Everyone's A Winner Baby, That's No Lie!

The Billion Dollar Business of Buried Clinical Trials

Audio and Video Overview

Not a robot reading the article. If you remember when investigative reporting had teeth — this is that. An AI-created podcast-style ‘deep dive’ that’ll make you question if it’s real. Think 60 Minutes in its prime, not morning-show banter…

We live in an age of participation trophies. Everyone’s a winner — even if you’re not. I don’t want to brag, but when it comes to making free throws one after the other after the other, I’m the best on the planet. Steph Curry’s lifetime 91%? Pathetic. Steve Nash and Mark Price? Punchlines to jokes. Truth is, I make Rick Barry look like the Pontiff in high heels. When it comes to free throws, I’ve got perfect mechanics, flawless muscle memory, a stroke as smooth as silk, and a release so automatic the rim has filed a formal complaint about lack of stimulation. My secret? Simple. I don’t count misses. And when you don’t tally your misses, it’s like Hot Chocolate sang back when Barry was building his legend from the line — Everyone’s a Winner.

And guess who else is notorious for not counting their misses? That’s right — Big Pharma. They've got my system down pat. A percentage that would make Curry weep, done the same way I do it — bricks don’t make the box score. The difference is that their airballs get FDA approval, direct-to-consumer ad campaigns, and quarterly earnings calls. Mine just hit the side of the garage. But nobody ends up dead.

If you’ve somehow never watched some of the great DTC pharmaceutical-ad parodies, I made it so you can fix that. There are far too many brilliant fako-not-fako commercials to be a coincidence-not-coincidence. Laugh now because when you’re finished there’ll be no more laughing the rest of the way (Annuale, Every Prescription Drug Commercial, Herpestopper, Heroin AM, Swiftamine, Chantix, Xentrex, Couplabeers, and that old standby, Progentatorivox).

To understand just how serious an issue we’re discussing today, you need to understand one of the biggest categories in all of biomedical science — Invisible & Abandoned Studies. Sit down, because this is going to blow your doors off.

Roughly half of all completed clinical trials never report their results — even though, in many cases, they are required by law to do so.

From here on out, we’re serious as a heart attack. No more jokes, because past this point, it stops being funny.

What “Invisible & Abandoned” Actually Means

A clinical trial is an experiment on human beings. People consent to risk — a new drug, unknown AE’s (adverse events) — on the understanding that whatever is learned will be shared so the next patient benefits. That’s the moral contract of medical research. Invisible & Abandoned studies are the trials where that contract was broken, and they break it at three separate points, each one worse than the last.

The first is invisibility… The trial is completed and then simply never published. No paper, no result, nothing. The patients took the risk, and the knowledge went into a drawer. The second is misreporting… The trial is published, but the question it was designed to answer gets quietly swapped out for a different one after the results are in — you pre-specify outcome A, A comes back ugly, so you publish outcome B and never mention A (like painting the bullseye around your bullet hole after the fact). The third is the subtlest and most widespread… The trial is published, the headline outcome looks fine, but the rest of the pre-specified outcomes — the safety signals, the secondary endpoints — get shaded, dropped, or reworded so the fine print no longer matches the protocol nobody reads.

And then there is the cleanest miss of all — the trial that was never finished. ‘Abandoned’ in RIAT's framing is not a metaphor. It is the trial the sponsor terminates, suspends, or quietly withdraws somewhere between first patient enrolled and final analysis, often with no reason posted at all, or a sterile ‘business decision’ entry that tells the reader nothing. Interim analyses are the tell. A sponsor peeks at the blinded data, sees the drug isn't separating from placebo — or worse, sees a safety signal — and the trial simply ends. No publication obligation, because there is no completed dataset to report. The patients who consented are sent home. The signal that would have surfaced never gets a denominator. It is the purest form of not counting a miss: the brick never reaches the rim because the shot was called off mid-air, and the box score records nothing happened at all.

There is a formal mechanism built to fix this. It is called RIAT — Restoring Invisible and Abandoned Trials — launched in 2013 and published in BMJ, the premise is blunt. If the people who ran the trial won’t publish or correct it, independent scientists will do it for them using FOIA and the underlying data. It is a good idea. It is also losing the battle. The RIAT Support Center FAQ page says so itself… “Few restorations of the trial literature have been done to date, because RIATing a single trial is an enormous, multi-year undertaking”. Sit with that.

