Between them sit the first two attempts medicine has ever been allowed to make on aging itself rather than on its wreckage. Neither will cure anything. Both could fail.
This is what they are testing, what would prove them wrong, and every document due between now and 2030 that will settle it. By Nicholas Thomas Sunday, September 27, 2026 · Special report · Two species, one question · Five figures Somewhere in Boston this summer, someone sat in an ophthalmologist's chair and got an injection in one eye. Then they went home with a bottle of doxycycline, the same tetracycline antibiotic a dermatologist hands out for acne, and they took it every day for eight weeks.
That is the whole ceremony: no cryogenics, no billionaire's clinic in a tax haven, no machine with a hum. An injection and a cheap old pill, and with them the first attempt ever made on a living human being to push a cell's clock backward instead of merely slowing its fall. I have been covering the money in this field for a while now, and I have learned to keep my hands in my pockets.
Longevity attracts a particular kind of promise, and the promises have a way of outliving the data. So let me say up front what this piece is not. Nobody has reversed human aging.
Nobody is close. The trial that started in March enrolls eighteen people, and its official purpose is to find out whether the treatment hurts them. And yet.
Two things happened this year that have not happened before in the history of medicine, and they happened quietly, in documents, the way real change usually arrives. A regulator allowed a therapy into human beings whose stated aim is to make old cells young again. And a different regulator, looking at a different species, has now accepted two-thirds of an application for a drug whose labeled purpose would be to make a life longer.
Neither is a cure. Both are permission slips. After a century in which aging was the one thing medicine agreed to treat only after it had already taken something from you, somebody has finally been allowed to try the other thing.
That is worth understanding properly, which means understanding what could go wrong, too. The scratched record Start with the idea, because the idea is older and stranger than the trial. Your DNA does not change much throughout a lifetime.
What changes is everything sitting on top of it: the chemical tags, the way the strand is spooled and packed, and the cast of proteins that decide which genes are read aloud, and which stay shut. Biologists call that layer the epigenome, and it is less a blueprint than a set of stage directions. The blueprint says what a cell could be.
The directions say what it is doing today. Those directions degrade. Not all at once, and not because anything is broken exactly, but the way a vinyl record degrades, accumulating scratches until the music smears and you can still tell what song it was but not quite hear it.
David Sinclair at Harvard has spent twenty years arguing that this smear is not a symptom of aging but a substantial part of its cause, and that if the information is merely obscured rather than erased, it might be recoverable. The tools for testing came from somewhere else entirely. In 2006, Kazutoshi Takahashi and Shinya Yamanaka reported in Cell that four proteins, OCT4, SOX2, KLF4 and c-MYC, could take an ordinary adult mouse cell and march it all the way back to the embryonic state, a cell that could become anything.
It won a Nobel Prize. In the same paper, it also produced tumors: transplanting the reprogrammed cells into mice "resulted in tumors containing a variety of tissues from all three germ layers." Run the process to completion, and a cell forgets who it is. Sometimes what it becomes is cancer.
So, the trick is to go partway and stop. Take three of the four factors, leaving out c-MYC, which, as the Sinclair lab noted in 2020, "is an oncogene that reduces the lifespan of mice and is not required for the initiation of cellular reprogramming." Run them briefly. Turn them off.
In principle, the cell sheds some of its accumulated noise and keeps its job. In 2016, Alejandro Ocampo, working in Juan Carlos Izpisua Belmonte's laboratory, showed the principle in living animals: cycling the factors on for two days and off for five "ameliorates cellular and physiological hallmarks of aging and prolongs lifespan in a mouse model of premature aging." Four years later, the paper that sent everyone to the eye came out. Yuancheng Lu and colleagues, in Nature in December 2020, put the question in a sentence I keep coming back to: Whether older individuals retain the information needed to restore these patterns—and, if so, whether this could improve tissue function—is not known.
