By Justin Yamashita, MSc. Benchtop, site, CRO: three levels of basic and clinical research, explained without spin.

Plain Talk is Root to Rx's plain-language teardown: the myth, the evidence, and the one question that beats it.

In Monday's Root Room, a 7-year-old asked the only question that mattered: was the dog in that video real? It probably was, and that's the trap. A true image and a false argument fit inside the same 4 seconds. The argument underneath it is that the people who make your medicine don't care what it costs an animal, and that a chip in a lab somewhere is a fig leaf for that. Here's the actual shape of the thing, starting further back than you'd expect.

Before any of this argument makes sense, you need the map, because almost nobody outside the industry has it. Here's the whole pipeline in one breath.

·        Somebody identifies a target, usually a protein doing something it shouldn't.

·        Chemists make thousands of candidate molecules and test them against that target in dishes and in software. A handful survive.

·        Those get tested in cells, then in animals.

·        If the animal data holds up, the company files an application with FDA and, if cleared, gives it to humans for the first time.

·        Phase 1, usually a few dozen people, checking safety and dose, not whether it works.

·        Phase 2 asks whether it does anything, in a few hundred patients.

·        Phase 3 asks whether it beats what we already have, in hundreds or thousands.

·        Then FDA reviews (or country equivalent).

·        Then it's on the market, and reporting continues for as long as people take it.

·        Phase 4 and VAERS (for vaccines) track long term safety data for things that the initial phases may not have had the time to catch.

Roughly ten to fifteen years, and the overwhelming majority of candidates die somewhere along that road. Now the question that actually matters here: why is the animal step in there at all, sitting right before the first human?

Not because animals predict whether a drug works. They're bad at that, and the next section is about exactly how bad. The animal step exists to answer a narrower question: what does a whole living body do to this molecule, and what does the molecule do to the body's organs. Where does it go, what does the liver turn it into, what comes out, what breaks first, and at what dose. ICH M3(R2), the international guideline governing what must happen before a first human dose, puts the goal as characterising toxic effects "with respect to target organs, dose dependence, relationship to exposure, and, when appropriate, potential reversibility." That's the basis for choosing the very first dose a human gets.

A dish of liver cells has no kidneys, no blood supply, no immune system, nothing to carry the drug somewhere unexpected and nothing to be surprised by what it finds there. For most of the last century, an animal was the only thing on the shelf that was a whole body. That is the actual reason, and it's worth saying plainly, because the argument you're about to hear on both sides depends on it.

Put it this way. Testing a drug on cells in a dish is bench-testing a car part. You can prove the fuel pump moves fuel, and that is worth knowing, but what it can't tell you is what happens once the pump is bolted into a running car, with heat coming off the engine block, vibration working things loose, and a fuel line routed past something it shouldn't touch. The part passed but the car still caught fire.

Now the number. More than 90% of drugs that pass safely through animal studies still fail once they reach a human, mostly because they turn out unsafe or they just don't work the way the animal data promised. That's not a figure some activist group produced. It's FDA's own stated reason, in its own 2025 roadmap, for spending federal effort replacing the method. I should say I defended animal testing as a gold standard myself, a few years back, having worked with mice and rats in neuroscience and biology labs in undergrad and grad school.

Read that number one way and it sounds like an indictment: they've been torturing animals for a test that barely works. Read it the other way and it says something less dramatic and more useful. For most of the history of drug development this was the best guess anybody had, and a guess that lands 1 time in 10 still beats no guess at all, which is what existed before it. The 90% isn't evidence of cruelty. It's evidence of a translation problem.

Picture this: a play gets written for one specific lead actor, the human body, but that lead actor is never available for the early rehearsals. So for decades the only option was casting an understudy, usually a mouse, sometimes a dog or a primate, and judging the whole production on how the understudy performed. The understudy is a real performer doing a real job, and plenty of the time the understudy nails it. But an understudy isn't the lead, and every director knows a strong understudy performance isn't proof the show works once the actual star steps in. That gap, the distance between the understudy's read and the lead's, is where 90% of drug candidates fall through.

What's new is not that anyone stopped valuing the understudy. What's new is that for the first time there's a way to get the actual lead into some of the early rehearsals: human liver cells living in a chip the width of a hair, a cluster of human cells grown into a miniature organ, models trained on real human data run in software before anything touches a cell. None of it replaces the whole show. It gets the right actor into more of the early dress rehearsal scenes so we can better anticipate how things will react.

Here's the part almost nobody explains, and it's the part that makes everything else click. Species don't just differ a little. Sometimes the same molecule does something completely different in one animal than in another, for reasons that have nothing to do with size or dose.

You already know three examples, and none of these are analogies. They're the literal thing.

Xylitol, the sweetener in sugar-free gum, is harmless to you. In dogs it triggers a rapid, dose-dependent release of insulin that can drop blood sugar to a dangerous low, and at higher doses it destroys the liver. The Merck Veterinary Manual is blunt about how species-specific that is: in most mammals, xylitol has no notable effect on insulin at all. Not a smaller effect, no effect. Same molecule, one species has a receptor-level response and the rest don't.