The mechanism specifically built to correct the scientific record has, over the course of a decade-plus three and counting, restored a literal handful of trials — against an ocean. That’s not a failure of the people doing the bailing. It’s that they have been handed a thimble and told to drain the Atlantic.

The Hard Numbers — As Hard As They Can Be Nailed Down

Start with invisibility, because it has the cleanest denominator anyone has ever assembled. In 2025, Cochrane — the single most respected name in evidence synthesis on Earth — published a review of reviews drawing on 204 separate studies covering (gulp) 165,135 clinical trials, the most comprehensive effort of its kind ever attempted. The finding: nearly half of all clinical trial results are never made public. Even when the analysis was restricted to the best-conducted studies, the publication rate rose only to 57% — meaning that under the most charitable calculations possible, more than four trials in ten still vanish. That 43% is not the scandal figure. That is the floor, after you let them count it their way.

The raw-data companion to the Cochrane finding is the Trials Tracker built by Ben Goldacre’s team at Oxford. It is not an estimate; it is a live audit of the world’s largest trial registry. Of 25,927 eligible trials completed by major sponsors since January 2006, results were missing for 11,714 of them — 45.2% — and roughly 8.7 million patients had enrolled in trials that never reported. Eight point seven million human beings. One registry. Major sponsors only. Since 2006. Hold that number; we are going to quantify the cost before we finish today.

Now misreporting. A systematic review in BMC Medicine found that the median proportion of trials with a discrepancy between the registered primary outcome and the published primary outcome was 31%. One trial in three changed the question after seeing the answer. That figure is corroborated by a brute-force cross-section that nobody can wave away: Goldacre’s group examined 89,204 registered studies on ClinicalTrials.gov and found that 28,229 of them — 31.7% — had their primary outcome altered, and that industry funding was statistically associated with the change. Same number, two completely different methods, and the money fingerprint is right there in the second one.

Then the deepest layer, the one almost nobody quotes. The same BMC Medicine body of evidence found that 64% of trials had discrepancies between pre-specified and reported outcomes for the non-primary endpoints — the safety data, the secondary measures, the stuff that tells you whether the drug hurts people. And the COMPare project put a human face on it by hand-auditing trials one at a time. Of 67 trials examined, only nine reported their outcomes correctly. The other 58 had 301 pre-specified outcomes that went unreported and 357 brand-new outcomes silently added after the fact. Roughly five trials in six were edited somewhere between the protocol and the page.

You will hear a defense of all this. The industry-adjacent line is that about half is a “zombie statistic” leaning on old data from before the reforms — before mandatory trial registration in 2005, the WHO policy in 2006, the FDA Amendments Act results-posting requirement in 2007. Here’s the answer, and it is fatal. The 2025 Cochrane review and the Oxford TrialsTracker both run years past every one of those reforms — and they still land at roughly half. The laws were written. The laws were passed. The misses still aren’t being counted. That is not an argument against the scandal. It is the scandal.

The Named Cases — Where “Not Counting Misses” Becomes a Body Count

Percentages don’t bleed. People do — and here is what the abstraction looks like with a name on it.

Tamiflu. Before the 2009 H1N1 outbreak, the United States alone had stockpiled nearly $1.5 billion worth of oseltamivir. The United Kingdom spent £424 million stockpiling 40 million doses, plus another £136 million on a sister drug — and the government’s own 2008 business case for that purchase assumed the drug would deliver a 40–50% reduction in influenza complications and mortality. That assumption was the sales pitch. It was not in the published evidence, because the evidence had not been published.

It took the Cochrane group a four-and-a-half-year campaign to pry the full clinical study reports out of the manufacturer, Roche. When they finally got them — 107 reports, analyzed against data from some 24,000 people — the conclusion was that there was no good evidence the drug reduced hospital admissions or complications, and that the original evidence presented to governments had been incomplete. Two governments. Roughly two billion dollars of public money. Spent on a benefit the buried data never showed. And the manufacturer?

Roche has benefited by more than $18 billion from oseltamivir through 2017 — and Roche stopped breaking it out as a separate line item years ago, burying it in aggregate antiviral revenue while post-COVID stockpile renewals almost certainly pushed the real take well past $20 billion. That is the entire machine in one drug: north of eighteen billion in (and counting, off the books), two billion of public money out, and the data that would have stopped it sat locked in a company filing cabinet for a decade.

