A Web-based Tool to Aid the Identification of Chemicals Potentially Posing a Health Risk through Percutaneous Exposure

Abstract

Occupational hygiene practitioners typically assess the risk posed by occupational exposure by comparing exposure measurements to regulatory occupational exposure limits (OELs). In most jurisdictions, OELs are only available for exposure by the inhalation pathway. Skin notations are used to indicate substances for which dermal exposure may lead to health effects. However, these notations are either present or absent and provide no indication of acceptable levels of exposure. Furthermore, the methodology and framework for assigning skin notation differ widely across jurisdictions resulting in inconsistencies in the substances that carry notations. The UPERCUT tool was developed in response to these limitations. It helps occupational health stakeholders to assess the hazard associated with dermal exposure to chemicals. UPERCUT integrates dermal quantitative structure-activity relationships (QSARs) and toxicological data to provide users with a skin hazard index called the dermal hazard ratio (DHR) for the substance and scenario of interest. The DHR is the ratio between the estimated ‘received’ dose and the ‘acceptable’ dose. The ‘received’ dose is estimated using physico-chemical data and information on the exposure scenario provided by the user (body parts exposure and exposure duration), and the ‘acceptable’ dose is estimated using inhalation OELs and toxicological data. The uncertainty surrounding the DHR is estimated with Monte Carlo simulation. Additional information on the selected substances includes intrinsic skin permeation potential of the substance and the existence of skin notations. UPERCUT is the only available tool that estimates the absorbed dose and compares this to an acceptable dose. In the absence of dermal OELs it provides a systematic and simple approach for screening dermal exposure scenarios for 1686 substances.

INTRODUCTION

The occupational hygiene efforts of the past 60 years have led to reductions in inhalation exposures to most hazardous substances in the workplace (Creely et al., 2007). As exposures by the inhalation route decrease, the relative contribution from other routes of exposure, particularly dermal contact, may become more important.

Although most health and safety regulatory agencies develop occupational exposure limits (OELs) for inhalation exposure, they do not propose dermal OELs. To manage dermal exposure risk, these agencies assign skin notations to substances that could result in health effects following dermal absorption. Such notations are intended to signal to health and safety professionals and employers that dermal exposure should be considered when dealing with these substances. However, this approach has many limitations. For example, dermal notations are either present or absent, they provide no indication of acceptable dose or exposure levels. They are also only assigned to substances that already have an inhalation OEL so substances that are hazardous by the dermal route but that are not inhalation hazards are not covered (Boeniger, 2003). Furthermore, there is no consensus on when a skin notation is required. The criteria for assigning skin notations differ by jurisdiction (Nielsen, 2004) and a study of dermal notations from six different countries and organizations (Nielsen and Grandjean, 2004) found that there was only 60–70% agreement between the chemicals that carried a skin designation. In a similar comparison, Lavoué et al. (2008) found an 87% agreement between the American Conference of Governmental Industrial Hygienists (ACGIH) threshold limit values (TLV) skin notations and the Swiss Accident Insurance Fund (SUVA) skin notations, but this relatively high agreement can partly be attributed to the fact that the analysis was restricted to chemicals that appeared in both databases. Lavoué et al. also noted that although the comparison was favourable, there were still discrepancies between the substances that were assigned skin notations, giving hygienists attempting to assess dermal risk a contradictory message. Chen et al. (2011) compared skin notations for 480 chemicals (all of which carried a notation in at least one jurisdiction). They found that only 3% of the chemicals carried a notation in all of the seven jurisdictions studied. In some jurisdictions [including Denmark, Norway, Sweden, and USA (OSHA)], a skin notation signals only that there is potential for dermal absorption but does not indicate whether this absorption might result in actual toxic effects.

In response to some of the above limitations, NIOSH has developed a new system for assigning notations (NIOSH, 2009). This method involves calculation of the ratio between the dose absorbed by the skin and the inhalation uptake of the same duration at the OEL (the SI ratio). A similar approach is used by the European Centre for Ecotoxicology and Toxicology of Chemicals (ECETOC) and the Netherlands (Dotson et al., 2011). The NIOSH method also classifies skin notations according to type of hazard (for example, systemic absorption, irritant, sensitizer) providing health and hygiene professionals with more information than is available from traditional notations. While this system is an improvement over traditional skin notations, at the time of writing, NIOSH notations have only been assigned to 30 chemicals.

