Abstract
OBJECTIVE
To investigate potential association between exposure to food coloring additives and type 2 diabetes incidence.
RESEARCH DESIGN AND METHODS
The study followed 108,723 participants (79.2% female, mean age 42.5 [SD 14.6] years) from the French NutriNet-Santé cohort (2009–2023). Dietary data were assessed using repeated 24-h dietary records, including industrial food brands. Cumulative time-dependent exposure to food additives was evaluated through multiple composition databases and ad hoc laboratory assays in food matrices. Associations between exposures to food coloring additives (sex-specific tertiles if proportion of exposed participants was more than two-thirds, or nonexposed/lower/higher exposed based on sex-specific median otherwise) and type 2 diabetes incidence were assessed using multivariable Cox proportional hazards models.
RESULTS
There were 1,131 incident type 2 diabetes cases diagnosed (median follow-up, 8.05 years). After false discovery rate correction, intakes of the following colors were associated with higher type 2 diabetes incidence: total food coloring additives (hazard ratio [HR]higher vs. non/lower consumers 1.38 [95% CI 1.17–1.63], P = 0.0002), total caramel (1.43 [1.21–1.67], P = 0.0002), plain caramel (1.46 [1.26–1.70], P = 0.0002), sulfite ammonia caramel (1.30 [1.07–1.59], P = 0.007), total carotene (1.27 [1.08–1.48], P = 0.007), carotenoids (1.39 [1.19–1.62], P = 0.0002), β-carotene (1.44 [1.23–1.68], P = 0.0002), paprika-capsanthin-capsorubin (1.26 [1.08–1.46], P = 0.004), lutein (1.20 [1.02–1.40], P = 0.0002), curcumin (1.49 [1.29–1.73], P = 0.0002), cochineal-carminic acid-carmines (1.27 [1.10–1.48], P = 0.003), and anthocyanins (1.40 [1.17–1.68], P = 0.0002).
CONCLUSIONS
Several positive associations were observed between exposure to natural and synthetic food coloring additives and type 2 diabetes incidence. Further studies are needed to gain insights into underlying mechanisms, and if confirmed, call for reevaluation of food coloring additives to protect consumer health.
Graphical Abstract
Introduction
Type 2 diabetes is a leading global public health challenge accounting for >95% of diabetes cases. Lifestyle-related factors contribute to its growing prevalence (1), in particular, ultraprocessed foods (UPFs), comprising >60% of daily energy intake in some countries, have been associated with increased type 2 diabetes incidence independent of energy intake and adiposity (2,3).
Food coloring additives, key markers of UPFs, are introduced to enhance or restore the color of food following exposure to light, temperature changes, and moisture, and to increase consumer appeal. Certain coloring additives showed adverse health effects in preclinical studies and short-term human trials, particularly azoic dyes (allura red, tartrazine, and sunset yellow) (4,5). These colors, along with carmoisine, ponceau 4R, and quinoline yellow, the “Southampton Six” colors, have been implicated in childhood behavioral effects (6). The European Food Safety Authority (EFSA) has lowered permitted levels for quinolone yellow, sunset yellow, and ponceau 4R (5), while titanium dioxide has been banned for use in food in the European Union. Furthermore, by the end of 2026, a number of synthetic coloring additives will be phased out from the U.S. market. In Europe, although some coloring additives, such as caramel colors are widespread, most synthetic additives are used sparingly, while natural colors are more frequently used. Public perception, accentuated by marketing that often uses this argument, tends to associate substances of natural origin with lower health risks, but the natural nature of a substance does not inherently guarantee its safety (7), as evidenced by numerous naturally occurring toxic substances. However, scientific data on the health impacts of natural food colors are lacking.
To date, evidence on metabolic effects of coloring additives is limited. Research suggests that coloring additives may alter key molecular pathways involved in insulin signaling and inflammation. For instance, in vivo studies have linked certain synthetic colors to elevated blood glucose levels, hepatic and gastrointestinal tissue damage, and potential alterations in gut microbiota composition (8–12). However, despite the rapid and manifold increase in use of coloring additives (>139,000 food products listed in Open Food Facts), no epidemiological study has quantified dietary exposures to these additives or investigated their association with diabetes incidence. This gap may be explained by limited detail on specific industrial food products consumed in previous cohort studies lacking brand-specific data and by variability in additive content between industrial food products and between countries. In this regard, the NutriNet-Santé cohort provides a wealth of information with extensive brand-specific dietary data collected through validated repeated 24-h dietary records (24HDRs) since 2009. Therefore, we aimed to investigate, for the first time, associations between real-life chronic exposure to food coloring additives, either natural or artificial, and incidence of type 2 diabetes.
Research Design and Methods
Study Population
The NutriNet-Santé e-cohort is a prospective French study initiated in 2009 to investigate the association between nutrition and health (13). Participants aged ≥15 are recruited through extensive multimedia campaigns and invited to enroll in the study via a dedicated web-based platform (https://etudenutrinet-sante.fr/). Upon enrollment, participants are invited to provide detailed information by completing self-administered questionnaires covering sociodemographic and lifestyle information (14), health status, dietary intake (15–17), anthropometric data (18,19), and physical activity level (20). Participants are then invited to update these data regularly via follow-up questionnaires.
Eligible participants (n = 134,076) were those who had completed two or more 24HDRs during the first 2 years in the cohort. We excluded energy overreporters (n = 54) and underreporters (n = 23,098) (21,22), and all prevalent diabetes type 1 (n = 299) and type 2 cases (n = 1,902); therefore, the final study population included 108,723 participants (Supplementary Fig. 1).
Ethics Approval
The study is registered at https://clinicaltrials.gov/ct2/show/NCT03335644, conducted according to the Declaration of Helsinki guidelines, and approved by the French Institute for Health and Medical Research Institutional Review Board (IRB-INSERM) and the “Commission Nationale de l’Informatique et des Libertés” (CNIL no. 908450/no. 909216). Each participant provides an electronic informed consent form in the NutriNet-Santé cohort before enrollment.
Dietary Assessment
Upon registration and every 6 months thereafter, participants were invited to complete sequences of validated web-based 24HDRs (15–17). At each period, 24HDRs were randomly assigned to 3 nonconsecutive days over 2 weeks (2 weekdays and 1 weekend day). Dietary intakes in energy, fiber, and macro- and micronutrients were assessed by merging with the NutriNet-Santé food composition table. Details on dietary data collection and identification of under-/overenergy reports can be found in Supplementary Method 1.
Assessment of Food Coloring Additive Exposures
Assessment of food additive intakes in the NutriNet-Santé cohort through brand-specific data of the 24HDRs has been previously described (23). Details are provided in Supplementary Method 2. Each participant’s date of consumption of each food or beverage was used to match the product to the closest composition data available, thus accounting for potential reformulations across time. Each food item consumed and reported in a specific dietary record was matched against three databases to assess presence of any food additive: Observatoire de la qualité de l’alimentation (OQALI), Open Food Facts, and Mintel Global New Products Database (GNPD). Next, the dose of food additive ingested with each food item was estimated based on 1) ad hoc laboratory assays quantifying additives in specific food items, 2) doses in generic food categories provided by EFSA, or 3) generic doses from Codex General Standard for Food Additives (GSFA). All coloring additives, defined as such by EFSA, GSFA or U.K. Food Standard Agency, with nonnull consumption, were accounted for in the calculation of total coloring additives and their classes Table 2.
Table 2.
