Research explained: prospective cohort study
A blood signature of ultra-processed eating was linked to heart disease and type 2 diabetes over 20 years
In 6,010 British civil servants, a pattern of 34 blood metabolites tied to ultra-processed food tracked with higher risk of both conditions, while a 50-metabolite pattern tied to minimally processed food tracked with lower risk.
People whose blood showed a chemical pattern typical of ultra-processed eating were more likely, on average, to develop heart and blood vessel disease and type 2 diabetes over about two decades, while a pattern typical of minimally processed eating was linked to lower risk, according to a study of 6,010 adults in BMC Medicine (published July 21, 2026). The blood patterns accounted for roughly half of the link between food processing and each disease. The study made headlines this week.
What the researchers did
Researchers at University College London and Imperial College London used Whitehall II, a long-running study of British civil servants. Participants filled in a 127-item food questionnaire twice, in 1991-1994 and 1997-1999, and each food was sorted by how processed it is using the Nova system. Fasting blood samples were tested for 233 small molecules, called metabolites, using NMR spectroscopy.
The researchers found a set of 34 metabolites that tracked with ultra-processed food intake and a set of 50 that tracked with minimally processed food. The ultra-processed pattern went along with more inflammation-related and artery-clogging molecules and less omega-3 fat and HDL cholesterol. The minimally processed pattern showed the reverse.
Over a median follow-up of 20 years for heart disease and 18 years for diabetes, there were 1094 and 853 new cases. People in the top third of the ultra-processed score had higher risk of both conditions than those in the bottom third (hazard ratios 1.19 and 1.20). A high minimally processed score was linked to lower risk (hazard ratios 0.81 for heart disease and 0.60 for diabetes). The blood patterns explained between 42.8% and 51.1% of the links between processing and disease.
How strong is the evidence?
This is an observational study, so it shows links, not cause and effect. Diet was self-reported decades ago, and civil servants are not a mirror of the whole population. About half of the link stayed unexplained, which the authors say points to other mechanisms. The authors declare no competing interests; the abstract does not describe funding.
What it means for you
The findings fit a simple habit: build more meals from foods close to their natural state. Cooked lentils, plain oats, salmon for omega-3 fat and vegetables like broccoli are easy places to start. You do not need to avoid every packaged food. Browse more ideas on our legumes page.