Researchers cross classified more than 200,000 US adults from three long running cohorts followed from the 1980s into the 2020s, separating overall diet quality (a plant based diet index) from processing level (NOVA). Roughly half of participants died or developed cardiovascular disease or type 2 diabetes over follow up.
Within each diet quality tier, risk was reported as close to identical whether the diet was ultra processed or minimally processed. The presenting author's framing was that people should focus on what they are eating overall, not just whether it is processed. No hazard ratios, confidence intervals, person years or covariate lists exist in the public record, so "almost identical" is the presenters' characterisation as reported, not a quantified equivalence.
The numbers
| Participants | more than 200,000 US adults |
| Follow up events | about half died or developed CVD or type 2 diabetes |
| Risk within each diet quality tier, by processing level | reported as "almost identical"; no hazard ratios or confidence intervals published |
What this does not show
- That processing does not matter. A wide, imprecise null is not the same as a demonstrated equivalence, and no effect sizes have been published, so nobody can currently assess how precise "almost identical" is.
- A causal separation of processing from what travels with it. Observational design cannot cleanly separate processing from cost, convenience, hyper palatability, eating rate, portion size, and the socioeconomic gradient underneath all of them.
- That this generalises to the public. The cohorts were not named in the release. They are very likely health professional populations, which is an inference, and health professionals are not the general public.
- A settled methodology. Cross classifying NOVA against a diet quality index is itself a contested choice; a plant based diet index may already absorb much of what NOVA is trying to capture.
- A citable finding. This is a conference abstract selected by committee but not journal peer reviewed. ASN says explicitly it should be treated as preliminary until published.
Caveats worth holding
- Conference abstract. ASN states that NUTRITION 2026 abstracts were selected by a committee but have not undergone journal peer review and should be treated as preliminary until published.
- No effect sizes exist in the public record whatsoever. The "almost identical" characterisation is attributed to the presentation as reported, not verified against published statistics.
- Observational design cannot cleanly separate processing level from cost, convenience, hyper palatability, eating rate, portion size and socioeconomic position.
- The three cohorts were not named. The inference that they are the Nurses' Health Study, NHS II and the Health Professionals Follow up Study is plausible but unconfirmed, and those are health professional populations.
- Cross classifying NOVA against a diet quality index is a contested methodological choice and the likely locus of critique.
- Wait for the peer reviewed paper before treating this as citable.