Do Ozempic and Zepbound Slow Aging? Reading Lilly and Novo's Aging-Clock Data

At the Aging Research and Drug Discovery meeting at Harvard on October 1 to 3, 2026, the makers of Ozempic, Wegovy, Mounjaro, and Zepbound brought a new kind of evidence. Novo Nordisk and Eli Lilly reported that their GLP-1 drugs made people look younger on molecular "aging clocks."

These clocks estimate biological age, a measure of how worn the body is on the inside, as opposed to chronological age, which is simply the years since birth. Novo said semaglutide, the drug in Ozempic and Wegovy, lowered a blood-protein estimate of heart age by about 2 to 4 years in five of its trials. Lilly said that in people taking tirzepatide, the drug in Mounjaro and Zepbound, 15 DNA-based clocks all moved forward by less than the 72 weeks that had passed.

Some outside researchers were impressed. "Two years is a pretty strong effect, in my book," Steve Horvath of UCLA, who published one of the first epigenetic clocks in 2013, told MIT Technology Review. Lilly was more careful and called its own results "preliminary and hypothesis-generating."

The real question is narrower than the headlines: did these drugs slow aging, or did they move a lab score that tends to track health? Most of the doubt sits in the gap between those two ideas.

An aging clock is a formula: researchers measure thousands of molecules in blood, then train a computer model to find the pattern that best predicts a person's age or risk of death.

Epigenetic clocks read DNA methylation, the small chemical tags on DNA that help switch genes on and off and that shift in a steady way as people get older. Horvath's 2013 clock used 353 of these tag sites and was built from about 8,000 samples. Later clocks were trained on health instead of birthdays: PhenoAge (2018) was trained on a composite of clinical measures linked to lifespan and health, and GrimAge (2019) uses DNA patterns that stand in for smoking and for blood proteins linked to lifespan. In the original studies, both predicted death and disease better than the first generation of clocks.

DunedinPACE (2022) works more like a speedometer than a clock. It was trained on how quickly 19 measures of organ health declined over two decades in people born in Dunedin, New Zealand, in 1972 and 1973, so a score of 1.0 means one year of biological aging per calendar year and a lower score means slower aging.

Proteomic clocks read proteins in the blood instead of DNA tags. A 2024 clock built from 204 proteins in 45,441 UK Biobank participants was linked to 18 chronic diseases and to death from any cause. Organ clocks go a step further by using proteins made mostly by one organ, and in a 2023 Stanford study of 5,676 adults, people whose blood proteins made their heart look old had a 250% higher risk of heart failure. Novo used organ clocks from a Harvard team led by Vadim Gladyshev, trained on UK Biobank protein data to predict time to death.

All of these links come from population studies, in which people with an "older" score tend to get sick and die sooner. That pattern does not prove that pushing the score down with a drug adds healthy years. A measure used this way is called a surrogate marker, meaning a stand-in for an outcome, such as death, that takes too long to measure directly.

Novo's analysis is the larger of the two, and it had a comparison group. According to its meeting abstract, the company measured blood proteins in 10,052 people from five randomized semaglutide trials. MIT Technology Review reported that about half took the drug and half took a placebo, a dummy shot. The placebo group served as a control group, meaning people treated the same way except for the drug, so that differences between the groups can be credited to the drug.

The fullest public version is a poster Novo showed at the American Diabetes Association meeting in June 2026. In a subset of SELECT, Novo's large heart trial, 791 people on semaglutide and 885 on placebo had their proteins measured. At week 20, heart and kidney clock scores were lower on semaglutide than on placebo. A combined multi-organ clock was lower at week 68 in two weight-loss trials, STEP 1 and STEP 2, and at week 104 in SELECT. The poster also reported that the broader whole-body clocks did not change.

It helps to look inside the heart clock, which is made of seven proteins. Two of the top contributors are NPPB, the protein behind the BNP blood test for heart strain, and TNNI3, which encodes cardiac troponin I, a marker of heart muscle damage. Part of what the clock calls a "younger" heart is therefore lower levels of familiar heart-stress markers, which is good news for the heart but something doctors can already measure in other ways.

Novo then ran a mediation analysis, a statistical method that tries to split an effect into parts. By that analysis, most of the drop in heart and kidney age came from the drug directly rather than through weight loss. The change in the heart clock accounted for about 36% of semaglutide's lower death rate in SELECT. Novo sponsored the analysis, and Gladyshev, whose lab built the organ clocks, is listed as a coauthor.

Novo's other headline number came from a model. Its researchers applied SELECT's results to a real-world UK cohort of 19,117 people and projected an average gain of 1.9 years of life, according to Longevity.Technology's report of the talk. That figure depends on assumptions about who would take the drug and for how long, and it was not measured in any trial. For more on reading claims like these, see our guide to drug-company comparisons.

Lilly's data came from SURMOUNT-5, a 72-week trial that randomly assigned 751 adults with obesity to tirzepatide or semaglutide. Both patients and doctors knew which drug was given, and the aging work used only the tirzepatide group. Of 82 people tested, 71 had blood samples from both the start and week 72, a span of 1.38 years, and by week 72, 91.5% were on the top 15 mg dose.

Kevin Duffin, Lilly's vice president of aging research, reported that all 15 epigenetic clocks measured in years moved forward by less than 1.38 years. Two of the 15 reached statistical significance, with median gains of 1.1 and 0.7 years. DunedinPACE fell from a median of 1.13 to 1.08. Lilly described the analysis as single-arm and not placebo-corrected, and said the results need "replication in larger, ideally randomized studies."