Daily food coloring additive exposures among study participants from the NutriNet-Santé cohort, 2009–2023 (N = 108,723)a,b
| All participants, mg/day | Consumers, mg/day | Consumers, mg/day/kg of body weight | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Additive | Mean | SD | P25 | P50 | P75 | Mean | SD | P25 | P50 | P75 | Proportion of consumers (%) | Mean | SD | P25 | P50 | P75 |
| Curcumin (E100) | 0.61 | 2.28 | 0.00 | 0.00 | 0.00 | 2.43 | 4.05 | 0.03 | 0.69 | 3.35 | 25.05 | 0.04 | 0.06 | 0.00 | 0.01 | 0.05 |
| Riboflavin (E101) | 0.35 | 2.61 | 0.00 | 0.00 | 0.00 | 5.37 | 8.85 | 0.12 | 3.21 | 6.70 | 6.45 | 0.08 | 0.14 | 0.00 | 0.05 | 0.10 |
| Tartrazine (E102) | 0.03 | 0.35 | 0.00 | 0.00 | 0.00 | 0.90 | 1.69 | 0.10 | 0.28 | 1.05 | 3.35 | 0.01 | 0.03 | 0.00 | 0.00 | 0.02 |
| Quinoline yellow (E104) | 0.00 | 0.03 | 0.00 | 0.00 | 0.00 | 0.32 | 0.41 | 0.10 | 0.20 | 0.35 | 0.26 | 0.00 | 0.01 | 0.00 | 0.00 | 0.01 |
| Sunset yellow FCF (E110) | 0.00 | 0.06 | 0.00 | 0.00 | 0.00 | 0.13 | 0.46 | 0.02 | 0.04 | 0.09 | 1.57 | 0.00 | 0.01 | 0.00 | 0.00 | 0.00 |
| Cochineal- carminic acid- carmines (E120) | 0.13 | 0.65 | 0.00 | 0.00 | 0.02 | 0.49 | 1.17 | 0.08 | 0.21 | 0.49 | 27.19 | 0.01 | 0.02 | 0.00 | 0.00 | 0.01 |
| Carmoisine (E122) | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.42 | 0.44 | 0.06 | 0.30 | 0.75 | 0.01 | 0.01 | 0.01 | 0.00 | 0.00 | 0.01 |
| Amaranth (E123) | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.18 | 0.15 | 0.13 | 0.18 | 0.23 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| Ponceau 4R (E124) | 0.00 | 0.01 | 0.00 | 0.00 | 0.00 | 0.15 | 0.13 | 0.07 | 0.11 | 0.22 | 0.36 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| Erythrosine (E127) | 0.02 | 0.44 | 0.00 | 0.00 | 0.00 | 3.43 | 4.24 | 1.04 | 2.34 | 4.52 | 0.65 | 0.05 | 0.07 | 0.02 | 0.03 | 0.07 |
| Allura red AC (E129) | 0.00 | 0.02 | 0.00 | 0.00 | 0.00 | 1.00 | 0.86 | 0.23 | 0.96 | 1.42 | 0.02 | 0.02 | 0.02 | 0.00 | 0.01 | 0.02 |
| Patent blue V (E131) | 0.01 | 0.29 | 0.00 | 0.00 | 0.00 | 0.34 | 1.57 | 0.01 | 0.04 | 0.14 | 3.39 | 0.01 | 0.03 | 0.00 | 0.00 | 0.00 |
| Indigotine (E132) | 0.00 | 0.11 | 0.00 | 0.00 | 0.00 | 0.74 | 2.52 | 0.11 | 0.26 | 0.71 | 0.18 | 0.01 | 0.05 | 0.00 | 0.00 | 0.01 |
| Brilliant blue FCF (E133) | 0.01 | 0.22 | 0.00 | 0.00 | 0.00 | 0.19 | 0.91 | 0.01 | 0.06 | 0.15 | 5.61 | 0.00 | 0.02 | 0.00 | 0.00 | 0.00 |
| Copper complexes of chlorophylls (E141) | 0.13 | 1.55 | 0.00 | 0.00 | 0.00 | 2.48 | 6.32 | 0.14 | 0.34 | 1.43 | 5.27 | 0.04 | 0.10 | 0.00 | 0.01 | 0.02 |
| Caramel colors without specification (E150) | 2.37 | 29.24 | 0.00 | 0.00 | 0.00 | 118.67 | 170.48 | 32.44 | 69.16 | 144.68 | 1.99 | 1.84 | 2.60 | 0.49 | 1.07 | 2.27 |
| Plain caramel (E150a) | 5.20 | 22.37 | 0.00 | 0.00 | 0.00 | 22.39 | 42.06 | 1.53 | 6.58 | 26.79 | 23.24 | 0.35 | 0.66 | 0.02 | 0.10 | 0.41 |
| Caustic sulfite caramel (E150b) | 0.09 | 1.82 | 0.00 | 0.00 | 0.00 | 8.00 | 15.60 | 1.30 | 2.44 | 6.51 | 1.09 | 0.12 | 0.24 | 0.02 | 0.04 | 0.10 |
| Ammonia caramel (E150c) | 5.27 | 65.03 | 0.00 | 0.00 | 0.00 | 99.97 | 266.06 | 12.06 | 26.79 | 63.09 | 5.27 | 1.57 | 4.38 | 0.19 | 0.41 | 0.99 |
| Sulfite ammonia caramel (E150d) | 145.58 | 477.11 | 0.00 | 0.00 | 7.14 | 556.37 | 800.89 | 89.23 | 291.70 | 683.57 | 26.17 | 8.49 | 12.32 | 1.37 | 4.49 | 10.35 |
| Brilliant Black BN (E151) | 0.00 | 0.01 | 0.00 | 0.00 | 0.00 | 0.47 | 0.39 | 0.17 | 0.27 | 0.59 | 0.01 | 0.01 | 0.01 | 0.00 | 0.01 | 0.01 |
| Brown HT (E155) | 0.00 | 0.01 | 0.00 | 0.00 | 0.00 | 1.48 | 0.90 | 0.95 | 1.23 | 1.87 | 0.01 | 0.03 | 0.02 | 0.01 | 0.02 | 0.04 |
| Carotenoids without specification (E160) | 0.21 | 0.86 | 0.00 | 0.00 | 0.02 | 0.57 | 1.35 | 0.02 | 0.05 | 0.62 | 37.00 | 0.01 | 0.02 | 0.00 | 0.00 | 0.01 |
| β-Carotene (E160a) | 1.38 | 8.55 | 0.00 | 0.00 | 0.03 | 3.24 | 12.85 | 0.02 | 0.05 | 0.21 | 42.67 | 0.05 | 0.20 | 0.00 | 0.00 | 0.00 |
| Annatto- bixin- norbixin (E160b) | 0.03 | 0.12 | 0.00 | 0.00 | 0.00 | 0.19 | 0.24 | 0.04 | 0.11 | 0.25 | 17.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| Paprika extract- capsanthin- capsorubin (E160c) | 0.21 | 0.49 | 0.00 | 0.00 | 0.20 | 0.44 | 0.64 | 0.08 | 0.21 | 0.54 | 47.94 | 0.01 | 0.01 | 0.00 | 0.00 | 0.01 |
| Lycopene (E160d) | 0.00 | 0.01 | 0.00 | 0.00 | 0.00 | 0.44 | 0.32 | 0.24 | 0.37 | 0.54 | 0.01 | 0.01 | 0.01 | 0.00 | 0.01 | 0.01 |
| β-apo-8′-carotenal (E160e) | 0.05 | 0.71 | 0.00 | 0.00 | 0.00 | 2.35 | 4.38 | 0.31 | 0.89 | 2.51 | 2.05 | 0.04 | 0.07 | 0.01 | 0.02 | 0.04 |
| Lutein (E161b) | 0.35 | 1.88 | 0.00 | 0.00 | 0.00 | 1.21 | 3.36 | 0.00 | 0.00 | 0.60 | 28.50 | 0.02 | 0.05 | 0.00 | 0.00 | 0.01 |
| Beetroot red (E162) | 0.47 | 2.59 | 0.00 | 0.00 | 0.00 | 5.21 | 7.07 | 1.19 | 2.86 | 6.43 | 9.02 | 0.08 | 0.11 | 0.02 | 0.04 | 0.10 |
| Anthocyanins (E163) | 1.35 | 7.54 | 0.00 | 0.00 | 0.00 | 10.10 | 18.37 | 2.57 | 5.36 | 10.37 | 13.37 | 0.16 | 0.30 | 0.04 | 0.08 | 0.16 |
| Calcium carbonate (E170) | 21.29 | 207.84 | 0.00 | 0.00 | 0.00 | 590.24 | 928.52 | 50.00 | 266.67 | 750.00 | 3.61 | 9.47 | 15.02 | 0.85 | 4.25 | 11.97 |
| Titanium dioxide (E171) | 0.54 | 4.48 | 0.00 | 0.00 | 0.00 | 7.69 | 15.22 | 1.53 | 3.43 | 7.68 | 6.99 | 0.12 | 0.25 | 0.02 | 0.05 | 0.12 |
| Iron oxides & hydroxides (E172) | 0.14 | 2.54 | 0.00 | 0.00 | 0.00 | 8.16 | 17.74 | 1.18 | 3.73 | 9.26 | 1.70 | 0.12 | 0.28 | 0.02 | 0.06 | 0.14 |
| Total azoic dyes | 0.03 | 0.37 | 0.00 | 0.00 | 0.00 | 0.86 | 1.67 | 0.09 | 0.25 | 0.92 | 3.93 | 0.01 | 0.03 | 0.00 | 0.00 | 0.01 |
| Total caramel | 158.50 | 484.42 | 0.00 | 0.00 | 56.59 | 349.31 | 671.20 | 12.62 | 85.29 | 399.48 | 45.38 | 5.33 | 10.32 | 0.20 | 1.30 | 6.10 |
| Total carotene | 2.23 | 8.99 | 0.01 | 0.18 | 1.08 | 2.87 | 10.11 | 0.09 | 0.39 | 1.70 | 77.73 | 0.04 | 0.16 | 0.00 | 0.01 | 0.03 |
| Total nanoparticle-based colors | 0.68 | 5.55 | 0.00 | 0.00 | 0.00 | 8.27 | 17.72 | 1.61 | 3.84 | 8.38 | 8.18 | 0.13 | 0.29 | 0.03 | 0.06 | 0.13 |
| Total food coloring additives | 185.81 | 531.31 | 0.20 | 6.71 | 90.06 | 213.47 | 564.27 | 0.95 | 12.93 | 142.26 | 87.04 | 3.28 | 8.74 | 0.01 | 0.20 | 2.17 |
