Literature review of existing approaches to aggregate exposure assessments
Historically, chemical exposure assessments carried out for regulatory compliance were aimed at the derivation and demonstration of safe use of a substance contained in an individual product (Existing Substances Regulation or ESR). In the early 1990s, the concept of aggregate exposure assessment began to receive more attention and dedicated methods and tools started evolving. The development of regulatory frameworks requiring aggregation of exposure supported these initiatives. One of the policy documents that instigated guidance on aggregate exposure assessment was the US Food Quality Protection Act (1996). This act introduced the concept of pesticide safety as "reasonable certainty that no harm will result from aggregate exposure to the pesticide chemical residue, including all anticipated dietary exposures and all other exposures for which there is reliable information”. The publication of this act raised a number of science policy issues, which resulted in the development of the General Principles for Aggregate Exposure and Risk Assessment (US-EPA, 2001). The legislation was also one of the triggers for the Health and Environmental Science Institute to organise a workshop on aggregate exposure (ILSI HESI, 2000), having an objective to evaluate the methodologies currently available for aggregate exposure assessment, with an emphasis on the practical scientific issues and data requirements for pesticides. In Europe aggregate dietary exposure assessment for pesticides is also required under the Regulation on maximum residue levels of pesticides in food (European Commission (EC No 396/2005), 2005).
Currently aggregate exposure assessments are executed in a tiered manner (Delmaar JE and van Engelen JGM, 2006; Meek et al, 2011). Recent examples of consumer aggregate exposure assessments include multiple case studies illustrating approaches developed for cosmetics and personal care products, some demonstrating applicability to cleaning and household care products, with a limited number of instances developed for a wider range of consumer goods covering coatings, medical devices and food contact materials.
In 2012, the Norwegian Scientific Committee for Food Safety (VKM) published a risk assessment on vitamin A in cosmetics (VKM, 2013). This included a low tier deterministic exposure estimate derived from summing worse case population estimates of dietary and cosmetic exposure.
More refined approaches providing more realistic estimates of exposure include examples such as Dudzina et al (2015), who developed and validated a person-oriented Probabilistic Aggregate Consumer Exposure Model (PACEM) using decamethylcyclosiloxane (D5) as a showcase compound. The aggregation of exposure in the model was performed at the individual level, making use of the detailed consumer exposure factor databases on product use and co-use and biometric details for 516 Dutch adults. The model allows estimation of product contributions to total internal dose as well as stratification of aggregate exposure by route, gender and age. By comparing the doses derived in different modelling tiers with relevant human biomonitoring data (Biesterbos et al, 2015) the authors could verify the applicability of the developed probabilistic model for risk assessment. The higher tier estimates were more realistic, but still reasonably conservative than those obtained following the deterministic worst-case approach. A pilot version of PACEM has been also tested by Gosens et al (2013) who estimated aggregate exposure to four parabens from baby care products for Dutch children between 0 and 3 years old. Additional validation of the PACEM tool was performed by Delmaar et al (2014) in a diethyl phthalate case study. It is worth noting that only the baseline end-exhaled air monitoring data for D5 acquired by Biesterbos et al (2015) provided a ‘true’ snapshot of aggregate exposure, since the measurements reflected the total systemic doses received collectively via all sources and pathways. Contrariwise, the modelling case studies focused on specific categories of consumer products overlooking other potentially relevant exposure pathways (e.g. via drinking water, ingestion of dust) that may contribute to aggregate exposure.
Manová et al (2015) also evaluated aggregate consumer exposure to ethylhexyl methoxycinnamate (EHMC) via the use of personal care products (PCP). The authors adopted a probabilistic approach to modelling aggregate exposure at an individual level. The products use data for 1196 adults and children in German-speaking part of Switzerland was fed into the model together with the analytical concentration data on EHMC in PCPs. The internal aggregate exposure estimates for the studied population were below the Derived No Effect Level (DNEL) for EHMC. However, it was shown that the predicted aggregate exposure may exceed the DNEL for thyroid-disrupting effects for children aged ≤4years, when an intense short-term exposure via sunscreen during a sunbathing day is accounted for. Considering the paucity of quantitative data on transdermal penetration of EHMC and the long-term effects of endocrine disruptors, comprehensive risk assessment could not be performed. The finding of the study highlighted the need for an alignment between advances in exposure modelling and the development of reference dose values for accurate risk evaluation.
Tozer et al, (2015) developed a probabilistic aggregate exposure model, using Creme Global software, to estimate consumer exposure from several rinse off personal cleansing products containing the anti-dandruff preservative zinc pyrithione. The model incorporates large habits and practices surveys from Europe and North America, containing data on frequency of use, amount applied, co-use along with market share, and combines these data at the level of the individual based on subject demographics to better estimate exposure.