Then they answered it in mice. OSK delivered into retinal ganglion cells "restores youthful DNA methylation patterns and transcriptomes, promotes axon regeneration after injury, and reverses vision loss in a mouse model of glaucoma and in aged mice." The conclusion is the most hopeful sentence in modern biology, written in the flat prose of a journal abstract: "mammalian tissues retain a record of youthful epigenetic information — encoded in part by DNA methylation — that can be accessed to improve tissue function and promote regeneration in vivo." The record is scratched. The music is still on it.
Figure 1. Go all the way and the cell forgets its job. The whole therapy lives in the decision to stop early.
Eighteen people and one eye each Life Biosciences turned that finding into a product candidate and called it ER-100. On January 28 of this year, the company announced that the FDA had cleared it for human testing, describing it as "the first ever cellular rejuvenation therapy using partial epigenetic reprogramming to receive FDA clearance to enter human clinical trials." On June 9, it said it had dosed the first participant. The therapy is a modified adeno-associated virus carrying the genes for OCT4, SOX2, and KLF4.
It goes into the vitreous of one eye in a single injection. The genes sit there, silent, until the patient swallows doxycycline, which throws the switch. Eight weeks later the antibiotic stops, and the switch is designed to fall open again.
Somebody thought hard about the off button, and the fact that the off button is an ordinary antibiotic tablet is the most reassuring engineering detail in the whole program. The trial is small, and the trial is honest about itself. Eighteen participants aged 40 to 85, twelve with open-angle glaucoma and six with a condition called non-arteritic anterior ischemic optic neuropathy, which strikes the optic nerve suddenly and for which no approved treatment changes its course.
Two dose levels in the glaucoma group. Follow-up for five years. The primary endpoints are adverse events, blood chemistry, eye pressure, and standard ophthalmology instruments—acuity charts, visual fields, OCT scans, and a test of retinal electrical response called pattern electroretinography.
The registry's estimated date for primary completion is May 2027. It is open-label, with no placebo group. Everyone knows who got what, and each patient is measured against their own earlier self.
For a first-in-human safety study, that is the right design, and there is nothing wrong with it. It also means that whatever the first numbers show, they cannot tell you that the treatment caused it, and somebody is going to report them as if they can. Write that down now, before the headline arrives.
Why the eye, of all places to begin? Because it is a room with a door. An injection into the vitreous stays in the vitreous.
One eye can be treated while the other sits there as a comparison you did not have to recruit. Vision is measurable in numbers that mean something to a patient and not only to a statistician. And if it goes badly, it goes badly in one eye rather than in a liver or a brain.
Something else about the eye has nothing to do with caution, and it's why I find this trial moving rather than merely interesting. Sight is not a biomarker. Nobody wakes up hoping their methylation age has improved.
People wake up hoping to read the label on a prescription bottle, recognize a grandchild across a room, keep their driver's license; that is the difference between a life and a waiting room. A therapy that gave an old eye back some of what glaucoma took would not have added a day to anybody's lifespan. It would have given back a piece of life.
Which brings up the strangest detail in the whole registry, and I haven't seen anyone else point it out. The first human test of the theory that aging is lost information carries no test of information. There is no methylation endpoint, no epigenetic clock, nothing that measures the thing the therapy is supposed to be doing.
The reason is practical: you cannot biopsy the retinal ganglion cells of a living person to see how their methylome is getting on. So, the grandest hypothesis in aging biology will be graded, first, on an eye chart. I think that is exactly right, and not only for want of an alternative.
The chart is the point. Figure 2. Eighteen people, one eye each, and an antibiotic for a switch.
What the first human test actually asks of them. The other experiment is asleep on your couch While that is happening in four American eye clinics, something with a much shorter timeline is happening in about seventy veterinary practices, and it may matter more. Loyal is a San Francisco company whose entire purpose is drugs that extend healthy life in dogs.
