1. Introduction
Children are uniquely susceptible to the myriad of environmental toxicants they are exposed to throughout development [
1], many of which have not been fully evaluated for developmental, neurological, and other toxicities [
2]. Consumer products represent an important exposure source for many toxicants due to their intended uses, which lead to direct contact with children [
2,
3,
4,
5,
6]. Examples of chemicals found in children’s products include reports of phthalates in baby bottles [
4], and brominated flame-retardants and lead in toys [
5,
7]. In addition to the extensive array of chemicals found in children’s products, the unique ways in which children interact with their environments and their increased biological susceptibility contribute to concerns about potential health impacts. Hand-to-mouth behavior is common among young children and increases the time a product may be in a child’s mouth, consequently, increasing oral exposure potential [
8]. Children also spend more time on or near the floor [
8], increasing exposure to inhaled or ingested house dust, which can act as a reservoir for chemicals often derived from consumer products [
9,
10]. Furthermore, because of their small body size, the dose associated with these exposures is proportionately greater than the dose adults receive [
8].
In addition to higher potential exposure, children also lack fully developed organ systems and detoxification pathways, greatly increasing their biological susceptibility to toxicants. Examples of this increased susceptibility include the adverse neurodevelopmental impacts of early life exposure to lead [
11] and mercury [
12]. It is estimated the 5%–20% of neurobehavioral disorders are attributable to environmental chemical exposures [
13]. Many of the effects of developmental exposure to toxicants can persist throughout the lifetime, limiting children’s abilities to reach their full potential. This has significant health and economic impacts. As of 2002, the United States’ annual cost for environmentally attributable neurobehavioral disorders was $9.2 billion [
13].
In the United States, the Consumer Product Safety Improvement Act of 2008 (CSPA) limits the use of some hazardous chemicals, including six phthalates, lead and cadmium in children’s products. Lead is not permitted in children’s products in concentrations greater than 100 ppm for total lead and 90 ppm for surface coatings. Three phthalates; diethyl hexyl phthalate, dibutyl phthalate (DBP) and butyl benzyl phthalate concentrations are restricted to no more than 1000 ppm per individual phthalate in children’s toys and product designed to care for children under age three. Diisononyl phthalate, diisodecyl phthalate and di-n-octyl phthalate are restricted in concentrations greater than 1000 ppm per individual phthalate in children’s toys that can be placed in a child’s mouth and in products designed for care of children under age three. While these laws help improve product safety, the permissible limits per phthalate are still relatively high and products may contain multiple restricted phthalates. Additionally, the Consumer Product Safety Improvement Act narrowly defines children’s products, excluding clothing, footwear and cosmetics. These factors have limited the Act’s effectiveness in protecting children.
In response to concerns over children’s exposure to hazardous chemicals found in consumer products, Washington State’s Department of Ecology (Ecology) implemented CSPA. Enacted shortly before the Consumer Product Safety Improvement Act of 2008, CSPA imposes more stringent regulatory limits on the concentrations of lead, cadmium and phthalates in children’s products sold in Washington State. Under CSPA total phthalate concentration in children’s products must be under 1000 ppm. Additionally, CSPA requires that children’s product manufacturers report the concentration range for 66 chemicals of high concern to children in any child’s product sold or manufactured in Washington State. Chemicals reported under CSPA were selected based on toxicity and potential for children’s exposure. Ecology established the 66 chemicals for required reporting based on a multi-phase prioritization process that highlighted carcinogenicity, reproductive and developmental toxicity, and endocrine disruption as toxicity endpoints [
14]. Phthalate and cadmium concentrations are reported under CSPA; however, because lead is tightly regulated at the federal level, it was not included as a chemical of concern under CSPA’s mandatory reporting requirement.
As of September 2015 there were over 33,000 reports in the CSPA database [
15]. Products reported include toys, children’s cosmetics, children’s jewelry, children’s clothing, child car seats and other products related to childcare. Within the CSPA database, products are classified in a hierarchical system with “segment” being the broadest and “brick” being the narrowest category. Segment examples include arts and crafts, baby care, beauty/personal care and clothing. Chemical concentrations are reported to fit within one of six ranges (<100, 100–500, 500–1000, 1000–5000, 5000–10,000 and >10,000 ppm). Along the concentration range, manufacturers report the function of the chemical in each product from a list including, coloration, pigments and dyes, surfactant, plasticizers and even chemicals found in the product that serve no function and are contaminants. Manufacturers also report whether the product is designed for a child under age three, or age three and above. This information is useful for characterizing how children’s exposures to these products may occur and in what capacity.
