Ultra-processed food tied to 32 separate health harms

Ultra-processed foods, as defined by the Nova food classification system, have been linked in individual studies to a widening list of health problems, from heart disease to depression. With meta-analyses multiplying across mortality, cancer, mental, respiratory, cardiovascular, gastrointestinal and metabolic outcomes, the field needed a single rigorous audit. Which associations hold up, how strong is each one, and how trustworthy is the evidence beneath them? An umbrella review, a systematic review of existing meta-analyses, is the tool built for exactly that question, and this is the one the field had been waiting for.

The team searched MEDLINE, PsycINFO, Embase and the Cochrane Database of Systematic Reviews for meta-analyses of studies that tracked people's diets and health, published from 2009 to June 2023. They identified 45 unique pooled analyses covering nearly ten million participants, including 13 dose-response associations. Each association was graded against pre-specified criteria running from convincing at the top down to no evidence at the bottom. Separately, the quality of the evidence behind each was assessed using the GRADE framework as high, moderate, low or very low.

Direct associations appeared between greater ultra-processed food exposure and 32 health parameters, 71% of those examined. The convincing tier included cardiovascular disease related mortality (risk ratio 1.50), type 2 diabetes (dose-response risk ratio 1.12), anxiety (odds ratio 1.48) and combined common mental disorders (odds ratio 1.53). Highly suggestive evidence covered all cause mortality (risk ratio 1.21), heart disease related deaths, depressive outcomes, poor sleep, wheezing and obesity (odds ratio 1.55). Quality ratings tempered the picture: 22 pooled analyses were rated low quality and 19 very low, with only four rated moderate, and 13 associations showed no evidence at all.

This is the broadest formal evaluation of the ultra-processed food literature to date, and it shows the strongest signals sit in cardiometabolic disease, common mental disorders and mortality rather than being spread evenly across every outcome studied. Because the included studies followed people rather than assigning diets, the review cannot establish that ultra-processed food causes these harms, and the mostly low quality ratings underline that limit. The authors argue the findings justify population level public health measures alongside urgent research into the mechanisms that might explain the links.

The consistency and sheer breadth of these associations give a reasonable basis for treating a diet heavy in ultra-processed food as a genuine health question rather than a passing worry, particularly where heart health, blood sugar and mood are concerned. At the same time, evidence of this kind cannot tell any individual how much their own risk would change by eating differently, and the quality behind most of the links was rated low or very low. It is a case for paying attention, not for alarm.