Science Lead @temple. Biomed Scientist in #longevity. Mentor & Investor in Health startups. Past CTO @BioVivaScience, Researcher @CASMIORG, Founders Health.
Oxford, England
Joined May 2008
- Tweets10.6K
- Following2.8K
- Followers14K
- Likes7.8K
Pinned Tweet
A weekly jab in the belly is generating more revenue than the entire AI industry.
Ozempic + Mounjaro: $71B in 2025.
OpenAI + Anthropic: $29B.
And they've barely started. ~2% of the 800 million eligible patients can currently access them.
h/t @DrSamuelBHume
🤖 Made with AI
Your tumor's own DNA can be turned into a vaccine against it.
One is being designed in public right now for a 34-year-old with stage 4 pancreatic cancer.
The team sequenced Yash's tumor, picked 24 targets, and wrote them into one strand of mRNA, the molecule behind COVID vaccines. The goal is to teach his immune system what to hunt.
Then they posted the full design online for outside experts to review and improve before a single dose is made.
Early results are encouraging. In a New York trial after pancreatic cancer surgery, 8 of 16 people mounted an immune attack, and 6 of those 8 had no relapse about three years later.
A 260-person trial of that vaccine is underway. Another personal mRNA vaccine kept 82% of melanoma patients cancer-free at two years, versus 70% on immunotherapy alone.
Yash's cancer has spread, which makes this a tougher case.
Bottom line: cancer medicine is becoming personal, and it is now being built in the open. Every expert who checks this design can help improve the next one.
We’re making a personalized cancer vaccine for Yash, a 34 year-old pancreatic cancer patient, in the open with his consent.
We want external experts to review our design and make their own contributions before we manufacture it at the RNA Foundry.
RNAFoundry.com/yash
All of your cells carry the same DNA, but each kind of cell reads different genes. So a one-letter DNA change can raise your risk of a disease through one kind of cell and do nothing in the rest.
Google DeepMind's AI, AlphaGenome 2, predicts which cell.
To test it, DeepMind gathered one-letter changes that genetic studies of people had already tied to turning a gene up or down, in 28 kinds of immune cell. The AI had to pick them out from a mix of changes that do nothing. It found 30 out of 100. The earlier version, retrained on the same data, found 16.
It also caught a known case, a change that raises the risk of bowel disease by turning up a gene called ETS2 in immune cells of the gut.
All of this was scored against known answers. DeepMind reports no lab tests yet.
Bottom line: knowing which cell a risky DNA change works in could tell researchers where to look for a treatment. This AI nearly doubles how many it links to the right cell.
Your heart, brain, and immune system share the same DNA sequence. So how can a single-letter change affect disease in some cells, while sparing the rest?
AlphaGenome 2 is our new AI model that zooms in on specific cells to help scientists understand what drives disease. 🧵
1
9
765
Ovarian cancer is often found too late.
By the time it’s diagnosed, more than half of cases have already spread. At that stage, about 3 in 10 women live five more years. When it’s found early, that number is over 9 in 10.
The tests we use today aren’t good enough.
In a trial of 200,000 women, screening did not clearly reduce deaths.
Now researchers have tested a new blood test that looks for a scrap of tumor RNA.
They used blood samples from 506 women, including 43 with the earliest stage: a tiny precancer in the fallopian tube, where the most common type often begins.
In one test, the new method told apart a woman with that precancer from one with a harmless lump 99 times out of 100.
CA-125, the blood test doctors use today, did it 31 times out of 100, worse than a coin toss.
Important caveat: it has not yet been tried in women before diagnosis.
Bottom line: if it holds up in women who feel fine, ovarian cancer could finally be caught while it is still curable.
Sources and Notes
The test: Tu-stroma, a long non-coding RNA carried in tiny vesicles in serum. Wilhelm, Kumar, Gogna, Madan and colleagues, Nature Medicine, 10 Nov 2026. doi.org/10.1038/s41591-026-0… @NatureMedicine
Who was tested: 506 people with tissue-confirmed diagnoses. 43 with STIC (serous tubal intraepithelial carcinoma, the precancer in the fallopian tube), 232 with high-grade serous ovarian cancer at stages I to IV, 143 healthy controls, 88 with benign masses. Groups were defined by diagnosis, so this is a case-control study, not screening.
Accuracy (area under the curve): STIC vs benign masses 0.99 against 0.31 for CA-125; STIC vs healthy controls 0.99 against 0.59; cancer vs controls 0.99 against 0.96; cancer vs benign masses 0.98 against 0.84. "99 times in 100" is the chance the test ranks a random precancer sample above a random benign-mass sample. No rise in 116 inflammatory or physiological comparators or 160 other cancers.
Limits: 43 precancer cases; the authors say prospective testing in consecutively enrolled women is required before the test's real-world accuracy is known. No sensitivity or cutoff is reported in the abstract.
Stage and survival: SEER, ovarian cancer, 54% diagnosed after distant spread; five-year survival 91.9% localized, 31.5% distant. seer.cancer.gov/statfacts/ht…
Screening today: UKCTOCS, Lancet 2021, 202,562 women, median 16.3 years; screening with CA-125 and ultrasound did not clearly reduce ovarian cancer deaths (no statistically significant reduction). doi.org/10.1016/S0140-6736(2…
3
215
A toothbrush that helps kill bacteria while you brush? That is the claim behind a new design that uses the electricity its own bristles make as they bend. No battery needed.
In a trial at Peking University, adults with plaque and bleeding gums used either the new brush or a standard one, chosen at random, for three months. With the new brush, plaque, swollen gums and bleeding improved by about 35% to 40%. The ordinary brush improved by about 15%.
The electricity comes from the plastic bristles themselves. Each bend builds a small charge, the same effect behind the spark when you click a lighter. In saliva, that charge makes a chemical that white blood cells also use to kill germs.
The study involved about 110 people at one hospital and did not track cavities or lost teeth.
Bottom line: it is still just brushing twice a day, but with bristles that may leave fewer bacteria behind. A bigger trial will show whether that protects teeth.
Sources and Notes
Spotted via @ScienceMagazine. Paper: Lu and colleagues, Peking University School and Hospital of Stomatology, "Self-charged toothbrush facilitates tooth healthcare maintenance", Science Advances, 4 Nov 2026. doi.org/10.1126/sciadv.aeg57… (PMID 42981395, PMC13733809, CC BY-NC 4.0). @ScienceAdvances
Trial: registered ChiCTR2300078294. 135 screened, 115 randomized by sealed envelope to one of four brushes (two bristle sizes of PVDF, two of nylon); 109 analysed at 3 months (30, 25, 28, 26). Adults with plaque index above 1.5 and at least 20 bleeding sites. The paper does not report who was blinded.
Results: Fig. 4B to D, month 3, plaque, gingival and bleeding index "recovery" read from the figure: PVDF about 34 to 41%, nylon about 13 to 16%, P<0.0001 between groups. Plaque bacteria linked to decay and gum disease fell with PVDF (sequencing, n=12).
Dish tests: after 9 h of contact with machine brushing, PVDF killed 78.21% (0.24 mm) and 56.57% (0.16 mm) of P. gingivalis against 8.41% for nylon; S. mutans 52.80% and 45.88% against 9.29%. The charge makes reactive oxygen (hydroxyl and superoxide, shown by EPR); an antioxidant blocked the killing.