Its lead candidate, LOY-002, is a flavored tablet given once a day to dogs ten years and older weighing at least fourteen pounds. The company says it targets the metabolic drift of old age—insulin resistance, chiefly—and aims to extend healthy lifespan and maintain quality of life. Press coverage has described it as borrowing some of the biology of eating less without asking the dog to eat less, which any dog would regard as the entire point.
Loyal has not disclosed the molecule. Chief executive Celine Halioua has said the average dog should be able to take it for under a hundred dollars a month. The trial is called STAY, and it is the largest interventional study ever run in veterinary medicine: about 1,300 senior dogs across roughly seventy clinics, randomized, placebo-controlled, nobody told who got the tablet.
The first dog was dosed in December of 2023. Enrollment closed in July 2025. It was designed to run about four years, because there is no shortcut.
You cannot hurry a survival curve. That is the hard, unglamorous arithmetic underneath every promise ever made about longevity, and it is why the field is so full of people selling supplements instead: a supplement can be sold on Tuesday, and a survival curve takes until 2028. Here is what I want you to notice about STAY.
Its endpoints are lifespan, healthspan, and what the owner says. Quality of life is measured by asking the person who feeds the dog. That is not a soft endpoint standing in for a hard one.
For a dog, and arguably for the rest of us, it is the actual outcome. The dogs in that trial are not laboratory animals; they are somebody's dog, with a name and a preferred spot on the rug, and the study was built around that fact rather than despite it. Three sections, two of them accepted The regulatory story is duller and more consequential.
A conditional-approval application at the FDA's Center for Veterinary Medicine has three major technical sections. Loyal has cleared two. In February 2025 the agency accepted the effectiveness section, determining, in the company's words, that the data "provides reasonable expectation of LOY-002's intended effect of extending canine lifespan." On January 13 this year, it accepted the safety section, behind which sat a study at one, three, and five times the intended dose with no clinically significant adverse events, plus field data from more than four hundred animals, some on the drug for a year.
What remains is manufacturing. On the company's own development page, it is still listed as in progress. Loyal said in February it hoped to file during 2026 with a review of roughly six months; the veterinary trade press expected 2027.
Nothing public says it has been filed. Until it is filed and accepted, LOY-002 is not approved or available, and anyone telling you otherwise is guessing. The word "conditional" is doing real work.
It lets a company sell before proving full effectiveness, on a lower evidentiary bar, while safety and manufacturing stay where they were. It runs for one year, renews annually for up to five, and the label has to say what it is. So, it is not a verdict.
It is a trapdoor with a five-year timer, and STAY is what decides whether Loyal walks off the platform or through it. Figure 3. Two sections down, one to go.
The one to go is manufacturing, and it has stalled better programs than this. Why a dog pill is a human event The molecule is not a human medicine. A veterinary approval creates no human indication, no shortcut past human trials, and no excuse for anybody to raid the dog's bottle.
I want to be very clear about that, because somebody will try. What it creates is a precedent, and precedent is the scarcest commodity in this field. For decades, the objection to treating aging was never that biology was implausible.
It was that aging is not an indication. You can get a drug approved for heart failure, for diabetes, for macular degeneration — for the wreckage. Not for the thing doing the wrecking, because a regulator wants a defined population, a measurable endpoint, and a trial that ends, and aging appeared to offer none of the three.
The most famous attempt to break that logjam in people is a trial called TAME, designed by Nir Barzilai at Einstein around a generic diabetes drug and more than three thousand participants at fourteen institutions. It has been ready for years. It has never enrolled a single patient, for want of something between $45 and $70 million, because metformin is generic and nobody stands to make money on it.
Barzilai told Science|Business last August, after another round of near misses, that "the promise versus the funding is ridiculous." That is the sound of a field stuck on a technicality rather than on a fact. Loyal has not solved that. It has done something smaller and more useful: produced a worked example.
A defined population, ten years and older, fourteen pounds and up. A measurable endpoint, which is how long the dog lives and how well. A safety package accepted in writing.
A manufacturing standard. A trial that ends. Turn all of that in, and the question stops being philosophical and starts being a yes or a no.
