The Planetary Health Diet arrives in consumer conversation with two promises folded into one name: a way of eating that is better for the person and better for the planet. EAT Forum, the non-profit that convened the commission behind it, describes the pattern as “a science-based dietary pattern designed to promote human health and protect the planet,” built around whole grains, fruits, vegetables, nuts and legumes, with fish, dairy and meat in moderate or small amounts.

A study published on 16 November in Annals of Medicine put the health half of that pitch to the longest test it has had: a prospective cohort of more than 100,000 British adults, followed through national health registries for a median of 12.5 years. The result is narrower than the framing. Diets that scored higher on the adherence index were followed by modestly lower rates of stroke and type 2 diabetes. They were not followed by measurably lower rates of ischemic heart disease, or of cardiometabolic multimorbidity — the state of carrying two or more of these conditions at once.

What the score actually counts

The study did not ask people to follow a menu. It scored what they had already reported eating against the Planetary Health Diet Index, a research instrument derived from the EAT-Lancet reference diet and following the method published by Bui and colleagues. The index has 14 dietary components, grouped as adequacy, optimum and moderation, each scored from 0 to 10 against targets set for a 2,500-kilocalorie day. One component, unsaturated fats, could not be scored from the available data, so the possible range in this study was 0 to 130 rather than 140.

That detail matters for anyone who has seen the diet presented as a simple plate graphic. Producing a score requires detailed intake data — here, two or more 24-hour dietary recalls from the Oxford WebQ questionnaire, which covers 206 foods and 32 beverages, averaged across assessments. Of the 502,128 people recruited into UK Biobank, only 126,743 completed at least two of these recalls, and after further exclusions the analysed groups ran from 108,596 people for ischemic heart disease to 112,323 for multimorbidity. The participants averaged about 55 years old and more than half were women.

The score also tracks social position and body weight closely. In the lowest adherence quartile, 67.6% of participants were overweight or obese and 46.6% were women; in the highest quartile those figures were 49.1% and 69.0%, with more university degrees and less deprivation. The statistical models adjusted for all of it — age, sex, ethnicity, education, employment, deprivation, body mass index, smoking, alcohol and physical activity, plus total energy intake — but a score that moves with so many other things is hard to isolate.

Twelve and a half years of diagnoses

Outcomes were not self-reported. Cases of ischemic heart disease, stroke and type 2 diabetes were assembled from primary care records, hospital inpatient data and death registers, coded to ICD-10 and followed to 1 September 2023. Over the follow-up period the cohort recorded 6,366 cases of ischemic heart disease, 2,319 strokes, 4,379 cases of type 2 diabetes and 2,231 cases of cardiometabolic multimorbidity.

On the fully adjusted model, each five-point increase in the index was associated with a 3% lower risk of stroke (hazard ratio 0.973, 95% confidence interval 0.951 to 0.997) and a 3% lower risk of type 2 diabetes (0.968, 0.952 to 0.984). For ischemic heart disease and for multimorbidity, the association did not reach significance.

Three percent per five points is a small number on a 130-point scale, and the authors do not dress it up. Comparing the top quartile with the bottom, type 2 diabetes risk was 12% lower (0.881, 0.805 to 0.965). The quartile pattern is not a clean gradient: the third quartile did slightly better than the fourth, and the authors report a significant non-linear relationship for diabetes, with the lowest risk at a score of about 84 points. They note that a separate UK analysis found cardiovascular risk plateauing at around 80 points.

A non-monotonic result is worth reading carefully rather than smoothing over. It can mean the index’s upper range is dominated by people whose other advantages are already captured, or that the scoring gives credit in ways the outcome does not distinguish. The paper does not settle which.