Limits: one hospital, 3 months, no cavity or tooth-loss outcomes, no comparison with electric or sonic brushes, long-term safety and cost not studied. Enamel roughness did not rise after simulated 3-month use.
Gum disease: WHO Global Oral Health Status Report 2022, severe periodontal disease in about 19% of people over 15, more than 1 billion cases.
2
190
Stopping an Ozempic-type diabetes drug was linked to about double the risk of heart attack, stroke, or death in a new study.
But before you panic, look at *who stops*.
People often stop a medicine because they get sick with something else. If that sickness is what drives the higher risk, the drug gets blamed for part of the story it did not cause.
Gym quitters work the same way: some of the risk is losing the exercise, and some is the illness/injury that forced them to stop.
That matters here.
Records from 57,000 Italians with diabetes fit that pattern. Most of the extra events in people who stopped were deaths: about 22 per 1,000 a year, versus 5 per 1,000 in those who stayed on the drug.
If the drug alone explained that gap, taking it would need to prevent about 4 out of 5 deaths.
The big trials found closer to 1 in 7.
The records did not say why people stopped. The authors point to frailty and cancer as possible reasons.
Bottom line: these drugs do protect people, and stopping gives up some of that protection.
But the doubled risk in this study far outweighs the known benefit, suggesting the main question is not just what happens when people stop, but *who* stops.
Stopping GLP-1 treatment linked to twice the risk of heart attack, stroke or death compared with continuing, in people with type 2 diabetes and no cardiovascular disease, according to a new study.
Find out more in @LancetRH_Europe: spkl.io/60137wUtS
ALT Research in context panel for a paper published in The Lancet Regional Health - Europe entitled \"GLP-1 receptor agonist discontinuation and cardiovascular events and all-cause mortality in people with type 2 diabetes: a retrospective cohort study\". Copyright: 2026 The Author(s). Published by Elsevier Ltd. CC BY 4.0.
Most of us will watch someone we love face cancer, Alzheimer's, or heart disease. In this clip, Anthropic's CEO says AI will make them "relics of the past" in under 10 years.
I hope he's right.
AI is already speeding up the first half of drug development: finding the drug. It can predict the shape of most proteins, work that once took a lab months or years.
The hard part is proving a drug works in people. That depends on the disease itself.
The trials behind the two Alzheimer's drugs that slow the disease ran about 18 months, because the decline takes that long to show. A Novartis heart drug was followed in 8,322 people for almost 5 years, and when the full results came out on Saturday, it had not prevented heart attacks or strokes.
No computer can make a disease move faster.
Today, a drug takes about 10 years from its first human dose to approval. About 9 in 10 fail along the way. That is where AI can help most. If better targets turn 1 success in 10 into 2, twice as many medicines reach patients from the same number of tries.
So my bet for 2036 is more medicines for these diseases, each one proven in people. Making them relics will take longer.
Dario Amodei says cancer, Alzheimer’s, heart disease, and mental illness will be relics of the past in less than a decade, and that "we can create an age of miracles"
“You know, if today we look at things like cancer, Alzheimer’s, heart disease, mental illness, these will be relics of the past in, I think, less than a decade."
"And I believe that the company which gets the risks right will also be the company that has the license to operate that society trusts and society believes can build the benefits in the right way."
"I think if we do this right, we can create an age of miracles where humanity is able to solve problems that have eluded us for millennia."
"We have to make sure that all the risks, all the ways that things can go wrong with this technology, we very carefully address those risks as we bring the benefits, and as we have a viable business at the same time.”
Cancer, Alzheimer's, heart disease, and aging itself all play out inside your cells. Today an AI built a 3D model of one, from code.
Anthropic's Claude Fable 5.2 placed the nucleus that holds your DNA, the mitochondria that fuel you and the machinery that builds your proteins.
That is impressive, and it is still a picture.
The goal is a working copy of a human cell inside a computer. Press play, and you could watch what a cancer drug does to it, or how the cell ages, before anything touches a person.
So far, only simple microbes have been run that way. In March, a team at the University of Illinois ran a bacterium with 493 genes through its whole life, until it split in two. Your cells have about 40 times as many genes.
Those simulations are code built from experiments. An AI that writes working biology code could help build one for a human cell.
Today's AI cell is a picture. The next step is a cell that runs, one where a way to slow aging could be tried in a computer before anyone tries it on you.
A thread of early experiments with Claude Fable 5.2:
A 3D model of a cell, built entirely from code one organelle at a time. Built by @kevin_t_ngo.
A cholesterol shot from China just posted eye-catching numbers: it cut LDL by about two thirds and Lp(a) by nearly half. But the hard truth is unchanged. Only lowering LDL has been shown to prevent heart attacks.
The drug is tafolecimab, sold in China as Sintbilo by Innovent. In 198 Chinese adults, most at low heart risk and off other cholesterol drugs, LDL fell about 65% in 12 weeks, versus 4% to 7% on placebo. Innovent says Lp(a) fell by nearly half. The results are from the company, and they have not yet been peer-reviewed.
Tafolecimab works the same way as Repatha, a shot approved in the United States in 2015. In a trial of 27,564 people with heart disease who were already on statins, Repatha cut LDL by about 60%. Over about two years, 15 fewer people per 1,000 had a heart attack, stroke, or heart death.
Lp(a) is a different story. It is an inherited cholesterol particle, high in roughly 1 in 5 people, and no approved drug is aimed at it.
On Nov. 7, a much larger Lp(a) drop still did not translate into better outcomes. Novartis's pelacarsen cut Lp(a) by nearly three-quarters in 8,322 people with heart disease. Over almost five years, heart events were essentially unchanged, 12.3% on the drug and 12.1% on placebo.
So LDL remains the number that matters most. This shot lowers it about as far as Repatha does. Lp(a) drugs still have to prove they can prevent heart attacks.
The FDA's new eye warning for Ozempic-type drugs amounts to 2 cases in a stadium full of people.
Picture MetLife Stadium near New York, where this summer's World Cup soccer final was played. Fill its 82,500 seats with people taking a GLP-1 drug like Ozempic for a year.
About 3 of them would develop NAION, a stroke of the eye's optic nerve that suddenly dims sight in one eye, even without the drug. The drug may add 2 more.
Those 2 cases sit on a very small base. Across 126 trials, mostly in people with diabetes or obesity who are already at higher risk, there were 10 cases on the drug and 5 on placebo. That gap could still be chance, though one large FDA study in diabetes found about 40% more cases than on an older diabetes pill.
Now fill the same seats with people who have heart disease and excess weight, like those in the SELECT trial. In a year, semaglutide would prevent about 370 heart attacks, strokes, or heart deaths.
If you take one of these drugs to protect your heart, the numbers still favor staying on it.
The US health regulator on Monday informed patients and healthcare providers of a potential increased risk of a serious eye condition associated with the use of GLP-1 drugs for weight loss and diabetes. cnn.com/2026/11/09/health/gl…
Nothing that will reverse skin aging next year is a secret today. The next wave of ideas is already out in the open, sitting in public trial registries. The catch is simple: the furthest along are still only at the first human step.