The UPERCUT tool described in this study was developed in response to the lack of information that can help health and hygiene professionals to assess the risk posed by dermal exposure and the limitations of the current skin notation system. The work follows on a study by Lavoué et al. (2008) who compared skin notations (in the ACGIH and SUVA lists), LD₅₀s and two dermal risk indices based on the dose received by inhalation at the OEL relative to the dose received through the skin [estimated using quantitative structure-activity relationships (QSARs)]. That study found that, as expected, substances with skin notations generally had lower LD₅₀s than substances without skin notations. They also found that the ratios between dermal and inhalation doses (estimated using the QSAR techniques) were generally higher for chemicals with skin notations than for those without. Lavoué et al. suggested that toxicological data and skin permeation QSARs could be integrated to create a systematic dermal risk assessment algorithm. This would allow risk assessment of substances for which there currently are no skin notations, and it would improve on some skin notation lists by considering both toxicity and skin permeability.

The objective of the current study was to use toxicological data, skin permeation QSARs, and existing OELs to develop a tool that can be used by occupational health and hygiene professionals to differentiate between dermal exposure scenarios that are unlikely to pose a risk to health, and those that require more extensive evaluation. We also aimed to include an uncertainty analysis in the tool to take into account variability in the parameters that are used to estimate risk.

METHODS

List of equation terms and their units

• J _max (mg cm⁻² h⁻¹): maximal flux

• S _expo (cm²): the skin surface exposed

• S_body (cm²): the whole body skin surface exposed

• T _expo (hours): exposure duration

• OEL (mg m⁻³): the occupational exposure limit

• AR (proportion): respiratory absorption rate

• V _8hrs (m³): the volume of air inhaled during an 8-h work shift

• K _p (cm h⁻¹): the permeation constant

• K _Henry (atm m³ mol⁻¹): the Henry constant

• MW (g mol⁻¹): the molecular weight

• K _ow (proportion): the octanol/water partition coefficient

• Sol_water (mg cm³): the water solubility

UPERCUT tool approach

The development of the UPERCUT tool was based on estimating a quantitative index called the dermal hazard ratio (DHR) for a wide range of chemical substances. The DHR is the ratio of the dose potentially absorbed through the skin and an ‘acceptable’ reference dose (for example, the dose received from 8h of exposure at the OEL). A DHR higher than one indicates that the received dose may exceed the acceptable limit so further evaluation using either dermal exposure monitoring or biomonitoring may be necessary. This index is similar to the SI ratio used for the development of NIOSH skin notations, but it allows further flexibility in the duration and surface of exposure. In addition, to address uncertainty associated with some of the parameters used in the calculation of DHR, a simulation method was also used to enable UPERCUT to present the uncertainty distribution surrounding the DHR estimate.

To calculate the DHR it was necessary for the tool to estimate both the received dose and the acceptable dose. The received dose was estimated using substance properties and the exposure assessor’s estimation of the exposure duration and exposed surface area of skin. The reference dose was based on either an OEL set by a regulatory or professional body (if available), or derived from animal toxicity data.

The DHR does not address all of the limitations associated with skin notations so the tool also contains additional information to help risk assessors evaluate the dermal risk associated with a substance. These include: (i) animal toxicology data; (ii) information on the potential for dermal permeability; (iii) the existence of a German MAK or ACGIH skin notation; and (iv) classification under the European carcinogenic, mutagenic, and reprotoxic substances (CMR) classification system.

The following sections detail the creation of the database behind the UPERCUT tool, the equations used for the calculation of the DHR, and the approach used to assess the associated uncertainty. The modelling approach is summarized in Fig. 1.

Figure 1.

UPERCUT modelling approach summary, with legend.

Database development

A preliminary list of substances of interest was based on high production volume lists from the Organisation for Economic Co-operation and Development (OECD) and United States Environmental Protection Agency (EPA).

Physico-chemical data were extracted from the PHYSPROP database (access was graciously provided by Syracuse Research Corporation, North Syracuse, NY, USA), which contains data for over 30 000 chemical substances. The properties recorded to the UPERCUT database were: vapour pressure (VP), molecular weight (MW), Henry’s constant (K _Henry), octanol/water partition coefficient (K _ow), and water solubility (Sol_water).