What happened after a first diagnosis

The more interesting analysis is the one that follows people through stages. Using a multi-state model — healthy, then a first cardiometabolic disease, then multimorbidity, then death — the authors found that higher adherence was associated with a lower risk of the first step and of dying without one. Top quartile versus bottom: 11% lower risk of a first cardiometabolic diagnosis (0.89, 0.84 to 0.94) and 19% lower risk of death from a healthy baseline (0.81, 0.73 to 0.89).

The transitions after a first diagnosis mostly did not reach significance. The authors read this as support for a role in primary prevention and not in secondary prevention, and they are candid about why the progression result may be hard to detect: participants who developed only one condition were censored, death competes with progression, and people who already have one cardiometabolic disease are at high risk of a second regardless of diet.

They also compute E-values, a measure of how strong an unmeasured confounder would have to be to explain the association away. Those values ranged from 1.11 to 1.67 — the authors call this moderate robustness, which is the right word. An observational cohort can show that a dietary pattern travels with better outcomes. It cannot show that changing the pattern causes them.

Why the name promises more than the data

The environmental half of the Planetary Health Diet’s appeal is not tested here at all. No emissions, land use or water outcomes were measured; this is a study of diagnoses in one national cohort.

It is also worth knowing how the reference diet was constructed. EAT’s own description states that the diet “is based entirely on the direct effects of different diets on human health, not on environmental criteria,” and that the name arose from evidence that adopting it would reduce the environmental impact of current diets, along with their nutritional shortfalls. The health targets came first; the sustainability modelling was layered on afterwards. A cohort like this one can speak to the first part only.

Stripped of branding, the eating pattern is familiar dietetic ground. The World Health Organization’s healthy-diet fact sheet recommends at least 400 grams of fruit and vegetables a day for everyone over 10, at least 25 grams of naturally occurring fibre, free sugars below 10% of energy, saturated fat at or below 10%, and protein at 10% to 15%. It adds that for many adults “a shift towards more plant-based sources of protein may bring health benefits, particularly when the shift is away from red meat.” That is a close description of what a high-scoring diet looks like — and none of it is new. The population-level shortfall in fibre, which we have covered separately, is one reason these patterns keep showing up in cohort data.

What actually has trial evidence behind it

If the practical question is how to lower type 2 diabetes risk, the strongest evidence is not a diet index. It is a structured lifestyle programme. The National Institute of Diabetes and Digestive and Kidney Diseases states plainly that type 2 diabetes “can be delayed or even prevented,” and points to the Diabetes Prevention Program: losing 5% to 7% of starting weight — about 10 to 14 pounds for someone at 200 pounds — through a reduced-calorie eating plan and activity most days of the week. It also suggests asking a clinician whether metformin is appropriate.

That is a randomised-trial finding with a defined target, and it does not require anyone to compute a 130-point score. What the new cohort adds is context for why such programmes emphasise plant-forward, fibre-rich eating: the people whose diets looked most like the reference diet had somewhat fewer strokes and diabetes diagnoses over more than a decade, without any single ingredient explaining it.

For readers weighing plant-based substitutes as a shortcut, the label gap is real — our look at plant-based meat and dairy found the products are not nutritionally interchangeable with what they replace — and the trial record on ultra-processed food is still being written.

What we did not find

No randomised trial has assigned people to raise their Planetary Health Diet Index and measured whether diagnoses fall; until one exists, the causal claim stays open, and the authors say so. This cohort found no association with ischemic heart disease or with multimorbidity, so the “heart-healthy” shorthand outruns the result. No environmental outcome was measured, despite the name. And the cohort is about 96% White, drawn from a volunteer sample with more education than the general population, which limits how far the estimates travel — the authors flag this themselves, along with the limits of 24-hour dietary recall and the absence of treatment data after a first diagnosis.

What remains is a modest, well-measured association in the direction the diet’s advocates would expect, arriving with more caveats than headlines usually carry. Anyone with existing heart disease, diabetes or kidney disease — or who is considering a big shift in how they eat — should take the question to a healthcare professional rather than to a score, as our medical disclaimer sets out.