Skip to main content

Assessing safety of food packaging through computational toxicology

Periodic Reporting for period 1 - SafePack (Assessing safety of food packaging through computational toxicology)

Reporting period: 2018-03-16 to 2020-03-15

Food packaging is generally recognized as contributing directly to product safety through prevention of biological (pathogens) and chemical contaminations. In addition, packaging indirectly impacts food safety by providing consumers with health related information on ingredients (e.g. nutritional, composition and allergen information), mode of preparation and shelf-life. It is important to ensure to packaging does not act as a major source of chemical exposure. Indeed, the application of continuously improving analytical methods has revealed that many chemicals may migrate from the food contact material into the food, resulting in undesirable consumer exposure. For most of these chemicals, no toxicological data is available. The most consistent data gap deals with migrating substances, which originate from impurities in starting raw materials and from reaction and degradation of substances formed during manufacturing (the so-called Non-Intentionally Added Substances, NIAS).
In this context, there is an urgent need to acquire more insight on the actual safety concern associated with the thousands of chemicals associated with food contact materials (FCM). It is imperative to avoid the use of laboratory animal experiments in such studies. Accordingly computational toxicology (a branch of toxicology concerned with the development and use of computer-based models to predict toxicological end-points, also called in silico methodology), represents a promising solution and is increasingly applied by academic and regulatory scientists to cope with the data gap in establishing safety of chemicals. Computational models have become popular in 2006 since the introduction of regulations, such as the European REACH regulation. Nowadays their use is strongly encouraged by regulatory authorities and agencies all over the world for reducing animal studies.

This project proposes a proactive approach to address the data gaps identified above, by proposing and applying a strategy to study rapidly and cost-efficiently the level of safety concern of large sets of food contact chemicals using in silico methods. Following a step-wise approach mimicking the standard chemical risk assessment workflow (exposure assessment, hazard identification, hazard characterization and risk characterization), in silico predictions were used for filling the experimental data gaps related to the key toxicological endpoints needed to perform food safety evaluation (mutagenicity, carcinogenic potency, chronic toxicity, developmental and reproductive toxicity). Then, applying the Margin of exposure approach (MoE, the ratio between predicted toxicity value and exposure estimate) will allow establishing the level of safety concern for chemicals in food without the need for toxicity testing.
The approach was applied on a compiled large set of ~3,400 curated food packaging chemicals. First, toxicity endpoints such as mutagenicity, developmental, reproductive and chronic toxicity [lowest-observed-adverse-effect level (LOAEL)] and carcinogenic potency (TD50), were sequentially screened, integrating for each single endpoint different models based on different algorithms. Next, individual predictions were evaluated in order to identify the most relevant/conservative toxicological value, to be then compared with exposure through a MoE approach:

Toxicity endpoints and quantitative reference values: All chemicals in the dataset were screened for Ames mutagenic potential by applying a previously developed integrated strategy (1). For mutagenic chemicals carcinogenic potency (TD50) was predicted. All chemicals in the dataset were then processed to predict chronic toxicity (LOAEL) using models based on different algorithms. A strategy for translating developmental and reproductive toxicity (DART) alerts in quantitative values has been developed on a small subset of food contact chemicals and published in open access (2).

Exposure: Exposure was calculated resulting from a theoretical migration level in food of 10 ppb and a food intake of 1 kg for a 60 kg individual (3). This level of 10 ppb is widely used as a pragmatic cut-off to prioritize management of migrating chemicals without toxicological information. This approach was compared with a less conservative one based on the application of Threshold of Toxicological Concern (TTC) Cramer Class III.

Margin of Exposure: A margin of exposure (MoE) was applied to investigate the level of safety concern. In the present work, the MoE refers to the ratio between a predicted toxicological endpoint (e.g. lowest observed adverse effect level for non-mutagenic chemicals and TD50 for mutagenic ones) and exposure. The size of the MoE determines the safety concern. It reflects the distance between a dose required to produce toxicity and exposure. So, a big MoE is associated with high safety confidence. There are ways to determine the minimum MoE required to ensure safety (4). Based on the application of this approach more than 95% of chemicals investigated in the present study were compatible with safety when present in food at a level of 10 ppb. For the remaining ones a deeper analysis will be performed to confirm or discard the presence of concern.

Each step of the strategy has been applied and validated to real case studies related to packaging safety assessment sequentially or independently, depending on needs and purposes. The work done has been disseminated via oral and poster communications in a number of conferences in the area of toxicology and risk assessment. The overall study results had been proposed and accepted for oral communication at the international QSAR2020 conference in the area of computational methods, which was cancelled due to Covid crisis.

Most of the predictive models developed and applied are freely available and/or implemented into the publicly available VEGA platform on the VEGAHUB website (www.vegahub.eu). Since VEGA is broadly used in different regulatory contexts and by academic and industry experts in different fields, this implementation will offer a broad access to the strategy. Furthermore, several additional models used and developed internally at Nestlé have been made available within the publicly available online LAZAR platform (https://lazar.in-silico.ch/predict). Within Nestlé this work is/will be applied on real case studies and will strengthen the collaborations between Nestlé and packaging suppliers.

1. Manganelli, S. et al. (2018) Integrated strategy for mutagenicity prediction applied to food contact chemicals. ALTEX 35,169-178. doi: 10.14573/altex.1707171
2. Manganelli, S., Schilter, B., Scholz, G., Benfenati, E., & Piparo, E. L. (2020). Value and limitation of structure-based profilers to characterize developmental and reproductive toxicity potential. Archives of Toxicology, 94:939–954.
3. EFSA (2016). Scientific opinion on recent developments in the risk assessment of chemicals in food and their potential impact on the safety assessment of substances used in food contact materials. EFSA Journal 14: 4357, 28 pp. doi:10.2903/j.efsa.2016.4357
4. Schilter B. et al. (2014). Establishing the level of safety concern for chemicals in food without the need for toxicity testing. Regul Toxicol Pharm 68, 275-296
The overall strategy developed will be published in open access and will serve as a tutorial to be applied by industries and this will warrant a broader exploitation of the project outcomes.
assessing safety of food packaging through computational toxicology