of randomized trials testing it have repeatedly found little or no difference between surgery and physical therapy. Even within orthopedics there is deep skepticism, and sham-controlled trials of the procedure are currently underway.

Meanwhile, as I discussed last week, in a recent MRI study of healthy Finnish people age 40 and over, fixable ‘injuries’ were seen in perfectly functional shoulders. And not just some of them: 96% of people with no shoulder complaints had potentially surgical lesions. Tendon tears, present in 80% of study participants in Kennedy’s age group, appear to be little more than a normal sign of aging. Like wrinkles in the rotator cuff.

In contrast, for years Kennedy has argued vaccines should not be recommended unless they’re supported by randomized trials. He has suggested the evidence base is shaky, with poor documentation of harms, and that vaccines are overused, driven by corporatized medicine and windfall profits.

To me, these are legitimate concerns. I see it as quite sensible to scrutinize the trials that underlie current practice. I believe skepticism about such evidence is admirable, and I believe his call for transparency is righteous. Medicine has adopted many treatments too quickly, often on the slimmest of data, with no avenue for patients to know or weigh in.

But RFK Jr.’s critiques of seminal vaccine data are, comparatively, small time. They boil down to methodologic details in trials that found benefits. Issues such as placebo selection, evolving formulations, and long-term follow-up are reasonable concerns worth airing and discussing—but they’re nit-picky.

In stark contrast, red flags in the data on rotator cuff surgery are big time: namely, the surgery didn’t work—there was no benefit. That ain’t a nit-pick.

To be fair, one man’s shoulder treatment does not carry the same implications as childhood vaccine policy. But this man made a name debating the fine print in vaccine trials. RFK Jr. having shoulder surgery is like Mohatma Gandhi having road rage.

Moreover, just last year the Secretary openly recognized that American healthcare wrongly incentivizes costly, often dubious elective surgeries—a crucial insight that helps explain why orthopedics is disproportionately lucrative and why unproven surgeries like rotator cuff repair are the cash cow of the specialty.

Based on this insight, Kennedy suggested shifting incentives away from such elective surgeries and “toward primary care and prevention.”

Yet now, when Kennedy became a patient, the calculus appears different. Faced with shoulder pain and an expert’s recommendation, he did not turn his own famously cutting eye on the data. Which may be a serious mistake. For a 72-year old man the risks associated with surgery (heart attack, stroke, pneumonia, blood clots, death) are greater than any benefit found in a trial of rotator cuff surgery. And in light of the Finnish data, the notion that whatever MRI finding his surgeon is targeting is the true cause of his shoulder problems is, at best, highly suspect.

All of which means that RFK Jr.’s decision to undergo surgery rests not on evidence, but on the advice of experts.

That’s interesting, too.

This same Secretary of Health recently said “Trusting the experts is not a feature of democracy and it’s not a feature of science. It’s a feature of religion and totalitarianism.”

I agree.

But evidence, it seems, is most powerful when it governs other people’s care.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

I used to do a lot of push-ups. Then, one day, my right shoulder abruptly said no más. Coincidentally, days later I saw my doctor. To be clear, I live by “See doctors, not too much, only when sick,” but my insurance policy insists on an annual visit. So once a year I skulk into the office and prove why doctors make awful patients—particularly when they’re washed-up, windbag, wiseacres who don’t practice medicine.

To his credit, however, my doctor is kind and attentive, and quickly noticed my wincing. After a brief shoulder exam he diagnosed a rotator cuff injury and began preparations for an MRI.

I demurred. “Oh, that’s alright doc, I’ll just do some PT. But thank you, really.”

Gently, and with concern, he offered “It’s important that you be imaged.”

“I’m imaged all the time!” I said, pulling out my phone and showing him selfies of me and my boys pretending to throw up.

Perhaps it was the cholesterol test (I declined), or the PSA (are you kidding??), or just my winning smile, but this was the final straw. He turned slowly from the phone to me.

“Uh-huh. Sure. You do PT.”

I concur!

RT is 100% reader-supported. To help ensure I can continue, please consider becoming a paid subscriber.

The problem, of course, is not my doctor, a man who suffers me with unparalleled grace. It is orthopedics—and a decades-long unmasking.