However, interpreting the complex, multilayered CSPA database requires an innovative framework that considers the lifestage-specific toxicity of the chemical and potential exposure routes. We have constructed a framework for the incorporation of these important factors into the toxicological prioritization of the CSPA data. The goals of this paper are to develop a framework for the prioritization and identification of high priority chemicals reported under CSPA and compare the results to other prioritization tools, such as the United States Environmental Protection Agency’s (EPA) ExpoCast and Toxicological Priority Index (ToxPi) and make recommendations to improve the collection of data relevant for prioritizing action on children’s products. The results of this work will help focus further efforts to protect children from potentially harmful chemicals found in consumer products.
3. Results
The average exposure score across all chemicals was 10.4 with a standard deviation of 2.6 (
Table 3). The three highest scoring chemicals for exposure were formaldehyde (average 14.2), octamethylcyclotetrasiloxane (average 13.9) and styrene (average 13.6). These three chemicals’ exposure scores were roughly average for most product characteristics, including lifestage, exposure duration, application to skin or body, and concentration. However, toxicokinetic scores, based on chemical properties and observed absorption rates were above average for all three chemicals for at least one exposure route. Across all records in the CSPA database, the maximum exposure score was 20.5 for formaldehyde in party horns and the minimum exposure score was 5 for molybdenum in drawing supplies. Toxicity scores were calculated by the sum of the products of the individual endpoint toxicities and potencies. The highest toxicity scores were for DBP (24), diethylhexyl phthalate DEHP (21) and formaldehyde (21). Four chemicals, phthalic anhydride, propyl paraben, methyl paraben and molybdenum, were not listed as endocrine disruptors, neurotoxicants, reproductive toxicants or carcinogens in any of the databases used in this study.
DBP, formaldehyde, DEHP, styrene and butyl benzyl phthalate scored relatively high for both exposure and toxicity scores and are thus found in the upper right-hand quadrant of
Figure 1, indicating the most concern. Chemicals not listed as toxicological concerns in any of the resources consulted are found on the Y-axis of
Figure 1, see dashed line box. Octamethylcyclotetrasiloxane was assigned an exposure score of 13.9, the second highest. However, the only resource consulted that identified octamethylcyclotetrasiloxane as a toxicological concern was the Global Harmonization System, which classified it as a suspected reproductive and developmental toxicant. Therefore, octamethylcyclotetrasiloxane is found in the upper left-hand quadrant of
Figure 1, indicating less concern than those chemicals with high exposure scores and high toxicity scores. Chemicals in the bottom left-hand quadrant are the least concerning because they have relatively low exposure and are not classified as toxicologically concerning by the resources consulted for this study. These chemicals are diethyl phthalate, phthalate anhydride and molybdenum.
The average total priority index is the product of the toxicity scores and the average exposure scores. Across all records, the average total priority index was 93.1 with a standard deviation of 79.4. The three highest total priority indices are attributed to formaldehyde (average 297.8), DBP (average 294.7) and styrene (average 231.2) (
Table 3).
In order to identify underlying relationships among the variables potentially driving the framework results, a principal component analysis (PCA) was conducted. The first two components of the PCA together explained ~56% of the variability in the priority index across chemicals (
Figure 2A). Principal component 1 (PC1) explained 33.6% of the variability and is associated with elevated toxicity scores for reproductive and developmental toxicity, carcinogenicity and neurotoxicity. Positive scores for PC1 are also indicative of products designed for children under age three with longer exposure durations and potential oral and inhalation exposure routes (
Figure 2B). Negative scores in PC1 are associated with concern over higher chemical concentrations, products applied directly to the skin or body, potential dermal exposure and higher scores for endocrine disruption. Principal component 2 (PC2) explained 23.1% of the variability between chemicals. Positive scores for PC2 are associated with higher reproductive and developmental toxicity and endocrine disruption scores. Positive PC2 scores were also associated with products with potential dermal exposures (
Figure 2B). As a result, the solvents (ethylene glycol, ethylene glycol monoethyl ester, methyl ethyl ketone) clustered together with formaldehyde and styrene with positive scores for PC1, indicating concern over neurotoxicity, reproductive and developmental toxicity and carcinogenicity and slightly negative scores for PC2 indicating concern over the concentration of chemicals reported and the product’s targeted lifestage. Butyl and ethyl paraben cluster together with negative scores for PC1 and relatively neutral scores for PC2. This indicates higher concern over endocrine disruption, and application directly to the skin or body and potential dermal exposure. Many of the phthalates, such as DEHP, BBP, DnHP, DBP and DIDP cluster together with positive scores for PC2, indicating concern over reproductive and developmental toxicity, endocrine disruption and potential dermal exposure. The phthalates generally cluster away from butyl and ethyl paraben. The separation is due to the presence of both endocrine disruption and reproductive and developmental toxicity for the phthalate cluster and solely endocrine disruption for butyl and ethyl paraben. The lower left-hand quadrant is negative for both PCs and characterized by chemicals that were not recognized as toxic in the sources considered for this study, such as molybdenum, propyl paraben, methyl paraben and phthalic anhydride (
Figure 2A). In this quadrant, scores are dominated by lifestage, concentration, and application variables related to exposure characterization.
