In order to enhance hazard identification and risk ranking most relevant hazards for infant food were identified: 34 microbiological hazards and 101 chemical hazards. Specific characteristics and knowledge rules for these hazards have been collected and stored in a data collection repository. This information can be used to further identify the hazards relevant in specific food products. Knowledge rules and a decision tree to classify the hazards into different empirical rankings was devised. Criteria for microbiological and chemical hazard risk ranking were defined. Data collected on the microbiological and chemical hazard databases have been used to develop the risk ranking prototypes for both types of hazards.
In order to improve hazard control and mitigation, SAFFI focused on mild processes and home practices. In addition to experiments showing that high pressure processing (HPP) mitigate chemical hazards in fruit puree, predictive models were used to define HPP conditions to achieve the performance criteria of non-thermal pasteurization. SAFFI also showed that pasteurization could be achieved with thermal radiofrequency (RF) and developed a predictive tool to simulate the thermal profile and associated microbial inactivation during infant food RF processing. Finally, high pressure thermal processing (HPTP) of vegetable-based infant foods enabled to inactivate spore-forming bacteria while limiting the neo-generation of chemicals like furans. At home, post-reheating stirring was shown to be particularly efficient for mitigating furans in vegetable based infant foods.
In order to enhance the detection and discovery of chemical hazards, several bio- and chemoanalytical methods have been developed and validated to ensure high throughput, cost-effective and robust monitoring of a wide range of priority chemicals. The sample pooling approach was also shown to drastically improve the efficiency of regulatory surveillance and industrial self-monitoring of priority hazards. For the discovery of unsuspected/unknown hazards, a 5-step action plan was proposed allowing safety assessment of infant foods from CALUX bioassay measurements while non-targeted methods coupling high-resolution mass spectrometry and bioinformatics were developed for the possible identification of suspect chemicals.
In order to enhance the detection of microbial hazards and the prediction of their behaviour, the presence, distribution and prevalence of target foodborne was analysed in samples collected over an extended period to capture seasonal fluctuations. Targeting infant cereals, samples were collected from raw materials, intermediate products, finished products, production environment to determine the spatial distribution and sources of contamination and to understand the overall structure of the microbiota in the substrates. In parallel, in order to promote prediction of microbial behaviour, Listeria monocytogenes was employed as a model organism and the mild acid adaptation and subsequent increased robustness to lethal acid pH were tested both in vitro and in situ in order to identify transcriptome and volatolome derived biomarkers.
An integrated, upgradeable Decision Support System (DSS) has been developed for assessing hazards in infant food chains. Initial stages focused on gathering data requirements and developing model prototypes across SAFFI activities. Data profiling mapped key variables, created metadata, and identified dataset commonalities, facilitating seamless integration of diverse models. Conceptual and logical data models were then built using entity relationships and data flow diagrams, and workshops and surveys provided feedback for refinement. The resulting beta DSS enables hazard identification, control, and detection through a structured SQL database and visualization tools. Designed for scalability, the DSS supports data uploads via templates, allowing future expansion beyond infant food chains to address broader food safety hazards.
In addition to the DSS and 7 key exploitable results, SAFFI results have been described in over 70 peer-reviewed open access scientific articles and in conference presentations, webinars, workshops, courses, datasets and quizz.