P25, 25th percentile; P50, 50th percentile; P75, 75th percentile.
aClasses of food coloring additives included the following substances: total caramel (E150, E150a, E150b, E150c, and E150d), total carotene (E160, E160a, E160b, E160c, E160d, E160e, E161b), total azoic dyes (E102, E104, E110, E122, E123, E124, E129, and E155), total nanoparticle-based color additives (E171 and E172), and total food coloring additives (E100, E101, E102, E104, E110, E120, E122, E123, E124, E127, E129, E132, E133, E140, E141, E150, E150a, E150b, E150c, E150d, E151, E155, E160, E160a, E160b, E160c, E160d, E160e, E161b, E162, E163, E170, E171, E172, E131, E153, E175).
bSpirulina extract (E134), copper complexes of chlorophylls (E140), fast green FCF(E143), vegetable carbon (E153), canthaxanthin (E161g), gold (E175), tannic acid (E181), jagua (E183), and calcium sulfate (E516) are allowed in the European Union but were not consumed by any participants.
Associations with type 2 diabetes incidence were investigated for coloring additives consumed by at least 10% of study population (to guarantee sufficient statistical power and avoid very low number of cases in exposure categories). These included individual coloring additives: curcumin (European [E] code E100), cochineal-carminic acid-carmines (E120), plain caramel (E150a), sulfite ammonia caramel (E150d), carotenoids without specification (E160), β-carotene (E160a), annatto-bixin-norbixin (E160b), paprika-capsanthin-capsorubin (E160c), lutein (E161b), and anthocyanins (E163). Coloring additives categories included total caramel (E150 [caramel colors without specification], E150a, E150b [caustic sulfite caramel], E150c [ammonia caramel], and E150d), total carotenes (E160, E160a, E160b, E160c, E160d [lycopene], E160e [β-apo-8′-carotenal], and E161b), and total food coloring additives (E100, E101 [riboflavin], E102 [tartrazine], E104 [quinoline yellow], E110 [sunset yellow FCF], E120, E122 [carmoisine], E123 [amaranth], E124 [ponceau 4R], E127 [erythrosine], E129 [allura red AC], E132 [indigo carmine], E133 [brilliant blue FCF], E140 [chlorophylls], E141 [copper complexes of chlorophylls], E150, E150a, E150b, E150c, E150d, E151 [brilliant black BN], E155 [brown HT], E160, E160a, E160b, E160c, E160d, E160e, E161b, E162 [beetroot red], E163, E170 [calcium carbonate], E171 [titanium dioxide], E172 [iron oxides and hydroxides], E131 [patent blue V], E153 [vegetable carbon], and E175 [gold]).
Incident Type 2 Diabetes Ascertainment
Type 2 diabetes was evaluated using a multisource approach. During the follow-up period, participants could report health events, medical treatments, and examinations through biannual health questionnaires or at any time via the health interface of their personal account. Additionally, the NutriNet-Santé cohort is connected to the national health insurance database to gather further details on medical treatments and consultations. The cohort is also linked to the French national mortality registry (Centre d'épidémiologie sur les causes médicales de décès [CépiDC]) to track the occurrence and causes of death. Detailed information can be found in Supplementary Method 3.
Statistical Analyses
Baseline participants’ characteristics are described as mean (SD) for quantitative variables and n (%) for qualitative variables overall and per sex-specific tertiles of total food coloring additive exposure. Intakes of coloring additives during first 2 years of follow-up were reported as mean (SD), 25th, 50th (median) and 75th percentiles, among the whole population and in consumers of each specific additive (mg/d and per kg of body wt). A correlation matrix was generated to visualize Spearman correlations between intakes of individual coloring additives. For each studied additive or group of additives, participants were categorized into low, medium, and high consumers, defined as sex-specific tertiles if the additive was consumed by at least two-thirds of participants, or else as nonconsumers, and lower or higher consumers separated by the sex-specific median (cutoffs provided in Supplementary Table 1).
The associations between exposure to coloring additives coded as cumulative time-dependent variables and type 2 diabetes incidence were investigated using multivariable Cox proportional hazard models with age as time scale. Hazard ratios (HR) and 95% CI were computed. Participants contributed person-time to the models from their age at enrollment in the cohort until their age at the date of diabetes diagnosis, death, last contact, or 31 December 2023, whichever occurred first. A counting process structure was used with cumulative time-dependent dietary variables updated every 2 years (food additive exposures and dietary covariates). Exposure during a specific period was computed using a weighted average of the most recent 2-year period and prior periods, thereby using all available dietary data (details in Supplementary Method 4).
Proportional hazard assumption was tested using Schoenfeld residuals. Restricted cubic splines were computed with three knots at 27.5th, 72.5th, and 95th percentiles of food coloring additive distribution to explore potential dose-response relationships. The covariates were determined using a directed acyclic graph informed by prior studies on their potential association with exposure and/or outcome (Supplementary Fig. 2). The first model was adjusted for age (time-scale), sex, number of dietary records (continuous), BMI (continuous, kg/m2), physical activity (categorical International Physical Activity Questionnaire [IPAQ] variable: low, moderate, or high), smoking status (never, former, or current smoker), number of cigarettes smoked in pack-years (continuous), educational level (less than high school, ≤3 years after high school, or >3 years after high school), family history of diabetes (yes or no), previous metabolic diseases (cardiovascular disease, dyslipidemia, or hypertension; yes or no), and mutually adjusted for total food coloring additive intake excluding the one studied.
