Key Takeaways
- A prospective study of 112,050 French adults found that several patterns of food additives commonly consumed together were associated with higher incidence of hypertension and cardiovascular disease.
- A mixture characterized by modified starches, pectin, guar gum, and carrageenan was associated with higher relative hazards of hypertension, cardiovascular disease, and coronary heart disease.
- The researchers examined combinations of additives rather than only individual substances, an approach intended to better reflect real-world dietary exposure.
- The study was observational and cannot establish that the additives caused cardiovascular disease or that avoiding individual ingredients would reduce risk.
- The findings need confirmation in independent populations because this and related studies were conducted within the same French NutriNet-Santé cohort.
Introduction
Carrageenan, guar gum and modified starches were among the additives in a dietary mixture associated with higher rates of some cardiovascular outcomes in a large French observational study.
People rarely consume food additives one at a time. A packaged food or drink may contain several stabilizers, emulsifiers, acidifiers, sweeteners, preservatives, or other additives together.
That raises a question that studies of individual additives cannot fully answer: are patterns of additives commonly consumed together associated with long-term health outcomes?
A prospective study published September 8, 2026, in BMC Medicine examined that question in 112,050 French adults. Researchers identified five recurring mixtures of food additives in participants’ diets and investigated their associations with hypertension and cardiovascular disease (Payen de La Garanderie et al., 2026).
Several mixtures were associated with higher incidence of cardiovascular outcomes. The results are notable because they examine real-world patterns of co-exposure, but they do not establish that any additive or combination of additives causes heart disease.
What the Study Examined
The participants were members of the French NutriNet-Santé cohort followed between 2009 and 2024. Their mean age was 42.8 years, and 79% were women.
Dietary exposure was assessed in unusual detail. Participants could complete up to three 24-hour dietary records every six months, including information about commercial brands of industrial foods.
Researchers combined those records with food-composition databases and laboratory analyses to estimate exposure to individual additives.
They then used a statistical method called non-negative matrix factorization to identify five patterns of additives that tended to be consumed together.
This does not mean that every participant consumed every additive in a mixture at the same level. The mixtures represent statistical patterns of co-exposure across participants’ diets.
The researchers used multivariable Cox proportional hazards models with time-dependent exposures to examine associations with subsequent hypertension and cardiovascular disease, including coronary heart disease and cerebrovascular events.
What Researchers Found
During a median follow-up of 7.92 years, 5,543 participants developed hypertension and 2,450 experienced cardiovascular disease events. These included 1,142 cerebrovascular events and 1,308 cases of coronary heart disease (Payen de La Garanderie et al., 2026).
The associations varied according to the additive mixture and cardiovascular outcome.
One Mixture Was Associated With Several Outcomes
Mixture 2 was characterized mainly by modified starches, pectin, guar gum, and carrageenan.
Compared with participants in the lowest tertile of exposure, those in the highest tertile had:
- 12% higher relative hazard of hypertension: HR 1.12, 95% CI 1.05 to 1.19
- 18% higher relative hazard of cardiovascular disease: HR 1.18, 95% CI 1.07 to 1.31
- 26% higher relative hazard of coronary heart disease: HR 1.26, 95% CI 1.10 to 1.45
These figures are relative hazard estimates comparing the highest and lowest exposure tertiles. A 26% higher relative hazard, for example, does not mean that 26% of highly exposed participants developed coronary heart disease or that an individual’s absolute risk rises by 26 percentage points.
Other Mixtures Showed More Selective Associations
Mixture 1, characterized mainly by sodium carbonates, diphosphates, glycerol, and ammonium carbonates, was associated with a 14% higher relative hazard of hypertension when the highest and lowest exposure tertiles were compared. The estimated relative hazard of cardiovascular disease was 12% higher, although that result was borderline, with the 95% confidence interval extending from 1.00 to 1.24 (Payen de La Garanderie et al., 2026).