Toxicological data and OELs were obtained from the Registry of Toxic Effects of Chemical Substances (RTECS), 2007 version (purchased from MDL Systems Inc., Hayward, CA, USA). Extracted animal toxicity data (restricted to mammals) included both acute (the median lethal doses, or LD₅₀ for oral, intra-peritoneal, sub-cutaneous, intravenous, and dermal administration; as well as Draize test results for dermal irritation) and chronic (lowest adverse effect levels, or LOAEL, for oral and cutaneous administration). Where multiple LD₅₀ values were available, the lowest was selected. LOAELs pertaining to gastric behavioural, weight loss, and local dermal effects were excluded.

The database was supplemented with data from the World Health Organisation Chemical Safety Cards, and from the list of substances in Annex I of the European Union Directive 67/548/EEC. OELs from the ACGIH TLV (ACGIH, 2010), and the German Maximale Arbeitsplatz Konzentration (MAK) values (Deutsche Forschungsgemeinschaft, 2009) were also added.

Substances were classified according to the United Nations Economic Commission for Europe (UNECE) Globally Harmonized System of Classification and Labelling of Chemicals (GHS) based on the data found in RTECS. Animal toxicology endpoints were classified as oral or dermal, and acute or chronic.

To provide the user with information on the intrinsic penetrative potential of the substance of interest, the substances were divided into three groups: poor, intermediate, and good penetrants. The groupings were developed by Magnusson et al. (2004), and are based on the substance MW and the logarithm of K _ow (log[K _ow]).

Data presented to users

The above data were collected for the development of the UPERCUT tool but not all of the data are presented to the user. Some of the data are used in the derivation of information that is presented by the tool. For example, the raw toxicological data and the OEL values are not presented but were used to derive the DHR that is presented. Data shown to users are as follows:

• Chemical Abstracts Service (CAS) number

• Substance name

• Risk phrase

• CMR classification

• Presence/absence of ACGIH or DE MAK skin notation

• Dermal penetration potential rating (according to Magnusson et al., 2004 ratings system) based on molecular weight and octanol-water partition coefficient

• GHS classification and labelling, based on toxicological data

• Estimated DHR and uncertainty distribution.

Dermal hazard ratio

The DHR was estimated using a cumulative dose approach, as used by Lavoué et al. (2008) and others (de Cock et al., 1996; Dotson, 2011; Walker et al., 2003). This approach compares the dose potentially absorbed through the skin, to the equivalent dose received through inhalation during a work shift at the OEL for inhalation. We used equations (12) and (13) from the publication by Lavoué et al. for the calculation of the DHR.

For solid and liquids (equation (1)) it was assumed that the skin is constantly exposed during the exposure period to a saturated aqueous solution of the substance of interest, and that dermal absorption followed Fick’s law of diffusion. DHR was calculated as follows:

$${\text{DHR}}_{\text{liq},\text{sol}}=\frac{J_{\text{max}}\times S_{\text{expo}}\times T_{\text{expo}}}{\text{OEL}\times \text{AR}\times V_{8\text{hrs}}}$$ (1)

For gases (equation (2)), exposure of the whole body to an aqueous solution in equilibrium with the atmospheric concentration (set to the OEL) was assumed.

$${\text{DHR}}_{\text{gas}}=\frac{K_{\text{P}}\times S_{\text{body}}\times 24.4\times T_{\text{expo}}}{V_{8\text{hrs}}\times \text{AR}\times {10}^9\times K_{\text{henry}}}$$ (2)

In the UPERCUT tool, T _expo and S _expo are selected by the user. S _expo is selected in the form of a choice of the body parts exposed, values were obtained from Fenske and Perkins (2003). The estimation of J _max and K _p (equations (3) and (4)) was described by Lavoué et al. (2008) using QSAR equations from Vecchia and Bunge (2003) [equation T1 in Table 1 pp. 88–89].

$$\begin{array}{l}{\text{Log}}_{10}\left(K_{\text{p}}\right)=-2.44+0.514\times \text{log}\left(K_{\text{OW}}\right)\\ -0.0050\times \text{MW}\end{array}$$ (3)

$$J_{\text{max}}=K_p\times So{l}_{\text{water}}$$ (4)

The parameters required to calculate K _Henry, K _ow, Sol_aq, and MW were taken from PHYSPROP. V _8hrs was set to 10 m³ (Paustenbach and Perkins, 2003) and AR was set equal to 75% (NIOSH, 2009).