The second model was additionally adjusted for time-dependent daily intakes of energy without alcohol (continuous, kcal/d), alcohol (continuous, g/d), saturated fats (continuous, g/d), sodium (continuous, mg/d), fiber (continuous, g/d), sugars (continuous, g/d), fruit and vegetables (continuous, g/day), dairy products (continuous, g/day), and processed meat (continuous, g/day). Specific coloring additives were further adjusted for nutrient intakes: total carotene, E160, E160a, E160b, and E161b were adjusted for dietary β-carotene; E160c was adjusted for dietary capsorubin, capsanthin, and β-carotene; and E100 was adjusted for curcumin.
The main model (model 3) was further adjusted for UPFs (continuous, percentage of weight of total food/day in g). P values with and without correction for multiple testing by the false discovery rate (FDR) were computed. Sensitivity analyses with additional adjustments based on model 3 are detailed in Supplementary Method 4. To test whether coloring additives and type 2 diabetes associations differed by sex, BMI, amount of UPFs, sugar-sweetened beverages, smoking, and physical activity, we used interaction terms and likelihood ratio tests. Additional sensitivity and exploratory analyses are detailed in Supplementary Method 4. Missing values for covariates were handled using multiple imputation via additive regression, bootstrapping, and predictive mean matching (n = 20) in the Harrell miscellaneous (Hmisc) R package (Supplementary Method 5).
All tests were two-sided, and P < 0.05 was considered statistically significant. All statistical analyses were conducted in R 4.5.1 software, except for restricted cubic spline, which was conducted in SAS 9.4 software.
Data and Resource Availability
Researchers from public institutions can submit a request to have access to the data for strict reproducibility analysis (systematically accepted) or for a new collaboration, including information on the institution and a brief description of the project to collaboration@etude-nutrinet-sante.fr. All requests will be reviewed by the steering committee of the NutriNet-Santé study. If the collaboration is accepted, a data access agreement will be necessary and appropriate authorizations from the competent administrative authorities may be needed. In accordance with existing regulations, no personal data will be accessible.
Results
Descriptive Characteristics
Over a median follow-up of 8.05 years (841,296.4 person-years), 1,131 incident type 2 diabetes cases were detected. Table 1 presents the baseline characteristics of the study participants, 79.2% of whom were women. The participants had a mean age at baseline of 42.5 years (SD 14.6). On average, they completed 18 (SD 17) 24HDRs during follow-up. The mean number of dietary records per 2-year study period is provided in Supplementary Table 2. Compared with low consumers, higher consumers of total food coloring additives were younger, had higher educational level, more likely to be current smokers, less physically active, more likely to have higher energy intake, and consume more UPFs but less alcohol.
Table 1.
Baseline characteristics of participants according to cumulative exposure to total food coloring additive, NutriNet-Santé cohort, 2009–2023 (N= 108,723)
| Sex-specific tertiles of total food coloring additives | |||||
|---|---|---|---|---|---|
| Characteristic | Overall (N = 108,723) | Tertile 1 (n = 36,241) | Tertile 2 (n = 36,241) | Tertile 3 (n = 36,241) | P valuea |
| Age (years) | 42.5 (14.6) | 44.2 (14.8) | 44.6 (14.5) | 38.5 (13.6) | <0.001 |
| Women, n (%) | 86,085 (79.2) | 28,695 (79.2) | 28,695 (79.2) | 28,695 (79.2) | NA |
| BMI (kg/m2)d | 23.6 (4.4) | 23.3 (4.2) | 23.6 (4.2) | 23.8 (4.6) | <0.001 |
| Family history of diabetesb,d, n (%) | 17,003.0 (15.7) | 5,609.0 (15.5) | 5,804.0 (16.1) | 5,590.0 (15.5) | 0.052 |
| Educational leveld, n (%) | <0.001 | ||||
| Less than a high school degree | 19,011.0 (17.7) | 6,803.0 (19.0) | 6,687.0 (18.6) | 5,521.0 (15.3) | |
| ≤3 years after high school | 52,775.0 (49.0) | 17,190.0 (48.0) | 17,343.0 (48.4) | 18,242.0 (50.7) | |
| >3 years after high school | 35,894.0 (33.3) | 11,835.0 (33.0) | 11,838.0 (33.0) | 12,221.0 (34.0) | |
| Smoking statusd, n (%) | <0.001 | ||||
| Never | 54,660.0 (50.4) | 17,751.0 (49.2) | 18,614.0 (51.5) | 18,295.0 (50.6) | |
| Former smoker | 35,231.0 (32.5) | 12,252.0 (34.0) | 12,233.0 (33.8) | 10,746.0 (29.7) | |
| Current smoker | 18,509.0 (17.1) | 6,057.0 (16.8) | 5,302.0 (14.7) | 7,150.0 (19.8) | |
| Smoking pack yearsd | 4.82 (10.7) | 5.26 (11.4) | 4.85 (10.6) | 4.35 (10.0) | <0.001 |
| IPAQ physical activity level,d n (%) | <0.001 | ||||
| Low | 22,544.0 (24.0) | 7,044.0 (22.4) | 7,316.0 (23.2) | 8,184.0 (26.4) | |
| Moderate | 40,448.0 (43.1) | 13,366.0 (42.6) | 13,544.0 (43.0) | 13,538.0 (43.6) | |
| High | 30,952.0 (32.9) | 10,990.0 (35.0) | 10,639.0 (33.8) | 9,323.0 (30.0) | |
| Energy intake without alcohol (kcal/day)c | 1,853.5 (456.5) | 1,786.5 (454.8) | 1,865.6 (433.6) | 1,908.4 (471.8) | <0.001 |
| Alcohol intake (g/day)c | 7.7 (11.8) | 7.6 (12.1) | 8.3 (11.8) | 7.3 (11.4) | <0.001 |
| Saturated fat (g/day)c | 33.2 (12.1) | 31.2 (12.2) | 33.9 (11.6) | 34.6 (12.4) | <0.001 |
| Sodium (mg/day)c | 2,721.9 (890.6) | 2,621.3 (913.6) | 2,767.1 (854.1) | 2,777.3 (894.6) | <0.001 |
| Fiber (g/day)c | 20.1 (10.0) | 21.2 (11.6) | 20.3 (9.3) | 18.9 (8.8) | <0.001 |
| Sugar (g/day)c | 93.2 (33.8) | 87.5 (33.8) | 93.2 (31.3) | 99.0 (35.0) | <0.001 |
| Fruit and vegetables (g/day)c | 465.8 (232.6) | 491.4 (257.1) | 475.2 (212.5) | 431.0 (221.7) | <0.001 |
| Dairy product (g/day)c | 158.5 (147.6) | 149.4 (149.9) | 160.7 (143.1) | 165.5 (149.1) | <0.001 |
| Processed meat (g/day)c | 34.5 (31.9) | 30.1 (32.0) | 35.9 (30.6) | 37.6 (32.6) | <0.001 |
| Ultra-processed food (% of weight intake)c | 17.3 (9.9) | 14.8 (8.6) | 15.7 (8.0) | 21.6 (11.3) | <0.001 |
| Total food coloring additive intake (mg/day) | 185.8 (531.3) | 0.1 (0.2) | 9.9 (9.0) | 547.4 (806.6) | <0.001 |
Data are mean (SD), unless indicated otherwise as n (%). NA, not applicable.
aKruskal-Wallis rank sum test for continuous variables; Pearson χ2 test for categorical variables.
bFamily history of diabetes in first-degree relatives.
cAll dietary intake data in this table were calculated as the mean daily intake across all records during the first 4 years of participation in the study (mean number of 24-h records per person during this period, 6 [SD 3]).
dMissing values: BMI, n = 3,001 in tertiles 1, 2, and 3 (1,032; 922; 1,047); family history of diabetes, n = 333 (162; 109; 62); education level, n = 1,043 (413; 373; 257); socioprofessional categories, n = 410 (209; 129; 72); monthly household income per consumption unit, n = 13,183 (4,721, 4,155, 4,307); smoking status, n = 323 (181; 92; 50); smoking pack years, n = 329 (182, 92, 55); IPAQ physical activity level, n = 14,779 (4,841; 4,742; 5,196).