The total priority index and exposure score can also be used to identify high priority product categories.
Table 4 summarizes the exposure scores and total priority indices across the product segments reported in the CSPA database. Kitchen merchandise, stationary/office machinery/occasion supplies and toys/games had the highest total priority indices. In each of these categories, formaldehyde, phthalates (as a group) and styrene had the highest priority indices. Together, kitchen merchandise, stationary/office machinery/occasion supplies and toys/games comprise approximately 16% of all CPSA reports. Almost half (44%) of reports fall under the clothing product segment. The three highest scoring chemicals in this category are formaldehyde (total priority index average of 264.8), styrene (total priority index average of 209.1) and phthalates (as a group, total priority index of 134.1). These results are shown in an expanded version of
Table 4 available as
Supplemental Table S3.
We compared the CSPA framework exposure and endocrine disruptor scores with ExpoCast and ToxCast, respectively. The CSPA endocrine disruptor score was calculated by the classification and the LOAEL reported in the European Chemical Agency’s Endocrine Disruptor Substances of Concern database (
Table S2) as well as the exposure score based on chemical and product properties. Filer
et al. (2014) [
33] applied ToxPi for the prioritization of endocrine disruptors based on the phase 2 ToxCast
in vitro assays and chemical properties that are associated with exposure potential. Eight of the ten chemicals identified as endocrine disruptors in this framework and six chemicals included in this framework but not identified as endocrine disruptors had ToxPi scores calculated by Filer
et al. (2014) [
33]. Butyl paraben scored high for endocrine disruption in both ToxPi and through the CSPA framework. However, other chemicals, such as DEHP and DBP, scored relatively higher through the CSPA framework than through the ToxPi predictions (
Figure 3A). Octamethylcyclotetrasiloxane and propyl paraben score high using ToxPi, but are not identified as endocrine disruptors in the resources consulted for this study.
ExpoCast predicts exposure to environmental chemicals at the population level [
30]. While many chemicals found in consumer products are have predicted exposure ranges in ExpoCast, other potential exposure routes are also considered. Octamethylcyclotetrasiloxane has one of the highest exposure scores and a higher predicted exposure in ExpoCast, relative to the other chemicals considered (
Figure 3B). Other chemicals, like styrene, have higher exposure scores relative to those predicted using ExpoCast. Two phthalates, diisononyl phthalate (DINP) and DEHP have higher median exposure predictions in ExpoCast relative to their exposure scores from the CSPA framework. This may be due to the fact that, in the United States, these chemicals are more tightly regulated in children’s products than in general consumer products.
Table 5 summarizes the top three chemicals according to each prioritization strategy: CSPA reporting frequency, CSPA total priority index, CSPA exposure score, ExpoCast, ToxPi endocrine disruptor score and CSPA endocrine disruptor score. When CSPA chemicals are prioritized based only on frequency of reports, cobalt and cobalt compounds, antimony and antimony compounds and ethylene glycol are prioritized. However, when the toxicity, toxicokinetics and exposure patterns are considered through the CSPA total priority index, formaldehyde, dibutyl phthalate and styrene are prioritized. The total number of reports of these three chemicals combined comprise approximately 15% of total CSPA reports during the time period assessed. Butyl paraben is identified as a high priority endocrine disruptor based on it scoring in the top three chemicals using both the CSPA endocrine disruptor score and the ToxPi score. ExpoCast and CPSA Exposure scores both identify octamethylcyclotetrasiloxane.
