Review Finds Ultra-Processed Food Subtypes Differ Sharply in Diabetes Risk

The paper is a systematic review of observational studies, combined with a new analysis of one large cohort. The review covered 6 cohort studies with 635,332 adults who did not have type 2 diabetes at the start. Follow-up ranged from 6 to 26.1 years. The authors screened 222 articles to find those six. Observational studies can show links. They cannot show cause.

The authors sorted ultra-processed foods into eight subtypes: breads and cereals; sauces, spreads and condiments; sweets, snacks and confectionery; dairy products; packaged snacks; artificially and sugar-sweetened beverages; meat-based products; and ready-to-eat dishes. They did not pool the results into one number. The studies measured intake in different units and adjusted for different factors, so the team looked at the direction of each association instead.

According to the review, most studies found that ultra-processed ready-to-eat dishes, meat-based products, and artificially and sugar-sweetened beverages were positively associated with type 2 diabetes. Ultra-processed dairy products, and sweets, snacks and confectionery, were generally inversely associated, although results varied. An inverse association means higher intake was linked to lower diabetes risk. It does not show that those foods protect against the disease.

The authors also caution about the evidence itself. They rated the overall risk of bias as moderate to high in all included studies. That means the individual findings should be read with care.

Their central argument concerns definitions. In causal reasoning, a question about an intervention should be clearly defined. If one label covers foods with different effects, results are hard to interpret and hard to apply elsewhere. The authors give an example from earlier research: a higher intake of total potatoes was linked to higher diabetes risk, but when split up, the higher risk appeared only for French fries, not for boiled, baked or mashed potatoes.

To show how grouping can hide patterns, the team ran a post-hoc analysis in the Danish Diet, Cancer and Health cohort, which included 52,201 participants. They built a fictive group called the "terrible five": sugar-sweetened beverages, processed meat, red meat, refined grains and vegetables. The first four have been linked to higher diabetes risk in earlier cohorts, and vegetables to lower risk. The combined group was positively associated with type 2 diabetes, while vegetables on their own were inversely associated. The authors stress that this group is not meant to represent ultra-processed food. It is an extreme example of how mixing very different foods can mask individual associations.

For personal choices, the study suggests that the category label may say less than what is actually in the food. For policy, the authors conclude that broad groupings like ultra-processed food could have unintended consequences and may hamper policies built on them unless the definition or the mechanistic understanding improves. The study does not test any diet and does not recommend one.

This adds to The Arising's October 20 report on a review of 46 studies, which found ultra-processed food and heart disease research methodologically weak. That was a different review on a different outcome. Both point to limits in the evidence, and this one adds a question about whether the category is defined well enough to study.

What to watch next

Studies that measure specific ultra-processed subtypes with better methods, and any work that clarifies the mechanisms behind the differences.

Faster Epigenetic Aging Linked to Higher Death Risk Across 381,000 Adults, Meta-Analysis Finds Sleep Shortened and Shifted Later as Electricity Reached Argentine Communities, Study Finds Exercise for Mild Cognitive Impairment: Network Meta-Analysis Ranking Could Not Be Confirmed FDA Clears AI Tool Estimating Alzheimer's Risk in Mild Cognitive Impairment Low Activity Plus Weak Grip Linked to Higher Cognitive Impairment Risk Review of 248 Studies Ranks Protein Plus Strength Training Top for Sarcopenia Amylin Drug Petrelintide Cut Weight Up to 9.2% in Diabetes Trial Hearing and Vision Loss Linked to Body-Mind-Memory Decline, Partly via Loneliness NAD+ and Calorie Restriction May Work Through Brain Cell Antennae in Mice Heart Amyloid Drug Tafamidis Linked to Healthier Blood Vessel Cells Eye Gene Therapy Linked to Vision Gains in Stargardt, Heads to Geographic Atrophy Trial Stair-Climb Test Gave Only Rough Estimate of Aerobic Fitness Body-Clock 'Circadian Depression' Subtype Found in a Quarter of Patients Epigenetic Age Acceleration Tracks Kidney Disease Progression in Type 1 Diabetes Most Epigenetic Aging Clocks Fail to Predict Dementia, Meta-Analysis Finds Review in Public Health Asks Whether Exercise Timing Affects Heart and Metabolic Markers Fasting Mimicking Diet Programme for Diabetes and Obesity Reaches UK Clinics AI Drug Firm Pathos Raises $237 Million, Shifts Toward Alzheimer's APA Survey: Most Americans Accept Chronic Stress, Nearly Half Withdraw Socially Tiny Peptide Alleviates Artery Calcification in Mice by Changing Cell Energy Use Hormone Therapy Linked to More Frequent, Longer Migraines After Menopause Meta-Analysis Examines Which Exercise, and How Much, Relates to Sleep in Older Adults 2,711 Autopsies Show Alzheimer's Brain Changes Are Rare Before 70 Does Pregnancy's Biological Aging Fully Reverse? A Hypothesis Awaits Testing Diet, Exercise and Empagliflozin Matched on Blood Sugar in Older Diabetics Review Finds Ultra-Processed Food Subtypes Differ Sharply in Diabetes Risk AI-Designed Peptide Targeting Zombie Cells Heads Toward First Human Testing Blocking Lactate Exporter MCT4 Improved Aged Egg Quality in Lab Study Aquatic and Dance Exercise Accompanied Lower Inflammation in Obese Older Women Aging Brain Cells Keep Their Identity, Challenging Epigenetic Drift Theory