The developed models for aggregate consumer exposure assessment are also being used in skin sensitisation risk assessment, as demonstrated by Nijkamp et al (2015) who investigated the fragrance ingredient, geraniol, in cosmetics and household cleaners using the PACEM model. The survey data, underpinning the model, allowed predicting body part specific aggregate dermal external exposure at an individual level and deriving the percentage of general population at risk. The authors, however, acknowledge that ideally the risk to sensitising agents should be assessed based on the internal exposure (e.g. the amount of substance that enters the epidermis and becomes available for recognition by Langerhans cells) rather than the external dermal load. Also, the timeframe of aggregation of exposures relevant for skin sensitisation is not known and may be longer than 24 hours assumed in this study, since the available test data are highly uncertain and suggest the induction phase may occur during both the acute/peak and chronic time periods.
Along the lines of this study, Safford et al (2015) also used a probabilistic aggregate exposure model to estimate consumer exposure to fragrance materials in personal care and cosmetic products using the Creme Global software Creme Care & Cosmetics. The model is described in detail in this report in the case studies for triclosan and phenoxyethanol and so will not be described in further detail here. However, it is worth noting that the model is now used as the standard approach in the safety assessment of fragrances, which is performed routinely by the Research Institute of Fragrance Materials (RIFM) when combined with surveys of use levels gathered in collaboration with the institute's member companies.
More recently, Dimitroulopoulou et al (2015a) have estimated aggregate exposure for a range of VOCs (formaldehyde, benzene, acrolein, d-limonene, a-pinene) being emitted from cleaning and surface treatment products including all-purpose cleaners, kitchen cleaners, floor cleaners, glass and window cleaners, furniture and floor polish products, combustible products, sprays, electric and passive air fresheners, coating products for leather and textiles, hair styling products, spray deodorants and perfumes. The modelling was carried out using CONC-CPM microenvironmental (ME) model. The simulations of indoor air concentrations and calculations of inhalation exposure were predicted from a single product use as well as from simultaneous use of multiple products that were documented in the form of ‘most representative worst-case scenarios’. The predictions of aggregate exposure took into account product co-use profiles developed for over 4,000 adults split into two specific consumer groups: housekeepers and retired people in different European regions (Dimitroulopoulou et al, 2015b). The questions considered for the development of these scenarios were related to the use of consumer products in the domestic environment resulted in acquisition of the information on frequency, the amount, the time and location of product use for every single individual.
With regards to chemicals occurring in products other than cosmetics and cleaning products, Koontz et al (2006) conducted a study on modelling occupational aggregate exposure for ethylene glycol butyl ether (EGBE) and dipropylene glycol methyl ether (DPGME) from sequential application of floor stripper, floor cleaning agent and floor protective finish using PROMISE and MCCEM exposure modelling tools. Although the models were run for professional use, the input parameters used were also valid for consumer applications (e.g. AER=1/hour, the applied amount relative to treated surface area). Aggregation of internal exposure was done over all routes and across products (where applicable) by simple summation. The paper also includes basic uncertainty analysis (for DPGME exposure only) and some validation. The toxicological endpoints of both compounds were not discussed; thus, it is not clear whether aggregation of exposure across compounds (potentially acting through a common MOA) would have been beneficial. Despite its limitations, the study exemplifies nicely considerations and the level of detail required for input data to model appropriately use scenarios for aggregate exposure. One of the conclusions that can be made is that the exposure may be aggregated across consumer products that are intended for use within a specific activity (e.g. wall painting, carpet installation, house cleaning).
Overall, most aggregate assessments published to date focus on specific substances and specific types of use (food, cosmetics, etc.). These are consistent with the current state of modelling tools, availability of exposure factors data, and understanding of data correlations needed to support higher tier predictions of aggregate exposure.
Recommended approaches for High Tier Aggregate Assessment of Chemicals
Background
Aggregate exposure should be estimated using a tiered approach (Delmaar JE and van Engelen JGM, 2006; Meek et al, 2011), which begins with a rough deterministic estimation of exposure and evolves, as needed, to a more complex person-orientated probabilistic approach. This is recently described, and applied to the exposure assessment of D5 and triclosan in CEFIC-LRI project ETHZ-B7 (Bakker 2014). This report introduces the concept that data can be refined where necessary at the highest tier, by incorporating data on chemical occurrence and product market share, to give a population-based aggregate exposure estimate to an ingredient that incorporates the best available data.
Aggregate exposure assessment is becoming a consideration in safety assessments in some sectors, whereas in other consumer product categories it is deemed to be less relevant. In the food sector, it is normal practice to look at a person’s daily exposure to a nutrient or food ingredient by considering their total exposure from the diet, which is in effect the daily aggregate exposure from all food sources, though it is not referred to as aggregate exposure. In other consumer products categories, such as household products, aggregate exposure is not considered, which may be because exposure to products is low, as many products are not directly applied to the skin, or because the products are not frequent, daily use products, for example household cleaning products that are used only on a weekly basis or less.