Vioxx. Here the hidden miss was not a stockpile. It was corpses. The VIGOR trial of Merck’s painkiller rofecoxib was published in The New England Journal of Medicine in Y2K. Three heart attacks, every one of them in the Vioxx group, were not in the data submitted to the journal. The NEJM editors later determined — from a computer diskette — that those data had been deleted from the manuscript two days before it was submitted, issuing a formal Expression of Concern in response, stating that the deletions called the integrity of the cardiovascular data into question. Had those three MI’s been counted, the cardiac risk would have been five times that of the comparison drug.

Meanwhile, Merck instructed its 3,000-strong sales force, in a written bulletin, “DO NOT INITIATE DISCUSSIONS ON... THE RESULTS OF THE... VIGOR STUDY”. The bill for those deletions came due in the morgue: based on the later APPROVe trial — an excess of roughly 16 cardiovascular events per 1,000 patients — there were tens of thousands of patients who suffered heart attacks or strokes attributable to rofecoxib. Tens of thousands of cardiac events, traceable to three heart attacks deleted from a manuscript two days before filing.

The SSRIs given to children. This is the canonical case, and yet another I wrote about back in the day (The R.I.A.T. Act & GlaxoSmthKline's infamous ‘Study 329’). Study 329 — paroxetine (Paxil) in adolescents — was published in 2001, claiming the drug was safe and effective for depressed teenagers. GSK then used that single publication to push it to more than two million American children by the end of 2002 — for an indication the FDA had never approved. When the RIAT team finally pried the data out of GSK in 2013, here is the scale of what had been buried: roughly 77,000 pages of de-identified individual case report forms — 200 to 300 pages per patient across all 273 trial subjects — that GSK had sat on for over a decade, accessible only through a ‘periscope’ remote-access portal that wouldn't let the reviewers print or download a single page.

The restored result, published in The BMJ in 2015, was the opposite of the original: no efficacy over placebo on the protocol-specified outcomes, and clinically significant suicidality and self-harm signals that the original paper had missed, miscoded, or misreported. The U.S. Department of Justice had already reached the same conclusion the hard way — in 2012, GSK pleaded guilty and paid $3 billion (then the largest healthcare-fraud settlement in U.S. history - and “Proof of Pattern”), with the DOJ citing Study 329 by name as evidence of fraud. This was not an isolated event.

The Treatment for Adolescents with Depression Study (TADS), examining fluoxetine/Prozac (August 2004), was reanalyzed under RIAT and published one year ago this week — Twenty-one years of standard-of-care prescribing in a burgeoning U.S. pediatric depression market. Two million American children prescribed off-label on the strength of the paper that buried them. Eleven children's suicide-related events still surfacing in 2025, by outsiders, fighting for the data. And in the seven years from 2016 through 2022 alone, U.S. adolescents and young adults aged 12–25 received 221 million antidepressant prescriptions — by the American Academy of Pediatrics' own count. The fraud didn't get punished. It scaled.

The restored data surfaced a marked increase in harms among the children taking the drug, including 11 suicide-related adverse events that differed from what the original investigators had reported. Seventy-seven thousand pages buried for a decade. Two million American children prescribed off-label on the strength of the paper that buried them. Eleven children's suicide-related events still surfacing in 2025, by outsiders, fighting for the data. And as if it’s even possible, there’s more. What if I told you that despite over 10% of americans take anti-depressants, more than 1 in 3 are taking meds that cause depression — many of them children.

Tamiflu’s own data-access investigators noted in print that Avandia is among the drugs for which previously unpublished trial data radically changed the public understanding of safety and efficacy. The pattern is not a series of exceptions. The exceptions are the trials that got reported honestly.

The Floor — What This Costs

So put the number on it. Not the whole thing — nobody can put an honest number on the whole thing, and I will get to why that is the point. Just the floor. Just the sliver we are actually allowed to see.

The peer-reviewed and industry literature on what it costs to run one human being through a clinical trial gives us hard anchors. The current figure for pivotal Phase III trials — the ones that actually get drugs approved and onto your pharmacy shelf — is an average per-patient cost of $113,030. That number has been climbing fast: between 2013 and 2023, the average per-patient cost in a Phase III trial rose by nearly 50%, with oncology trials routinely topping $100,000 per patient and cell-and-gene-therapy trials exceeding $1 million per patient. A deliberately conservative anchor for a lower bound — older, broader, and including non-pivotal work — runs in the tens of thousands per head. A rock-bottom academic-trial anchor lands lower still, on the order of a few thousand dollars a head.