Rubedo is testing a cream drug designed to clear senescent cells, the damaged cells that stop dividing and pile up in aging skin, in its first human study.
Krystal's spinoff Jeune injected a collagen gene into skin, saw a modest result, and in 2025 moved to a new version.
A Shanghai team plans the first trial of collagen mRNA, in 10 people, in January 2027. Cell reprogramming for skin, from Turn Bio, has not reached people. None is near a pharmacy.
And the timeline is brutal. From first human dose to approval takes about 10 years, and only about 8 in 100 drugs make it. GLP-1 drugs took nearly 40 years from hormone discovery to the weight-loss shot, in public the whole way.
Two approaches already have trial proof. In an Australian trial of 903 adults, daily sunscreen showed no detectable increase in skin aging over 4.5 years, and tretinoin cream has been approved for sun damage since 1995.
Bottom line: the new skin drugs are years from approval. The ones with proof are sunscreen and tretinoin.
Heart attacks in women are often missed because not all of them look like chest pain.
In a US registry of more than 1.1 million heart attack patients, 42% of women had no chest pain, versus 31% of men. The gap was widest under 45.
The warning signs can start early and look ordinary: in a survey of 515 women after a heart attack, 95% recalled new symptoms a month or more before, most often unusual tiredness, poor sleep, or shortness of breath.
And being believed is part of the problem. Of women under 55 who saw a doctor for these symptoms before their heart attack, 53% said the doctor did not think it was their heart, against 37% of men.
Bottom line: chest pain still matters. So do new breathlessness and exhaustion, especially if they are severe or do not settle.
Most stomach cancers start with a common infection, and a breath test can spot it.
The culprit is H. pylori. It usually begins in childhood, stays for life, and may explain about 9 in 10 cancers of the stomach outside the part nearest the gullet. Most people carry it without symptoms, though some have recurring acidity.
Treating it lowers risk. In a Korean trial of 1,838 carriers with a parent, sibling or child with stomach cancer, a week of two antibiotics plus an acid-suppressing pill cut stomach cancer over 9 years to 1.2%, versus 2.7% with placebo. Across seven trials in 8,323 healthy adults, cases fell by almost half.
Most of the evidence comes from East Asia, where stomach cancer is common.
Bottom line: if acidity keeps coming back, ask your doctor about an H. pylori breath or stool test. Weight loss, black stools, repeated vomiting or trouble swallowing need a doctor soon.
STOMACH CANCER never forgives anyone. I received him in ER at 40, already at Stage 4.
He had 5 warning SIGNS for months, but ignored them until one endoscopy changed everything.
1. ACIDITY THAT KEPT COMING BACK
1
17
2,549
For the first time ever, scientists filmed DNA moving from one human cell to another.
Dividing cells do not always split their chromosomes cleanly. Sometimes they leave loose bits of DNA outside the nucleus, and tumor cells seem especially prone to that mistake. Peter Ly's lab in Dallas wanted to know where those stray bits go.
So the team tagged DNA green in some cells and red in others. If green DNA showed up inside a red cell, it had to have come from a neighbor. To generate more loose DNA, they used drugs that make the split go wrong more often.
Then they saw it happen. One green fragment slid through a thin tube connecting the cells and entered a red one. It took about 6 hours.
With the drugs, about 4 in 100 cells passed DNA. Without them, about 1 in 100 did.
The team also tested whether the moved DNA actually works. In one experiment, cancer cells passed a gene that blocks an antibiotic to nearby cells, and those cells survived. That was rare, and everything happened in lab dishes.
Bottom line: the big question now is whether tumors can spread cancer-drug resistance this way.
Sources and Notes
Human cells in lab dishes passed pieces of DNA to neighboring cells along thin membrane threads, more often when chromosomes were sorted wrongly during division. Some transferred genes stayed active. Nobody has shown this in a tumor yet.
· · · · ·
1. THE PAPER
- Maurais EG, ... Ly P. Genome instability triggers intercellular DNA transfer between human cells. Cell 2026;189(15):4548-4561. PMID 42161273, open access (CC BY 4.0), PMC13193222.
doi.org/10.1016/j.cell.2026.…
· · · · ·
2. WHAT THEY FILMED
- Retinal epithelial cells (RPE-1) with green (H2B-GFP) or red (H2B-mCherry) labeled chromatin were grown 1:1 and treated with CENP-E and Mps1 inhibitors, which cause chromosome mis-segregation and micronuclei. Both cell sets got the drugs.
- Figure 1F follows one green DNA fragment along a nanotube-like thread into a red cell over 375 minutes.
- The authors describe the threads as nanotube-like, reminiscent of tunneling nanotubes, about 10 to 60 micrometers long.
· · · · ·
3. HOW OFTEN
- Figure 1C: about 4.2% of cells had DNA transferred between nanotubes with the inhibitors vs about 1.2% without (3 experiments, P = 0.0013). Cells were held in place with cytochalasin D over a 16-hour window for this count.
- The number of nanotubes per 100 cells did not change; the DNA traffic did.
· · · · ·
4. DOES THE DNA WORK
- Male DLD-1 donor cells carrying a neomycin-resistance gene on the Y chromosome (Yq11.221) were mixed with female HeLa cells. After weeks of drug selection, some female cells carried Y-derived DNA as extra fragments and made the resistance gene's RNA and protein.
- Transfer was rare and needed selection to detect.
· · · · ·
5. LIMITS
- Lab cell lines only. The cancer link is the authors' open question, not a finding.
- Cas9 cuts and radiation (Figure 2) also increased transfer, in separate experiments.
Weight loss can change your face, whether you use Ozempic or not.
In a survey of 1,226 people in one weight-loss program, 13% noticed a thinner face after losing less than 10% of their weight. After losing more than 20%, that jumped to 50%.
The scans tell a similar story. In 20 people taking GLP-1 drugs, the group that includes Ozempic, the middle of the face lost about 7% of its volume for every 22 pounds lost. Most of that came from fat under the skin.
Among people who lost 10% to 20% of their weight, 38% of those on the drugs noticed a thinner face, compared with 19% of those not on them. The survey cannot separate the effect of the drug from the effect of faster weight loss.
Bottom line: if you lose a lot of weight, some of it may show up in your face.
7
10
1,246
Sources and Notes
The evidence is small and indirect: a scan study of 20 adults and a self-reported survey from one program. Together they point to facial volume loss tracking the amount of weight lost, but neither tests whether slower weight loss spares the face.
· · · · ·
1. THE SCANS
- Midface volume fell a median 9% overall, 11% in the fat just under the skin and 7% in the deeper fat pads.
A regression put it at about 7% per 10 kg (22 lb) lost.
doi.org/10.1002/ohn.1209
- Sharma and colleagues, Vanderbilt facial plastic surgery, Otolaryngology Head and Neck Surgery, 2025.
20 adults (median age 54) with head CT or MRI before and after a GLP-1 drug. Average loss 11.0 kg in about 321 days.
doi.org/10.1002/ohn.1209
· · · · ·
2. THE SURVEY
- More weight loss went with more people reporting decreased facial volume: 48 of 380 (13%) who lost under 10% of body weight, 147 of 395 (37%) at 10 to 20%, 163 of 325 (50%) above 20%.
