Food additive mixtures and type 2 diabetes incidence: Results from the NutriNet-Santé prospective cohort

Mixtures of food additives are daily consumed worldwide by billions of people. So far, safety assessments have been performed substance by substance due to lack of data on the effect of multiexposure to combinations of additives. Our objective was to identify most common food additive mixtures, and investigate their associations with type 2 diabetes incidence in a large prospective cohort.

Participants (n = 108,643, mean follow-up =  7.7 years (standard deviation (SD) =  4.6), age =  42.5 years (SD =  14.6), 79.2% women) were adults from the French NutriNet-Santé cohort (2009-2023). Dietary intakes were assessed using repeated 24h-dietary records, including industrial food brands. Exposure to food additives was evaluated through multiple food composition databases and laboratory assays. Mixtures were identified through nonnegative matrix factorization (NMF), and associations with type 2 diabetes incidence were assessed using Cox models adjusted for potential socio-demographic, anthropometric, lifestyle and dietary confounders. A total of 1,131 participants were diagnosed with type 2 diabetes. Two out of the five identified food additive mixtures were associated with higher type 2 diabetes incidence: the first mixture included modified starches, pectin, guar gum, carrageenan, polyphosphates, potassium sorbates, curcumin, and xanthan gum (hazard ratio (HR)per an increment of 1SD of the NMF mixture score = 1.08 [1.02, 1.15], p = 0.006), and the other mixture included citric acid, sodium citrates, phosphoric acid, sulphite ammonia caramel, acesulfame-K, aspartame, sucralose, arabic gum, malic acid, carnauba wax, paprika extract, anthocyanins, guar gum, and pectin (HR = 1.13 [1.08,1.18], p < 0.001). No association was detected for the three remaining mixtures: HR =  0.98 [0.91, 1.06], p = 0.67; HR =  1.02 [0.94, 1.10], p = 0.68; and HR =  0.99 [0.92, 1.07], p = 0.78. Several synergistic and antagonist interactions between food additives were detected in exploratory analyses. Residual confounding as well as exposure or outcome misclassifications cannot be entirely ruled out and causality cannot be established based on this single observational study.

This study revealed positive associations between exposure to two widely consumed food additive mixtures and higher type 2 diabetes incidence. Further experimental research is needed to depict underlying mechanisms, including potential synergistic/antagonist effects. These findings suggest that a combination of food additives may be of interest to consider in safety assessments, and they support public health recommendations to limit nonessential additives.

The NutriNet-Santé cohort is registered at clinicaltrials.gov (NCT03335644). https://clinicaltrials.gov/study/NCT03335644.