Study on preservatives and cancer has shortcomings Federal Institute for Risk Assessment issues critical assessment

Link to the study: External Link:https://www.bmj.com/content/392/bmj-2025-084917

In the study, the preservatives contained in the food were inferred from the participants’ nutrition reports. At the start of the study, the participants (a total of 105,260) were free of malignant tumours. During the study period, 4,226 of them developed cancer.

Eleven of the 17 preservatives examined were not associated with the occurrence of cancer. For the remaining six substances, there was a statistically significant association with an increased incidence of cancer. Should the result for these preservatives be confirmed, the study authors believe this would be a reason to reconsider their current regulatory status.

The BfRshort forGerman Federal Institute for Risk Assessment has conducted an assessment of the publication and, in this context, points to a number of weaknesses in the study:

• Causality questionable: The NutriNet-Santé study is an observational study that investigated statistical correlations (associations) between preservatives and the occurrence of cancer. However, statistically significant associations cannot automatically be interpreted as causal relationships.
• Multiple testing: The study examines associations between a large number of preservatives and the occurrence of cancer. This makes it susceptible to the phenomenon of multiple testing: if several statistical tests are carried out, it is to be expected that some will, by chance, yield a false-positive result. As many preservatives were investigated in this particular publication, positive associations may be due to chance.
• Confounders: A major problem in observational studies is the occurrence of confounders. These must be taken into account during data analysis. This is done, amongst other things, using statistical methods (adjustment). Despite extensive adjustment, residual confounding may remain. For example, it is difficult to separate the effect of certain preservatives from that of the food in which they are prevalent. For instance, the intake of sulphites (a preservative in wine) is linked to wine consumption. The observed cancer risk associated with sulphites may therefore be attributable more to alcohol consumption than to the preservative itself. Regular alcohol consumption is an established risk factor for certain types of cancer.
• Unclear intake: The data used to estimate the intake of preservatives (exposureClose dialogexposureExposureTo glossary) in the study was based on self-reported information from the participants. Inaccurate or incorrect answers lead to uncertainty in the estimates of the preservatives ingested through food.
• Questionable plausibility: For acetic acid and acetates, the study reports a 12 per cent (acetic acid) and 15 per cent (acetates) increased cancer risk (hazard ratios of 1.12 and 1.15 respectively). This is of the same order of magnitude as the increased cancer risk identified in the study for sorbates (14 per cent), sulphites (12 per cent), sodium erythorbate (12 per cent) and potassium nitrate (13 per cent). Acetic acid and acetates are natural components of many foods and are also produced in significant amounts by the human metabolism. This calls into question the plausibility of the associations identified for these preservatives and suggests that they should be considered of little relevance.
• Re-assessed: The preservatives addressed in the study were re-assessed by the European Food Safety Authority (EFSAshort forEuropean Food Safety Authority) as part of the programme for the re-assessment of authorised food additives. With regard to potential mutagenic (genotoxic) and carcinogenic properties, the EFSAshort forEuropean Food Safety Authority had no health concerns regarding the additives in question.

Conclusion: For the majority of the preservatives examined in the study, no association with the occurrence of cancer was found. The observed associations are subject to uncertainties and should be interpreted with caution. It is unclear whether the observed associations are the result of multiple testing. Even if the observed associations were to describe actual effects, it remains unclear whether these are truly attributable to the additives. As the authors of the publication also emphasise, the results would need to be confirmed independently. From the BfR’s perspective, this publication does not justify changing the risk assessments of the preservatives in question carried out by EFSAshort forEuropean Food Safety Authority.