- At the same weight loss, people on a GLP-1 drug alone reported it more often than people on non-GLP-1 regimens: 38% vs 19% at 10 to 20% loss, 52% vs 41% above 20%.
The authors say it can't separate a drug effect from the amount or speed of loss, and speed wasn't recorded.
- Rao, Kourosh and colleagues, Dermatology and Therapy, online 11 August 2026.
Anonymous survey of 1,226 patients in one virtual weight-management program (Mass General Brigham review board), on GLP-1 drugs, other drugs or both.
doi.org/10.1007/s13555-026-0…
· · · · ·
3. THE TWINS
- A BMI 4 points higher went with an older look before 40 and a younger look after 40.
Guyuron and colleagues, Plastic and Reconstructive Surgery, 2009.
186 pairs of identical twins were photographed at the Twins Day Festival in Twinsburg, Ohio, with a panel judging each twin's apparent age.
doi.org/10.1097/PRS.0b013e31…
- Bariatric surgery gives a small, mixed signal. Of 14 patients photographed before and after, five had a change in perceived age that passed a statistical test: three looked older, two younger.
The authors read substantial loss as older but more attractive.
doi.org/10.5999/aps.2015.42.…
· · · · ·
4. THE LONGEVITY PART
- On the longevity jab: heart attack, stroke or heart death hit 6.5% on semaglutide vs 8.0% on placebo over a mean 39.8 months (hazard ratio 0.80).
SELECT enrolled 17,604 adults aged 45 or older with heart disease, overweight and no diabetes, and its main outcome was heart events. Novo Nordisk funded it.
doi.org/10.1056/NEJMoa230756…
· · · · ·
5. LIMITS
- 20 patients were scanned at one hospital (deep fat did not track weight loss), the survey is self-reported from one program with mostly white women, and the twins were photographed once.
The twin result links BMI to apparent age and says nothing about facial fat. None of it tests whether a slower loss spares the face.
- A 14 October 2026 review of 46 reports reads it as weight-loss deflation on top of baseline anatomy and aging, with speed, menopause and hormone therapy still uncertain.
doi.org/10.1093/asj/sjag230
· · · · ·
6. WORDS USED
- GLP-1 drugs are weight-loss and diabetes drugs such as semaglutide (Ozempic).
- BMI is body mass index, a weight-for-height measure.
- Bariatric surgery is weight-loss surgery.
- A regression is a statistical fit that estimates how much one thing changes as another rises.
- Hazard ratio below 1 means fewer events on the drug than on placebo.
· · · · ·
7. THE PHOTOS ON THE CARD
- The woman on the card is a published case: 39 years old, on a GLP-1-based drug, 18 lb lighter in the 4 months between the photos and 40 lb lighter after 12 months. Between the two photos she had only wrinkle-relaxing injections in the neck, frown lines and crow's feet. Filler came after all the weight was off.
Moradi and colleagues, Aesthetic Surgery Journal Open Forum, 2026, Figure 3. The patient consented to publication of her images.
doi.org/10.1093/asjof/ojag01…
1
446
Headline from a big trial: stopping a statin after 70 was called "generally safe." But the trial's own numbers tell a different story.
STREAM followed 1,880 people age 70 and up. They all had other illnesses but no heart disease. Half stopped their statin.
What happened? Over 4 years, 12.7% of the people who stopped had a heart attack, stroke, or died. That compared with 9.2% of people who stayed on the pill.
That is about 1 extra heart attack, stroke, or death for every 40 people who stopped within 2 years.
Muscle pain did not improve.
After the trial ended, the independent safety board told people who had stopped to ask about restarting.
Bottom line: if someone over 70 is thinking about stopping a statin, ask the doctor about the 1 in 40.
Presented at #ScientificSessions26:
In older adults with at least two chronic conditions and no known atherosclerotic cardiovascular disease, statin discontinuation was noninferior to continuation for a composite of myocardial infarction, stroke, or death at 2 years. Full STREAM trial results: nej.md/47U1X9N
@AHAScience
ALT A graph from the Original Article "Discontinuing Statins in Older Adults without Cardiovascular Disease." The trial name, STREAM, is in parentheses after the title. The graph shows estimated cumulative incidence of primary-outcome events. The trial results are being presented at the American Heart Association Scientific Sessions 2026. The logo of the New England Journal of Medicine is at the bottom.
Sources and Notes
STREAM is one open-label trial in Europe. Stopping met its main goal at 2 years, yet more people who stopped had a heart attack or stroke or died. The trial says nothing about people who already have heart disease.
· · · · ·
1. THE TRIAL
- 1,880 adults aged 70 and older (average age 77), each with at least 2 other chronic conditions, no known heart or artery disease, and on a statin for at least a year to prevent a first event.
- 938 were assigned to stop and 942 to keep taking the statin, at 24 centers in Switzerland, France and the Netherlands. Open-label, with outcomes judged by assessors who did not know the group.
- NEJM, 7 Nov 2026, presented at the American Heart Association Scientific Sessions.
doi.org/10.1056/NEJMoa260663…
clinicaltrials.gov/study/NCT…
· · · · ·
2. THE MAIN GOAL (WHY IT WAS CALLED SAFE)
- The test was time free of heart attack, stroke or death over 2 years: 96.5 weeks after stopping vs 98.5 weeks on the statin, a gap of 2.0 weeks. The trial had allowed up to 5.2 weeks before calling stopping worse.
- In numbers: difference -2.0 weeks (95% CI -3.7 to -0.3).
- The American Heart Association release headline: "Discontinuing statins generally safe for older adults without heart disease."
newsroom.heart.org/news/disc…
· · · · ·
3. THE EVENT COUNTS
- Over a median of 103 weeks (up to 4 years for some), 119 of 938 who stopped (12.7%) had a heart attack or stroke or died, vs 87 of 942 who continued (9.2%).
- In numbers: hazard ratio 1.38 (95% CI 1.05 to 1.81).
- The authors estimate 1 extra heart attack, stroke or death within 2 years for each 40 people who stop. The longer follow-up gap, 3.5 in 100, works out to about 1 in 29.
- In a subgroup of 235 people, LDL cholesterol a year later was about 176 mg/dL (4.6 mmol/L) after stopping vs 99 mg/dL (2.6 mmol/L) on the statin.
- Quality of life, muscle pain, falls and muscle loss did not differ.
- After the trial ended, the data and safety monitoring board recommended that people in the stopping group be advised to discuss restarting with their doctor (reported by the presenter, Nicolas Rodondi).
acc.org/latest-in-cardiology…
hcplive.com/view/stream-stat…
· · · · ·
4. STAREE, THE OTHER SIDE OF THE QUESTION
- 9,971 Australians aged 70 and older without heart disease or diabetes took atorvastatin 40 mg or placebo for a median of 5.9 years.
- Major cardiovascular events (heart attack, stroke, artery procedures and heart death) were 6.0% vs 8.3%, a 30% relative cut (hazard ratio 0.70). Heart deaths did not differ, and the second main goal, years alive without disability, was not changed.