The most valuable thing Loyal ships may not be a tablet. It may be a category. And dogs, unlike mice, age in public.
They sleep in our houses, breathe our air, eat too much of what we eat, and get our diseases on a clock about seven times faster than ours. A conditionally approved canine longevity drug would generate real-world evidence on metabolism, on several illnesses at once, on quality of life in ordinary homes, at a scale and speed human aging research has never had and cannot buy. The gap that is actually the target Now the numbers explain why any of this is worth the trouble.
An American born today can expect about 79.0 years, according to the National Center for Health Statistics in January, which was up six-tenths of a year from 2023 and the best news that dataset has carried in a while. But the last stretch of that expectancy is not made of the same stuff as the middle. A 2024 study in JAMA Network Open measuring 183 countries put the global distance between how long people live and how long they live well at 9.6 years, widening by 13 percent since 2000.
The United States had the widest gap on Earth: 12.4 years. Twelve years. That is not a statistic; it is a decade and a bit of somebody's actual life, spent alive and diminished, and we lead the world in it.
That gap is what these programs are aiming at, and it is also, by the most cited estimate in the field, the largest prize in medicine. Andrew Scott, Martin Ellison and David Sinclair calculated in Nature Aging in 2021 that slowing aging enough to add one year to life expectancy would be worth $38 trillion, and ten years $367 trillion. The sentence of theirs that ought to govern this whole enterprise is the quieter one: "a compression of morbidity that improves health is more valuable than further increases in life expectancy." More years is not the goal.
Better years are the goal, and the arithmetic agrees. The institutions have started to move. ARPA-H put up to $144 million behind seven teams in February under a program called PROSPR, whose stated aim is extending "the number of years aging adults live healthy lives." The $101 million XPRIZE Healthspan, which asks competitors to restore muscle, cognitive, and immune function by at least a decade in people aged 50 to 80, named ten finalists on August 11, each with a million dollars, with trials running to 2029 and the grand prize in 2030.
Money that used to go to supplements is going to endpoints. Figure 4. Twelve and a half years, on average, of being alive and diminished.
That is the number to beat. Everything that could go wrong, at the same size as everything that could go right Hope that cannot survive contact with the evidence is not hope; it is marketing. So here is the other column.
AAV vectors provoke immune responses, and once a patient has antibodies to the vector, you generally cannot give it again. An inducible switch that leaks is a therapy nobody can turn off, which is precisely the nightmare in a system built on genes that make cells divide. A 2024 review in Nature Communications is blunt about what happens when reprogramming factors are left on: sustained expression "may have lethal effects in mice" and "may result in liver and intestinal failure." Partial reprogramming that goes a step too far does not produce a young cell; it produces the wrong cell.
An eighteen-person open-label study cannot establish that anything worked, and the first encouraging number out of it will be reported as though it can. Life Biosciences' primate data — the results that got it into humans — exist as conference presentations at ARVO in 2023 and the American Academy of Ophthalmology in 2024. They report improved retinal electrical responses and healthier axon bundles.
They have not been published in a peer-reviewed journal, and the animal counts and follow-up durations are not public. That is not unusual for a company protecting an IND, and it is also not the same as evidence you can check. On the other side of the ledger, a dog drug that adds a few good months to a selected trial population may not generalize to dogs in general, never mind people.
And manufacturing can kill a program the science has already blessed, which is exactly the door LOY-002 is standing in front of right now. Then there is the failure mode nobody puts in a slide deck. The therapy can work completely and reach almost nobody.
The pill can be cheap while gene therapy, the specialist center, the imaging, and five years of monitoring are not. If this field builds rejuvenation medicine the way it built gene therapy — brilliant, priced in the millions, available in a dozen cities — then the phrase "longer, healthier lives for all" will have been a lie told in good faith, which is still a lie. I would put access into the development problem at the start rather than at the end.
Can it be scaled? Can it be delivered outside an academic hospital? Can a person three hours from the nearest one be followed up?
