SOURCE: www.nature.com
Abstract
Experimental studies suggested potential adverse effects of preservative food additives, but epidemiological data are lacking. We aim to investigate associations between exposure to these compounds and type 2 diabetes incidence in the NutriNet-Santé prospective cohort (n = 108,723; 79.2%women; mean age=42.5 (SD = 14.6); France, 2009-2023). Dietary intakes are assessed using repeated 24h-dietary records. Exposure is evaluated through multiple composition databases and ad-hoc laboratory assays in food matrices. Associations between cumulative exposures to preservatives and diabetes incidence are characterised using multi-adjusted Cox models. The sum of total preservatives encompasses 58 substances. Among those, 17 are consumed by at least 10% of the study population and thus individually investigated. Thirteen (12 after multiple test correction) widely used individual preservatives are associated with higher diabetes incidence (n=1131cases): potassium sorbate, potassium metabisulfite, sodium nitrite, acetic, citric and phosphoric acids, sodium acetates, calcium propionate, sodium ascorbate, alpha-tocopherol, sodium erythorbate, and rosemary extracts. These findings call for their safety re-evaluation and support recommendations to favour fresh and minimally processed foods without superfluous additives. Trial registration: The NutriNet-Santé cohort is registered at clinicaltrials.gov (NCT03335644).
Introduction
Food additives encompass a wide range of substances which can be divided into different categories based on their functional properties, e.g., colours, preservatives, sweeteners, emulsifiers, etc. Our group already investigated artificial sweeteners, emulsifiers, nitrites and nitrates, and most recently, food additive mixtures in association with type 2 diabetes incidence in the NutriNet-Santé cohort. The present study focuses on preservative food additives, which are massively used by the food industry globally. Among the three and a half million foods and beverages listed in the Open Food Facts World database in 2024, more than 700,000 contain at least one of these additives. Preservatives (European codes usually within the E200s and E300s ranges) are added to prolong the shelf-life of foods. They protect them against deterioration caused by micro-organisms and/or growth of pathogenic micro-organisms, and by oxidation, such as fat rancidity and colour changes.
From 2004 to 2024, about a sixth (n = 25) of food additive re-evaluations conducted by the European Food Safety Authority (EFSA) concerned preservatives. Sixteen of them led to the derivation of Acceptable Daily Intake (ADI) reference values for these additives or their respective groups. ADIs were based on a range of toxicological endpoints, including behavioural, carcinogenic, developmental, haematological, reproductive, and thyroid toxicity, as well as growth retardation, increased blood methaemoglobin level, and increased mortality, all relying on experimental data. Moreover, in vivo and in vitro studies highlighted potential metabolic-related generally adverse effects of preservatives on pancreatic tissue, insulin disruption, the tricarboxylic acid (TCA) cycle, inflammation, and advanced glycation end products (AGEs) activation, questioning their potential impacts on metabolic disorders such as type 2 diabetes, with direct metabolic aetiology. EFSA suggested that metabolisation of these additives implies conversion into substances, such as acetate into acetyl-CoA, that play a role in the body’s metabolisM. For some of these additives, corresponding substances are naturally present in foods and beverages (e.g., certain antioxidant vitamins such as vitamin C – ascorbic acid). In this case, epidemiological studies investigated the link between their dietary intake, supplementation or nutritional status and the risk of type 2 diabetes (suggesting non-to-low protective evidence), but to our knowledge, none of these have considered food additives sources, while the impact of the substance may depend on the matrix in which it is integrated. Plus, no consumption data is available for other preservative food additives due to the important variability in additive composition between commercial products and the lack of brand-specific data in previous cohorts. Thus, this study aimed to comprehensively quantify the time-dependent exposure to preservative food additives available on the French/European market and examine any potential associations with the incidence of type 2 diabetes in a large prospective cohort.