- NEJM, 28 Aug 2026, presented at the European Society of Cardiology.
clinicaltrials.gov/study/NCT…
hcplive.com/view/staree-stat…
· · · · ·
5. LIMITS
- Open-label, so people knew their group. About 14% crossed over (131 who stopped restarted; 112 who continued stopped).
- People recruited by family doctors lost more event-free time after stopping (3.4 weeks) than those recruited in hospitals (1.1 weeks).
- STREAM tested stopping in multimorbid people; STAREE tested starting in healthier people. Neither covers anyone who has had a heart attack, stroke or stent.
- PREVENTABLE, a US placebo trial of about 20,000 adults aged 75 and older, is still running.
clinicaltrials.gov/study/NCT…
2
348
Two of this year's biggest heart trials lowered what they aimed at, and heart attacks and strokes did not fall. I think timing is the lesson.
1. ZEUS tested an inflammation shot. It cut a key inflammation marker by 64% in 6,376 people with artery and kidney disease, but events stayed flat: 17.6% vs 17.7% on placebo.
2. Lp(a)HORIZON lowered Lp(a), an inherited cholesterol particle, by 78% in 8,323 people with artery disease. Events were 12.3% vs 12.1%.
3. Paul Ridker, who led ZEUS, said the horse may already have been out of the barn.
4. Plaque starts young. In autopsies of young Americans, 1 in 10 teenagers had fatty plaque in a heart artery. People born with genes that keep LDL low get about 3 times the protection of statins started later.
5. Stroke survivors treated to a top blood pressure under 120, not 140, had fewer heart and stroke events: 10.9% vs 13.5%.
Age is the biggest risk factor because it adds up years of cholesterol, blood pressure, blood sugar, and inflammation.
Bottom line: know your LDL and blood pressure, check Lp(a) once, and treat early.
Sources and Notes
Both trials were presented on Saturday 7 November 2026 at the American Heart Association Scientific Sessions in Chicago and published the same day in the New England Journal of Medicine. Both enrolled people who already had artery disease. Neither tested what happens if treatment starts earlier in life, so point 4 comes from other kinds of evidence, named below.
· · · · ·
1. ZEUS, THE INFLAMMATION SHOT (POINT 1)
- Ziltivekimab, a monthly injection from Novo Nordisk that blocks interleukin-6, a signal that drives inflammation, against placebo.
- 6,376 people with artery disease, chronic kidney disease and a raised inflammation marker (CRP 2 mg/L or more). Mean age 69.4. 66 in 100 had diabetes and 42 in 100 heart failure. Median CRP at the start 4.5 mg/L. Median LDL 72 mg/dL, so most were already on cholesterol drugs.
- CRP fell 64% and interleukin-6 activity 76% against placebo.
- Heart attack, stroke or heart death: 561 people on the drug (17.6%) and 565 on placebo (17.7%). Hazard ratio 0.99 (95% CI 0.88 to 1.11).
- Serious infections: 20.4% on the drug, 17.9% on placebo.
- Ridker et al., NEJM 2026. doi.org/10.1056/NEJMoa261212…
· · · · ·
2. Lp(a)HORIZON, THE Lp(a) DRUG (POINT 2)
- Pelacarsen, an Ionis and Novartis antisense drug that cuts Lp(a), an inherited cholesterol particle, against placebo.
- 8,323 people who had had a heart attack, a stroke or symptomatic leg artery disease, with Lp(a) of 70 mg/dL or more. Mean age 60. Median LDL 65 mg/dL. Median CRP 0.8 mg/L.
- Lp(a) fell 77.7%, from a median of 234 to 64 nmol/L.
- Heart attack, stroke, heart death or urgent revascularization: 12.3% on the drug, 12.1% on placebo. Hazard ratio 1.02 (0.90 to 1.15). In people with Lp(a) of 90 mg/dL or more, hazard ratio 1.00.
- More people stopped the drug for side effects (20.6% against 7.5%). 19 cases of anaphylaxis on the drug, none on placebo.
- Cho et al. (presented by Steven Nissen), NEJM 2026. doi.org/10.1056/NEJMoa261015…
- Extended analyses, Nissen et al., NEJM Evidence 2026. doi.org/10.1056/EVIDoa260037…
· · · · ·
3. WHY THEY MISSED (POINT 3)
- Paul Ridker, who led ZEUS, showed the slide "The horse may already have been out of the barn" and called the treatment too little, too late for people this sick. He suggested moving upstream in the inflammation pathway (NLRP3).
- Steven Nissen on HORIZON: measuring Lp(a) is still worthwhile.
- Lp(a)HORIZON is the harder result to explain. HORIZON patients were younger, less inflamed and had a 78% fall in the target, and still saw no benefit. Genetic studies estimated that a fall of roughly 66 to 101 mg/dL in Lp(a) would be needed to show a clear benefit in a trial of this length (Burgess 2018, Lamina 2019), and some experts think the drug did not lower it enough in absolute terms. Others think Lp(a) lowering alone may not help as much as hoped. This is not settled.
- Eric Topol @EricTopol: blood CRP barely tracks inflammation in the heart arteries themselves, and starting in the seventh decade of life is too late.
- Ethan Weiss, cardiologist: both trials selected people by an abnormal blood marker, unlike SELECT (semaglutide), which did not and did cut events (hazard ratio 0.80). His baseline comparison table and the three survival curves are the clearest view of the difference.
- Marios Georgakis: absolutely zero effect.
- Next Lp(a) trials: olpasiran (Amgen, OCEAN(a)-Outcomes, 7,297 people, primary completion March 2028, NCT05581303) and lepodisiran (Lilly, ACCLAIM-Lp(a), 17,300 people, March 2029, NCT06292013). ACCLAIM also enrolls people aged 55 or older at risk of a first heart attack or stroke.
· · · · ·
4. TIMING, LDL AND GENES (POINT 4, SECOND HALF)
- People born with gene variants that lower LDL had 54.5% lower heart risk per 1 mmol/L (39 mg/dL) of LDL, about three times the effect of statins started later in life. This is a genetic estimate (Mendelian randomization), not a trial. Ference et al., J Am Coll Cardiol 2012. PMID 23083789. pubmed.ncbi.nlm.nih.gov/2308…
- In 21 trials of statins, ezetimibe and PCSK9 drugs, the risk cut per 1 mmol/L lower LDL grew from 12% in year 1 to 20%, 23% and 29% in years 3, 5 and 7. Wang et al., Circ Cardiovasc Qual Outcomes 2022. PMID 35430872. pubmed.ncbi.nlm.nih.gov/3543…
· · · · ·
5. PLAQUE STARTS YOUNG (POINT 4, FIRST HALF)
- PDAY, autopsies of 2,876 Americans aged 15 to 34 who died of accidents, homicide or suicide. Raised fatty streaks or plaque in the right coronary artery were found in about 1 in 10 at ages 15 to 19 and 3 in 10 at ages 30 to 34. McGill et al., Arterioscler Thromb Vasc Biol 2000. PMID 10938023. pubmed.ncbi.nlm.nih.gov/1093…
· · · · ·
6. INFLAMMATION STILL MATTERS (CONTEXT, NOT IN THE POST)
- Colchicine 0.5 mg daily: COLCOT, after a recent heart attack, hazard ratio 0.77 (Tardif, NEJM 2019, doi.org/10.1056/NEJMoa191238…). LoDoCo2, stable heart disease, hazard ratio 0.69 (Nidorf, NEJM 2020, doi.org/10.1056/NEJMoa202137…). CLEAR SYNERGY, started right after a heart attack, no benefit, hazard ratio 0.99 (Jolly, NEJM 2025, doi.org/10.1056/NEJMoa240592…).