A total of 87.0% of participants had a nonnull intake of coloring additives in the first 2 years of follow-up (Table 2). Among consumers, the mean (SD) intake of total food coloring additives were 213.5 (564.3) mg/day compared with 185.8 (531.3) mg/day in the total population (186.4 [524.5] mg/day in women and 183.6 [556.6] mg/day in men). Supplementary Table 3 shows the daily food coloring additive exposures among consumers by follow-up periods. Overall, participants did not exceed the EFSA Acceptable Daily Intakes (ADI) (Supplementary Table 4), except for lutein, for which three participants exceeded the ADI (i.e., 1 mg/kg body wt/day), with a mean intake of 0.02 (SD 0.05) mg/kg body wt/day (25th–75th percentiles 0.00002–0.10500). These participants had on average higher intake of milk, yogurt, milk-based desserts, biscuits/cookies, sweet products, and sauces, which could explain higher observed lutein exposure.
Of the total number of coloring additives detected in the study population, most were synthetic colors (20 of 34 colors; i.e., 58.8% vs. 20.6% each for natural colors and ambivalent). However, apart from caramel colors, synthetic colors were consumed only by a small or very small proportion of the population (Table 2), as opposed to colors of natural origin. By proportion of consumers, the main coloring additives were paprika, capsanthin, capsorubin (47.9%), β-carotene (42.7%), carotenoids without specification (37.0%), lutein (28.5%), cochineal-carminic acid-carmines (27.2%), sulfite ammonia caramel (26.2%), curcumin (25.1%), plain caramel (23.2%), annatto-bixin-norbixin (17.0%), and anthocyanins (13.4%). No strong correlation between intakes of coloring additives was identified (Supplementary Fig. 3). Coloring additives consumed by <10% of the study population are listed in a footnote to Fig. 2. Furthermore, nine coloring additives are allowed in the European market but were not consumed by participants in this study (seldom used in the French market). Their list is provided in footnote to Table 2.
Figure 2.
Associations between exposure to food coloring additives and type 2 diabetes incidence among study participants from the NutriNet-Santé cohort, 2009–2023 (n = 108,723).a,b,c,d 1Natural; 2Artificial; 3Ambivalent (natural or artificial). aMultivariable Cox proportional hazard models (model 3) adjusted for age (time-scale), sex, number of dietary records, baseline BMI (continuous, kg/m2), physical activity (categorical IPAQ variable: high, moderate, or low), smoking status (never, former, or current smoker), number of smoked cigarettes in pack-years (continuous), educational level (less than high school, ≤3 years after high school, or >3 years after high school), family history of diabetes (yes or no), previous metabolic diseases (yes or no), time-dependent daily intakes of energy without alcohol (continuous, kcal/d), alcohol (continuous, g/day), saturated fats (continuous, g/d), sodium (continuous, mg/d), fiber (continuous, g/d), sugars (continuous, g/d), fruit and vegetables (continuous, g/day), dairy products (continuous, g/day), processed meat (continuous, g/day), UPF (continuous, percentage weight of total food/day in g), and mutually adjusted for total food coloring additive intake excluding the one studied. Specific food coloring additive models were further adjusted for natural sources: total carotene, E160, E160a, E160b, and E161b were adjusted for dietary β-carotene; E160c was adjusted for dietary capsorubin, capsanthin, and β-carotene; and E100 was adjusted for curcumin. bThe three food coloring additive categories were defined as follows: sex-specific tertiles for total food coloring additives and total carotene and otherwise (if less than two-thirds of consumers): 1/nonconsumers, 2/lower, and 3/higher consumers, the latter two being separated by the sex-specific median. Cutoffs were recalculated for each follow-up period and are available in Supplementary Table 1. cAssociations with diabetes risk were investigated for food coloring additives consumed by at least 10% of the study population. dWhen the log-linearity assumption was not rejected (P for nonlinearity ≥0.05 assessed by restricted cubic splines models), the P for linear trend was retained (obtained by coding the exposure as an ordinal categorical variable 1, 2, 3). When the assumption of log-linearity was not met (P for nonlinearity <0.05), it was not adapted to calculate a P for linear trend; thus, the likelihood ratio overall P value was retained (obtained by coding the exposure as a nonordinal categorical variable and calculating likelihood ratio test between model with food additive exposure and model without food additive exposure). The P value displayed in the forest plot is *P for trend (when P for nonlinearity was ≥0.05) or the overall §P (when P for nonlinearity was <0.05). eModel 3 (main model): corresponding P values after FDR adjustment: total food coloring additives: 0.0002; total caramel: 0.0002; E150a: 0.0002; E150d: 0.007; total carotene: 0.007; E160: 0.0002; E160a: 0.0002; E160b: 0.03; E160c: 0.004; E161b: 0.0002; E100: 0.0002; E120: 0.003; and E163: 0.0002.
The contribution of food groups to food coloring additive intakes is presented in Fig. 1 and Supplementary Table 5. Unsweetened (49.6%) and sweetened drinks (32.2%) were the main contributors to total food coloring exposure. Key sources varied, such as drinks for caramels, confectionery for carotenoids and anthocyanins, dairy for curcumin and lutein, and fish/seafood for paprika-derived colors. For compounds that also occur naturally in the diet, the relative contribution from food additives were 0.07% for β-carotene, 22.0% for capsanthin, 33.0% for curcumin, and 39.6% for capsorubin.
Figure 1.
Dietary sources of food coloring additive intakes among study participants from the NutriNet-Santé cohort, 2009–2023 (N = 108,723).a,b,c aFor food additives consumed by at least 10% of the study population. These included individual food colors curcumin (E100), cochineal-carminic acid-carmines (E120), plain caramel (E150a), sulfite ammonia caramel (E150d), carotenoids without specification (E160), β-carotene (E160a), annatto-bixin-norbixin (E160b), paprika-capsanthin-capsorubin (E160c), lutein (E161b), and anthocyanins (E163). Food color categories included total caramel (E150, E150a, E150b, E150c, and E150d), total carotene (E160, E160a, E160b, E160c, E161b, E160d, and E160e), and total food coloring additives (E100, E101, E102, E104, E110, E120, E122, E123, E124, E127, E129, E132, E133, E140, E141, E150, E150a, E150b, E150c, E150d, E155, E160, E160a, E160b, E160c, E160d, E160e, E161b, E162, E163, E170, E171, E172, E131, E153, and E175). Food coloring additives consumed by <10% of the study population were included in total and category-specific intakes but could not be evaluated individually. These included riboflavin (E101), tartrazine (E102), quinoline yellow (E104), sunset yellow FCF (E110), carmoisine (E122), amaranth (E123), ponceau 4R (E124), erythrosine (E127), allura red AC (E129), patent blue V (E131), indigotine (E132), brilliant blue FCF (E133), copper complexes of chlorophylls (E141), caramel subclasses (E150b, E150c), brilliant black BN (E151), brown HT (E155), lycopene (E160d), β-apo-8′-carotenal (E160e), beetroot red (E162), calcium carbonate (E170), titanium dioxide (E171), iron oxides and hydroxides (E172), azoic dyes (E102, E104, E110, E122, E123, E124, E129, E155), and total nanoparticle-based colors (E171, E172). bDetailed percentages are presented in Supplementary Table 5. cThe “Others” food group includes food items such as herbs and spices, sugar-free sweets and chewing gums, unprepared powders (e.g., instant coffee) and condiments (e.g., mustard, hot pepper sauce, or wasabi).