Results
Descriptive characteristics
The population study included 108,723 participants (Fig. 1), 79.2% women. On average, they completed 21 24HDRs (SD 18); median = 17; 25th–75th percentiles = 6–32; maximum = 84. Participants had a mean age at baseline of 42.5 years (SD 14.6) (range = 15.2–99.0). Compared to lower consumers (tertile 1), higher consumers of food additive preservatives (tertile 3) tended to be younger, slightly less exposed to a family history of diabetes, with a lower prevalence of metabolic diseases (cardiovascular disease, dyslipidaemia, and hypertension), a lower physical activity level, they were more likely to be current smokers, consumed less alcohol and had higher energy intakes (Table 1—descriptive unadjusted analysis).
Intakes of preservative food additives are displayed in Tables 2 and 3. A total of 99.7% of participants exhibited a non-null intake of food additive preservatives in the first 2 years of follow-up. Out of the 58 preservative food additives detected and quantified in our databases, 17 were consumed by at least 10% of the participants and thus were individually investigated in relation with type 2 diabetes incidence (Tables 2 and 3), to allow sufficient statistical power in Cox models (but all 58 additives were accounted for in sums of preservative categories). In terms of proportion of consumers, the main preservative food additives were: citric acid (E330) (91.8% consumers), lecithins (E322) (87.1%), total sulfites (83.6%), ascorbic acid (E300) (83.5%), sodium nitrite (E250) (73.7%), potassium sorbate (E202) (65.5%), sodium erythorbate (E316) (52.7%), sodium ascorbate (E301) (50.2%), potassium metabisulfite (E224) (44.4%), and potassium nitrate (E252) (32.6%). No strong correlation between preservative food additives was identified (Supplementary Fig. 1).
Manufacturers use preservatives ubiquitously across a wide range of food groups (Fig. 2 and Supplementary Table 1). Some are nonetheless more specific to given food groups: 22.2% of sorbates were found in fruit- and vegetable-based foods; 85.5% of sulfites in alcoholic drinks; 53.7% of nitrites, 80.4% of nitrates, and 42.2% of erythorbates in processed meat; 43.6% and 30.5% of propionates in refined and whole grains and cereals, respectively; 50.6% of ascorbates and 25.7% of citrates in fruit- and vegetable-based foods; and 30.3% of tocopherols in breakfast cereals. In all, 34.6% of food additive preservatives were consumed through ultra-processed foods in this population study.
For food additive substances that also naturally occurred in the diet, the relative contribution of the food additive source varied depending on the compound: from 1% for tocopherols or 5% for acetates on average, to 17% for citric acid, 29% for ascorbates, and 63% for sulfites.
No participant exceeded the ADIs set by EFSA7 for sorbates, erythorbates or nitrates. However, 90 participants exceeded the ADI set for sulfites with a mean intake of 0.88 mg of sulfur dioxide/kg body weight per day (SD 0.33, median = 0.80, 25th–75th percentiles [0.75–0.92]) and 55 exceeded the ADI for nitrites with a mean consumption of 0.10 mg nitrite ion/kg body weight per day (0.03, 0.08, [0.08–0.10]).
Associations between preservative additives and incidence of type 2 diabetes
Participants’ median follow-up time was 8.05 years (841,296 person-years). Between 2009 and 2023, 1131 incident type 2 diabetes cases were identified. Schoenfeld residuals did not refute the proportional hazard assumption (Supplementary Fig. 2).
Restricted cubic spline plots (Supplementary Fig. 3) did not indicate departure from linearity for sodium acetates (E262), total tocopherols, tocopherol-rich extract (E306), alpha-tocopherol (E307), phosphoric acid (E338), and extracts of rosemary (E392) (p values for non-linearity ≥0.05), in this case p values for trend are provided in the following paragraph. For others, restricted cubic splines suggested a plateau effect (p values for non-linearity <0.05); therefore, the likelihood ratio overall p values (requiring no underlying hypothesis of linearity) are displayed thereafter. The retained p values are provided for all tested additives in Supplementary Table 2 for the main model and in Supplementary Table 3 for sensitivity analyses.
