Involved in the project were researchers from Wageningen University, the University of Veterinary Medicine Budapest, the Syreon Research Institute, and the Czech University of Life Sciences. The full publication detailing the researchers’ development of CHEFS and trends identified in their analysis can be accessed here.

The CHEFS database can be accessed here.

Analysis of the Data Reveals Food Safety Trends

Using their CHEFS database, the researchers examined trends in food safety monitoring between 2000 and 2024. Of the 392 million analytical results in the CHEFS database, 0.025 percent (97,294) were non-compliant. Sampling strategies represented included random sampling (53.1 percent), risk-based sampling (23.4 percent), suspect sampling (5.3 percent), and “other” sampling (18 percent).

Overall, the number of analytical results increased between 2000 and 2024, except for a dip in 2020 and 2021 attributable to the COVID-19 pandemic. Not only has the number of analytical results grown over time, but the quality of the metadata reported to EFSA has improved; for example, the percentage of samples with an unknown origin decreased tenfold between 2017 and 2024.

The most frequently monitored hazards were toxic heavy metals and mycotoxins in the chemical contaminant hazard category; organophosphates chlorpyrifos, diazinon, and pirimiphos-methyl in the pesticide residues hazard category; and antibiotics erythromycin, danofloxacin, and doxycycline in the veterinary medical product residues (VMPRs) hazard category. Chemical contaminants had the highest rate of non-compliant results, followed by pesticide residues and VMPRs. Even despite the near-total use of selective sampling strategies for VMPRs, the hazard category had the lowest non-compliance rate, suggesting the effective regulation of veterinary drug use for food animals in the EU.

Of the foods most frequently analyzed, milk and dairy products were most often found to be above legal limits in the chemical contaminant group, animal feed was most often above legal limits for pesticides, and cereal and grains were most often above legal limits for VMPR.

The number of analytical results reported to EFSA differed greatly between countries. Germany and France had the highest reporting rates, which are also the second and first most productive food-producing countries in the EU, respectively. On the other hand, in countries like Poland and Spain, analytical results reported to the EU were relatively few, considering their comparatively high food production volumes.

Additionally, the analysis showed that contaminated samples more often originated from countries outside of the EU.

The Importance of Traceability for More Useful Data

The researchers underscored the need for better traceability to meaningfully connect the CHEFS database with external datasets, because it is crucial to have accurate metadata on the origin of food products. Sometimes, country of origin is missing from entries in the database or it is unclear whether country of origin refers to the place of production, processing, packaging, import, or even sampling location. Accurate origin information would help identify sources of contamination and enable targeted interventions. Knowing the date of an analytical result could also add great value, making data usable for study of climate conditions on contaminants of crops, for example.

In the future, the researchers envision CHEFS being integrated with external datasets (e.g., climate data, economic and geopolitical indicators, and legislative records) to investigate indirect influences on food safety; for example, associations between changing temperatures and humidity on the presence of toxins or the prevalence of a certain contaminant following a regulatory adjustment to legal limits. Additionally, CHEFS shows promise for integration with comparable food safety monitoring systems from across the globe, enabling cross-regional comparisons and strengthening global food safety surveillance. 

Opportunities for AI Applications in Food Safety Research and Policymaking

CHEFS also opens up new opportunities for advanced AI applications in food safety research and policymaking. The Rapid Alert System for Food and Feed (RASFF) is an important source of food safety data in Europe, containing notifications on food and feed that pose a risk to public health. Although the sensitive nature of RASFF data makes it not feasible to include in CHEFS, the researchers see an opportunity to use an AI approach called federated learning to compare CHEFS and RASFF data. Federated learning is a type of machine learning in which models are trained across multiple decentralized devices or servers holding local data, without requiring the exchange of sensitive data.