4. Discussion
The goal of this framework was to identify and prioritize chemicals in the CSPA database for further consideration and to compare the results with other prioritization methods, such as ToxCast and ExpoCast. In order to do this, it was necessary to understand the context surrounding the potential exposure and the toxicity and potency of the chemical. We used the target age group and product segment descriptions to identify potential exposure routes and durations and combined this information with the concentration to provide context surrounding the exposure. Chemical properties and absorption parameters were used to incorporate toxicokinetics. Toxicity and potency were calculated using a wide array of curated databases (
Table 1,
Table 2 and
Table S2). By combining these parameters in a multi-attribute utility function, we were able to calculate a total priority index for each of the ~33K CSPA records related to the most frequently reported chemical groups, about 88% of all CSPA records to date.
Two methods were used to identify priority chemicals in the CSPA database from this framework. Exposure scores and toxicity scores were plotted to identify chemicals notable in both dimensions. This method identified formaldehyde, styrene, DBP, BBP, DEHP, DIDP and butyl paraben as priority chemicals. The second method for identifying priority chemicals was through the calculation of a total priority index, which is the product of the exposure and toxicity scores. This method identified formaldehyde, DBP, styrene, BBP and DEHP as the highest priority chemicals. With the exception of BBP, which is considered a reproductive and developmental toxicant and an endocrine disruptor, the other five highest priority chemicals were considered toxic for three out of the four endpoints considered in this framework. A PCA confirmed the observation that toxicity drives a substantial part of the variability between chemicals. Neurotoxicants, such as formaldehyde, styrene, methyl ethyl ketone and ethylene glycol clustered together while the phthalates known for both endocrine disruption and reproductive and developmental toxicity clustered together, separate from compounds known for endocrine disruption alone. Formaldehyde did not cluster as close to styrene as was expected. Both chemicals are characterized by reproductive and developmental toxicity, neurotoxicity and carcinogenicity, however, they have varying scores related to dermal and oral toxicokinetics. This exemplifies the importance of including exposure routes, toxicokinetics and toxicity in one framework. The other prioritization frameworks examined in this study focused on toxicity (ToxPi) or exposure (ExpoCast), therefore combination of exposure and toxicity is a unique and important feature of this framework.
The CPSA framework can be used to identify individual chemicals or chemical groups of high priority to children’s environmental health. In this analysis, the three chemicals with the highest total priority indices, when phthalates and parabens were grouped were formaldehyde, phthalates and styrene. Individually, formaldehyde, styrene and DBP had the highest total priority indices. While analysis of individual chemicals can help identify potential issues related to regrettable substitutions, consideration of phthalates and parabens as groups may be relevant to regulations that may approach chemicals as groups, taking a more holistic view of toxicity since many phthalates and parabens have similar mechanisms of toxicity.
The chemical groups with the highest average total priority indices were formaldehyde, phthalates and styrene. While the CSPA framework is not weighted by reporting frequency, reports for these chemicals comprised approximately 15% of total reports. Thus, the magnitude of exposure potential from these products is not inconsequential. Formaldehyde, phthalates and styrene were also identified as high priority chemicals when average total priority indices were compared across product segments. Kitchen merchandise, stationary/office machinery/occasion supplies and toy/games were the three product segments with the highest average total priority indices. Within each of these segments, formaldehyde, phthalates and styrene were the highest priority chemicals.
The results of this framework were compared to other prioritization tools such as ExpoCast and ToxPi. While the CSPA framework relies on curated databases for toxicological assessment, ToxPi uses high throughput data from the in vitro assays publicly available through the ToxCast database. The comparison demonstrated the benefits and drawbacks to both approaches. For example, ToxPi was able to calculate scores for more endocrine disrupting chemicals than the curated Existing Substance Endocrine Disruptor Database (ECHA) that was used to calculate the CSPA endocrine disruption score. Octamethylcyclotetrasiloxane is an example of a chemical that was poorly characterized in available databases, but scored high for endocrine disruption using in vitro assays. However, regulatory decisions for future action on CSPA chemicals rely on the presence of a substantial body of evidence. Therefore, the benefit to using the CSPA framework with curated databases, allows for a stronger degree of confidence in the toxicological assessments. As the ToxCast assays continue to be more widely applied and more adverse outcome pathways are created, this high-throughput approach will provide added value.
