PCs—Assessment of algorithms and scenario independent parametersInhalation

The tool structure is such that substances are assigned to one of 4 volatility bands based upon their vapor pressure (VP). For all bands, air concentrations are calculated based upon the following formula:

This equation is consistent with the instantaneous release, lowest tier algorithm in the RIVM ConsExpo model [15]. The full TRA equation provides exposure in mg/kg body weight/day units; event time, inhalation rate and body weight variables are omitted for air concentration. As compared to ECHA [14] inhalation equations for low tier assessments, the TRA algorithm considers a dilution fraction due to normal non-ventilated air flow between residential rooms and does not consider the respirable fraction for spray releases. TRA also includes a modifying factor (fraction released to air) for VP < 10 Pa and provides an upper bound air concentration based upon saturated vapor concentration (SVC). The SVC applies only to non-aerosols. Each parameter of the algorithm is assessed in detail in the Supplementary Information (SI). Here, we focus on the VP banding approach because it is the unique aspect of the TRA inhalation algorithm.

The VP band approach applies to non-aerosol products. For aerosols, the tool assumes that 100% of the substance in the product is instantaneously released to air. For non-aerosols, for the highest VP band it is assumed that 100% of the substance present in the product instantaneously volatilizes and becomes well-mixed within the room air. Thus, the algorithm prediction for this vapor pressure band will be the highest concentration possible based upon the input values of use amount and weight fraction. The assumption of complete mixing may underestimate air concentrations near the emission source in the first few minutes. The effect of this assumption, however, is greater with greater room size; for a 20 m3 room (the TRA default room size), the assumption was indicated to be reasonable [16].

At lower VP bands, for each order of magnitude decrease in VP, a factor of 10 reduction to the fraction released and therefore predicted air concentration is implemented. Modeling analysis was done to evaluate if these reductions provide conservative estimates for amount released for the scenario duration (Table 1, details in SI). The results for the painting scenario (Fig. 1) indicate that the TRA VP approach predicts release fractions 0.5–3.5 orders of magnitude higher than ConsExpo for the range examined; other scenarios gave similar results. Additional details (Table 1, SI) support an overall finding that the release fraction using the TRA VP band approach is conservative as compared with higher tier modeling and measured data. This is a result of the comparatively low cut-off of 10 Pa for the assumption of complete instantaneous release. Other authors have also indicated the conservative nature of the inhalation algorithm [7, 13]

Table 1 Data/Analyses to address assumptions of the TRA approach.Fig. 1: TRA Release fractions compared to ConsExpo release fractions.

TRA release fractions in red, ConsExpo in blue. Inset boxes give differences in orders of magnitude.

Dermal route

The dermal algorithm is:

The dermal transfer factor (TF) in the TRA represents the fraction of the substance in the thickness layer (TL) in contact with the skin that is transferred to the skin [3]. The default TF remains at 1 for all scenarios within the TRA. The TF does not refer to or account for the fraction of material that might be subsequently absorbed through the skin into systemic circulation.

The algorithm is consistent with that provided in the EU Technical Guidance Document (TGD) [17], and ECHA [14], although frequency of use and the potential to consider a transfer factor is included. In TRA default mode, both factors are set to 1 and so have no impact on the prediction. The TRA algorithm assumes that there is a uniform TL on skin across the whole contact area and that the total amount of substance in this uniform layer is available for absorption.

For PCs, a value of 0.01 cm is used as the TL for all scenarios with the exception of 0.001 for air care, continuous action, solid (the latter will be addressed with articles since it is a solid). Reduced thickness of a (uniform) layer of 0.001 cm for some ACs has been set in TRA to account for the reduced mobility of substances in the article matrix and is applied here to the solid air freshener. This approach has been challenged for articles and will be discussed in detail in that section; here we provide the basis for the approach for liquid products. A default TL value of 0.01 cm has been used for years for liquids and is identical to the TGD default TL of 0.01 cm for non-solid media in contact with the skin [14, 17,18,19,20,21,22]. Data on liquid thickness layer on skin are summarized in Table 1 and support that 0.01 cm is a conservative value [23]. The TL approach implies a dermal load of 10 mg/cm2 across the entire exposed skin surface area, which is close to the maximum that may be reasonably assumed (ca. 12–14 mg/cm2) according to the EU TGD [17].

Additional discussion of the dermal algorithm can be found in SI Section 4.

Oral route

The ingestion algorithm is:

The algorithm is comparable with that in the EU TGD [17] and ECHA [14] for initial tier oral assessments, with the addition of the possibility to consider a transfer factor (TF). As the TRA default TF = 1 for all scenarios, this capability does not impact the conservative nature of default assessments (SI).

