Biological Age Claims Need More Than One Clock

Biological age is the longevity industry's favorite number: one blood draw, one algorithm, and a verdict in years. Younger is a triumph, older is a sales pitch. A Yale team has now run the obvious experiment nobody selling clocks wanted run: score the same people on 12 different epigenetic clocks and see whether the clocks agree. Mostly, they don't.

The reality check

The paper, published in Aging Cell on October 8, revisits three intervention datasets that had each reported a significant drop in biological age on a single clock, including an 8 week diet and supplement program and a 2 year folic acid and vitamin B12 trial. The team ran 11 more clocks on the same data. Not one moved significantly in the same direction. In the B12 data, 10 of the 11 extra clocks leaned the other way.

Part of the explanation is plain noise. The same lab had shown earlier that testing one person twice on standard clocks can produce swings of 0.9 to 2.4 years; its retrained PC clocks cut the median gap between technical replicates to 0.3 to 0.8 years. Against that background, the authors' demand is modest and devastating: before anyone claims rejuvenation, show several reliable clocks moving together.

How the test was built

To learn what real change looks like, the team turned the question around: if clocks can see a body getting younger, they should certainly see one getting older. They took three events known to raise mortality risk: major surgery in 30 patients averaging near 78, radiotherapy with or without chemotherapy in 72 patients, and coronavirus infection in 35 young adults. Each time, at least 4 of the 12 clocks registered significant acceleration, and three mortality linked measures, PCPhenoAge, PCGrimAge and DunedinPACE, rose in all three. The classic clocks trained on calendar age flagged none of the events.

That yields three criteria: reliable PC clocks respond, clocks trained on mortality or pace of aging respond, and several clocks move the same way. Some interventions pass. Both arms of a 2 year Mediterranean diet trial and a 3 week alcohol abstinence program in 24 chronic drinkers lowered at least five clocks each, with 14 of 18 reliable clock tests significant against 5 of 18 for the noisy versions. The bar is high, not impossible.

The limits the authors admit

The paper covers nine intervention datasets of 14 to 92 people each. No intervention has ever been proven to extend human lifespan, so there is no gold standard for rejuvenation, and passing the criteria shows a signal is probably real, not that it adds a year of life. Some of what clocks see is a shift in the mix of blood cells: adjusting for 12 immune cell types erased roughly three quarters of the surgery and chemoradiation effects.

Keep that test in hand. Two claims are about to need it.

Claim one: GLP-1 drugs rewind the clock by 2 to 3 years

That is the New York Post's headline after the Aging Research & Drug Discovery meeting at Harvard, where Novo Nordisk and Eli Lilly presented. The primary documents tell a more specific story, and the two companies ran very different studies.

Novo's work is now a preprint. It applies proteomic organ clocks, which read blood proteins rather than DNA methylation, to 10,052 participants in five placebo controlled semaglutide trials, with follow up of up to 156 weeks. Against placebo, heart biological age fell by roughly 2 to 4 years, and kidney, brain, pancreas, lung and multiorgan clocks by 1 to 4 years. These are high risk patients from trials such as SELECT, FLOW and SOUL, not healthy adults, and the authors, most of them Novo employees, write that the results do not establish biological rejuvenation or surrogate validity.

Lilly's epigenetic analysis is smaller and unpublished. According to the slides as reported by Longevity.Technology, it followed 71 people on tirzepatide in the SURMOUNT-5 trial for 72 weeks, about 1.38 years. All 15 clocks showed less aging than the calendar time that passed, but only two reached significance, with median gains of 1.1 and 0.7 years. There was no placebo or comparison arm, and Lilly called the results preliminary.

Under the Yale rule, Novo has what most clock claims lack: a placebo group, thousands of people and several organ clocks pointing the same way, though proteomic clocks are not what the Yale paper tested, so the rule transfers as logic, not verdict. Lilly has clocks agreeing in direction, but two significant results out of 15 and no control. Neither supports a blanket "2 to 3 years younger."

Claim two: a cream turns skin 2.1 years younger

Rubedo Life Sciences announced phase 1b results for RLS-1496, an investigational topical drug tested on sun damaged skin in patients with atopic dermatitis. Across 11 paired samples, measured skin epigenetic age after four weeks of treatment averaged 2.1 years below baseline (p=0.032), on a single clock, MitraBio's. The release leads with a best responder at 11.7 years and describes no untreated comparison group. Eleven samples and one clock is exactly the configuration the Yale paper warns about: an early signal, worth a follow up study, not a demonstrated rejuvenation.

One clock is a story. Several clocks agreeing is evidence. Most of what is sold to you is the first one wearing the second one's clothes.

What comes next

Novo's preprint still faces peer review. Lilly plans to extend the work to semaglutide and other weight management studies, per the same conference report. Rubedo says a phase 2a study in pattern hair loss starts this month, a chance to report more than one clock. The stronger test sits further out: a federally funded trial of up to $38 million will enroll 726 adults in their 60s and compare rapamycin, dapagliflozin and semaglutide against placebo, with enrollment due in 2027. For a quick checklist on the next headline, see our guide to reading a health study in 5 minutes; the sponsor question also runs through Lilly's oral GLP-1 head to head.

None of this says biological age is useless. It says the unit is noisier than the marketing, and any claim smaller than that noise, from any company, including the famous ones, should be priced accordingly: as a hypothesis.