Associations Between Food Coloring Additive Intakes and Type 2 Diabetes Incidence
Schoenfeld residuals did not refute the proportional hazard assumption (Supplementary Fig. 4). Restricted cubic spline plots are presented in Supplementary Fig. 5. No departure from linearity was observed (P values for nonlinearity ≥0.05) for total caramel, sulfite ammonia caramel, carotenoids, β-carotene, annatto-bixin-norbixin, paprika-capsanthin-capsorubin, and curcumin, cochineal-carminic acid-carmines, and in this case, P values for trend are provided in Fig. 2. For some associations, the dose-response curve suggested a plateau effect (P values for nonlinearity <0.05), for total food coloring additives, plain caramel, total carotene, lutein, and anthocyanins. In these cases, overall likelihood ratio P values are presented. Both P trends and overall P values are provided for all tested additives in Supplementary Table 6.
Associations of food coloring additive exposure with type 2 diabetes incidence are presented in Fig. 2 and Supplementary Table 6. Following FDR correction, in the fully-adjusted model (model 3), intakes of following coloring additives were associated with higher type 2 diabetes incidence: total food coloring additives (hazard ratio [HR]higher vs. non/lower consumers 1.38 [(95% CI 1.17–1.63], P = 0.0002), total caramel (1.43 [1.21–1.67], P = 0.0002), plain caramel (1.46 [1.26–1.70], P = 0.0002), sulfite ammonia caramel (1.30 [1.07–1.59], P = 0.007), total carotene (1.27 [1.08–1.48], P = 0.007), carotenoids (1.39 [1.19–1.62], P = 0.0002), β-carotene (1.44 [1.23–1.68], P = 0.0002), paprika-capsanthin-capsorubin (1.26 [1.08–1.46], P = 0.004), lutein (1.20 [1.02–1.40], P = 0.0002), curcumin (1.49 [1.29–1.73], P = 0.0002), cochineal-carminic acid-carmines (1.27 [1.10–1.48], P = 0.003), and anthocyanins (1.40 [1.17–1.68], P = 0.0002). These results were similar across all sensitivity analyses testing further adjustments (Supplementary Table 6).
We found no evidence of interaction between food coloring additive intakes and sex, BMI, amount of UPFs, sugar-sweetened beverages, smoking, or physical activity for type 2 diabetes incidence (all interaction FDR P value > 0.05). Naturally occurring β-carotene (HRhigher vs. non/lower consumers 1.02 [95% CI 0.85–1.22], Ptrend = 0.85), curcumin (0.95 [0.79–1.15], Ptrend = 0.60), capsanthin (0.99 [0.85–1.15], Ptrend = 0.94), and capsorubin (0.99 [0.85–1.15], Ptrend = 0.94) were not associated with increased type 2 diabetes incidence. In our study, UPF consumption was associated with a higher incidence of type 2 diabetes (HR 1.20 [95% CI 1.11–1.29], P < 0.001). Mediation analysis was conducted to assess the proportion of this association mediated by total food coloring additives. Indirect effect through total food coloring intake was statistically significant (pure natural indirect effect HR 1.02 [95% CI 1.00–1.03], P = 0.03). Overall, 8.6% (95% CI 1.6–18.5) of the association between UPF consumption and type 2 diabetes incidence was mediated by total food coloring additives (P = 0.03).
Conclusions
Our findings revealed associations between several food coloring additives and higher type 2 diabetes incidence for plain caramel, sulfite ammonia caramel, carotenoids, β-carotene, paprika-capsanthin-capsorubin, lutein, curcumin, cochineal extract-carminic acid-carmines, and anthocyanins, as well as total food coloring additives, total caramel, and total carotene, after adjustment for most known or potential risk factors or confounders.
To our knowledge, this is the first epidemiological study to investigate associations between coloring additives and incidence of type 2 diabetes in a large prospective setting. The qualitative and quantitative exposures to coloring additives were assessed for the first time in the NutriNet-Santé cohort by considering different commercial brands of products, which is necessary to identify specific colors consumed by participants at the individual level, given the huge variability of food additives content in commercial products. Thus, directly comparing our findings with previous epidemiological literature is not possible. For coloring additives with available EFSA data, lower mean intakes were observed in NutriNet-Santé versus EFSA estimates, reflecting differences in assessment dates, study populations, and methods—NutriNet-Santé used precise brand-specific repeated 24-h records versus generic items and limited data on which EFSA estimates are based.
Our findings for several coloring additives are supported by mechanistic evidence. In a recent in vitro study (24) from our consortium, cochineal-carminic acid-carmines—red colors used in confectionery, beverages, and processed meats—significantly increased DNA damage markers and reduced cell viability. Moreover, paprika-capsanthin-capsorubin—dark red color in sauces, snacks, and spice blends—caused decreased cell viability at higher levels. These findings suggest coloring additives may induce cellular stress in metabolically active tissues, providing biological plausibility for their role in chronic disease pathogenesis.
Total food coloring additives include azo dyes, synthetic compounds with diazotized amine combined with another amine or phenol characterized by azo linkages and aromatic amine precursors. Aromatic amines are biotransformed by azo reductases of intestinal microflora, with adverse health effects largely associated with this biotransformation (25). In rats, tartrazine—a yellow color used in bread, beverages, candies, snacks, dairy products, and sauces— increased serum glucose by 15.8% versus control individuals and elevated protein kinase C (9), which is linked to secondary insulin resistance via hyperinsulinemia and/or hyperglycemia (26).
Recent evidence shows that tartrazine induces mitochondrial dysfunction in zebrafish (27), notable given that type 2 diabetes has been linked to impaired mitochondrial genome integrity, mitochondrial RNA expression, and mitophagy in pancreatic β-cells (28). Evidence also links dysbiotic gut microbiota to metabolic disease, with certain food additives exacerbating this via increased intestinal permeability; sunset yellow, in particular, caused intestinal histopathological alterations in mice, compromising mucosal integrity (10).
Some coloring additives may have potential obesogenic effects by disrupting cellular redox balances critical for energy metabolism (29). Certain colors, notably erythrosine—red color used in candies, bakery products, popsicles, and beverages— and tartrazine, show endocrine-disrupting effects in in vivo studies (30). These effects may have potential to trigger insulin resistance and dysregulation of metabolic homeostasis, with implications in the pathogenesis of type 2 diabetes.
In our study, azo dyes—brilliant blue, tartrazine, sunset yellow, erythrosine, carmoisine, and allura red—counted toward total food coloring additives but could not be assessed individually due to low consumption reflecting limited use in the French market. Although numerous artificial food colors are permitted, many are rarely used as the food industry increasingly favors natural colors for functional and marketing purposes. Natural colors are often perceived by consumers as inherently safer, though this assumption is not consistently supported by toxicological evidence, and regulatory and public health attention (particularly in the U.S.) remains primarily focused on artificial colors. We observed significant associations between both natural and artificial food colors and incident type 2 diabetes, consistent with literature suggesting potential health risks across both classes (31).
Coloring additives may exert adverse metabolic effects via embedded chemical constituents. In vivo evidence shows 4-methylimidazole, a byproduct of caramel coloring (notably ammonia caramel and ammonia sulfite caramel) induces pancreatic β-cell hyperplasia and hyperinsulinemia, leading to enhanced glucose clearance and reduced fasting glucose and HbA1c (32). This persistent overstimulation suggests potential for long-term β-cell exhaustion or insulin resistance, hallmarks of type 2 diabetes. Furthermore, chronic exposure to allura red AC—used in beverages, confectionery, breakfast cereals, and baked goods—in mice at doses present in diet produces colitis by promoting serotonin synthesis and inflammation via microbiota-dependent and -independent pathways (12). Nevertheless, many common coloring additives remain unexplored for metabolic effects, warranting further experimental and epidemiological research.