Single-arm means there was no control group, and without one, no reader can tell how much of the change came from tirzepatide, how much from losing weight by any route, and how much from chance. Lilly said it plans to extend the work to semaglutide and to other weight-management studies.

Several things can push a clock score down without changing how fast someone ages.

Eating less and losing weight is one of them. In the CALERIE trial, 220 adults without obesity were randomly assigned to cut calories by 25% or to eat as usual for two years. Most fell short of the goal and averaged about 12% fewer calories, yet DunedinPACE slowed by 2 to 3% compared with the control group, while PhenoAge and GrimAge did not change. GLP-1 drugs cut food intake sharply, so some movement in the clocks would be expected from that alone. Novo's mediation analysis tried to account for weight loss, while Lilly's design had no way to do so.

Chance is another. People who score high on a first test tend to score closer to average on the next one, a pattern called regression to the mean, and with no control group that drift can look like a drug effect. Clocks are also noisy. A 2022 study in Nature Aging found that technical noise alone made six popular epigenetic clocks disagree by up to 9 years on repeat runs of the same sample. The authors built steadier versions, such as PCGrimAge, that agreed within about 1.5 years for most repeats.

Even the way a score is reported can change how strong it looks. A hazard ratio compares how often an event happens in one group with how often it happens in another. In a separate abstract at the same meeting, researchers from Cambridge, Cologne, and Yale showed that simply rescaling clock scores pushed hazard ratios as high as 300 billion, even though the clocks got no better at predicting who would die.

The strongest controlled evidence so far comes from a smaller, independent trial. Researchers led by a University of California San Diego team reanalyzed blood from a 32-week, placebo-controlled trial of semaglutide 1.0 mg in people with HIV and abnormal fat buildup. Of 108 people randomized, 84 had samples from both time points, and DunedinPACE was 9% slower on semaglutide than on placebo, and several second-generation clocks also improved. The analysis was post hoc, meaning it was planned after the trial was done, and it was funded mainly by the NIH. The authors wrote that the small size, the HIV-specific group, and the short follow-up limit how widely the results apply.

Clocks matter only if they track outcomes people care about, and for semaglutide, large trials already measure some of those outcomes directly.

SELECT enrolled 17,604 adults aged 45 or older who had heart disease and overweight or obesity but no diabetes. Over an average of 39.8 months, 6.5% of the semaglutide group had a heart attack, a stroke, or a heart-related death, compared with 8.0% on placebo. That is a hazard ratio of 0.80, or a 20% lower rate. Deaths from heart causes were lower but missed statistical significance (hazard ratio 0.85, p=0.07). SELECT tested its secondary goals in a fixed order, so that miss meant the trial could not formally test the next ones, including death from any cause. A 2024 analysis reported 19% fewer deaths from any cause (hazard ratio 0.81), with heart and non-heart deaths contributing about equally and fewer people dying of infections, including COVID-19.

FLOW tested semaglutide 1.0 mg in 3,533 people with type 2 diabetes and chronic kidney disease. Its main outcome, a combination of serious kidney events and deaths from kidney or heart causes, fell 24%, and death from any cause was 20% lower (hazard ratio 0.80), a result that passed the trial's formal testing. Our kidney guide covers FLOW in more detail.

These results show fewer deaths and disease events in people who were already sick. They do not show that the drugs slow aging itself, and they tell us nothing about lean, healthy adults. Horvath made the same point: "I did not yet see sufficient evidence that they will benefit skinny people." Gladyshev told MIT Technology Review, "In an unhealthy population, I do think it's an anti-aging drug," but cautioned that the results may not apply to healthy people.

For tirzepatide, the outcome trial is SURMOUNT-MMO, which enrolled 15,374 adults with obesity. Its main goal counts death from any cause along with heart attack, stroke, artery-opening procedures, and heart failure events. ClinicalTrials.gov lists its estimated primary completion date as October 2027.

A convincing answer needs several things the current data lack. The clocks should be chosen before a trial starts, so researchers cannot pick the best-looking ones afterward. The trial needs a placebo group and enough people to detect small changes, and it should include healthier older adults, since they are the people the longevity pitch targets. It also needs long follow-up showing that people whose clocks slowed went on to have fewer illnesses and deaths.

Novo's mediation analysis is an early try at that last step, because it links clock changes to real deaths in SELECT. It is still a company analysis of a subgroup, shown on a poster rather than in a peer-reviewed paper.

No GLP-1 drug is approved to slow aging; the FDA approved them for specific conditions, such as type 2 diabetes and obesity, and they have real side effects. In SELECT, 16.6% of people on semaglutide stopped the drug because of side effects, compared with 8.2% on placebo.

Home "biological age" tests use the same kinds of clocks described here. Because a single result can swing by years from measurement noise alone, one test says little about whether a drug is working. Some people also take small "microdoses" of GLP-1 drugs as a longevity plan; our microdosing explainer covers what is known about that practice.

If you have obesity, diabetes, heart disease, or kidney disease, the outcome trials give solid reasons to talk about these drugs with a clinician. If you are lean and healthy, the aging-clock data shown so far do not establish a benefit, and a doctor who knows your history is the right person to weigh the risks.