California’s Proposition 65 requires warning labels for chemicals associated with carcinogenicity or reproductive toxicity, affecting global cosmetic and fragrance formulations. While safrole has an established safe harbor level (No Significant Risk Level, NSRL) of 3 µg/day, its structural analogues, estragole and methyleugenol, lack specific safe harbor thresholds. To address this data gap using non-animal methodologies, this study evaluated a Next-Generation Risk Assessment (NGRA) workflow combining read-across and in silico modeling. Structural, physicochemical, and skin-permeation profiling supported category homogeneity (Tanimoto Tc ≥ 0.82 and comparable log Kp values) in accordance with OECD and ECHA guidelines. Multi-platform QSAR consensus (ToxTree, Danish QSAR, and VEGA QSAR) indicated that while parent phenylpropenes lack direct bacterial mutagenicity, they share a hepatic bioactivation pathway leading to mammalian clastogenicity and carcinogenicity predictions. Molecular docking with human Cytochrome P450 1A2 (CYP1A2; PDB: 2H14) identified safrole as the category’s conservative toxicological benchmark, exhibiting the highest binding affinity (−8.026 kcal/mol) and closest proximity to the catalytic HEME iron. Consequently, applying safrole’s 3 µg/day NSRL to estragole and methyleugenol provides a quantitative threshold to evaluate consumer exposure in cosmetic products, ensuring public health protection and regulatory compliance.
The increasing prohibition of animal testing for cosmetic products has driven the development of alternative approaches to ensure consumer safety. Skin sensitization is one of the most critical toxicological endpoints to evaluate, requiring rigorous assessment to ensure the safety of cosmetic ingredients. This study proposes an Integrated Testing Strategy (ITS) that combines in chemico (Direct Peptide Reactivity Assay—DPRA) and in silico approaches (a six-platform computational panel) to evaluate the sensitization potential of substances. Initially, the in chemico methodology was validated through a partial proficiency demonstration to ensure experimental reliability. Subsequently, this ITS was applied to the Baccharis trimera extract and its major marker, 3-Caffeoylquinic acid (chlorogenic acid/3-CQA). Our results demonstrate that the DPRA alone is insufficient to classify the sensitization potential of complex mixtures, as recommended by OECD guidelines. The integration of in silico data proved essential to interpret the reactivity of the botanical matrix, revealing that the sensitization potential observed in the extract does not stem solely from 3-CQA, but likely results from the synergistic contribution of more lipophilic caffeoylquinic acid isomers. This approach demonstrates that integrating experimental and computational methods is fundamental for a robust safety assessment, offering an efficient, animal-free strategy for the early screening of cosmetic ingredients and for refining the interpretation of toxicological data in complex chemical environments.