Isolating the effect of the studied food coloring additive from its food vector is challenging. Sensitivity analyses adjusting for UPF characteristics (overall UPF intake, other additives associated with diabetes, trans fats) and main food vectors (e.g., beverages for caramel colors) found attenuated (which may be due to overadjustment) but overall similar results, supporting a potential effect of the studied substance itself, that should be confirmed in future experimental studies. Extensive evidence shows substances exert different effects depending on the matrix in which they are ingested (33). Indeed, the food matrix properties—composition, structure, pH—affect absorption and metabolism by microbiota and host, and physical and chemical interactions of substances with other elements of the environment play modulatory roles (34,35). For instance, while no association was detected for natural β-carotene in fruit and vegetables, high dose β-carotene isolated in dietary supplements, in interaction with tobacco/asbestos exposure, causally increases lung cancer risk (36). In vitro experiments showed that β-carotene upregulates inflammation-related genes under hyperglycemic conditions, with histone modifications suggesting epigenetic regulation (37). Similarly, while dietary fibers within complex natural fruit, vegetable, and legume matrices are protective, isolated purified fibers (e.g., inulin) exacerbated colitis in mice models (38). In recent studies, we observed associations of multiple food additive preservatives, but not natural food-born substances, with higher type 2 diabetes and cancer risks (39,40). Thus, experimental studies should explore differential health effects of chemicals by source and nature (additives vs. naturally occurring). Moreover, coloring additives–type 2 diabetes associations exhibited both linear and nonlinear patterns. Departures from linearity were observed for several colors, suggesting threshold or plateau effects. Further mechanistic and epidemiological investigations are needed to better understand these findings.
Our study’s strengths include a large prospective population-based cohort with a long follow-up, regularly updated brand-specific 24HDRs linked to multiple food composition databases, ad hoc laboratory assays for the most frequently consumed additive-food pairs in food products, and dynamic matching accounting for reformulations.
There are also some limitations. Although we controlled for multiple confounding factors, unmeasured or residual confounding cannot be excluded. Measurement errors in dietary assessment and additive exposure might occur; however, records were validated against dietitian interviews and blood and urinary biomarkers for energy and nutrients. Nevertheless, specific exposure to coloring additives has not been validated against blood or urine assays because of absence of specific biomarkers for most additives. Coloring additives exposure was restricted to substances permitted in the European Union, and several coloring additives were not ingested by a sufficient number of participants for individual investigation. The cohort consists of a high proportion of women and health-conscious individuals; thus, caution is warranted when extrapolating findings. Last, although it may be inherently challenging to disentangle the independent effects of coloring additives from those of other food additives and food composition parameters, our results were stable when adjusting models for artificial sweeteners, nitrites, emulsifiers, preservatives, trans fats, energy, and nutrients.
In conclusion, our findings revealed positive associations between widely consumed food coloring additives and type 2 diabetes incidence. To our knowledge, this is the first study to provide insights into the role of coloring additives in the development of diabetes. Further long-term epidemiological and experimental research is needed to understand underlying mechanisms. If confirmed, these results call for a reevaluation of regulations governing the use of food coloring additives by the food and beverage industry to better protect public health.
This article contains supplementary material online at https://doi.org/10.2337/figshare.31807444.
Article Information
Acknowledgments. The research question developed in this article corresponds to a strong concern of the participants involved in the NutriNet-Santé cohort and of the public in general. The results of the current study will be disseminated to the NutriNet-Santé participants through the cohort website, public seminars, and a press release. The authors thank Thi Hong Van Duong, Régis Gatibelza, Amelle Aitelhadj, and Aladi Timera (computer scientists) and Selim Aloui (IT manager); Julien Allegre, Nathalie Arnault, Nicolas Dechamp, Laurent Bourhis (data managers/statisticians); Erwan Louveau and Mathéo Le Floch (statistician trainees); Maria Gomes and Mirette Foham (participant support); Laure Legris and Laura Chaud (dietitians), and Marie Ajanohun, Tassadit Haddar (administration and finance) and Nadia Khemache (administrative manager) for their technical contribution to the NutriNet-Santé study (Université Sorbonne Paris Nord and Université Paris Cité, INSERM, INRAE, CNAM, Centre for Research in Epidemiology and StatisticS (CRESS), Nutritional Epidemiology Research Team (EREN), F-93017 Bobigny, France). Finally, the authors warmly thank all the volunteers of the NutriNet-Santé cohort.
The NutriNet-Santé study was supported by the following public institutions: Ministère de la Santé, Santé Publique France, Institut National de la Santé et de la Recherche Médicale (INSERM), Institut National de la Recherche pour l’agriculture, l’alimentation et l’environnement (INRAE), Conservatoire National des Arts et Métiers (CNAM), and University Sorbonne Paris Nord. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant agreement No 864219, ADDITIVES), the French National Cancer Institute (INCa_14059), the French Ministry of Health (arrêté 29.11.19), the IdEx Université de Paris (ANR-18-IDEX-0001), and a Bettencourt-Schueller Foundation Research Prize 2021. S.S. benefited from a postdoctoral fellowship from the CANCEPT network, DA 2022-020 from Institut national du Cancer. A.H. is funded by a PhD grant from the French Ministry of Health/Sorbonne Paris Nord University. This project was awarded the NACRe (French Network for Nutrition And Cancer Research) Partnership Label. B.C.’s laboratory is supported by a Consolidator Grant (grant agreement InterBiome No. ERC-2024-CoG-101170920) from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program, grant for the AFA Crohn RCH France and the national program “Microbiote” from INSERM.
This work only reflects the authors' view, and the funders are not responsible for any use that may be made of the information it contains. Researchers were independent from funders. Funders had no role in the study design, the collection, analysis, and interpretation of data, the writing of the report, and the decision to submit the article for publication.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.S. drafted the manuscript. S.S. and G.J. performed statistical analysis. S.S., B.S., and M.T. designed the study. S.S., B.S., and M.T. had full access to all the data in the study. F.S.d.E., C.A., A.D.S., and M.T. developed the additives composition database and matched consumption/composition data. P.Y. led the validation of cases. B.S. and M.T. supervised the writing. B.S. and M.T. supervised statistical analysis. All authors contributed to the data interpretation and revised each draft for important intellectual content. All authors read and approved the final manuscript. S.S. and M.T. attest that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. M.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Cuilin Zhang.
Funding Statement
Where authors are identified as personnel of the International Agency for Research on Cancer/World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy, or views of the International Agency for Research on Cancer/World Health Organization.
Footnotes
Clinical trial reg. no. NCT03335644, clinicaltrials.gov
Contributor Information
Sanam Shah, Email: sanam.shah@eren.smbh.univ-paris13.fr.
Mathilde Touvier, Email: m.touvier@eren.smbh.univ-paris13.fr.