Mixture 4, characterized by ammonium and sodium carbonates, diphosphates, and alpha-tocopherol, was associated with a 12% higher relative hazard of hypertension, a 13% higher relative hazard of cardiovascular disease, and a 16% higher relative hazard of coronary heart disease.
Mixture 5, characterized by citric acid, sodium citrates, phosphoric acid, and acesulfame-K, was associated with a 19% higher relative hazard of hypertension.
Mixture 3, characterized mainly by magnesium carbonates, riboflavin, alpha-tocopherol, and ammonium carbonates, was not associated with the cardiovascular outcomes studied.
The differences among mixtures are important. The findings do not support the broader conclusion that all food additives, or all combinations of additives, have the same relationship with cardiovascular health.
Why Studying Additives as Mixtures Matters
Food-additive research has often evaluated substances individually. Consumers, however, encounter combinations of additives in actual foods.
The mixture approach attempts to capture that complexity.
The researchers also conducted exploratory analyses looking for statistical interactions between individual additives. Some patterns were compatible with possible synergistic or antagonistic interactions.
Those results should be interpreted carefully.
A statistical interaction does not prove a biological “cocktail effect.” It cannot establish that two additives become harmful when combined or identify which ingredient, if any, is responsible for an association.
There is another complication. Additive mixtures may partly reflect the foods in which they occur.
People with greater exposure to a particular mixture may consume different products and have different overall dietary patterns from people with lower exposure. The observed association could therefore reflect individual additives, combinations of additives, characteristics of the foods containing them, other dietary behaviors, or some combination of these factors.
The study cannot separate all of those possibilities.
How the Findings Compare With Earlier Research
Previous NutriNet-Santé studies have also reported associations between food additives and chronic disease.
In 2023, a prospective analysis of 95,442 adults found that higher intake of several emulsifiers, including certain celluloses and mono- and diglycerides of fatty acids, was associated with modestly higher incidence of cardiovascular outcomes (Sellem et al., 2023).
Researchers later moved from individual additives toward mixtures. A 2025 PLOS Medicine study of 108,643 NutriNet-Santé participants identified five commonly consumed additive mixtures and found that two were associated with higher incidence of type 2 diabetes (Payen de la Garanderie et al., 2025).
Some mixture patterns overlap with those examined in the new cardiovascular study. The earlier diabetes analysis included one mixture containing modified starches, pectin, guar gum, carrageenan, and other additives, and another containing citric acid, sodium citrates, phosphoric acid, artificial sweeteners, and additional substances.
This makes the new cardiovascular findings relevant to an emerging line of research on additive co-exposure.
But there is an important qualification: these studies are not independent replications.
All three used the French NutriNet-Santé cohort and involved overlapping or related participant populations. Seeing associations with different health outcomes within the same research cohort can generate useful hypotheses, but confirmation in independent populations would provide much stronger evidence that the relationships are reproducible.
What the Results Mean for Consumers
The study does not provide a list of additives that consumers need to eliminate.
An additive appearing in an associated mixture should not automatically be considered responsible for the association. Some substances appeared within complex combinations, and some additives occurred in more than one mixture.
Nor can the results tell an individual how much consuming a particular additive changes their personal likelihood of developing hypertension or heart disease.
The more useful implication is for research and food-safety assessment. If future independent studies produce similar findings, evaluating realistic combinations of additives could complement research that examines substances individually.
For now, cardiovascular dietary decisions should continue to be based on the broader evidence for healthy dietary patterns and established cardiovascular risk factors rather than on avoiding individual ingredients because they appeared in this study.
People should not stop prescribed medicines or make major dietary changes solely because of these observational findings. Those with hypertension or elevated cardiovascular risk can discuss individualized dietary advice with a clinician or registered dietitian.
Limitations
The study’s observational design is its most important limitation. Participants were not randomly assigned to different additive exposures, so the research can identify associations but cannot establish causation.