Do the trials include the people who will need the drug? None of those questions is answered by anything described above, and all of them are answerable if somebody decides to answer them early. One more thing, and then I will stop hedging.
A definition of success that sorts people into those who age well and those who fail at it is not worth having. Nobody becomes less valuable when they need help. The reason to compress the sick years is to reduce suffering that did not have to happen and to widen the range of what a person can still choose — not to hand out grades at the end of a life.
What would settle it? I do not forecast. I read calendars.
Here is the one that matters. Between now and the end of 2027: whether Loyal files its manufacturing section and whether the FDA accepts it. The first safety data from ER-100, with primary completion estimated for May 2027 and the possibility of something sooner.
Whether Life Biosciences publishes its primate work somewhere a reviewer has been. And NewLimit, the reprogramming company, funded with $435 million in June, is putting its first medicine into people, which it says will be next year. Between 2028 and 2030: STAY reports.
Four years on from December 2023, the first placebo-controlled comparison of lifespan and healthspan ever run to a design an American regulator signed off on. If LOY-002 has been conditionally approved by then, that is also when it either converts to a full approval or runs out its five-year clock. Whether ER-100 reaches a randomized trial in either disease.
The XPRIZE grand prize in 2030 and the PROSPR trials reporting through 2029. And the question sitting behind all of them, which nobody has scheduled: whether any human regulator, watching an animal drug get approved to extend a life, decides to write the pathway for the human version. My line of inquiry, not a prediction: the thing standing between us and this medicine is not biology.
Biology has produced more credible results in the last five years than in the fifty before. The obstacle is trial design — how to prove something about time in less time than you are trying to measure. Every program in this piece is an attempt to solve that puzzle, with a contained tissue that reports fast, or a species that lives faster, or a surrogate somebody will accept.
The interesting question for the rest of this decade is which of those shortcuts a regulator takes first. Figure 5. Nobody has to guess.
Everything in this story reports on a date somebody has already stated. What we are actually hoping for I want to be careful about the size of the hope, because the field's besetting sin is inflating it. Nobody in this story is promising immortality.
The most optimistic honest reading of ER-100 is that a handful of people with damaged optic nerves might see a little better and that nothing bad happens to them. The most optimistic honest reading of LOY-002 is that old dogs get a few more good seasons. Set against the fantasies that circulate in this business, those outcomes are modest.
They are also, if they land, the first two bricks of something we have never had. Think about what medicine currently offers a person at seventy-five. It offers to manage things.
A statin for the arteries. Drops for the pressure in the eye. A stent when the artery closes anyway, a walker when the hip goes.
All of it good, all of it late, all of it addressed to the consequences of a process nobody is permitted to treat. The entire apparatus is built around the assumption that the underlying thing is weather and all we can do is put up shutters. What those four eye clinics and seventy veterinary practices are testing is the other assumption.
That the underlying thing is a process. That processes have mechanisms. That those mechanisms can be modified, and the modification can be measured, and the measurement can be argued about in public by people with instruments.
That is not a cure for death. Death is not the problem this work is aimed at, and I would be suspicious of anyone who said it was. The problem is the confiscation that comes first — the reading lamp that stops being enough, the stairs that turn into a decision, the keys somebody quietly takes away.
Twelve and a half years of it, on average, in this country. That is the number to beat. So: a person in Boston with a sore eye and a bottle of antibiotics on the kitchen counter.
Thirteen hundred dogs on rugs across America, half of them on a pill and half on nothing, and not one of their owners knowing which. Somewhere in that unglamorous scene is the first serious attempt our species has made to treat time as a variable rather than a sentence. It may not work.
It will certainly take longer than anybody wants. And it will be oversold, by people with creams and by people with newsletters, which is why the sentences above carry dates and sources and why the failures are printed at the same size as the promises. But the trials are running now.
The dogs are living their years, with the pill or without it, and in a few years, somebody will open the envelop