Ten healthy diets compared against biological age. The gap between them is smaller than it looks

In brief

The write-ups put it like this: «several very different diets are linked to slower biological ageing». That is true. But the interesting part is not the headline — it is the two things the retelling usually drops: how small the effect is and how badly people fit inside any one diet's frame.

Let us go through it from the source (PMID 42668319^[1], Nature Communications, August 2026).

What was done

6470 participants of the German Rhineland Study — 57% women, mean age 56, ranging from 30 to 95. For each of them ten different diet quality scores were calculated: Mediterranean, DASH, MIND, Nordic, the Alternate Healthy Eating Index, three plant-based indices, the dietary inflammatory index and EAT-Lancet.

Then DNA methylation was read across roughly 850,000 points in the genome, and three estimates of biological age were computed. The findings were re-checked in an independent cohort of 1034 people.

Finding one: «healthy eating» is not one thing

The first thing the authors did was check how much the top quarters of the different scores overlap. You would expect someone who eats well to sit near the top on all of them.

It came out the other way: the overlap is minimal. The same person can be in the top quarter on one score and near the bottom on another.

Finding two: all ten track — but look at the numbers

All ten scores were significantly associated with DunedinPACE, which is not an «age» but a pace of ageing: the measure is scaled to 1, above 1 means ageing faster than a year per year, below 1 means slower.

Only two scores held up across all three clocks — DASH and the Nordic score. Here are their numbers in full.

A word on why only two. The three clocks correlate weakly with each other — coefficients of 0.22–0.28. They are different instruments, not three copies of one: they measure different aspects of ageing, and a diet may well move one without moving another.

This is what the retellings leave out. The numbers run in hundredths of a standard deviation. Not «five years younger», not «minus a decade» — three hundredths.

Now look at the opposite pole of the same table. The dietary inflammatory index, per one standard deviation, shifts the pace of ageing by +0.07 SD — almost as much as DASH shifts it the other way (−0.10). A step towards bad eating costs about what a step towards the best of the healthy diets gains. And the distance between the healthy diets themselves is hundredths.

The study's first author puts it plainly: the differences between the diets are measurable, but modest in absolute terms.

Why that is not a reason to shrug

A modest effect at population level is not the same as a useless one. First, this is one standard deviation on a diet score — a fairly moderate shift in eating, not a turn to asceticism. Second, we are talking about a feature that changes across decades.

But the reverse holds too: numbers like these do not support a promise of «minus years». If a guide or an advert tells you a diet will make you five years biologically younger, it is not talking about this study.

Finding three: different marks, shared pathways

The most elegant part. Different diets move different sets of methylation sites — each leaves its own trace. But once those sites are grouped into biological pathways, the pathways turn out to be shared: cell signalling, metabolism, neurogenesis.

And one detail worth noticing

EAT-Lancet showed no association with either GrimAge or PhenoAge — that is the authors' verbatim finding. But stated in full it is more interesting: no association with accumulated biological age, while the association with the pace of ageing is there and is significant. The authors suggest this dietary pattern may influence the rate of ageing without showing up in markers of age already accumulated.

No conclusions about it can be drawn from a single study. But the detail is instructive: the louder a diet is sold as the one correct answer, the more useful it is to look at what the measurements show.

What the study does not say

Listing it honestly, because these are the places where retellings straighten the curve.

• It does not say diet slows ageing. A cross-section shows coincidence, not sequence.
• It does not say one diet beats another. The spread between the scores is smaller than the distance between eating well and eating badly.
• It promises no years. Every effect is expressed in fractions of a standard deviation, not in years of life.
• It was not tested across populations. Two German cohorts, participants predominantly German.

On top of that, adjusting for body mass index and for objectively measured physical activity attenuated the associations — meaning part of the effect runs through weight and movement rather than around them.

The bottom line

The practical meaning of this work is the opposite of what people usually read diet research for. It does not crown a winner — it shows that choosing between good diets matters less than the fact of eating well.

And if one line is to be carried away, let it be this: the gap between ten healthy diets is measured in hundredths; the gap between healthy eating and its absence is a different conversation entirely.

On how epigenetic clocks relate to ordinary lab tests, and what is actually worth checking, we have a guide to four systems.

This material is for information only. Changing your diet with chronic illness, in pregnancy, or alongside drug therapy is a conversation to have with a physician.