Short term win. 100%
Semaglutide lowered the biological age of the heart by about 2 to 4 years versus placebo in a new preprint. The measuring tools were built to predict death, and they moved in five separate trials covering 10,052 people.
Why it matters: Many aging biomarkers (measurable signals in the body that are meant to track how fast you age) have never been shown to respond to a drug that improves real outcomes, so it is unclear whether they track anything that can be changed. This paper tests whether protein-based aging scores shift under a drug already known to cut heart attacks, strokes, and deaths, which could help decide which aging measures are worth using in future trials.
To test that, the paper used stored blood samples from five trials where people were randomly given either semaglutide or a placebo. The trials covered different groups: SELECT (people with obesity and heart disease), STEP-HFpEF and STEP-HFpEF DM (obesity-related heart failure, without and with type 2 diabetes), FLOW (diabetes with kidney disease), and SOUL (diabetes with high cardiovascular risk). Follow-up lasted up to 156 weeks, or three years.
Those blood samples were then run through proteomic clocks. Proteomics means measuring the hundreds of proteins circulating in your blood. A clock is a formula that takes those protein levels and turns them into an estimated age for a specific organ, such as the heart or kidney.
The clocks were built using the UK Biobank, a large health study that followed hundreds of thousands of people over time. Researchers fed the formula each person's proteins and told it to find the patterns that predicted who died sooner. So the score is better read as a mortality risk estimate than a true age. A higher score than expected for your age means your proteins look like those of people who were more likely to die earlier.
Key points:
- Heart biological age fell in all five trials, by roughly 2 to 4 years versus placebo
- Kidney, brain, pancreas, lung, and multiorgan clocks fell by roughly 1 to 4 years, but results varied by trial and timepoint
- A blood-based frailty score dropped after the first year, with the clearest declines at week 52 or later
- Predicted risk scores for type 2 diabetes, cardiovascular disease, and some lung diseases improved
- Changes were not fully explained by weight loss, HbA1c (average blood sugar), or CRP (an inflammation marker)
Start with what happened in the placebo groups. Their clock scores crept upward over follow-up, which is what aging looks like in these models. Semaglutide shifted scores down relative to that drift. The heart clock was the most consistent, and the protein behind much of it was NPPB, a marker of cardiac strain that the heart releases when its walls are stretched.
The frailty score followed a different timeline. Frailty here means a blood-based estimate of how vulnerable your body is to decline. In FLOW, the score rose at week 12, then fell significantly at weeks 52 and 156. In SOUL, it was unchanged at week 13 and lower by week 104. Gains in this measure appear to take a year or more.
The paper then asked a harder question: how much of semaglutide's benefit runs through the clock? This is called mediation analysis, a statistical method that estimates how much of a treatment's effect travels through a given variable. In SELECT, the change in the heart clock statistically accounted for 36% of the effect on death from any cause and 40% of the effect on death from cardiovascular causes. In SOUL, it accounted for 49% of the effect on major cardiovascular events (heart attack, stroke, or cardiovascular death), 36% on all-cause death, and 34% on kidney outcomes. These analyses were exploratory, and they show a statistical link only. They do not show the clock causes the benefit.
One result went the wrong way, and it shows how these clocks can fail. The adipose (fat tissue) clock rose early on. Semaglutide lowers leptin, a hormone made by fat that signals how much energy you have stored. In the UK Biobank, low leptin in older adults marked people who were losing weight because of illness, so the model learned to read low leptin as a bad sign. In these trials, the leptin drop reflected intentional fat loss. The clock misread a healthy change as aging, and the adipose clock later fell in several trials.
The limitations are worth noting:
- This is a preprint that has not been peer reviewed
- The analysis was funded by Novo Nordisk, which makes semaglutide and also funded all five trials, and company employees are among the contributors
- Participants all had obesity, heart disease, diabetes, or kidney disease, so results may not apply to healthy people
- The SELECT proteomic sample was enriched for people who had cardiovascular events, so it was reweighted to better match the full trial, and fewer organ results stayed significant
- Clocks trained in one population can be miscalibrated in another
- A few proteins drive much of the heart and adipose clocks
- A lower clock score is not a clinical outcome, and the paper itself states that its results do not establish biological rejuvenation, causal mediation, or surrogate validity
So nothing here supports taking semaglutide as an anti-aging drug. The finding is narrower. Organ-specific protein clocks respond to a drug with proven benefits, the heart responds most reliably, and the response is not fully explained by weight loss. Whether these clocks can stand in for hard outcomes in shorter trials is still untested.
Two questions remain open for me.
The first is whether the benefit comes from the molecule or from eating less. Semaglutide works partly by cutting appetite, so people on it consume fewer calories, and caloric restriction is one of the best-studied ways to slow aging in animals. The paper adjusted for weight loss, blood sugar, and inflammation, and the clock changes held up. But statistical adjustment is not the same as separating the two. Weight loss, lower calorie intake, and the drug's direct actions on the heart, kidney, and brain are tangled together. A trial comparing semaglutide with the same weight loss achieved by diet alone would answer this, and I am not aware of one with proteomic clocks.
The second is whether any of this applies to people without metabolic dysfunction. Every participant had obesity, diabetes, heart disease, or kidney disease. In those groups, a drug that improves glucose control, blood pressure, and inflammation has a lot of damage to reverse. A healthy person with normal metabolism starts from a different place, and the same shift in a clock score may mean something smaller, or nothing.