The results of Cox models are presented in Fig. 3 and Supplementary Table 2. Higher intakes of the following additives were associated with a higher incidence of type 2 diabetes: total preservatives (HRcategory 3 vs. 1 = 1.47, 95% CI [1.26–1.73], p < 0.001), total preservatives (non-antioxidant) (1.49[1.27–1.74], <0.001), total sorbates (1.85[1.57–2.18], <0.001), potassium sorbate (E202) (2.15[1.76–2.63], <0.001), potassium metabisulfite (E224) (1.29[1.10–1.51], <0.001), total nitrites (1.50[1.26–1.78], <0.001), sodium nitrite (E250) (1.50[1.26–1.78], <0.001), total acetates (1.36[1.16–1.59], <0.001), acetic acid (E260) (1.31[1.08–1.60], <0.001), sodium acetates (E262) (1.23[1.01–1.50], 0.003), total propionates (1.43[1.21–1.68], <0.001), calcium propionate (E282) (1.44[1.22–1.70], <0.001), total preservatives (antioxidant) (1.40[1.19–1.65], <0.001), total ascorbates (1.17[1.00–1.38], 0.01), sodium ascorbate (E301) (1.41[1.19–1.66], <0.001), total tocopherols (1.27[1.07–1.50], <0.001), alpha-tocopherol (E307) (1.30[1.05–1.60], 0.001), total erythorbates (1.42[1.20–1.69], <0.001), sodium erythorbate (E316) (1.43[1.20–1.69], <0.001), citric acid (E330) (1.29[1.11–1.51], 0.003), phosphoric acid (E338) (1.39[1.10–1.75], <0.001), and extracts of rosemary (E392) (1.20[1.02–1.41], 0.01). Overall, these results were stable across all sensitivity analyses testing mutual adjustment for other preservative food additives, additional adjustment for the proportion of ultra-processed foods in the diet, diagnosis and/or treatment for at least one prevalent non-diabetes metabolic disorder, time-dependent intakes of total emulsifier and total artificial sweetener food additives, vitamins C and E supplements, trans fatty acids, dietary patterns rather than individual food groups, polyunsaturated fatty acids and heme iron; models with start of follow-up after the first 2-year period; models not excluding energy under-reporters; marginal structural model; and model using splines for covariates (Supplementary Table 3). An exception was for total ascorbates, which became non-significant with p = 0.4 in the marginal structural model, maybe due to loss of statistical power (however, the specific ascorbate E301 remained stable with this model, p < 0.001). As expected, no association was detected with the hip fracture negative control model (613 incident cases, all p values > 0.05).
We also tested associations between intakes of acetic acid, citric acid, nitrates, nitrites, sulfites, vitamin C, and vitamin E from naturally occurring sources and type 2 diabetes incidence. Naturally occurring acetic acid and nitrates from drinking water were associated with increased type 2 diabetes incidence, while natural sulfites were associated with lower incidence (Supplementary Table 4).
In this population study, ultra-processed food exposure (% weight in the diet) was associated with higher type 2 diabetes incidence (HR = 1.20, 95% CI [1.11–1.29], p value < 0.001). We computed a mediation analysis to assess the proportion of this association mediated by preservative food additives that were related to type 2 diabetes in this study (i.e., potassium sorbate (E202), potassium metabisulfite (E224), sodium nitrite (E250), acetic acid (E260), sodium acetates (E262), calcium propionate (E282), sodium ascorbate (E301), tocopherol-rich extract (E306), alpha-tocopherol (E307), sodium erythorbate (E316), citric acid (E330), phosphoric acid (E338), and extracts of rosemary (E392)). In all, 17% of the association between ultra-processed food and type 2 diabetes was mediated by exposure to these preservatives (p value of the mediated proportion = 0.01).
