TL;DR
This study presents a comprehensive hazard assessment of over 500 antioxidants (AOs), revealing that a significant portion exhibits concerning environmental persistence and toxicity. The findings highlight the urgent need for more experimental data and the development of safer AO alternatives, particularly for synthetic compounds.
What Problem Does This Paper Address?
Antioxidants (AOs) are ubiquitous in modern life, found in everything from plastics and fuels to pharmaceuticals and food. Their widespread use means they inevitably enter the environment and can be detected in water, soil, sediment, and even human biosamples. While AOs are designed to prevent oxidative damage, their own environmental fate and potential toxicity are not fully understood. This poses a challenge for regulatory bodies and researchers aiming to assess and mitigate risks associated with these chemicals. Specifically, there's a critical data gap regarding the comprehensive hazard profiles of the vast array of AOs in use, including their persistence, bioaccumulation potential, toxicity, mobility, and the formation of potentially harmful transformation products (TPs). This paper aims to fill this gap by systematically evaluating a large number of AOs and identifying those that warrant the most concern.
Methods
The researchers employed a multi-pronged approach to assess the hazards of AOs. First, they curated a comprehensive list of over 500 natural and synthetic AOs from scientific literature and regulatory databases. This list was then structurally grouped to identify common chemical classes. The core of the study involved a hazard assessment based on five key properties: recalcitrance to mineralization (not readily biodegradable, NRB), persistence (P), bioaccumulation (B), toxicity (T), and mobility (M).
To overcome significant data gaps, they combined existing experimental data from various sources (ECHA, EPA ECOTOX, safety data sheets, PubChem, IARC, NITE) with in silico predictions. For properties with limited experimental data, they utilized predictive tools like OPERA, EpiSuite, and Toxtree. Crucially, experimental data was prioritized over predictions when available. They also investigated the potential toxicity of transformation products (TPs) by using tools like enviPath and ECOSAR. Finally, they used dimensionality reduction techniques (PCA and t-SNE) to visualize structural similarities and explore relationships between chemical structure and hazard profiles.
Datasets
The study utilized several publicly available datasets and databases for hazard information:
- ECHA database: For experimental and predicted data on AO properties.
- USEPA ECOTOX database: For experimental data on aquatic toxicity.
- Safety Data Sheets (SDS): Provided by vendors for experimental data.
- PubChem: For both experimental and predicted data.
- International Agency for Research on Cancer (IARC) Classified Agents List: For experimental data on carcinogenicity.
- Japanese GHS classifications list: For experimental data.
- Peer-reviewed literature: For experimental data.
Additionally, they employed in silico prediction tools (OPERA, EpiSuite v 4.11, Toxtree 3.1.0, BIOWIN, Körner et al. prediction workflow, enviPath, ECOSAR) to fill data gaps. The study also generated a comprehensive list of AOs, their TPs, and their hazard assessments, which is available as a supplementary file.
Key Insights and Ideas
- Widespread Hazard: A significant majority of AOs (60%) are classified as both not readily biodegradable (NRB) and toxic (T). This suggests that many commonly used antioxidants pose a dual threat to the environment and potentially human health.
- PBT/PMT Concerns: Nearly 5% of the AOs studied meet the criteria for being Persistent, Bioaccumulative, and Toxic (PBT) or Persistent, Mobile, and Toxic (PMT) substances, which are of high concern under EU regulations. This highlights a critical subset of AOs requiring immediate regulatory attention.
- Synthetic Origin of High-Concern AOs: All PBT and PMT classified AOs are synthetic, and predominantly primary antioxidants (phenolic or amine compounds). This strongly suggests that the development of safer alternatives should focus on synthetic routes and chemical classes.
- Transformation Products (TPs) are a Major Concern: The study emphasizes that AOs can form toxic TPs, which can be more hazardous than the parent compound. This is particularly true for phenolic and amine structures, underscoring the need to assess not just the parent AO but also its degradation products.
- Natural AOs Show Promise (with caveats): Natural or nature-identical AOs, as well as certain secondary AOs, are less frequently classified as hazardous. While this suggests they could be safer alternatives, a full hazard and risk assessment still requires considering production volumes and usage patterns.
- Data Gaps Hamper Assessment: The study reveals substantial data gaps, particularly for persistence (P), bioaccumulation (B), and mobility (M) properties. This necessitates the use of in silico predictions, but also highlights the critical need for more experimental data to refine hazard assessments and regulatory decisions.
- Structural Complexity and Hazard: The t-SNE visualization reveals clustering of hazardous AOs, with primary phenolic and amine structures being prominent. Conversely, safer AOs tend to be natural primary AOs or secondary AOs with ester or sulfur-containing groups. However, even small structural differences can lead to vastly different hazard classifications, emphasizing the sensitivity of these assessments.
Key Results
- 60% of AOs are NRB and T: This indicates a broad environmental and toxicological concern.
- 31.6% of NRB and T AOs are also P: This subset represents a significant portion of potentially problematic compounds.
- 4.8% are PBT and 2.1% are PMT: These categories align with EU high-concern substance criteria.
- Only 19 AOs identified as non-hazardous: This small number underscores the pervasive nature of AO hazards.
- Synthetic AOs dominate PBT/PMT categories: All PBT/PMT AOs are synthetic.
- Data availability: Experimental data is most available for T (74%) and NRB (44%), with significant gaps for P (0.05%), B (15.7%), and M (29.5%).
- TP toxicity: 186 out of 506 predicted TPs may harm aquatic species, stemming from 98 parent AOs.
Limitations and Future Work
Limitations: * Reliance on in silico predictions: While necessary to fill data gaps, in silico predictions introduce uncertainties and can lead to systematic errors (e.g., misclassification of phosphite structures). * Experimental data limitations: For reactive AOs, experimental toxicity data can be uncertain due to the presence of both parent compounds and TPs. * Limited structure-hazard analysis: The current dataset, despite its size, is not comprehensive enough for a fully detailed structure-hazard analysis, especially concerning the interplay of electronic properties (HOMO-LUMO gap) and hydrophobicity (log Kow) with toxicity. * TP prediction uncertainties: While enviPath and ECOSAR were used, the toxicity assessments of predicted TPs carry inherent uncertainties, and only a fraction of predicted TPs are actually found in the environment. * Use case data limitations: Publicly available data on AO use cases is insufficient to fully link functionality to hazard profiles.
Future Work: * Generate more experimental data: This is crucial for refining hazard assessments, especially for AOs currently estimated as non-hazardous, and for machine learning applications. * Develop safer AO alternatives: The study strongly advocates for research into alternative AO structures with more favorable safety profiles. * Further investigate TP formation and toxicity: Future research should focus on identifying the structures and hazards of TPs, particularly those derived from primary AOs. * Utilize Novel Approach Methodologies (NAMs): The paper suggests using NAMs like cumulative toxicity equivalents (CTE) and persistent toxicity equivalents (PTE) for a more effect-based determination of AO and TP hazards. * Conduct comprehensive structure-hazard analyses: With more data, more rigorous analyses exploring the relationship between molecular properties and hazard profiles can be performed.
Summary updated