Residual confounding is particularly relevant. People who consume greater amounts of foods containing particular additive mixtures may differ in their overall diet, degree of food processing, lifestyle, socioeconomic circumstances, or other characteristics that influence cardiovascular health. Statistical adjustment reduces some of these differences but cannot guarantee their complete removal.
Exposure measurement was detailed but not perfect. Food formulations can change, databases may not capture every product precisely, and dietary records depend partly on participant reporting.
The cohort also was not representative of the general population. About 79% of participants were women, their mean age was 42.8 years, and they were recruited in France. Food formulations, dietary habits, and baseline cardiovascular risks can differ substantially in other populations.
The exploratory additive-interaction analyses cannot establish biological synergy.
Finally, the results have not yet been independently replicated in another cohort. This matters because the earlier studies used for comparison also came from NutriNet-Santé.
Funding and Researcher Independence
The NutriNet-Santé cohort and this research received support from multiple public institutions, research programs, and foundations, including French public-health and research organizations, the European Research Council, and the French National Cancer Institute (Payen de La Garanderie et al., 2026).
According to the paper, the researchers were independent from the funders. The funders had no role in study design, data collection, analysis, interpretation, manuscript preparation, or the decision to submit the research for publication.
The authors declared no competing interests.
Final Thoughts
The study addresses an important limitation of traditional food-additive research by examining combinations of substances that tend to be consumed together.
Among 112,050 French adults followed for a median of 7.92 years, several additive mixture patterns were associated with higher incidence of hypertension or cardiovascular disease. The associations varied by mixture and outcome, while one of the five mixtures showed no association with the cardiovascular outcomes studied (Payen de La Garanderie et al., 2026).
The findings justify further investigation, not conclusions that individual additives cause heart disease.
Independent prospective cohorts and experimental research will be important for determining whether these associations can be reproduced, whether particular substances or combinations contribute directly to cardiovascular disease, and whether mixture-based evidence should eventually play a larger role in food-additive safety assessment.
References
Payen de La Garanderie, M., Hasenböhler, A., Yvroud-Hoyos, P., Szabo de Edelenyi, F., Bourhis, L., Dechamp, N., Agaësse, C., De Sa, A., Huybrechts, I., Pierre, F., Audebert, M., Coumoul, X., Chassaing, B., Julia, C., Kesse-Guyot, E., Allès, B., Fezeu, L. K., Blacher, J., Hercberg, S., . . . Touvier, M. (2026). Food additive mixtures and hypertension and cardiovascular diseases incidence in the NutriNet-Santé prospective cohort. BMC Medicine. Advance online publication. https://doi.org/10.1186/s12916-026-05132-z
Payen de la Garanderie, M., Hasenbohler, A., Dechamp, N., Javaux, G., Szabo de Edelenyi, F., Agaësse, C., De Sa, A., Bourhis, L., Porcher, R., Pierre, F., Coumoul, X., Kesse-Guyot, E., Allès, B., Fezeu, L. K., Cosson, E., Tatulashvili, S., Huybrechts, I., Hercberg, S., Deschasaux-Tanguy, M., . . . Touvier, M. (2025). Food additive mixtures and type 2 diabetes incidence: Results from the NutriNet-Santé prospective cohort. PLOS Medicine, 22(4), e1004570. https://doi.org/10.1371/journal.pmed.1004570
Sellem, L., Srour, B., Javaux, G., Chazelas, E., Chassaing, B., Viennois, E., Debras, C., Salamé, C., Druesne-Pecollo, N., Esseddik, Y., Szabo de Edelenyi, F., Agaësse, C., De Sa, A., Lutchia, R., Ancel, P.-Y., Hercberg, S., Galan, P., Julia, C., Kesse-Guyot, E., . . . Touvier, M. (2023). Food additive emulsifiers and risk of cardiovascular disease in the NutriNet-Santé cohort: Prospective cohort study. BMJ, 382, e076058. https://doi.org/10.1136/bmj-2023-076058
