Sweet Drinks Drive Most of the Ultra-Processed Food Link to Diabetes in UK Biobank Study

A large prospective analysis of UK Biobank participants has found that the well-documented association between ultra-processed food and type 2 diabetes is not spread evenly across the diet. Instead, it appears to be driven primarily by sweet-tasting beverages, with savoury snacks making a smaller contribution and considerably weaker evidence for the remaining food groups typically bundled into ultra-processed food indices.

The study, published in BMC Medicine, included 188,405 UK Biobank participants who had completed Oxford WebQ dietary assessments. Rather than relying on a single broad measure of ultra-processed food, the researchers built a study-specific UPF-like food-group index composed of five questionnaire-observable categories: processed meat, sweet snacks, savoury snacks, sweet-tasting beverages and ready or fried foods. They also examined sweet-tasting beverage subtypes separately and constructed dietary difference indices, including a minimally processed plant-food minus UPF-like score.

Participants were followed for a median of 14.4 years for incident type 2 diabetes and 14.3 years for cardiometabolic multimorbidity, the co-occurrence of multiple cardiometabolic conditions. The researchers used Cox proportional hazards models to estimate hazard ratios per 1-standard-deviation higher score, applying sequential adjustments for demographic factors, conventional nutrients, body mass index and a nuclear magnetic resonance biomarker score.

In the base model, the total UPF-like index was associated with type 2 diabetes with a hazard ratio of 1.20 (95% confidence interval 1.17–1.22) and with cardiometabolic multimorbidity with a hazard ratio of 1.13 (95% CI 1.11–1.15). These estimates describe associations observed in observational data and should not be read as causal effects.

The associations attenuated as the models became more fully adjusted. After adjustment for conventional nutrients, the hazard ratios were 1.18 (95% CI 1.15–1.20) for type 2 diabetes and 1.11 (95% CI 1.09–1.13) for multimorbidity. Adding body mass index to the models lowered the estimates further to 1.09 (95% CI 1.06–1.11) and 1.06 (95% CI 1.04–1.08) respectively. With additional adjustment for the NMR biomarker score, the corresponding hazard ratios were 1.07 (95% CI 1.05–1.10) and 1.05 (95% CI 1.03–1.07). This stepwise attenuation suggests that part of the raw association travels through nutrient profiles and body mass, although a residual association remained after all adjustments.

The component-level analysis was the most revealing part of the study. In models adjusted for nutrients and body mass index, the sweet-tasting beverage subtype composite was associated with type 2 diabetes with a hazard ratio of 1.12 (95% CI 1.10–1.15) and with cardiometabolic multimorbidity with a hazard ratio of 1.08 (95% CI 1.07–1.10). Savoury snacks showed smaller positive associations with both outcomes. By contrast, the remaining food groups in the index contributed much weaker evidence, indicating that a single aggregate ultra-processed food score can mask substantial heterogeneity among the products it combines.

The difference index told a consistent story from the opposite direction. The minimally processed plant-food minus UPF-like difference index was inversely associated with both outcomes, with a hazard ratio of 0.92 (95% CI 0.90–0.94) for type 2 diabetes and 0.95 (95% CI 0.93–0.96) for cardiometabolic multimorbidity in the nutrient- and body-mass-index-adjusted models. Higher scores on this index reflect a dietary pattern weighted toward minimally processed plant foods relative to the UPF-like groups.

The authors conclude that although the overall UPF-like dietary index was associated with incident type 2 diabetes and cardiometabolic multimorbidity, the observed associations appeared to be driven primarily by sweet-tasting beverages and, to a lesser extent, savoury snacks. They suggest that reporting component food groups alongside ultra-processed food-related indices may help distinguish food-specific associations from broader dietary patterns when studying cardiometabolic risk.

As an observational cohort study, the analysis identifies associations rather than proven causal effects, and residual confounding cannot be excluded. The findings apply to UK Biobank participants with Oxford WebQ dietary data and rely on self-reported dietary information. Nevertheless, the study offers a methodological lesson for nutrition research: broad processing-based classifications may obscure which specific foods carry the cardiometabolic signal, and disaggregating them could sharpen both future research and the interpretation of existing evidence.

Subject of Research: Associations between ultra-processed food indices, component food groups and incident type 2 diabetes and cardiometabolic multimorbidity in the UK Biobank

Article Title: Ultra-processed food indices and component food groups in relation to incident type 2 diabetes and cardiometabolic multimorbidity: a prospective cohort study

Article References: Wei, L., Han, Y., Ji, Y., & Fang, Y. (2026). Ultra-processed food indices and component food groups in relation to incident type 2 diabetes and cardiometabolic multimorbidity: a prospective cohort study. BMC Medicine. https://doi.org/10.1186/s12916-026-05353-2

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05353-2

Keywords: ultra-processed food, type 2 diabetes, cardiometabolic multimorbidity, UK Biobank, sweet-tasting beverages, savoury snacks, prospective cohort study, dietary indices, BMC Medicine, nutrition epidemiology, hazard ratios, body mass index