When things work the way they’re supposed to in medicine, the sequence is simple: First, researchers prove a treatment works, then doctors adopt it widely. In many cases, orthopedics has skipped step one.

Since the advent of X-rays, orthopedists have used imaging as a crucial, and often primary, determinant of the need for surgery. This has led to miraculous cures for broken and deformed bones. But as imaging advanced well beyond bones, so too did the ambitions of orthopedic surgery. MRIs, CTs, and arthroscopes now detect endless ‘abnormalities’—a frayed meniscus, a bulging disc, a torn ligament, a ruptured tendon.

For orthopods, seeing such pictures is a primal call to action: Ug. Thing broken. Me operate.

They attack the imperfections by trimming, stitching, shaving, plating, pinning, and reinforcing. But in terms of scientific method, this approach skips step one—the part where trials show surgery helps. Instead, orthopedists have often published a few rosy case reports and voilà—Mikey liked it!—an entire surgical industry was formed.

Eventually, curious minds began performing the studies that should have come first. And one by one, some of orthopedics’ most common procedures began to fall.

Arthroscopic surgery for meniscus tears—performed hundreds of thousands of times each year—was tested against sham surgery, where patients were anesthetized and the surgeon simply pretended to operate. The real surgery was no better.

Arthroscopic ‘cleanup’ of arthritic knees collapsed the same way.

Vertebroplasty and kyphoplasty—cement injections meant to stabilize spinal compression fractures—followed the identical arc: widely adopted, then disproven by sham-controlled trials.

Even ACL reconstruction, long considered inevitable, now faces trials suggesting rehabilitation alone typically leads to comparable outcomes.

Now a study published last month in JAMA Internal Medicine is raising new questions about shoulders. In it, 602 Finnish adults aged forty-one and older underwent MRI scans of their rotator cuffs, the ring of muscles that powers the shoulder joint. About a third had shoulder pain, the rest had no shoulder problems at all.

Among those with shoulder complaints, MRI found potentially surgical abnormalities—tears, degeneration, tendinopathy—in 98 percent, a stunning number. But the number for shoulders with no problems dropped jaws even more: 96 percent.

Awkward.

These results align with sham surgery trials, which overwhelmingly suggest that exercise and physical therapy may often be a wiser, safer approach than the costly, invasive surgeries that millions of Americans still undergo each year.

Importantly, none of this means orthopedic surgery isn’t a profoundly important specialty. When bones are shattered, joints are deformed, or trauma disrupts anatomy, surgery can be miraculous. No randomized trial is needed to prove that repairing a fractured femur matters.

But those injuries represent only a fraction of modern orthopedic practice.

Much of the field now revolves around treating subtle structural abnormalities found on imaging under the assumption that fixing them will help.

But assumptions aren’t science—randomized trials are.

And once a treatment becomes routine, even if it rests on shaky evidence, it develops its own ecosystem. Training programs teach it. Hospitals invest in the equipment. Surgeons build careers around it. Patients come expecting it.

In theory, when a randomized trial shows a procedure doesn’t work, the practice should disappear. In reality, it often doesn’t.

Arthroscopic knee surgery continued for years after the sham trials. Meniscus operations still number in the hundreds of thousands annually. Rotator cuff repairs remain a thriving industry. Procedures that randomized evidence has debunked continue to generate billions in revenue and fill operating rooms across the country.

Medicine prides itself on being evidence-based. But surgical traditions, once established, can be remarkably resistant to evidence.

Why don’t these surgeries work? Probably because the intrinsic healing capacity of the human form has been perfected by a million years of evolution. Modern orthopedic surgery, at about a century, has some catching up to do.

In medicine, the order matters: test first, adopt later. Much of orthopedics did it the other way around, and we are still sorting out the consequences.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

Get full access to Research Translation at researchtranslation.substack.com/subscribe

When Arthur Conan Doyle wanted Sherlock Holmes to show off, he didn’t give him a bloody fingerprint or a dropped monocle. He gave him a dog that didn’t bark.

In Silver Blaze, the silence was the clue. The horse didn’t vanish because of a stranger—the stable dog knew the intruder. The absence of noise told Holmes everything.