The TRA does not include a separate dust ingestion pathway for non-intentional mouthing exposure. This pathway would primarily apply to semi-Volatile Organic Compounds (SVOCs) and non-VOCs, as VOCs would partition to air rather than dust. The TRA assumes that at least 0.1 % of any compound (lowest VP band) evaporates immediately and is inhaled in the standard room with standard ventilation. In addition, depending upon the scenario, oral and/or dermal exposures may occur as well, in which exposure is calculated based upon 100% of the substance present in the ingested or dermal contact mass. Total direct exposure is intended to exceed the potential exposure contribution via dust ingestion.

Comparison of TRA predictions with those of the recently developed DustEx dust specific model and exposure estimates based upon measured data (Table 1) [24,25,26] support that, if used with default assumptions, TRA predictions should cover exposure via the pathway of dust ingestion as well. If the TRA is run with refined assumptions in a case where dust ingestion may occur, then the user should assess if there is a need to evaluate dust exposure independently (for example via the DustEx framework) [24]. A recent analysis [27] indicates that dust-mediated transfer is most notable for substances with intermediate octanol-air partition coefficients (\(}}}}}_}}}}}^-^\)) where indoor partitioning is mediated by air.

PCs—total exposure

Within the TRA, each route equation operates independently, i.e., even if the inhalation algorithm indicates 100% of the substance is present in air, the dermal (and oral) algorithms are unchanged and provide additional exposure predictions for substance present on skin or ingested when relevant. Mass balance is exceeded in these cases, which will result in additional conservativeness in the tool.

PCs—scenario dependent parameterization

Default values for TRA scenario dependent parameters (Table SI-6.1) were compared with alternate values found within published Specific Consumer Exposure Determinants (SCEDS) or other public sources of these data (Table SI-6.2). Comparison of values for individual parameters, however, is of limited utility as it is the combination of parameters and the model algorithm that determine the relative conservatism in the exposure assessment. For example, in the CONCAWE SCEDs the default weight fraction has been raised to 1 in all cases, yet the exposure estimates are lower than those based upon the TRA defaults due to refinements in other values (Fig. SI-6.1) [28]. Overall, most SCEDS or other sources of defaults were similar or less conservative than those in the TRA. In some cases, data were available for only one parameter and its impact on the scenario as a whole could not be assessed (for example, paint use amount was provided but without indication of room size or exposure duration [29]). For the SCEDS, which provide complete scenario information, predicted total exposures were lower than those of the TRA with the exception of CEPE scenarios in which exposure time was increased or, in the case of spray painting, a dermal route was added in the SCED [30]. The significance of the dermal route for this scenario will be assessed in the next section.

PCs—comparison to modeled predictions or measured data

Table 2 and Fig. 2 (and SI) present results from various studies that have compared TRA predictions with other models and/or measured data [3, 7,8,9, 11, 16, 31, 32]. In some cases, the TRA was run with defaults; when measured data were available it was generally modified to match the exposure scenario conditions.

Table 2 Summary of benchmarking for TRA predictions with models and/or data.Fig. 2: PCs: Comparison of TRA exposure predictions with predictions of other models and available data.

Note log scale of coloring; the same scale applies to both mg/kg.day and mg/m3. Gray color indicates not assessed (no information). Vertical axis provides product category, scenario basis (def = default or spec=specific modifications; if only weight fraction was modified the modified value is indicated as ‘x’WF) and source (reference number, table number, or supplemental information section). Horizontal axis provides the name of the model used to generate the prediction or if the exposure values are data-based (if so, whether a typical or upper bound value). Additional details for all are found in Table 2 and/or supporting information.

Overall, these analyses do not cover all PC codes, but the general trend to provide the most conservative results across publications and models and measured data support the finding that the TRA is an appropriate Tier 1 tool. In the few cases where other tools provided greater route-specific predictions, generally they were within a factor of 2.5 of those of the TRA, were also intended to be conservative predictions and/or there was uncertainty in the defaults applied in other tools. In all cases with measured data, the TRA default predictions far exceeded measurements. In the one case where TRA inputs refined by the authors provided predictions less than measured data [11], information was insufficient to address all of the refinements for appropriateness with TRA use.

The comparison to measured data was, however, limited to air concentrations. No monitoring data for consumer dermal or ingestion routes were identified for PCs and ACs included in the TRA. To address this gap, recent dermal occupational data were evaluated for relevance (Tables 2, SI) [33, 34]. These occupational data support that the tile glue and brush painting dermal estimates in TRA are appropriate. They also suggest that for completeness, particularly in cases where the dermal route is of specific interest, it may be useful to consider a dermal route i.e., for spray paint. Note, however, that in general the TRA is not meant to provide route-specific values but overall systemic exposures (RCRs are added across routes). The expectation would be that when route specific exposures are of interest, for example because of acute effects, more refined tools may be available to address those situations.