ExpoCast was the other high throughput prioritization tool included in this analysis. The relationship between the CSPA exposure score and ExpoCast exposure prediction is highly variable. While some chemicals, such as octamethylcyclotetrasiloxane and molybdenum are relatively consistent between the two scores, other chemicals, such as styrene, had vastly different exposure scores between the CSPA framework and ExpoCast. Styrene has a high exposure score from the CSPA framework and a much lower ExpoCast prediction. This is partially due to differences in how the exposure scores are calculated relative to ExpoCast. The exposure score is only based on the potential for exposure from children’s products reported in the CSPA database, while the ExpoCast prediction includes multiple exposure sources. Additionally, while frequency of chemical reporting was not included as a variable in the CSPA exposure score, chemicals were selected for inclusion in the framework based on the number of reports. Therefore, the CSPA exposure score is not explicitly weighted for production volume. ExpoCast, on the other hand, relies on chemical use estimations [
30]. DINP and DEHP have moderate CSPA exposure scores and high ExpoCast predictions. This could be related to US consumer product laws, which limit the permissible concentration of some phthalates in children’s products, but not consumer products as a whole.
This is the first framework developed for the toxicological interpretation of the CSPA data. The benefits to using this framework include the relatively high amount of context regarding exposures and the detailed chemical and toxicological properties, including potency considered. Because the CSPA database comprises over 33,000 records as of September 2015, the relatively high throughput capacity of the framework is important. Lifestage, exposure duration, exposure route, application to skin or body and concentration were all derived directly from the fields in the CSPA database. This allowed for the relatively quick processing of the extensive database. Additionally, because all information was derived directly from the CSPA database or based on chemical properties that were widely available, there were no missing data for the exposure score. This allows for the identification of chemicals with high exposure potential and less well characterized toxicity. Some chemicals, such as phthalic anhydride, propyl paraben, methyl paraben and molybdenum, were not classified as toxic for the endpoints considered in this framework in any of the databases and resources consulted. These chemicals were included in the CSPA database but did not receive toxicity scores because they were either (1) toxic to biological systems not considered in this framework; (2) toxicologically characterized by databases not included in either this framework such asREPROTEXT [
34]; or (3) were included as part of a larger group of chemicals. This lack of data can lead to lower total priority indices that are not necessarily indicative of safety. However, because the exposure score is complete in all cases, it can be used to identify chemicals, such as octamethylcyclotetrasiloxane, that have high exposure potential from children’s products but may be poorly characterized in the databases considered. Octamethylcyclotetrasiloxane is an example of a chemical in need of further characterization in the curated databases considered for this study. Both ExpoCast and the CSPA exposure score identify octamethylcyclotetrasiloxane as having high potential for exposure. Additionally, its high ToxPi score suggests that octamethylcyclotetrasiloxane could be a potent endocrine disruptor.
Washington State was among the first to require reporting of chemicals of concern in children’s products. Since then, other states have begun to implement similar requirements. While Washington has developed an extensive database to help guide future regulatory action, improvements to the reporting structure of the CSPA database could expand the toxicological interpretation of the data. For example, metals such as molybdenum, cobalt and antimony are reported by total elemental amount. There are significant differences in toxicities and toxicokinetics between metal compounds. Because metal compounds were not specified in the CSPA database, the unique features of specific metal compounds are not reflected in the CSPA framework and may compromise the ability to accurately assess the toxicities and toxicokinetics associated with the presence of antimony, cobalt and molybdenum in children’s products. Additionally, more information regarding when the laboratory tests were performed by the manufacturers would help determine whether volatile chemicals reported as “contaminants” are likely to off-gas by the time the product reaches the consumer. As chemical reporting requirements become more common and more consumer product databases are developed, prioritization of the data based on both exposure potential and toxicity, as can be done using this framework, will be critical. This increase in reporting frequency requirements will also be met with a need for more stakeholder engagement through focus groups and value based decision analytics to ensure that the models are answering the right regulatory questions [
35,
36,
37].
The CSPA framework presented here provides a method for processing large amounts of consumer product data in a relatively high-content manner. However, one limitation of this approach is that it does not calculate a comparable risk between chemicals but, rather, ranks the chemicals in the CSPA database. Thus a total priority index of 100 is not 10 times more concerning than a total priority index of 10. Instead, it indicates a difference in rank. This decision was made to allow the database to be processed in a relatively high throughput manner. The total priority index focuses on exposure potential from product use as well as exposure potential from house-dust as the product disintegrates. While it includes chemical properties to account for absorption, bioaccumulation is not included. No persistent organic pollutants were among the chemicals used in this analysis. However, persistent organic pollutants are included in the CSPA database as a whole. Inclusion of these additional data and bioaccumulation factors, may modify the high priority chemicals identified using this framework. To account for some of the limitations of the CSPA framework, it is recommended that this approach be employed, along with other prioritization tools, such as ToxPi and ExpoCast.