That’s where we are with the newest trial of PCSK9s, a class of drug that dramatically lowers cholesterol levels.

At first glance, the study looks like a triumph of modern pharmacology. It’s a massive, multinational, double-blinded trial done at hundreds of sites in a dozen countries, enrolling nearly three thousand people. The paper trumpets spectacular reductions in LDL (half or more!) and a power analysis north of 99%. Reviewers nodded. Editors smiled. Headlines followed.

And yet: The dog never barked.

This drug class exists for one reason: to reduce cardiovascular events—heart attacks, strokes, and deaths. Cholesterol is not the goal, it is the surrogate. Human outcomes are the prize.

So where are they?

Nowhere in the paper. Nowhere in the abstract. Nowhere in the results. Nowhere, even, in the discussion of limitations. Not a single Kaplan–Meier curve. No exploratory table. Not even the typical cowardly offering—“CV outcomes will be reported separately.” Just, silence.

That is not an accident.

Holmes would first note the power analysis (a window to the soul of a study): The authors insist the trial was designed to have more than 99% power to prove the drug can lower LDL. But that is absurd on its face. Sniff test: failed. No regulator requires 99% power to prove a biomarker outcome. Eighty percent is standard. Ninety if you’re nervous. But as I’ve pointed out before, increasing power, by upping the number of participants, can be a cover story for ulterior motives.

And it’s pricey, to the tune of $50,000 or more per enrollment. Fiscally responsible companies don’t add study participants without a good reason. Which makes it notable that a standard power analysis for this trial suggests the researchers didn’t need to enroll three thousand people—they needed about twenty. Yup. Twenty. For the whole study. Both arms. The PCSK9 drugs drop cholesterol so powerfully that it takes very few participants to demonstrate it mathematically. Which means the company paid more than $140 million for seemingly extra, unnecessary enrollments.

So, what’s really happening here?

There is a clue that may at first seem small (Dr. Watson once said, "I am glad of all details, whether they seem to you relevant or not"): According to an online recruitment posting for this exact trial, the goal was not to lower cholesterol, but rather “to reduce major cardiovascular events, including cardiovascular death, heart attack and stroke, by lowering LDL-cholesterol.”

Ahaa.

There it is, in black and white. It takes thousands of subjects to find the tiny reductions in cardiac events that cholesterol lowering drugs can (under the best of conditions for the highest risk people) theoretically achieve. This explains the study’s size. Moreover, in a trial this expensive, CV events were not just possibly collected, they were inevitable. Logged. Coded. Reviewed. Sitting somewhere in a database.

Which brings us back to the dog.

Why not include the cardiovascular outcomes in the paper? Because they weren’t reduced. If CV outcomes had favored the drug, even weakly, they would have been mentioned or, more likely, headlined. But nary a mention. Not even, “Numerically fewer events, ” or “a favorable trend.” Journals allow that sort of nonsense. Regulators tolerate it. Marketing departments adore it.

Instead, nothing. That is not sloppiness, it’s subterfuge.

The brilliance of the design is that it allowed them to pivot. By claiming a primary endpoint of LDL reduction (a cakewalk for PCSK9s), and then massively inflating sample size, they could quietly chase their true goal, cardiovascular benefit—while still guaranteeing ‘success’.

Heads they win. Tails they don’t lose.

The drug succeeded at what was always easy: moving a lab value. But it failed at its secret raison d’être: helping humans.

Sherlock Holmes would have closed the file already. The most telling result of this enormous, costly, meticulously planned trial is not what it reported—but what it very carefully did not.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

When readers sent me media articles touting baloxavir (‘Xofluza’), I started looking. In the interim, Sensible Medicine, a wonderful Substack by evidence-based doctors, published a podcast, and today a written piece, recommending the drug. I believe they have it deeply wrong. Fun! Check it out, tell me what you think.

When I was seven years old I tried hard to see through a frosted glass window. My pediatrician, a silly and smiley man with a nasal voice, stood behind me.

“You see that, across the street?” he quacked. Then, leaning forward, as if to sync his line of sight with mine, “See?” His hand was on my shoulder, squeezing gently.

I squinted, and tried hard. “Wait… is that it?”