Determination of OEL values

The OELS in the database included the 2009 MAK values, the 2010 TLVs, and OELs found in the RTECS database, which includes international OELs, NIOSH recommended exposure limits (RELs), and OSHA permissible exposure limits (PELs). For many substances there were several OELs available, and for some there were none.

In situations where there was at least one OEL available, and to take into account variable OELs for one chemical, the log-transformed OELs were fitted to a random effect model with the chemical name as a random effect (Supplementary Appendix 1 is available at Annals of Occupational Hygiene online). From this model it was possible to predict a single OEL value for each chemical that takes into account the existence of significant within-chemical variability. The resulting OELs were labelled ‘calculated’ OELs.

In situations where there were no OELs available the animal toxicology data were used to predict OELs (Supplementary Appendix 2 is available at Annals of Occupational Hygiene online). The method was similar to that described by Suda et al. (1999) and Whaley et al. (2000). Linear regression models were built from chemicals with OELs for each of the seven types of available toxicological data (acute and chronic, by different exposure routes) with the ‘calculated’ OELs as the outcome variable. These models were then used to predict OEL values for substances for which there were no available OEL. These were labelled ‘predicted’ OELs.

Uncertainty analysis—input parameters

There is uncertainty associated with the DHRs estimated using the above approach. In order to help UPERCUT users to correctly interpret DHRs as approximations, uncertainty estimates around the DHR are also calculated.

The uncertainty surrounding the DHR was estimated using Monte Carlo simulation based on estimates of the uncertainty of individual parameters used to calculate it. Uncertainty distributions were used for the values of OELs, physico-chemical parameters (Sol_water and log[K _ow]), K _p and J _max. For the other parameters, S _body, T _expo are chosen by the user, and MW is associated with low uncertainty. AR and V _8hrs can vary across individuals and agents, but were set constant at typical population average.

Uncertainty surrounding OELs was estimated differently for ‘calculated’ and ‘predicted’ OELs. For ‘calculated’ OELs, uncertainty was modelled as a lognormal distribution with a geometric standard deviation (GSD) equal to within-chemical variability estimated from the random effect model described in Supplementary Appendix 1 is available at Annals of Occupational Hygiene online. For ‘predicted’ OELs that were estimated using regression models built with toxicological data, uncertainty was also modelled as lognormal but with a GSD calculated with the residual modelling errors (Supplementary Appendix 2 is available at Annals of Occupational Hygiene online).

To model the uncertainty around the physico-chemical parameters (Sol_water, log[K _ow], and K _Henry), the distributional shape of the values of each parameter in the database was examined. They approximated a normal distribution for log(K _ow) and a lognormal distribution for Sol_water and K _Henry, so these distribution shapes were used to model uncertainty. The variability parameters used to describe the distributions were taken from Marino (2006). A coefficient of variation of 70% was used for K _Henry. Sol_water and log(K _ow) are related to each other and this was evidenced in our database where there was a Pearson’s correlation coefficient of −0.81 between the two. Due to this strong correlation, a binormal distribution was used in the Monte Carlo simulation for these two parameters, with coefficients of variation of 50% for Sol_water (before log transformation), and 40% for log(K _ow).

Uncertainty in DHR—Monte Carlo analysis

The uncertainty associated with K _p was estimated by jointly generating 500 000 values of log(K _ow) and Sol_water in Monte Carlo simulation according to the distributions defined above. The log(K _ow) values were used with a fixed value for MW to calculate K _p values using Vecchia and Bunge’s equation (equation (3)). To reflect the QSAR uncertainty, an additional randomly generated value with normal distribution and mean of 0 and an SD of 0.8 was added to each K _p value (in the logarithmic scale).

The SD of 0.8 is the residual error of the QSAR equation reported by Vecchia and Bunge. The 500 000 simulated values of K _p were then used with the simulated Sol_water values to generate 500 000 values of J _max. The simulation was performed for each substance the actual standard deviation used to simulate values depends on each chemical’s log(K _ow) value, uncertainty of log(K _ow) was expressed as a CV(%) and not log-transformed.

DHR values were then calculated 500 000 times for each substance using the distributions obtained according to the methods described above. The simulation was performed assuming S _expo = 2 hands and T _expo = 8h.