- CANTOS, canakinumab, another inflammation blocker, cut events 15% (hazard ratio 0.85) in 2017. Ridker, NEJM 2017. doi.org/10.1056/NEJMoa170791…
- CaRi-Heart, an AI from Caristo, an Oxford spin-out, reads inflammation in the fat around the heart arteries on a routine heart CT scan. FDA De Novo authorization announced 29 July 2026. biospace.com/press-releases/…
· · · · ·
7. BLOOD PRESSURE AFTER STROKE (POINT 5)
- OPTIMAL-Stroke, Brazil, 4,368 people after an ischemic stroke or TIA (mini-stroke). Target top number under 120 against under 140. Heart death, heart attack, stroke or hospital stay for unstable angina or heart failure: 10.9% against 13.5% over about 4 years. More kidney problems in the intensive group. Published in NEJM on 7 November 2026. American Heart Association release: newmediawire.com/news/intens…
· · · · ·
8. WHAT TO DO
- Know your LDL and blood pressure. Lp(a) is mostly set by your genes, so one test is usually enough. Ask your doctor about treating early, before a first heart attack or stroke. None of the trials above tested early treatment of Lp(a) or inflammation in healthy people.
Most people who died of heart disease after a heart scan had no major blockage.
In a study of 40,091 patients in England's NHS, 8 in 10 had no major blockage, yet they accounted for nearly two-thirds of heart deaths.
Why? The standard scan report shows narrowing, but not inflammation. That inflammation predicted heart deaths even when there was no blockage.
CaRi-Heart, an AI tool from Oxford spin-out Caristo, reads inflammation from the same scan and combines it with plaque and risk factors to score risk.
In an NHS England pilot, one hospital reviewed 135 patients. For those with no major blockage, doctors advised more preventive drugs for 25 in 100 on the usual report. With CaRi-Heart, that jumped to 86 in 100.
The study measured advice, not heart attacks or deaths.
If you are told there is no major blockage, ask about inflammation, plaque, and whether a statin makes sense.
Sources and Notes
The death numbers come from a study of 40,091 patients at eight NHS hospitals in England. The NHS is England's public, tax-funded health service. The advice numbers come from a separate, smaller study at one of those hospitals, which looked back at 135 patients and counted what doctors advised, not what happened to them.
· · · · ·
1. WHO DIED OF HEART DISEASE AFTER THE SCAN (LINES 1 AND 2)
- 40,091 consecutive patients had a clinically indicated CT scan of the heart arteries at eight hospitals in England (Oxford, Bath, Papworth, Royal Brompton, Harefield, Leicester, Milton Keynes and Leeds) and were followed for a median of 2.7 years. Most of these scans are done to check chest pain.
- 32,533 of them (81 in 100) had no major blockage, defined as no narrowing of 50 percent or more in the left main artery and no narrowing of 70 percent or more in the other main heart arteries. They had 1,118 of the 1,754 cardiac deaths (64 in 100), and 66 in 100 of all heart attacks, new heart failure and cardiac deaths.
- Because this group is most of the patients, each person in it has a lower risk than a person with a major blockage. The point is where most deaths happen.
- Chan, Antoniades and colleagues (the ORFAN study), The Lancet, 2024. Several authors work for Caristo Diagnostics, which makes the AI score. Funded by the British Heart Foundation, an NHS AI award, Innovate UK, NIHR and the Oxford Biomedical Research Centre.
doi.org/10.1016/S0140-6736(2… (PMID 38823406)
· · · · ·
2. WHAT THE USUAL REPORT MISSES (LINE 3)
- The usual report grades how narrowed the arteries are and describes any plaque. It does not measure inflammation.
- In 3,393 of these patients from two of the hospitals, followed for a median of 7.7 years, a CT measure of inflammation in the fat around the heart arteries (the FAI score) predicted cardiac death in people with and without a major blockage. About half of this group had no or minimal plaque.
Same paper as above.
· · · · ·
3. THE CaRi-HEART AI (LINE 4)
- CaRi-Heart is an AI-based risk score from Caristo Diagnostics, founded in 2018 as a spin-out from the University of Oxford. It combines the inflammation measure (the FAI score), plaque burden and four risk factors (diabetes, smoking, high cholesterol and high blood pressure) into a chance of cardiac death within 8 years.
biospace.com/caristo-diagnos…
- NHS England is piloting a chest pain pathway that uses it at five hospital sites.
· · · · ·
4. THE US GUIDELINE
- The 2021 AHA/ACC chest pain guideline recommends making preventive treatment as good as it can be for patients with stable chest pain and known plaque that narrows an artery by less than half. Class 1 recommendation, based on expert opinion (level C-EO). It points to the cholesterol and prevention guidelines, which lead with statins.
Gulati and colleagues, Journal of the American College of Cardiology, 2021.
doi.org/10.1016/j.jacc.2021.…
· · · · ·
5. THE NHS PILOT HOSPITAL STUDY (LINE 5)
- 135 consecutive patients (of 137) at the chest pain clinic of Milton Keynes University Hospital, about 50 miles north of London, one of the five sites in the NHS England pilot, January to September 2024. Mean age 61, and 47 in 100 were women.
- Each patient's advice was compared two ways: on the usual scan report, and with the AI score report (CaRi-Heart, from Caristo Diagnostics in Oxford, which ran the analysis). Advice was recorded for 133.
- "No major blockage" in this study means narrowing under 50 percent. In those patients, more preventive drugs were advised for 25 in 100 on the usual report and 86 in 100 with the AI score.
- Across all patients it was 30 against 84 in 100 (McNemar p < 0.0001). By drug: statins 23 against 65 in 100, ezetimibe 8 against 56, aspirin 8 against 27, colchicine 5 against 15. More preventive drugs means starting or raising any of these. No one was advised to stop a drug.
- Sanfeliu Garces, Alsinbili, Kardos and colleagues, Milton Keynes University Hospital and the University of Buckingham, Journal of Clinical Medicine, 6 November 2026. No external funding; the project was supported by an NHS AI award. The authors declare no conflicts of interest.
doi.org/10.3390/jcm15218639
· · · · ·
6. LIMITS
- One hospital, records reviewed afterwards, 135 patients, and advice only. There are no data yet on heart attacks, deaths, whether patients took the drugs, or side effects.
- The authors list the risk of overtreatment, that colchicine is not licensed in the UK for heart prevention, that aspirin has a limited role when a person has not had a heart attack or stroke, and that cholesterol results were not recorded.
- The FAI score rises steeply with age. Here the median roughly doubled every 11 to 12 years, and the high-risk group was on average 13 years older (66 against 53).
- The score was built and checked mostly in white European patients (82 in 100 of this group).
