Of 204,494 UK adults, 848 later turned up in hospital records with a myeloid blood cancer. Those eating more ultra-processed food were slightly more likely to be among them

Ultra-processed food has already been tied to a list of chronic diseases. Much less is known about myeloid neoplasms, a group of blood cancers and related conditions that begin in the bone marrow. A team led from the Jiangsu Institute of Hematology at Soochow University in Suzhou went looking in the UK Biobank, and published what they found in The American Journal of Clinical Nutrition on 26 October.

They followed 204,494 people who had no hospital record of these diseases when the clock started. Each person's share of energy from ultra-processed food was averaged across the online diet questionnaires they had completed, and follow-up began at their last valid one. Over a median of 10.5 years, 848 people developed a myeloid neoplasm recorded in hospital inpatient data. Allowing for demographic, behavioural, socioeconomic, dietary, physical activity and body size factors, each one standard deviation higher share of ultra-processed energy went with a 12% higher hazard (hazard ratio 1.12, 95% CI 1.05 to 1.21).

The clearest signal among the subtypes was in myeloproliferative neoplasms, in which the marrow makes too many red cells, white cells or platelets (hazard ratio 1.15, 95% CI 1.04 to 1.27, and still significant after correcting for multiple tests). Measuring ultra-processed food by its share of food weight instead of energy also showed the association. The team then looked at clonal haematopoiesis of indeterminate potential, or CHIP, acquired blood-cell mutations that become common with age and raise the odds of a later blood cancer. In people without detectable CHIP at the start the link remained, at 1.10 (95% CI 1.01 to 1.20), and there was no evidence that CHIP status changed it. So the association was not limited to people who already carried those mutations.

Keep the scale in view. A 12% rise is relative, and it applies to something uncommon: across the whole group, about 1 person in 240 developed one of these conditions during follow-up. Cases came from hospital inpatient records, so anyone whose illness never reached those records would be missed. Diet was self-reported, and this is an observational study, where people eating more ultra-processed food may differ in ways that adjustment cannot fully catch. The team found lipid, inflammatory, blood and immune markers that lined up with intake, some of them also seen in outside datasets, but they say plainly that this offers biological context and does not establish a causal pathway. The full paper is behind a paywall, so we checked these figures against its abstract. The abstract does not say how many percentage points one standard deviation represents, which is why we cannot turn the 12% into a practical amount of food.