A more detailed analysis of the impact of TRA not including a dermal route for the spray paint exposure scenario is also provided in Tables 2. Adding a dermal exposure estimate based upon the SysDEA data to the ConsExpo inhalation estimate for a 0.1 weight fraction (based upon the scenario in Oltmanns et al. 2015) provides a total exposure lower than the TRA inhalation prediction for a 0.1 weight fraction.

ACs—assessment of algorithms and scenario independent parameters

Inhalation, oral and dermal algorithms remain the same for articles.

Inhalation route

In general, the inhalation algorithm will be more conservative when applied to articles than products. Whereas liquid products are generally applied to surfaces in thin layers, substances present within articles need to diffuse to the article surface to become airborne. Thus, the TRA assumption of instantaneous release at the start of an exposure scenario is a further departure from reality for articles as compared to liquids.

As many substances in articles will be non-VOCs or SVOCs, it is useful to consider if inhalation estimates for the lowest VP band (<0.01 Pa, 0.001 of total amount in product instantaneously released) are conservative. The modeling analysis provided earlier indicates that it is, and that as VP is reduced further beyond the 0.1 Pa cutoff for the lowest VP band, the conservatism increases (Fig. 1).

Oral route

The assessment presented earlier applies to articles as well. The algorithm’s assumption that 100% of the amount placed in the mouth is ingested is more conservative for articles than products, as it assumes total ingestion or migration out of the solid item placed in the mouth. The overall level of conservatism in the prediction will depend upon the scenario dependent values of amount ingested.

Dermal route

While the algorithm remains the same, for articles the default TL in the TRA is reduced to 0.001 except for the following scenarios which retain 0.01: toys (cuddly toy); car seat, chair, flooring; diapers; sanitary towels; tissues, paper towels, wet tissues, toilet paper. The reduced thickness of a uniform layer of 0.001 cm for some article scenarios has been set in the TRA to account for the reduced mobility of substances in the article matrix (based upon expert judgement and stakeholder consensus). For some ACs intended for prolonged/intensive contact and/or articles where it was reasoned that moisture could be present, the default TL for nonsolid media of 0.01 cm was retained; it was also assumed that the contact might more closely resemble a liquid layer.

Several alternate approaches for estimating dermal exposure via article contact are summarized in Table 3 [10, 35,36,37]. Delmaar et al. [12] has indicated that the TRA dermal algorithm is not sufficiently conservative for articles, as it neglects replenishment of the substance in the TL considered to be in direct contact with skin from the reservoir within the article. By utilizing an approach that considered diffusivity within the article matrix to be the controlling factor for dermal exposure, these authors generated predictions orders of magnitude higher than the TRA in some cases [12]. However, it is also recognized that this method does not take into account mass transfer to the skin nor uptake within the skin [12, 38]. For the purposes of assessing the TRA, only the mass transfer to the skin is relevant as predictions are for external exposures (as per REACH requirements [14]).

Table 3 Article Categories Summary Table—Algorithm Approaches, Model and Data Benchmarking.

Huang et al. [38] reviewed models for near-field exposure pathways of chemicals in consumer products. This assessment included both the TRA and Delmaar approaches for the pathway of transfer of chemicals from within a solid object to skin surface, and authors concluded that this pathway was considered immature as few models were available to predict this transfer or existing models required chemical specific parameters for which adequate prediction methods are not currently available.

Alternate proposed methods are discussed further in the comparison with other modeled results.

ACs—total exposure

As for PCs, total exposure also considers dermal, inhalation and oral without conservation of mass balance (i.e., all released to air, yet exposure via dermal contact or ingestion also occur in relevant scenarios).

ACs—scenario dependent parameterization

Overall data for refining or evaluating parameters used for prediction of article exposures is limited (Table SI-9.1). No SCEDs were identified for any ACs. One study provided some information regarding exposure duration and frequency and use amount for 2 articles [29]. RIVM has a fact sheet for toys which also includes several scenarios included in the TRA, but this fact sheet is from 2002 and most model parameters derive from estimations [39].