· · · · ·
7. THE EARLIER TRIAL
- SCOT-HEART randomized 4,146 people with stable chest pain to standard care plus a CT scan of the heart arteries, or standard care alone. Over 5 years, death from coronary heart disease or a non-fatal heart attack was 2.3 percent against 3.9 percent (hazard ratio 0.59, 95% CI 0.41 to 0.84). More preventive treatments were started in the scan group.
Newby and colleagues, University of Edinburgh, NEJM, 2018.
doi.org/10.1056/NEJMoa180597… (PMID 30145934)
· · · · ·
8. WORDS USED
- A heart scan here is a CT scan of the heart arteries (coronary CT angiography), an X-ray scan with dye that shows narrowing and plaque.
- Plaque is a fatty build-up in the artery wall. A major blockage is a narrowing big enough to limit blood flow; each study's cut-off is given above.
- Statins and ezetimibe lower LDL cholesterol. Colchicine is an old anti-inflammatory drug.
2
626
In the 1660s, the man behind Boyle's law wrote a to-do list for science. He put aging at the top.
Line one read "The Prolongation of Life." Line two, "The Recovery of Youth."
Robert Boyle, the chemist whose gas law is still taught in school, was a founding Fellow of the Royal Society, formed in London in 1660, later led by Isaac Newton, and still meeting today.
Science has done much of his list. Flight. Diving. Longitude at sea. Drugs to ease pain. Even "varnishes perfumable by rubbing", now scratch-and-sniff.
Aging is the big one still open. Life expectancy has roughly doubled, much of it from saving children, yet no treatment has been shown to slow aging itself in people.
In worms and mice, aging has been slowed in the lab. An AI at MIT has sifted 800,000 molecules for ones that kill worn-out cells.
Boyle put aging first about 360 years ago. It should be first now.
Sources and Notes
The animal results are not human results. No treatment has yet been shown in a trial to slow aging in people.
· · · · ·
1. BOYLE'S LIST, IN THE ROYAL SOCIETY ARCHIVE
- The manuscript: Royal Society Archives RB/1/8/30, Boyle Papers volume 8, folios 208-9, "List of scientific projects/desiderata", also known as Boyle's wish list.
catalogues.royalsociety.org/…
- Photographs of both pages and a transcription, from the Royal Society:
royalsociety.org/blog/2010/0…
- It is in a scribe's hand dated to the 1660s. Michael Hunter's annotated edition:
eprints.bbk.ac.uk/id/eprint/…
- The Royal Society was founded in London on 28 November 1660 and received its royal charter from Charles II in the early 1660s. Boyle was a founding Fellow. Isaac Newton was its President from 1703 to 1727. It still meets today. It is often called the oldest scientific academy in continuous existence, though Germany's Leopoldina (1652) is older.
britannica.com/topic/Royal-S…
en.wikipedia.org/wiki/List_o…
- Boyle's law (1662): at a fixed temperature, the pressure of a gas rises as its volume falls. It is still taught in school chemistry and physics.
britannica.com/science/Boyle…
· · · · ·
2. THE FULL LIST, AS WRITTEN
1. The Prolongation of Life.
2. The Recovery of Youth, or at least some of the Marks of it, as new Teeth, new Hair colour'd as in youth.
3. The Art of Flying.
4. The Art of Continuing long under water, and exercising functions freely there.
5. The Cure of Wounds at a Distance.
6. The Cure of Diseases at a distance or at least by Transplantation.
7. The Attaining Gigantick Dimensions.
8. The Emulating of Fish without Engines by Custome and Education only.
9. The Acceleration of the Production of things out of Seed.
10. The Transmutation of Metalls.
11. The makeing of Glass Malleable.
12. The Transmutation of Species in Mineralls, Animals, and Vegetables.
13. The Liquid Alkaest and Other dissolving Menstruums.
14. The making of Parabolicall and Hyperbolicall Glasses.
15. The making Armor light and extremely hard.
16. The practicable and certain way of finding Longitudes.
17. The use of Pendulums at Sea and in Journeys, and the Application of it to watches.
18. Potent Druggs to alter or Exalt Imagination, Waking, Memory, and other functions, and appease pain, procure innocent sleep, harmless dreams, etc.
19. A Ship to saile with All Winds, and A Ship not to be Sunk.
20. Freedom from Necessity of much Sleeping exemplify'd by the Operations of Tea and what happens in Mad-Men.
21. Pleasing Dreams and physicall Exercises exemplify'd by the Egyptian Electuary and by the Fungus mentioned by the French Author.
22. Great Strength and Agility of Body exemplify'd by that of Frantick Epileptick and Hystericall persons.
23. A perpetuall Light.
24. Varnishes perfumable by Rubbing.
- In Boyle's day, "transplantation" (line 6) meant moving a disease into another body or object, not organ transplants.
- "Scratch-and-sniff" for line 24 is a modern gloss on a 17th-century wish.
- Pain relief (line 18) sits inside a longer wish for "Potent Druggs", which also asked to "appease pain".
- "About 360 years": the surviving copy is dated to the 1660s, so 357 to 366 years before 2026.
· · · · ·
3. WHERE THE FIRST TWO LINES CAME FROM
- Francis Bacon opened his own list, the "Magnalia Naturae" printed with New Atlantis in the 1620s: "The prolongation of life. The restitution of youth in some degree. The retardation of age."
thomasmorestudies.org/wp-con…
- The Royal Society's archivist writes that Boyle's list was largely copied from Bacon's.
royalsociety.org/blog/2018/0…
· · · · ·
4. THE QUOTE
- From Boyle's Memoirs for the Natural History of Human Blood (1684), not from the list. He hoped an "enlightened Posterity" would have "the generous gratitude to remember the Difficulties this Age surmounted, in breaking the Ice, and smoothing the way for them".
Quoted in Hunter (above), from The Works of Robert Boyle, vol. 10, pp. 96-7.
· · · · ·
5. LIFE EXPECTANCY
- England, 1600 to 1800: about 35 to 40 years at birth, held down by deaths in infancy.
campop.geog.cam.ac.uk/blog/2…
- England and Wales today: about 79 for men and 83 for women.
ons.gov.uk/peoplepopulationa…
· · · · ·
6. WORMS, MICE AND AI
- Worms: mutations in the gene daf-2 let roundworms live more than twice as long. Kenyon and colleagues, UCSF, Nature, 1993.
doi.org/10.1038/366461a0
- Mice: rapamycin started at 600 days of age raised the age at 90% mortality by 14% in females and 9% in males. Harrison and colleagues, Nature, 2009.
doi.org/10.1038/nature08221
- AI: graph neural networks trained on a lab screen of 2,352 compounds scored more than 800,000 molecules. Three killed senescent (worn-out) cells selectively, and one reduced them in the kidneys of aged mice. No lifespan result. Wong, Collins and colleagues, MIT and the Broad Institute, Nature Aging, 2023.
doi.org/10.1038/s43587-023-0…
1
368
Your thymus does most of its work before you can walk.
Hidden behind the breastbone, this small gland trains immune cells called T cells to attack viruses and cancer. Then, starting around age one, its working tissue fades as fat moves in.
For years, doctors thought the adult thymus barely mattered. Two new studies challenge that.
In one, 8 in 100 adults who had their thymus removed died within 5 years, compared with 3 in 100 matched chest surgery patients who kept theirs.
