High-throughput human exposure estimation and data curation methods for impact assessment and chemical substitution

Our modern society relies on chemical substances to enable the diversity and functionality of the wide range of consumer products that we use on a daily basis. However, chemical ingredients across consumer products and their life cycles have the potential to reach and expose humans and ecosystems, and potentially pose adverse effects on their health. Characterizing chemical exposures and quantifying related potential impacts on human and environmental health is thus crucial for a sound management of chemicals in consumer products. Various tools already exist for assessing chemicals exposures and their effects, such as approaches for alternatives assessment and chemical substitution, risk screening and prioritization, life cycle toxicity characterization, and consumer exposure assessment. However, available tools have various limitations and are currently not able to efficiently and consistently quantify exposures for tens of thousands of marketed chemicals used in thousands of consumer products. Thus, high-throughput methods are urgently needed to consistently assess the broad range of chemicalproduct combinations. In addition, to deliver reliable results, exposure and impact assessment tools require a considerable amount of highquality data covering physicochemical properties, exposure, and toxicity effect information that are consistent with the boundary conditions of the different assessment contexts. Such data, however, are currently missing for the wider range of chemicals found in consumer products.

The work presented in this PhD thesis addresses these research gaps by focusing on three research objectives: (i) to develop operational high-throughput exposure, hazard, and risk screening tools for chemicals in consumer products with a focus on plastic-based products, (ii) to develop models for complementing missing pathways to estimate exposure and related health risks for chemicals in children’s products, and (iii) to develop semi-automated data curation methods to provide chemical data from available data sources as input for modeling human and ecological exposures and related health effects for use in different assessment tools, such as life cycle impact assessment and risk screening.

After an introductory chapter, the second chapter presents an overview of the state of knowledge of chemicals in plastic materials and provides a dataset of 𝑛 = 6010 chemicals reported in plastics. Challenges and gaps in assessing such chemicals in terms of their impacts on humans and the environment via their specific product application are discussed, and possible solutions and ways forward to address these challenges and gaps are provided. As a next step, a set of 𝑛 = 1518 chemicals of concern (CoCs) are identified using available regulations and hazard classifications as criteria, highlighting that a relevant amount of the chemical substances found in plastic materials might pose non-negligible adverse effects on humans and ecosystems. In the third chapter, the focus moves from plastic materials in general to plastic-based children’s products as specific product application. Chemical emissions from plastic toys and subsequent human exposures are estimated using a series of mass balance models integrated into a coupled near-field/farfield exposure assessment framework. Applying these methods, 𝑛 =126 CoCs in plastic toys are identified considering current regulations, exposure, and related cancer and non-cancer risk as chemical prioritization criteria. The last part of this chapter presents a new model developed for quantifying chemical exposure estimates for children via mouthing as a highly relevant yet often neglected pathway.

The second part of this PhD thesis presents novel methods developed to address the critical need for chemical input data in support of providing reliable and interpretable exposure and toxicity results for thousands of chemical substances. The fourth chapter introduces two different methods for semi-automatically curating and selecting chemical data from online data sources that are used in human and environmental exposure and toxicity characterization models. The first method aims to derive a harmonized dataset for specific assessment contexts where the highest-quality data are strictly needed. The second method aims to be more inclusive by systematically harmonizing all available data in a given data source to arrive at representative model input values with quantified uncertainty for application in various screening-level contexts. The fifth chapter presents a semi-automated workflow to derive points of departure (PODs) as input for quantifying human toxicity effects, applied to in vivo data from the ToxValDB, for 𝑛 = 10,055 chemicals. The recommended PODs drastically increase the coverage of substances for which human toxicity values can be derived and have potential applications across numerous chemical assessment frameworks.

In conclusion, the work conducted in this PhD project contributes to research within the emerging field of high-throughput exposure and computational toxicity assessment. The proposed quantitative and comparative methods support the evaluation of chemicals emitted along a wide range of consumer product life cycles in terms of multi-pathway exposure and related effects on humans and ecosystems as basis for science‐based and informed decision support.