Oral route

For objects, 10 cm2 is a common value used for the surface area of the object placed in the mouth [40], based upon mouth size. The volume of material ingested in the TRA article scenarios ranges from 0.01–0.3 cm3, which would be equivalent to ingestion/absorption from a thickness of 0.001–0.03 cm for a 10 cm2 area. In comparison with the TRA assumption that 100% of the substance present is ingested, article-to-saliva leaching data for several article types indicates that for nano silver only a fraction of the total content is leached (Table SI-9.1). A recent comprehensive review of measured migration rates of substances from articles into saliva reported a range from 1.7 × 10–6–33 μg/10 cm2/min [41]. Using these values and a constant 10 cm2 contact, to reach the TRA exposure estimates of 0.1–4.3 mg/kg/day across article categories for a 10 kg child, mouthing would need to occur ≫ 24 hours/day based upon the lowest migration rate, and 0.5–22 hours/day for the highest migration rate. The highest migration rate is for a PVC article, whereas the TRA scenario with the lowest predicted exposure is for bedding. The next highest TRA exposure would be reached in about 2 hours with the highest migration rate. As daily mouthing times for individual article categories are typically <1 hour/day [23], these comparisons indicate the TRA should provide conservative values for most substance-material combinations, and the highest substance-material migration rate may yield an estimate similar to the TRA prediction dependent upon the mouthing time associated with the particular scenario. Using predicted migration rates for each substance-material datapoint (N = 437), the review authors developed oral exposure predictions for dolls and pacifiers. Highest predicted exposures were 22–253 μg/kg body weight/day for dolls and 6–224 μg/kg/day for pacifiers. The lowest TRA prediction of 100 μg/kg/day for mattress bedding falls within the range of these estimates, and the range of predictions for all other TRA scenarios (430 μg–4300 μg/kg/day) fall above these values.

Dermal route

Minimal data were available to evaluate or refine the dermal parameters. For articles, in the TRA the TL is multiplied by the density of the article to assess the amount of substance released per unit area of the article over time. As experimental data for TL is limited for articles, Spaan et al. [10] summarized data on the amount of substance released over time (Table SI-9.1).

Spaan et al. [10] concluded transfer from surface to skin after application of substances to surfaces can be high: up to 100% based upon their measurements of application to glass and aluminum. They estimated transfer from surfaces to skins or gloves for applied substances with unknown binding to be 10–60%. For substances within articles, experiments of wiping show < 10% if expressed as the amount present in the first 10 um. Data were not available to address the effect of longer wiping durations. They conclude that using a factor of 1 (i.e., 100%) with a TL of 10 µm, as is currently used in the TRA, should be a precautionary approach, and that values below 0.1 (i.e., 10%) would likely be more realistic for PVC and printed paper for a 10 µm TL.

Data in Table SI-9.1 for leaching of nano-silver into sweat and for transfer from article surface to wipes supports that only a fraction of total mass present is released.

ACs—comparison to modeled predictions or measured dataInhalation route

TRA air concentration predictions were compared to SVOC indoor air concentrations in general and two ACs where data were identified. In all cases the TRA predictions exceeded measured concentrations, sometimes by orders of magnitude (Table 3) [42, 43].

Oral route

Table 3 and Fig. 3 include comparison of TRA oral exposure estimates for 4 ACs. In all cases the TRA prediction exceeds the worst-case estimate.

Fig. 3: ACs: Comparison of TRA exposure predictions with predictions of other models and available data.

Note log scale of coloring. Gray color indicates not assessed (no information). Vertical axis provides article category and source (reference number, table number, or supplemental information section). Horizontal axis provides the name of the model used to generate the prediction, specific model adjustments (PI = product ingredient which is the weight fraction, TF = transfer factor), if the exposure predictions are data-based (if so, whether a typical or upper bound value), and for one substance present in flooring the aggregate exposure estimate of total exposure from all sources based upon NHANES biomonitoring data. Additional details for all are found in Table 3 and/or supporting information.

Dermal route

Modeling and data from several studies are summarized in Table 3 and Fig. 3. The approach of Delmaar provides higher dermal exposure predictions than TRA, generally by two orders of magnitude [12]. Other approaches, however, provide lower predictions with the exception of the Spaan mass balance approach for flooring, based upon consideration of surface area of flooring contacted rather than skin contact area. TRA dermal predictions in all cases exceed absorbed concentrations estimated based upon skin absorption data and total exposure (all sources) based upon biomonitoring data [44] by orders of magnitude (Table 3) [44,45,46]. While it is recognized that there are a number of studies reporting migration rates into sweat for specific chemical-article combinations, it was beyond the scope of this effort to develop a comprehensive database of migration rates. The TRA scenario-relevant studies which were identified support that the TRA approach provides a conservative exposure prediction.

Overall, limited measured data specific to consumer exposure scenarios are available to help evaluate the TRA article exposure predictions. The comparison of model estimates provides a relative order of performance. For inhalation exposures the TRA has been shown to be conservative (Fig. 1). For oral exposures TRA predictions tend to be most conservative but for a limited analysis (Table 3, Fig.