I am not extrapolating in a vacuum….

Now do the arithmetic, out loud, so anyone can check it. Take the most conservative count anyone has — 8.7 million patients, one registry, major sponsors only, and only since 2006. At the rock-bottom academic figure of roughly $4,000 a head, the floor is about $35 billion. At a mid-range conservative figure of $20,000 a head — well under the cost of a single emergency room visit in modern American medicine — about $174 billion. At the current industry per-patient median for the trials that actually get drugs onto your pharmacy shelf — $113,030 — that same narrow sliver is roughly $983 billion. Nearly a trillion dollars.

A Trillion Dollars. And read what that figure does not include. It does not count a single one of the 31% of trials that were published with the primary outcome switched. It does not count the 64% with the safety data quietly reworded. It does not count one trial completed before 2006. It does not count minor sponsors. It does not count a single trial in any registry other than the single American database. It is the floor of a fraction of a fraction.

So who could possibly put a number on the whole of it? Nobody. It is unknowable — and it is unknowable by design, because any industry that does not count its misses will try to make certain the misses cannot be counted. But here is the honest end of that sentence: once the visible sliver alone runs to the better part of a third of a trillion dollars, the question is no longer whether this is a trillion-dollar scam. The question is how far past a trillion it goes. It is genuinely hard to believe the true cost of this thing — in dollars, and in people — is anything less than catastrophic.

Why You’re Reading This Here And Not in Your Doctor’s Inbox

Read the last three sections again, because the most important thing about them is who found this.

The people who pried Tamiflu loose from Roche weren’t fringe cranks — they were the Cochrane Collaboration, the single most respected name in evidence-based medicine. The people who caught Merck deleting heart attacks from the VIGOR manuscript two days before submission weren’t bloggers — they were the editors of the New England Journal of Medicine, who put their own names on an Expression of Concern accusing the authors of compromising the integrity of the data. The TADS reanalysis that surfaced eleven buried suicide-related adverse events ran through peer review and sits on PubMed Central right now.

Every one of these is the establishment’s own machinery — peer review, expert consensus, the gold-standard journals — turned around and pointed at the industry that captured it. And that is precisely the material that gets you throttled, deranked, demonetized, and quietly disappeared from search. Not because it’s false. Because it’s true, it’s sourced to the very institutions the public is told to trust, and it’s load-bearing for a multi-trillion-dollar legal and commercial edifice that cannot survive the public connecting these dots. When the suppression apparatus comes for you, it doesn’t come because you’re wrong. It comes because you used their own peer-reviewed receipts to prove you’re right.

Data Laundering: The Part That Should Make You Furious

Strip away the dollar figures for a second, because the money — the fines are simply the price of doing business — is not actually the worst of it.

Eight point seven million people — and that is the floor, the visible sliver — rolled up their sleeves and let someone inject them with, or dose them with, something whose effects were not fully known. Every one of them signed a consent form. Every one of them was told, in effect, that their risk would mean something — that whatever was learned would be shared so the next patient was safer. For roughly half of them, that was a lie at the moment it was spoken. The knowledge they bought with their bodies went into a drawer.

But the part almost nobody understands is that it doesn’t stay in the drawer.

An invisible or misreported trial is not inert. It feeds the systematic reviews. It feeds the Cochrane meta-analyses. It feeds the clinical practice guidelines your own doctor follows — guidelines written by a clan-like network of "superstar" authors, typically funded by the very industries whose products those very guidelines recommend. When half the data is missing, and the missing half is the unflattering half, the "best available evidence" is not neutral — it is bent, in a single direction, toward industry. The buried miss does not just disappear. It gets laundered through the literature until it is scored as a make, and then it is handed to your physician as the standard of care.

That is the scam. Not one bad trial. A system of bad trials that doesn’t count misses, can’t be made to count them, costs somewhere between a trillion dollars and a number nobody is allowed to know, and ends — every single time you follow the thread far enough — with a patient who was never told.

So, when it comes to Big Pharma, everyone’s a winner, baby. Even if you’re not.