A study published on February 2, 2017 in the peer-reviewed journal Food and Chemical Toxicology proposes a strategy for identifying the substances of highest concern among the non-evaluated substances potentially present in food contact materials (FCMs). Melissa Van Bossuyt and colleagues from the Scientific Institute of Public Health and Vrije Universiteit Brussel, Brussels, Belgium, suggested using (quantitative) structure-activity relationship ((Q)SAR) models as a means for evaluating the potential toxicity of large numbers of chemicals. Based on the obtained ranking of potential toxicity, the substances can then be prioritized for an in-depth safety evaluation. The presented case study focused on substances from printed paper and board FCMs, and evaluated their potential mutagenicity as assessed by the in vitro Ames test. This endpoint is well characterized, with several (Q)SAR models available.

The non-evaluated substances to be screened were selected from the inventory containing all substances which may be used in printed paper and board FCMs, compiled by the same group in 2016 (FPF reported).  Out of the 6,073 unique substances present in the inventory, 4,690 are non-evaluated compounds. From these, 1,769 single substances were screened in the current study. The remaining 2,921 non-evaluated substances could not be assessed with the proposed strategy because of their chemical structure (e.g. polymers, mixtures, complexes, inorganic substances). Furthermore, 46 more substances had to be excluded because no definite chemical identification (CAS or SMILES) could be located for them. Therefore, in the final analysis, 1,723 substances were evaluated.

The global (Q)SAR models for Ames mutagenicity included two freely available programs, Toxtree (SAR) and VEGA platform (two QSAR and one SAR model), as well as two commercial models, Derek Nexus™ (SAR)  and Sarah Nexus™ (QSAR). In addition, two local QSARs for aromatic azo compounds, CORAL (freely available) and istKNN (commercial) were used. The output of each model was coded as positive, negative, or undefined, according to set rules. The compounds were further checked for being present in the Flavours, Additives and food Contact materials Exposure Tool (FACET) database, to obtain a first indication of their actual use.

In the four global (Q)SARs, up to 366 compounds were predicted to be mutagenic (366 in Toxtree, 350 in VEGA, 255 in Derek and 229 in Sarah). Between Toxtree and VEGA, 269 substances overlapped, and 119 compounds overlapped between Derek and Sarah models. The authors suggested that the non-overlapping compounds with contradictory prediction results should be examined in detail, as this could help revealing chemical classes in need of improvement with regard to prediction of Ames mutagenicity.

Overall, from the 1,723 studied substances, 106 were predicted mutagenic by all four (Q)SAR tools, and 572 were predicted non-mutagenic. A substantial part of the remaining 1,045 study substances were either positive in at least one of the models, or negative in all tools but outside of the applicability domain in at least one of them, therefore these were assigned to the ‘undefined’ category.

The 106 compounds predicted mutagenic by all four tools were considered of highest priority for further investigation of potential mutagenicity. The majority of these compounds (99) were components of printing inks, and most of them are found in the FACET database, suggestive of an actual use. Further, the authors showed that the use of local QSARs allowed further refinement of predictions for specific chemical classes, shown on the example of 25 compounds with an aromatic azo bond found among the 106 substances of highest priority.

Among these 106, 53 compounds already had in vitro Ames data available, which could be located in the training sets of the used (Q)SAR tools. Hence, the authors suggested that these 53 compounds should now be followed with in vivo testing, followed by migration testing in case the mutagenic potential is confirmed in vivo. Of note, physico-chemical properties of these compounds suggest that they are likely to be bioavailable after oral intake. For another 53 substances from the 106 chemicals of highest priority, no experimental in vitro data could be found, and the authors suggested that these should be urgently investigated further. Most of these compounds are also likely to be bioavailable after oral intake.

The authors concluded that their prioritization strategy can also be applied to other groups of chemicals, such as impurities or degradation products found in chemical formulations, or industrial chemicals lacking a (recent) safety evaluation.


Van Bossuyt, M., et al. (2017). “(Q)SAR tools for priority setting: a case study with printed paper and board food contact material substances.Food and Chemical Toxicology (published February 2, 2017).

Van Bossuyt, M. et al. (2016). “Printed paper and board food contact materials as a potential source of food contamination.” Regulatory Toxicology and Pharmacology 81:10-19.