In another, AI scored thymus health on chest scans from 25,031 older heavy smokers. Over 12 years, 26 in 100 people in the lowest-scoring quarter died, versus 13 in 100 in the top quarter.
That still does not prove the thymus is the cause. Sometimes illness is why it is removed. Sometimes sickness makes it shrink.
Even so, companies are racing to bring it back. Intervene Immune is testing three drugs. Tolerance Bio is planning trials. United Therapeutics is making thymus cells from stem cells.
No study has yet shown that regrowing the thymus adds years to life. But for an organ once dismissed in adults, that question suddenly looks worth asking.
Sources and Notes
The surgery and scan studies are observational, so they show a link, not cause. Regrowing the thymus in adults is early work: one small uncontrolled trial, one registered trial with no results, and no approved treatment for adults.
· · · · ·
1. WHEN THE THYMUS DOES ITS MOST WORK (LINE 1)
- Thymus tissue from 204 people aged 1 month to 107 years was measured.
The working tissue (the thymic epithelium, where T cells are made) was largest in the first year of life and shrank steadily with age after that.
Fat replaced the tissue extensively after about age 15. The organ's overall size changed little with age.
Steinmann, Klaus and Muller-Hermelink, Kiel, Scand J Immunol, 1985.
doi.org/10.1111/j.1365-3083.…
- Textbooks often say the thymus peaks at puberty. That refers to its weight, which includes fat and connective tissue.
Some blood studies of new T cell output (TRECs) find output holding up until early adulthood, so the first-year peak rests on tissue measurements and output models.
· · · · ·
2. THE SURGERY STUDY
- Adults who had the thymus removed were compared with demographically matched patients who had heart or chest surgery without removal.
There were 1,146 matched pairs. Matching was on age, sex and race, plus infection, cancer and autoimmune disease before surgery.
- At 5 years, all-cause mortality was 8.1% against 2.8%. Cancer was 7.4% against 3.7%.
In numbers: mortality relative risk 2.9 (95% CI 1.7 to 4.8). Cancer relative risk 2.0.
- About half the thymectomy group had a thymus condition such as a thymoma or myasthenia gravis.
The higher death rate held after those with thymoma, myasthenia gravis or cancer were excluded.
- In a subgroup of 22 against 19 people followed for about 14 years, the thymectomy group made fewer new T cells and had more inflammatory signals in the blood.
- Kooshesh, Scadden and colleagues, Massachusetts General Hospital, NEJM, Aug. 3, 2023.
Funded by the Tracey and Craig A. Huff Harvard Stem Cell Institute Research Support Fund and others.
doi.org/10.1056/NEJMoa230289…
· · · · ·
3. THE SCAN STUDY
- A deep learning model scored thymic health on routine chest CT in two cohorts, 27,612 people in all.
25,031 came from the US National Lung Screening Trial (ages 55 to 74, at least 30 pack-years of smoking) and 2,581 from the Framingham Heart Study.
- In the lung screening cohort, 12-year Kaplan-Meier mortality was 13.4% in the top quarter of the score against 25.5% in the bottom quarter. These are unadjusted rates; the score falls with age, so age explains part of the gap.
The link held after adjusting for smoking and stratifying by sex and age.
In numbers: hazard ratio 0.49, 95% CI 0.45 to 0.53.
- The authors state that the retrospective design does not allow conclusions about causality and that disease may drive thymic decline.
- Bernatz, Aerts, Birkbak and colleagues, Mass General Brigham and Harvard Medical School, Nature, published online March 18, 2026.
Licence CC BY-NC-ND 4.0.
Funders include the US National Institutes of Health and the European Research Council.
doi.org/10.1038/s41586-026-1…
· · · · ·
4. REGROWING IT
- Intervene Immune, TRIIM pilot: 10 men aged 51 to 65 took growth hormone, DHEA and metformin for one year, and 9 were analysed.
On MRI, fat in the thymus was replaced by working tissue, a clear gain in 7 of the 9 men.
There was no control group. The authors also reported shifts in epigenetic clock estimates, which are blood biomarkers, not lifespan.
The lead authors hold shares in Intervene Immune.
Fahy and colleagues, Aging Cell, 2019.
doi.org/10.1111/acel.13028
- TRIIM-X, NCT04375657, Intervene Immune: the same three drugs against an active control, ages 40 to 80, about 85 people.
Primary outcomes include thymus density on MRI or CT. Phase 2, listed as recruiting since Nov. 2020, last updated May 2025, no results posted.
clinicaltrials.gov/study/NCT…
- Tolerance Bio licensed efineptakin alfa (NT-I7), a long-acting form of the immune signal IL-7, from NeoImmuneTech on Aug. 19, 2026.
It plans Phase 2 trials of thymic function, starting with people with HIV whose immune systems do not recover on treatment.
The company says the drug has been given to more than 800 people in earlier studies, mostly in cancer.
tolerancebio.com
- United Therapeutics bought Thymmune Therapeutics in July 2026 for $140 million up front and up to $160 million in milestones.
Thymmune turns induced pluripotent stem cells into thymus cells. Its lead product, THY-100, formed a new thymus in animals and is aimed first at children born without a thymus.
ir.unither.com/press-release…
- Already approved: RETHYMIC, processed donor thymus tissue implanted in children born without a thymus (congenital athymia).
FDA approval Oct. 8, 2021, based on 105 patients treated from 1993 to 2020. Without treatment, these children usually die by age 2 or 3.
fda.gov/vaccines-blood-biolo…
- NCT07224763, United Therapeutics at NYU Langone Health: a pig kidney carrying pig thymus tissue (the thymokidney) for people with kidney failure. Phase 1/2, about 50 people.
clinicaltrials.gov/study/NCT…
- A field overview: "Thymus renaissance poised to boost health and longevity," Nature Biotechnology news, May 2026.
doi.org/10.1038/s41587-026-0…
· · · · ·
5. THE PAST
- Jacques Miller's 1961 Lancet paper showed the thymus is an immune organ, in mice.
doi.org/10.1016/S0140-6736(6…
- Switching on the gene FOXN1 in old mice regrew the thymus. Bredenkamp, Nowell and Blackburn, Development, 2014.
doi.org/10.1242/dev.103614
- In 22 adults with HIV, growth hormone raised thymus density on CT. Napolitano, McCune and colleagues, J Clin Invest, 2008.
doi.org/10.1172/JCI32830
· · · · ·
6. LIMITS
- Both mortality studies are observational. Illness can be the reason a thymus is removed or wastes away.
- No treatment is approved to regrow the thymus in adults, and no trial has shown that regrowing it adds years.
· · · · ·
7. WORDS USED
- The thymus is a small gland behind the breastbone where T cells, a type of immune cell, mature.
- Thymectomy means surgical removal of the thymus.
- IL-7 (interleukin 7) is a signalling protein that helps T cells form and survive.
- Induced pluripotent stem cells are adult cells reprogrammed so they can become other cell types.
- TRECs are small DNA circles made when new T cells form, a marker of new T cell output.
- Kaplan-Meier mortality is the share of people who have died by a set time, estimated from follow-up data.
- Pack-years measure smoking: packs per day times years smoked.
1
764