400 Drugs Screened in Half a Million People — Only a Handful Tracked With Longer Life
A researcher behind a UK Biobank analysis screened more than 400 prescribed medicines against long-term survival in half a million people. Most tracked with worse mortality — largely because of the diseases they treat — but 14 stood out, including a statin, sildenafil and several hormone replacement therapies, with a strong signal for SGLT2 inhibitors. The conversation then moves to model-organism screening, dog aging, chemical reprogramming and why GDF15 may become the LDL cholesterol of aging.
Overview
The idea began with metformin: an earlier UK study suggested diabetics taking it appeared to outlive non-diabetics, a finding later studies did not confirm. Rather than obsess over one molecule, the research team asked a broader question — across all 400+ drugs people are actually prescribed, which ones track with longer survival? Using the UK Biobank's half a million participants and up to 20 years of follow-up, they compared each drug against matched controls adjusted for age, sex, disease history, smoking, BMI and socioeconomic status.
Most drugs tracked with worse mortality, an expected artefact called confounding by indication: the underlying illness, not the pill, shortens life. Fourteen drugs went the other way with high statistical power, including atorvastatin, sildenafil and multiple estrogen-based hormone therapies, with SGLT2 inhibitors showing the strongest class-level signal despite shorter follow-up. Because human trials on lifespan are impractical, the follow-up work moves into worms, flies, fish and mice, using lab automation and machine learning to screen thousands of compounds and look for effects conserved across species.
The discussion covers dog aging as a faster commercial path, chemical reprogramming as a pill-based alternative to gene-therapy approaches, and the view that PI3 kinase may matter as much as mTOR. On measurement, the guest argues the field needs one workable blood marker rather than only composite clocks — and nominates GDF15, the protein most tightly correlated with age and most predictive of mortality. A validation study across six large cohorts on three continents is on the way.
Key quotes
5Diabetic people that were taking metformin, they appeared on average to live longer than even non-diabetic people.
That's very interesting, but why focus only on metformin?
Most of the drugs we looked at were associated with worse mortality. This doesn't mean that the drugs are bad — it's called confounding by indication.
GDF15 is by far the best, the most correlated with age, but also the most predictive of mortality.
The parallel I draw is with LDL cholesterol — one single marker that you can very easily measure, and lowering it leads to better outcomes.
Key ideas
9One drug became a whole screen
Metformin's reputation in aging research came from a long-term study where treated diabetics appeared to outlive non-diabetics. That result was not confirmed later, but it prompted a far bigger question: why not test everything at once?
Half a million people, 400 drugs, 20 years
The UK Biobank offers open-access data on roughly 500,000 people followed since 2006, including exactly which medications each person takes. That made a systematic one-by-one survival comparison across more than 400 drugs possible.
Most drugs look bad, and that's expected
The majority of medicines tracked with worse mortality even after adjusting for age, sex, disease, smoking, BMI and socioeconomic status. The illness behind the prescription drives that, not the medicine — the classic confounding-by-indication problem.
Fourteen drugs went the other way
A statin (atorvastatin), sildenafil in men, and several estrogen-based hormone replacement therapies were among the standouts. Notably, some of these are not prescribed for a specific disease at all, which weakens the usual confounding explanation.
SGLT2 inhibitors gave the strongest class signal
Grouping drugs into classes showed SGLT2 inhibitors performing strongly, even with fewer patients and shorter follow-up than older diabetes drugs. Independent mouse work has also reported lifespan effects for this class.
Human trials are impractical, so worms come first
Prospective lifespan trials in people are prohibitively long and costly — the aspirin trial that was run came back negative. The alternative is testing candidates in worms, flies, fish and mice with automated, high-throughput screening.
Dogs as the shorter road
Because aging is not a recognised indication, a compound that extends mouse lifespan has no obvious clinical path. Dogs, living around twelve years, allow realistic trials — and one company has pursued conditional approval on that route.
Aging as accumulated wear and tear
The guest's working model combines DNA damage, epigenetic change and plain wear and tear, like a car accumulating damage through use. Rate of DNA damage accumulation correlates strongly with lifespan across species.
PI3 kinase may outrank mTOR
After retesting nearly every compound published in the literature across multiple species, mTOR's role was confirmed — but the PI3 kinase pathway looked even clearer. Many PI3K inhibitors already exist as cancer drugs, though toxicity at higher doses is the obstacle.
Practical takeaways
6- 1
Read drug-and-mortality headlines with the indication in mind 12:10
When a study says a medicine 'is associated with' higher mortality, the disease it treats is usually doing the work. Ask what the comparison group looked like before drawing conclusions.
- 2
Association is not a prescription 18:00
Every finding here comes from observational data on people who were already prescribed these medicines. None of it says a healthy person should start taking them — that is a conversation for your own physician.
- 3
Newer, more targeted drugs are easier to interpret 24:00
SGLT2 inhibitors act on one clear mechanism, while metformin's mechanism is still debated even by specialists. Cleaner mechanisms make cleaner signals — worth remembering when comparing study results.
- 4
Track one marker you can actually act on 58:40
The guest's case for GDF15 rests on it being measurable in a normal clinic and tightly tied to mortality. Whatever markers you follow, favour ones that are repeatable and interpretable over exotic panels.
- 5
Composite pictures beat single numbers today 1:02:00
Most of the field combines clocks, blood markers, insulin sensitivity, blood pressure and functional capacity. Looking at your markers together, in context, is closer to how researchers actually think.
- 6
The unglamorous basics still win 1:06:30
Asked what he wished he had adopted sooner, the guest lands on nothing exotic: sleep, nutrition and movement remain what the evidence supports today. The value of the research is understanding why.
Topics & chapters
15Introduction
Siim Land introduces his guest and the papers on drugs, mortality and biological age that prompted the conversation.
The metformin origin story
Why one diabetes drug became the aging field's obsession, and how a long-term UK study produced a result nobody expected.
From one drug to all of them
The reasoning that turned a single-molecule question into a systematic screen of everything people are prescribed.
Inside the UK Biobank
Half a million participants, follow-up since 2006, and complete medication records make the analysis possible.
Why most drugs look harmful
Confounding by indication explained — the disease behind the prescription dominates the survival signal.
The fourteen standouts
Atorvastatin, sildenafil and multiple hormone replacement therapies emerge with high statistical power.
SGLT2 inhibitors and drug classes
Class-level analysis, dose response and sex-specific effects, plus why newer drugs have less statistical power.
Wealth, healthcare and confounding
Whether cost differences between drug classes could explain the results, and how socioeconomic status was handled.
Testing candidates in model organisms
Worms, flies, fish and mice as the practical alternative to impossibly long human lifespan trials.
Dog aging and the regulatory wall
Aging is not a recognised indication, so companies either target an age-related disease or move to dogs.
Chemical reprogramming
From Yamanaka factors to compound cocktails — why a pill-based approach may sidestep gene-therapy safety issues.
What aging actually is
DNA damage, epigenetics and wear and tear, plus the pathways that look most conserved across species.
SGLT2 versus metformin
Different mechanisms, different signals: glucose excretion via the kidney versus a still-debated AMPK story.
Clocks, proteomics and GDF15
Why proteomic markers may beat methylation clocks in the clinic, and the case for one workable blood protein.
Composite markers and what's next
The composite-endpoint debate, the LDL analogy, and a validation study across six cohorts on three continents.