Acetaminophen, the active ingredient in Tylenol, is safe for you at normal doses and can kill a cat at a fraction of what a dog tolerates. The reason is in the genome (DNA). The gene for UGT1A6, the enzyme that packages acetaminophen for disposal, is a broken pseudogene in cats and other felids. It carries inactivating mutations and produces nothing useful. Cats push the drug down a backup pathway, that pathway saturates, and what's left is the metabolite that does the damage. It's like the episode of I Love Lucy where Lucy and Ethel end up on the chocolate factory line, overwhelmed and overloaded by chocolate, except this is a lot less fun for cats.

Speaking of chocolate, here's the one everybody knows. Theobromine's half-life in a dog is about 17.5 hours against roughly 6 to 10 in a person, so it stacks up in the dog while it's clearing in you. Worth being honest that this one is a two-to-threefold difference rather than the all-or-nothing gap in the first two examples.

In March 2006, at Northwick Park Hospital in London, eight healthy volunteers were dosed in a Phase 1 trial of an antibody called TGN1412. Six got the drug, two got placebo. Within an hour the six were in trouble, and within sixteen hours all six were critically ill. Every one survived, but several needed intensive organ support for over a week, and one of them later lost all of his toes and the tips of several fingers.

The drug had been through monkeys. On a milligram-per-kilogram basis the dose those volunteers received was hundreds of times below the highest dose the macaques had tolerated without harm. The reason the monkeys were fine turned out to be one detail: the specific immune cell population that drives that reaction in humans doesn't carry the receptor the drug targets in macaques. The antibody had nothing to grab in the monkey. In a person it had plenty.

That is what the 90% looks like from the inside. Thursday's Informed carries two more cases, including a 1993 NIH hepatitis trial where the drug had passed clean through mice, rats, dogs and monkeys, and the reason it didn't kill any of them comes down to a transporter that human mitochondria have and rodent mitochondria don't.

Here's the turn, and it's the reason this issue exists.

If you accept that animal data is a weak predictor of what happens in a human, and you should, that conclusion doesn't only cut in the direction people expect. It cuts at the vial too.

Four issues ago we took apart BPC-157, the peptide sold as a natural alternative to medicine people fear. Look at what holds that pitch up. A 2025 systematic review in HSS Journal, looking at BPC-157 in orthopaedic sports medicine, found 36 studies. Thirty-five of them were preclinical. The single clinical item was a retrospective series of twelve patients. Its authors wrote that the in-human safety of BPC-157 was not assessed in the included studies and therefore remains unknown.

So the same person can, in one breath, tell you animal studies are worthless because 90% of them fail to translate, and in the next breath tell you a rat study proves a peptide will heal your tendon. Both of those cannot be true. If rodent data is too weak to approve a drug that has been through a decade of scrutiny, it is nowhere near strong enough to justify injecting something that has been through none.

The evidence is identical. What changes is who benefits from which conclusion. That's the whole trick, and once you see it you can't unsee it. Keep this in mind next time you see a Dr Anecdote or an Ivy the Influencer bragging about how the peptide they took because the animal model research was so powerful and it's changed their life. The Skeptic's Toolkit shows the data isn't there to qualify that for humans yet, and the Influencer sample size of 1 that has ulterior motives with sales proceed profit is biased. Whereas there are peptides like Semaglutide for type 2 diabetes that are tested in humans and approved, so this isn't an indictment on peptides. This is an indictment on basing treatment on immature data.

There's a serious version of the abolitionist argument and it deserves a serious answer, not a dismissal. It goes: animals suffer, they suffer for a test that fails nine times in ten, non-animal methods now exist, so stop. Every step of that is either true or defensible, and people who hold it are not being stupid.

Here's the honest answer. The number should keep falling, and it is falling. The field has had a formal framework for this since 1959, when Russell and Burch published the three Rs: replacement, reduction and refinement. That's older than most of the people arguing about it, and it came from inside the field, not from outside pressure.

But zero on Monday morning means one of two things. Either you stop developing new medicines, or you go straight from a dish to a person. Look at the second one honestly, because it's the one nobody says out loud. It doesn't remove the risk. It moves it, onto the first humans dosed, who are usually healthy volunteers being paid a few hundred pounds or patients who have run out of options. TGN1412 is what a bad first-in-human looks like with a full monkey study behind it. Nobody arguing for zero has explained what that trial looks like with nothing behind it.

That's not an argument for keeping the number where it is. It's an argument for why the honest question is how fast, not whether, and why anyone promising you a date is either selling something or hasn't read the roadmap.

Two different questions keep getting blurred into one. Is animal testing still legally required before a new drug reaches human trials? No, and it hasn't been since December 2022, when the FDA Modernization Act 2.0 removed that requirement. Is animal testing therefore over? Also no. Conflating those two is exactly how you end up with a video insisting nothing has changed sitting next to a press release implying everything has.

FDA has been unusually specific about the middle. Its April 2025 roadmap does not promise the end of animal testing. It promises a direction: FDA "will aim to make animal studies the exception rather than the norm" within 3 to 5 years, and the language throughout is "will aim" and "could," not a deadline written into law. Anybody quoting it as more settled than that, on either side, is misquoting it.