Aggregate exposure assessment is becoming an area of interest in the sector of cosmetics and personal care products in Europe. For example, in the SCCS Notes of Guidance for the Testing of Cosmetic Ingredients and their Safety Evaluation, the aggregate exposure assessment of preservatives is estimated using a simplistic approach of adding deterministic exposures from all the individual product types in which the chemical might be present (SCCS, 2012). In addition, the SCCS has requested the consideration for aggregate exposure for a number of chemicals including citral, farnesol, and phenylacetaldehyde, silver (SCCS, 2008), ethyl lauroyl arginate (SCCS, 2014), cetyl pyridinium chloride (SCCS, 2015) and decamethylcyclopentasiloxane (D5). Furthermore, in the European Cosmetics Regulation (EC 2009), substances classed as carcinogenic, mutagenic or toxic to reproduction (aka CMR) class 1A/1B should be assessed for total (aggregate) exposure, considering their simultaneous presence in cosmetics, foods, medicines, and in products legislated under REACH (i.e., Registration, Evaluation, Authorisation and Restriction of Chemicals). Therefore, there is a requirement to assess aggregate exposure across consumer product categories, although no published guidance is available.
There are no standard methods recognised for aggregate exposure assessments, although it is recommended to estimate it using a tiered approach (Delmaar JE and van Engelen JGM, 2006; Meek et al, 2011), which begins with a rough deterministic estimation of exposure and evolves to a more complex person-orientated probabilistic approach. Deterministic additive methods, such as the SCCS preservative method (SCCS, 2012), assume that everybody in the population uses all the products each day, and that all of the products contain the chemical of interest, which is not a realistic scenario. This technique may be sufficient for a low tier screening level assessment, or for chemicals with a wide margin of safety, but as it does grossly exaggerate the aggregate exposure, a more refined approach will be needed for some risk assessments. An approach has been described for refining a deterministic aggregate exposure assessment to the paraben preservatives (i.e., methy-, ethyl- and isopropyl paraben) in personal care products by incorporating data on co-use and non-use patterns of product usage, and the occurrence of the ingredient (Cowan-Ellsberry and Robison, 2009). This has led to considerable refinement in exposure (51-92%). Co-use is the term describing the combination of products used by the same subject and by applying the co-use statistics, a more refined aggregate exposure model can be developed that better reflects population exposure. Since product use data are readily available for many cosmetic products (Hall et al, 2007, 2011; Loretz et al, 2005, 2006, 2008) the co-use approach offers a practical method to refine aggregate exposure assessments.
Recently co-use data from European and US subjects has been incorporated into high tier exposure estimates for chemicals in consumer products using subject-oriented probabilistic models with Creme Global software (Tozer et al, 2015, Comiskey et al, 2015, Safford et al, 2015).
Another refinement to more accurately reflect aggregate exposure estimations in populations is the incorporation of chemical occurrence data or market share data, which describe the likelihood the chemical is present in a product, since only the consumers using products containing the ingredient will be exposed. This factor (usually expressed as a value between zero and one or a percentage) can be used in probabilistic modelling to estimate the likelihood of co-exposure to a given substance that is potentially present in a given category. Incorporation of chemical occurrence data into exposure assessments is being done already in the area of food safety (Mistura et al, 2013). To give a cosmetics example, consumers using only "fragrance/perfume free” cosmetics would not be exposed to perfume raw materials through the use of these products. When chemical occurrence data is combined with reliable market share data for the products it can be used to determine the probability of exposure, which can be incorporated to refine the exposure assessment. Incorporation of chemical occurrence data including non-use data into exposure modelling brings refinement by discounting exposures where the chemical is not present in the product of interest. In Tozer et al, (2015) the exposure of zinc pyrithione was modelled by incorporating chemical occurrence data on the proportion of the population who are users of antidandruff shampoos, as zinc pyrithione is only present in anti-dandruff shampoos. For infrequently used substances, what is often the most conservative assumption in an assessment is that a substance is always present in every product category it can be used in. For aggregate exposure assessments, this assumption has an additive effect giving rise to a very conservative estimate of exposure.
As there is little guidance on how to refine high tier exposure assessments using chemical occurrence data and more realistic data on the concentrations of chemicals in product, two case studies are presented to demonstrate the technique, using triclosan and phenoxyethanol, where the exposure assessments are conducted at tier 0, 1 and 2. As tier 3 requires very detailed exposure input data, such as raw data sets on specific product use including ingredient concentration and presence in product, which can be difficult to attain, these examples will only be taken to Tier 2:
Tier 0 Qualitative Exposure Assessment
The purpose of the tier 0 exposure assessment is to provide a preliminary overview of all possible exposure sources, pathways and routes for the chemical of interest, in order to determine whether an aggregate exposure assessment is appropriate.
Tier 1 Worst-Case Scenario Assessment
The aim of the tier 1 assessment is to determine a realistic upper bound of the aggregate consumer exposure to the chemical in a population.
Tier 2 probabilistic assessment
The aim of the tier 2 assessment is to determine more realistic estimates of aggregate consumer exposure to the chemical, by increased use of measured data, using probabilistic methods.
To note that these case studies are not intended to provide definitive exposure assessments for these chemicals, but rather have been selected for illustrative purposes to demonstrate the refinement techniques.