Supporting information
References
- 1. Ong KL, Stafford LK, McLaughlin SA, et al. Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet 2023;402:203–234 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Touvier M, da Costa Louzada ML, Mozaffarian D, Baker P, Juul F, Srour B. Ultra-processed foods and cardiometabolic health: public health policies to reduce consumption cannot wait. BMJ 2023;383:e075294. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Monteiro CA, Louzada ML, Steele-Martinez E, et al. Ultra-processed foods and human health: the main thesis and the evidence. Lancet 2025;406:2667–2684 [DOI] [PubMed] [Google Scholar]
- 4. Miller MD, Steinmaus C, Golub MS, et al. Potential impacts of synthetic food dyes on activity and attention in children: a review of the human and animal evidence. Environ Health 2022;21:45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. European Food Safety Authority (EFSA). EFSA updates safety advice on six food colours . 2009. Accessed 20 August 2025. Available from https://www.efsa.europa.eu/en/press/news/091112
- 6. McCann D, Barrett A, Cooper A, et al. Food additives and hyperactive behaviour in 3-year-old and 8/9-year-old children in the community: a randomised, double-blinded, placebo-controlled trial. Lancet 2007;370:1560–1567 [DOI] [PubMed] [Google Scholar]
- 7. U.S. Food and Drug Administration. Use of the Term Natural on Food Labeling, 2024. Accessed 13 October 2025. Available from https://www.fda.gov/food/nutrition-food-labeling-and-critical-foods/use-term-natural-food-labeling [Google Scholar]
- 8. Amin KA, Abdel Hameid H, Abd Elsttar AH. Effect of food azo dyes tartrazine and carmoisine on biochemical parameters related to renal, hepatic function and oxidative stress biomarkers in young male rats. Food Chem Toxicol 2010;48:2994–2999 [DOI] [PubMed] [Google Scholar]
- 9. El‐Desoky GE, Wabaidur SM, AlOthman ZA, Habila MA. Evaluation of nano‐curcumin effects against tartrazine‐induced abnormalities in liver and kidney histology and other biochemical parameters. Food Sci Nutr 2022;10:1344–1356 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Di Vincenzo F, Del Gaudio A, Petito V, Lopetuso LR, Scaldaferri F. Gut microbiota, intestinal permeability, and systemic inflammation: a narrative review. Intern Emerg Med 2024;19:275–293 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Şensoy E. Comparison of the effect of sunset yellow on the stomach and small intestine of developmental period of mice. Heliyon 2024;10:e31998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Kwon YH, Banskota S, Wang H, et al. Chronic exposure to synthetic food colorant allura red AC promotes susceptibility to experimental colitis via intestinal serotonin in mice. Nat Commun 2022;13:7617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hercberg S, Castetbon K, Czernichow S, et al. The Nutrinet-Santé Study: a web-based prospective study on the relationship between nutrition and health and determinants of dietary patterns and nutritional status. BMC Public Health 2010;10:242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Vergnaud A-C, Touvier M, Méjean C, et al. Agreement between web-based and paper versions of a socio-demographic questionnaire in the NutriNet-Santé study. Int J Public Health 2011;56:407–417 [DOI] [PubMed] [Google Scholar]
- 15. Touvier M, Kesse-Guyot E, Méjean C, et al. Comparison between an interactive web-based self-administered 24 h dietary record and an interview by a dietitian for large-scale epidemiological studies. Br J Nutr 2011;105:1055–1064 [DOI] [PubMed] [Google Scholar]
- 16. Lassale C, Castetbon K, Laporte F, et al. Validation of a Web-based, self-administered, non-consecutive-day dietary record tool against urinary biomarkers. Br J Nutr 2015;113:953–962 [DOI] [PubMed] [Google Scholar]
- 17. Lassale C, Castetbon K, Laporte F, et al. Correlations between fruit, vegetables, fish, vitamins, and fatty acids estimated by web-based nonconsecutive dietary records and respective biomarkers of nutritional status. J Acad Nutr Diet 2016;116:427–438.e5 [DOI] [PubMed] [Google Scholar]
- 18. Touvier M, Méjean C, Kesse-Guyot E, et al. Comparison between web-based and paper versions of a self-administered anthropometric questionnaire. Eur J Epidemiol 2010;25:287–296 [DOI] [PubMed] [Google Scholar]
- 19. Lassale C, Péneau S, Touvier M, et al. Validity of web-based self-reported weight and height: results of the Nutrinet-Santé study. J Med Internet Res 2013;15:e152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Craig CL, Marshall AL, Sjöström M, et al. International Physical Activity Questionnaire: 12-country reliability and validity. Med Sci Sports Exerc 2003;35:1381–1395 [DOI] [PubMed] [Google Scholar]
- 21. Black AE. Critical evaluation of energy intake using the Goldberg cut-off for energy intake:basal metabolic rate. A practical guide to its calculation, use and limitations. Int J Obes Relat Metab Disord 2000;24:1119–1130 [DOI] [PubMed] [Google Scholar]
- 22. Black AE. The sensitivity and specificity of the Goldberg cut-off for EI:BMR for identifying diet reports of poor validity. Eur J Clin Nutr 2000;54:395–404 [DOI] [PubMed] [Google Scholar]
- 23. Chazelas E, Druesne-Pecollo N, Esseddik Y, et al. Exposure to food additive mixtures in 106,000 French adults from the NutriNet-Santé cohort. Sci Rep 2021;11:19680. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Recoules C, Touvier M, Pierre F, Audebert M. Evaluation of the toxic effects of food additives, alone or in mixture, in four human cell models. Food Chem Toxicol 2025;196:115198. [DOI] [PubMed] [Google Scholar]
- 25. Chung KT, Stevens SE, Cerniglia CE. The reduction of azo dyes by the intestinal microflora. Crit Rev Microbiol 1992;18:175–190 [DOI] [PubMed] [Google Scholar]
- 26. Schmitz-Peiffer C. Signalling aspects of insulin resistance in skeletal muscle. Cell Signal 2000;12:583–594 [DOI] [PubMed] [Google Scholar]
- 27. Haridevamuthu B, Murugan R, Seenivasan B, et al. Synthetic azo-dye, tartrazine induces neurodevelopmental toxicity via mitochondria-mediated apoptosis in zebrafish embryos. J Hazard Mater 2024;461:132524. [DOI] [PubMed] [Google Scholar]
- 28. Walker EM, Pearson GL, Lawlor N, et al. Retrograde mitochondrial signaling governs the identity and maturity of metabolic tissues. Science 2025;388:eadf2034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Corkey BE, Shirihai O. Metabolic master regulators: sharing information among multiple systems. Trends Endocrinol Metab 2012;23:594–601 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Paramasivam A, Murugan R, Jeraud M, Dakkumadugula A, Periyasamy R, Arjunan S. Additives in processed foods as a potential source of endocrine-disrupting chemicals: a review. J Xenobiot 2024;14:1697–1710 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Renita AA, Gajaria TK, Sathish S, et al. Progress and prospective of the industrial development and applications of eco-friendly colorants: an insight into environmental impact and sustainability issues. Foods 2023;12:1521. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Rekha B, Velmurugan G, Freddy AJ, et al. Chronic intake of 4-methylimidazole induces hyperinsulinemia and hypoglycaemia via pancreatic beta cell hyperplasia and glucose dyshomeostasis. Sci Rep 2018;8:17037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Weaver CM, Givens DI. Overview: the food matrix and its role in the diet. Crit Rev Food Sci Nutr 2025;65:6880–6897 [DOI] [PubMed] [Google Scholar]
- 34. van Het Hof KH, West CE, Weststrate JA, Hautvast JG. Dietary factors that affect the bioavailability of carotenoids. J Nutr 2000;130:503–506 [DOI] [PubMed] [Google Scholar]
- 35. Yang M, Koo SI, Song WO, Chun OK. Food matrix affecting anthocyanin bioavailability: review. Curr Med Chem 2011;18:291–300 [DOI] [PubMed] [Google Scholar]
- 36. Gallicchio L, Boyd K, Matanoski G, et al. Carotenoids and the risk of developing lung cancer: a systematic review. Am J Clin Nutr 2008;88:372–383 [DOI] [PubMed] [Google Scholar]
- 37. Kondo S, Suzuki R, Nakashima Y, Mochizuki K. β-Carotene enhances the expression of inflammation-related genes and histone H3 K9 acetylation, K4 dimethylation, and K36 trimethylation around these genes in juvenile macrophage-like THP-1 cells. Biochem Biophys Rep 2022;31:101325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Miles JP, Zou J, Kumar M-V, et al. Supplementation of low- and high-fat diets with fermentable fiber exacerbates severity of DSS-induced acute colitis. Inflamm Bowel Dis 2017;23:1133–1143 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Hasenböhler A, Javaux G, Payen de la Garanderie M, et al. Intake of food additive preservatives and incidence of cancer: results from the NutriNet-Santé prospective cohort. BMJ 2026;392:e084917. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Hasenböhler A, Javaux G, Payen de la Garanderie M, et al. Associations between preservative food additives and type 2 diabetes incidence in the NutriNet-Santé prospective cohort. Nat Commun 2026;16:11199. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.