Food consumption by degree of food processing and risk of type 2 diabetes mellitus: a prospective cohort analysis of the European Prospective Investigation into Cancer and Nutrition (EPIC)

Abstract

Background: It is unknown whether the association between ultra-processed food (UPF) intake and type 2 diabetes mellitus differs from other degrees of food processing. We examined the association between degree of food processing and incident type 2 diabetes mellitus.

Methods: This was a prospective cohort analysis of the European Prospective Investigation into Cancer and Nutrition (EPIC). Dietary intake was assessed at baseline using dietary questionnaires and classified according to the Nova classification into unprocessed/minimally processed food (MPF), processed culinary ingredients (PCI), processed food (PF) and UPF. Type 2 diabetes mellitus cases were verified through multiple methods. Cox regression and statistical substitution analysis was used to estimate associations between MPF + PCI, PF and UPF intake and incident type 2 diabetes mellitus. To investigate heterogeneity in the association between UPF and incident type 2 diabetes mellitus, UPF sub-group analysis was conducted. Different reference groups were used in each analysis.

Findings: Over an average 10.9 years follow-up of 311,892 individuals, 14,236 type 2 diabetes mellitus cases were identified. Each 10% increment of total daily food intake from UPF (%g/day) was associated with 17% (95% confidence interval (95%CI): 1.14-1.19) higher incident type 2 diabetes mellitus. Each 10% increment in MPF + PCI or PF intake was associated with lower incident type 2 diabetes mellitus (MPF + PCI hazard ratio: 0.94 (95%CI: 0.92-0.96); PF hazard ratio: 0.92 (95%CI: 0.89-0.95)). Replacing UPF with MPF + PCI or PF was associated with lower incident type 2 diabetes mellitus. However, heterogeneity was observed across UPF sub-groups, with breads, biscuits and breakfast cereals, sweets and desserts, and plant-based alternatives associated with lower incident type 2 diabetes mellitus.

Interpretation: These findings support recommendations to focus on reducing intake of specific UPF for lowering type 2 diabetes mellitus risk.

Funding: International Agency for Research on Cancer.

Keywords: Diet; Europe; Food processing; Nova classification; Type 2 diabetes mellitus; Ultra-processed food.

Fig. 1

Hazard ratios and 95% confidence intervals of the association between Nova group intake and incident type 2 diabetes mellitus. Associations between Nova group intake and incident type 2 diabetes mellitus in EPIC. Error bars indicate 95% confidence intervals. Nova group intakes expressed as a percentage of daily dietary intake (%g/day). Hazard ratios expressed per 10%g/day increase in Nova group intake, in three separate models. Cox proportional hazard models, with age as the underlying time variable. Time at entry was age at recruitment, and exit time was age at type 2 diabetes mellitus diagnosis, end of follow-up, loss to follow-up, or death, whichever came first. Models were adjusted for sex, occupation, study centre, education level, smoking status and intensity, physical activity level, alcohol intake, family history of diabetes and total energy intake. Abbreviations: MPF: unprocessed/minimally processed food; PCI: processed culinary ingredients; PF: processed food; UPF: ultra-processed food.

Fig. 2

Hazard ratios and 95% confidence intervals of the association between UPF sub-group intake and incident type 2 diabetes mellitus. Associations between UPF sub-group intake and incident type 2 diabetes mellitus in EPIC. Sub-group intake expressed as a percentage of daily dietary intake (%g/day). Hazard ratios expressed per 10%g/day increase in sub-group intake. Cox proportional hazard models, with age as the underlying time variable. Time at entry was age at recruitment, and exit time was age at type 2 diabetes mellitus diagnosis, end of follow-up, loss to follow-up, or death, whichever came first. Model 2 was adjusted for sex, occupation, study centre, education level, smoking status and intensity, physical activity level, alcohol intake, family history of diabetes and total energy intake. Abbreviations: CI: confidence interval; HR: hazard ratio; SD: standard deviation; UPF: ultra-processed food.