“Ah, forget it” he said, smiling at my mother. “You missed it.”

Which was true—I’d completely missed the shot in my shoulder.

Research Translation is fully reader-supported. Like it? Consider becoming a paid subscriber to help me keep doing it.

In addition to being a beloved and brilliant pediatrician, Dr. Gribbetz was a master of distraction. Unfortunately, that’s a skill drug company scientists have also mastered—for different ends—and they played it to perfection with a flu drug called baloxavir.

In recent weeks The Atlantic and the Wall Street Journal, noting the rough flu season and weak effects of this year’s vaccine, ran articles extolling the virtues of baloxavir, a single-dose drug hoping to replace 5-day Tamiflu. Both media articles rely heavily on experts who have fallen hard for a 2018 trial published in the New England Journal of Medicine.

To be fair, the study also convinced FDA reviewers who approved the drug. Though clearly, that ain’t tough, and the study is a crowded swamp of data. Reading it feels like trying to see through frosted glass.

I’ll try to be efficient here, but for the geeks I’ll get into more detail on the podcast.

The ‘study’ is actually two separate trials. One compared three doses of baloxavir to a placebo. The other compared baloxavir vs. Tamiflu vs. placebo. Which is remarkable because Tamiflu is a debunked drug, years ago revealed by many independent investigations as not just unhelpful but, to some critics, a premeditated fraud.

As I’ve written and podcasted before, Tamiflu trades side effects for flu symptoms, adding as much misery as it subtracts, while failing to prevent serious illness. Meanwhile, symptoms—the only measure either drug even claims to affect—were identical with baloxavir. So the first and most obvious headline should be: BALOXAVIR WAS NO BETTER THAN TAMIFLU, A PROVEN FAILURE.

That should end all discussion.

But it won’t. The most interesting quirk about the baloxavir paper is that the first trial is an “FDA, Phase 2, dose-finding study”—a fancy way of saying it was done to find a dose, not prove efficacy. The trial had three drug groups of differing doses, plus a placebo group. And no group, when compared to placebo, had enough participants to say whether the drug works. That isn’t a flaw, it’s the point of Phase 2 studies. But it means efficacy claims drawn from them are, by definition, a mistake.

Why? Because with three drug groups and one placebo, the drug is getting multiple shots on goal. Efficacy trials get one shot on goal—no more. Ever. And that’s why Phase 2 studies are, well, phase 2. Only Phase 3 studies can say whether a drug works.

And in fact, it's weird to publish two studies of different design, from two separate FDA phases, together. It’s weird because the studies have different goals and are interpreted differently. Which is deeply distracting.

So what were readers being distracted from?

The paper’s much more pivotal Phase 3 study compared baloxavir to Tamiflu and placebo. And here’s where a red flag pops up, bright and tall: Why wasn’t the study a non-inferiority trial comparing the two drugs?

‘Non-inferiority’ is a special design statistically and methodologically focused on finding equivalence between drugs. That’s different from a classical superiority trial, designed to find a difference between drugs. For years non-inferiority has been the FDA’s favored design for testing when a new drug claims to be as effective as an older drug. Typically, in such cases the new drug offers some convenience upgrade, like being a pill instead of an injection. Or, perhaps, 1-day versus 5-day dosing.

Like baloxavir.

Baloxavir is a picture-perfect candidate for a non-inferiority study, but that’s not what the company did. Which means they probably expected more than equivalence with Tamiflu. They expected superiority. When that failed, they settled for (and pretended to celebrate) sameness.

How can we know the trial aimed to show baloxavir was better, when the authors claim superiority based on dose-finding, placebos, and virologic results? The Methods section is the paper’s guilty conscience, confessing. The researchers used classical (non-parametric) superiority testing techniques to compare the drugs.

Besides this, there is circumstantial evidence. For starters, early lab studies found baloxavir had better effects on viral load and biologic markers than Tamiflu. But that only matters when it translates into people feeling better or being safer, which didn’t happen. This probably surprised and dismayed company researchers, who—like the literati of medical evidence—surely know that Tamiflu is a failed drug.

So when baloxavir failed, they changed the narrative: “It’s as good as Tamiflu—yay!”

