Did a month of new menus make older Australians biologically younger?

In a Sydney trial, people aged 65 to 75 ate one of four fully supplied menus for a month. A new analysis compared a "biological age" score, built from routine blood, urine and body measurements, across the menus, and headlines said ageing had been reversed. The paper itself warns against that reading, and the trial wasn't designed to measure ageing.

8 October 2026 · How we rate evidence

For four weeks in 2017 and 2018, a meal-delivery company brought all the food for about 130 people aged 65 to 75 in Sydney. Each person had been randomly assigned one of four menus. Nobody counted calories: they could eat as much as they liked, as long as it came from the delivery. In 2026, researchers at the University of Sydney published a new analysis of that trial's tests, asking a question the trial hadn't registered. Did any menu make people's bodies look younger?

The claim going around. The answer was published in Aging Cell in April. In May, Inc. ran "Want to Reverse Your Biological Age? A New Study Says to Change What You Eat." In September, ScienceDaily reprinted the university's press release under the headline "Scientists reversed biological age in older adults in just 4 weeks." The release itself is mostly careful. It quotes lead author Caitlin Andrews: "It's too soon to say definitively that specific changes to diet will extend your life."

Behind the headlines is a human question. Can changing what you eat in your late sixties make your body younger, and can a blood test tell?

What they did. All four menus got 14% of their energy from protein. Two were omnivorous, with half the protein from animals. Two were semi-vegetarian, with 70% from plants. Each came in two versions. The higher-fat one got about 40% of its energy from fat. The higher-carbohydrate one got about 28% from fat and 53% from carbohydrates, mostly whole, minimally processed foods. The higher-fat omnivore menu was closest to what participants normally ate, so it served as the comparison.

The team used a formula called the Klemera-Doubal method to turn routine measurements into a "biological age": 7 measurements in one version of the score, 15 in the other. They then subtracted each person's real age. The press release says the team combined "20 different biomarkers." In fact the two versions use 7 and 15, with 15 different measurements in all. The team had complete data for 104 people, 23 to 27 per menu.

What they found. After four weeks, the team compared each group's age gap with the higher-fat omnivore group's, allowing for where each person started. The higher-carbohydrate omnivore group's gap was 3.5 years lower on one version of the score and 4.1 years lower on the other. Those are averages with wide ranges. By our calculation from the paper's tables, the true difference could plausibly lie anywhere from about 0.3 to 6.7 years on the first version and 0.8 to 7.5 on the second. The higher-fat omnivore group itself showed no clear change.

The menu that cut both fat and animal protein had the smallest and least certain difference, although you might expect it to do best. So the release's line that people who "reduced either dietary fat or animal-based protein showed signs of a lower biological age" is on firmest ground for the lower-fat omnivore menu.

The trial itself is well run: menus were randomised using sealed envelopes, all food was supplied, and weighed records showed high compliance.

What it doesn't show. First, the trial wasn't built to test this score. Its registered main outcomes were how much people ate and a hormone called FGF21. Biological age appears nowhere in the registry, the paper doesn't say it was named in advance, and the measurements that make up the score had already been analysed and published in 2024, before this analysis was submitted. Our rating turns on that: it puts the main finding and the higher-fat semi-vegetarian result at Speculative, close to Early, rather than Early. The trial also stopped recruiting at 129 people of a planned 220. The paper makes six comparisons against the same reference group without correcting for multiple tests. With a standard correction for the three menus on each score, only one result stays clear of chance, and only narrowly.

Second, a lower score after four weeks isn't the same as slower ageing. The paper is cautious here; the headlines were not: "Caution is warranted in interpreting such changes as evidence of biological age reversal." Cholesterol and blood pressure are ingredients in both scores, and they respond to recent meals. In the supplement, changes in the score were weakly linked to changes in LDL cholesterol and, for one version, waist size. Nobody was measured after the menus ended.

Third, the score may say little about most of these participants. The team checked it against deaths in more recent US survey data. In men, a higher score predicted earlier death: on the 15-measure version, risk was 44% higher for each standard step up the scale. In women, it didn't clearly predict death, and 70 of the 104 participants were women. The group was also healthy. On average, they already scored younger than their real age at the start.

What it could mean for people. The three other menus all had more fruit, vegetables, legumes and fibre than the comparison, and each cut fat, animal protein or both. That's the direction dietary guidelines already point. The main trial paper, published in 2024, reported better readings on several health measures across all participants, whichever menu they had. This analysis adds that a composite score moved within a month. It can't say whether the change would last, or whether it would mean fewer heart attacks, less frailty or a longer life. For now, the honest answer to the human question is that it hasn't been shown.

Why it's still interesting. Trials in which researchers supply every meal let them change one part of a diet at a time, and this one did it in older adults. This result cuts two ways. If these scores respond within weeks, they could one day let researchers test diets without waiting decades for disease to appear. But the same speed is a warning. A biological-age number may reflect last month's meals more than how fast someone is ageing.

Open questions.

Does it last, and does it matter? Measure the score again months after people return to their usual food, and follow illness and survival in longer trials. If the drop fades within weeks, the score is tracking meals. If it lasts and predicts health, it becomes a fast way to test diets.

What moved the score? Recalculate it leaving out one measurement at a time. If cholesterol and waist size do most of the work, a standard cholesterol test would tell you as much as the "biological age." If many measurements shift together, the score is picking up something broader.

Does it work for women? Check the score against later illness and death in women in another large study, and report the diet results for men and women separately. Two-thirds of this trial's participants were women, so the answer decides how much it tells us about them.

A note on timing. The paper came out online on 27 April 2026, and the meals were delivered in 2017 and 2018. The first coverage we found came in May. ScienceDaily's repost, four and a half months after the paper came out, started a second round in September. The study hadn't changed.