Some of it has already shipped, which is worth separating from the promise. Per FDA's own year 1 progress report in April 2026: draft guidance now exists to cut routine 6-month primate toxicology for antibody drugs down to 3 months where earlier testing plus lab-based data show no concerning signals. Guidance has moved testing away from blood drawn from horseshoe crabs, sparing by FDA's own estimate more than 1,000,000 animals a year on that change alone. And there's now a searchable database so a company can check, before it starts, whether a non-animal method will actually be accepted for its specific question. None of that is "no more animal testing." All of it is real, dated and checkable, which is a different category of claim than a video with 4 million views and no source under it.

Q6. Am I seeing the full picture, or only part of the evidence?

Run this one in both directions or it doesn't count as Q6. The "they still torture animals and always will" version sees only the drugs still tested the old way and skips the phase-out guidance, the database, the horseshoe crab change, all of it dated and public. The "this is basically solved" version, the one in the press releases, sees only the roadmap's ambition and skips FDA's own admission that this is a 3-to-5-year direction, not a current state. And the version that says animal data is worthless right up until a rat study is selling something skips the part where it's the same evidence either way.

Both halves are true at once. The switch is real and it is not finished. Anyone who needs you to believe only one half is asking you to stop looking at the other.

I used to think this argument had two sides: people who care about animals and people who care about medicine. It doesn't. Nearly every person in that fight wants fewer animals used and safer drugs reaching people faster, and for the first time in decades the technology is starting to let both of those point the same direction instead of trading against each other. Reply or leave a comment: what's a scientific process you assumed was permanent that you'd want to see the actual timeline on?

Forward this to the person who needs it, and you are on your way to a Series 1 set.

This one runs long because the argument does, and I still cut it short. Thursday's Informed carries the rest: the 1993 fialuridine trial, what thalidomide's animal testing did and didn't cover, the Alzheimer's translation record including the two amyloid drugs that finally worked, and every figure above with its citation ranked by how much weight it can hold. If any claim in this issue made you want to check it yourself, that's where the receipts are.

1. FDA. "Roadmap to Reducing Animal Testing in Preclinical Safety Studies." April 2025. Source of the 90% figure, in FDA's words: "Over 90% of drugs that appear safe and effective in animals do not go on to receive FDA approval in humans, predominantly due to safety and/or efficacy issues." fda.gov

2. Marshall LJ, et al. Alternative Laboratory Animals. 2023. The review FDA cites as the source of that figure.

3. FDA Modernization Act 2.0. Public Law 117-328, signed December 29, 2022.

4. FDA. "FDA Achieves Year 1 Goals in Reducing Animal Testing in Drug Development." April 20, 2026. Source of the horseshoe crab figure: guidance "which could spare more than one million animals per year." fda.gov

5. ICH M3(R2), Guidance on Nonclinical Safety Studies for the Conduct of Human Clinical Trials. ICH Step 4, June 2009. Source of what nonclinical safety work is actually for.

6. Eastwood D, Findlay L, Poole S, et al. "Monoclonal antibody TGN1412 trial failure explained by species differences in CD28 expression on CD4+ effector memory T-cells." British Journal of Pharmacology. 2010;161(3):512-526.

7. Horvath CJ, Milton MN. "The TeGenero incident and the Duff Report conclusions." Toxicologic Pathology. 2009;37(3):372-383. Source of who was dosed, what happened to them, and the dose margins.

8. Vasireddi N, Hahamyan H, Salata MJ, et al. "Emerging Use of BPC-157 in Orthopaedic Sports Medicine: A Systematic Review." HSS Journal. 2025;21(4):485-495.

9. Merck Veterinary Manual, "Xylitol Toxicosis in Dogs" and "Chocolate Toxicosis in Animals." Source of the xylitol and theobromine figures. A professional reference work, not peer review.

10. Shrestha B, Reed JM, Starks PT, et al. "Evolution of a Major Drug Metabolizing Enzyme Defect in the Domestic Cat and Other Felidae." PLOS ONE. 2011;6(3):e18046. Source of UGT1A6 being a pseudogene in cats.

11. Russell WMS, Burch RL. The Principles of Humane Experimental Technique. London: Methuen, 1959. The origin of the three Rs.

New here? Root to Rx runs three pieces a week. Monday you watch a real health claim try to work on someone. Wednesday you get the tools to run it yourself. Thursday you get the full evidence so you can check my work. Written by a clinical researcher, free, at RootToRx.com.

For educational purposes only. Nothing in this newsletter is medical advice. Talk to your doctor before making any health decision.

Instagram: instagram.com/roottorx  |  LinkedIn: linkedin.com/company/roottorx  |  Facebook Lab: facebook.com/groups/theroottorxlab  |  Subscribe: RootToRx.com

You are receiving this because you subscribed at roottorx.com.

Disclosure: Root to Rx is published independently by Open Label Media LLC. Views expressed are personal views of Justin Yamashita and do not represent his employer or any affiliated organization. No employer resources or proprietary information are used. Every claim is sourced from publicly available materials.

Root to Rx | 228 Park Ave S, #29976, New York, New York 10003, United States