And sadly, most flu ‘experts’ are experts in virology, but not evidence. So when the red flag appeared—a failed superiority trial dressed up as success—they did what seven-year-olds do when asked to see through frosted glass.

They squinted, nodded, and missed the shot.

Tamiflu is a failed drug. Now we know baloxavir is too. Moreover, both represent a profound risk to public health, because both must be started within two days of first symptoms. If everyone with a sniffle runs immediately to their doctor, or ER, or clinic—resources already dangerously over-burdened—people in dire need of these resources will be harmed.

So as I’ve written and podcasted about before, if you want one day’s relief from the flu, try silicon nasal sprays. They’re better, safer, and cheaper, and they’re over-the-counter.

Get full access to Research Translation at researchtranslation.substack.com/subscribe

Below marks my first critique of politics, and you’ll see why. For background: I have no political affiliation and do not care for party politics. Years ago I became deeply disenchanted with parties of all stripes. I do not seek a packaged bundle of beliefs or positions—I have my own. I care what works, one issue at a time, and I’m open. What I cannot stand, the attribute that repels me most, is the abuse and distortion of truth.

On Tuesday the New England Journal of (oddly partisan) Medicine published two opinion pieces on autism—a favored topic of RT! The first addressed Trump and RFKjr’s September press conference on prenatal Tylenol, while the second discussed vague claims by FDA officials implying a chemotherapy drug can now treat autism.

To jump into this cesspool of exaggerators and distortionists, we must first acknowledge an uncomfortable but obvious truth: When, on the campaign trail, you promise frenzied crowds that, once in office, you will quickly discover a cause and treatment for autism—a poorly defined neuro-developmental condition that has eluded researchers for decades—you have two options. Fail and concede, or fail and pretend.

Guesses, anyone?

After months of fishing through a swamp of poorly supported but rhetorically convenient pseudo-theories, RFKjr landed on the one entertained by stoner neuroscience grads munching funyuns at 3am. (Duuude. It’s totally the Tylenol.)

I walked through the Tylenol-autism data in October, and last week a review published in a prominent journal confirmed: Put simply, Tylenol has—provably—zero connection to autism. This was self-evident to anyone who actually read the relevant studies.

And yet, September 22nd, 2025 the Health Secretary, FDA head, NIH director, and President, acted out an SNL-ready skit, a tragicomedy of performance art. Speaking in the imperious tones of concern for their people, the four most powerful authorities in public health made a claim that wasn’t just weak—it was palpably b******t.

The irony is thick. After years of furiously and publicly condemning Anthony Fauci and others for a crisis of public trust, they offered something far more corrosive than technocratic arrogance: A display of knowing dishonesty. This wasn’t confusion or misplaced confidence. It was the deliberate use of scientific authority to launder a claim that everyone involved knew was indefensible.

That isn’t a mistake. It’s contempt—for evidence, for the public, and for the idea that science is anything more than a prop.

Which is why recent FDA maneuvers around leucovorin, folate autoantibodies, and ‘folate deficiency associated with autism’ cannot be viewed in isolation. They sit in the same epistemic ecosystem. Once again a small, messy batch of weak studies became a nonsense narrative.

To be fair, the folate-autism story is not proven nonsense—yet. It is, um, unfinished. Hypotheses deserve testing, not regulatory implication or political theater. The recent Lancet review, like mine months ago, showed what happens when you slow down and let the data breathe (scary headlines evaporate). And I have little doubt the brain-vitamin narrative will dissolve. But there’s work to be done.

Bottom line? Autism doesn’t yield to deadlines. It yields to patience, rigor, and a tolerance for ambiguity—three necessary tools for science, ones in desperately short supply right now. When science is forced to perform certainty on command, what it produces isn’t truth—it’s theater.

And this is the kind of theater that doesn’t merely fail to rebuild trust, it finishes it off. This administration didn’t set out to heal the public wound of scientific overreach. They inherited that wound and began poking at it deliberately. Trust wasn’t restored, because restoration was never the point. Control was. And a government that treats science as a stage prop shouldn’t be surprised when the audience stops believing anything it says.

Get full access to Research Translation at researchtranslation.substack.com/subscribe