One can of soda a day. That, according to science, is the exact line beyond which type 2 diabetes risk starts to climb in earnest. But the most interesting finding in this study isn’t that threshold, it’s that not all ultra-processed foods are equally harmful.
The study, published in Diabetes & Metabolism Journal by a team led by Yujin Kim, pools data from 12 prospective cohort studies covering a total of 714,199 participants and 38,308 type 2 diabetes cases recorded during follow-up. It is, by far, the most complete analysis to date on the link between ultra-processed food and diabetes, because it includes populations outside Europe and North America (South Korea, Brazil and Australia) that earlier meta-analyses had left out.
Risk climbs 48% when comparing extremes. People with the highest ultra-processed food intake had a 48% higher risk of developing type 2 diabetes than those with the lowest intake (relative risk of 1.48). The association held up almost unchanged even after adjusting for diet quality, total caloric intake and body mass index, suggesting the effect isn’t simply explained by eating worse or eating more overall.
Researchers calculated the risk per gram consumed. Beyond comparing extremes, the team translated the finding into more manageable figures: every 10% increase in the share of ultra-processed food in the daily diet raised risk by 14%, every additional 100 grams a day raised it by 5%, and every extra daily serving raised it by 4%. And here’s the study’s most striking finding: when the risk curve was plotted in absolute grams, it wasn’t a straight line. Below 300 grams a day, risk rises moderately, but beyond that amount, equivalent to one standard can of soda, the slope becomes much steeper. And there’s no ceiling: risk keeps climbing the more you consume, with no upper limit detected.
Not all ultra-processed foods weigh the same on the scale. This is perhaps the most useful part of the study for anyone looking to apply it to daily life. Breaking consumption down by food type, processed meat (cold cuts, sausages, deli meats) was the subgroup most strongly associated with diabetes (34% higher risk), followed by sugar-sweetened or artificially sweetened beverages (5% higher risk). But ultra-processed cereals and breads, and packaged savory snacks, showed an inverse association, meaning slightly protective, likely because some of those products include whole grain, and in the case of packaged sweets and desserts, thanks to fiber or cocoa flavanols, which improve insulin sensitivity.
The link was stronger in Europe and North America than in the rest of the world. The authors attribute this, at least in part, to higher consumption of processed meat, the subgroup most tied to risk, in those regions compared with others. It’s a reminder that “ultra-processed” isn’t a uniform category: what kind of ultra-processed food you’re eating matters a great deal.
As with any study of this kind, it’s worth remembering this is an observational association, not an experiment proving direct causation. The authors themselves note limitations such as possible measurement error from estimating intake through food frequency questionnaires not originally designed to classify ultra-processed foods, and the impossibility of fully ruling out unmeasured confounding factors. Even so, the fact that the result holds after adjusting for diet quality, total calories and BMI, and replicates across very different populations, makes this one of the strongest pieces of evidence published so far on this link.
Source: Kim Y, Cho Y, Kim JE, Lee DH, Oh H. “Ultra-Processed Food Intake and Risk of Type 2 Diabetes Mellitus: A Dose-Response Meta-Analysis of Prospective Studies.” Diabetes & Metabolism Journal, 2025. Open access (CC BY-NC). DOI: 10.4093/dmj.2024.0706. PMID: 40490026. PROSPERO registration: CRD42024503079.