The data distributions obtained from the 500 000 iterations approximated a log normal distribution and a GSD was estimated for each substance. This GSD, specific to each agent, is invariant to changes in surface or duration of exposure. As a result, the tool does not need to carry out an independent simulation with each use. The calculated GSDs were used in UPERCUT to graphically express the distribution surrounding the estimated DHR. Users are also shown the probabilities of: DHR >1 (dermal dose = inhalation dose); 0.1 <DHR <1 (dermal dose is between 10 and 100% of inhalation dose); and DHR < 0.1 (dermal dose < 10% of inhalation dose).

Validation

A partial validation exercise was carried out in which the K _p values used by the UPERCUT tool (estimated from Vecchia and Bunge, equation (3)) were compared to experimental skin permeation data from the EDETOX database (Williams, 2004). We extracted 231 available K _p values for 53 substances. These data were used to generate a unique K _p value for each of the 53 substances using the same approach used to create a single OEL value for substances with more than one OEL (Supplementary Appendix 1 is available at Annals of Occupational Hygiene online). The values were then compared to the UPERCUT K _p estimates and associated uncertainty distributions.

RESULTS

The UPERCUT tool is publically available online (http://www.i-s-t.ch/outils-en-ligne/modele-percutane/upercut/, see also the “other tools” tab at http://www.i-s-t.ch/outils-en-ligne/modele-percutane/upercut/, see also the “other tools” tab at http://www.expostats.ca). It contains data on 1686 substances. The proportions of these 1686 substances for which the UPERCUT output information types are available is provided in Table 1.

Table 1.

Availability of information on agents

Type of information	Percentage of 1686 substances for which the information is available
Risk phrases	58.3%
No risk phrase available, but there is a positive Draize test for dermal irritation listed in RTECS	11.8%
MAK skin notation	31.5%
TLV skin notation	26.3%
GHS acute oral toxicity classification	85.8%
GHS acute dermal toxicity classification	43.9%
GHS chronic-oral toxicity classification	41.6%
GHS chronic dermal toxicity classification	7.5%
Dermal penetration potential according to Magnusson et al. (2004)	100%
Carcinogen, mutagen, or teratogen (CMR) classification	39.2%
At least one existing OEL	36.7%
It is possible to calculate a DHR	72.1%

‘Calculated’ OELs

A total of 618 substances in the database had at least one OEL (in total, 5789 OEL values). Among these, there was a median of 10 OELs per substance (range 1–23), with 80% of agents with >1 OEL value. Within-substance variability was GSD_intra = 2.9 and between substance variability was GSD_inter = 19.7. Figure 2 shows the calculated OELs for the 618 agents, along with the actual range of available values in our databank of OELs (vertical segment) and an 80% confidence band estimated with GSD_intra.

Figure 2.

Illustration of variation in OELs for the same substance. The centre line represents the estimated values for each substance. The two outer lines represent the estimated range in which 80% of OELs would fall (according to the within-substance GSD).

‘Predicted’ OELs

We created models for the seven types of toxicological values. Basic features of the regression models between the calculated OELS and the seven types of toxicological values are presented in Table 2.

Table 2.

Characteristics of the calculated OEL prediction regression models

Type of toxicity endpoint	n a	α b	β c	R 2d	σ e
Acute-intravenous	142	−2.30	0.85	0.45	2.26
Acute-subcutaneous	139	−2.98	0.77	0.42	2.15
Acute-cutaneous	256	−2.96	0.69	0.33	2.09
Acute-intra-peritoneal	324	−1.65	0.68	0.29	2.24
Acute-oral	520	−2.21	0.67	0.26	2.38
Chronic-oral	345	1.27	0.28	0.12	2.44
Chronic-cutaneous	77	−1.46	0.61	0.39	2.11

^aSample size.

^bRegression intercept.

^cRegression slope.

^dCoefficient of determination.

^eResidual error.

When compared to models based on chronic toxicity data, models based on acute toxicity were based on larger sample sizes and stronger correlations were observed between acute toxicological endpoints and the available OELs. As a result, when several types of toxicity were available for one substance, the models based on acute toxicity data were prioritized. Higher priority was then given to models with higher R ² except for chronic-cutaneous LOAEL, which were given lowest priority because of the small associated sample size. The ranked model priorities were as follows (the number of substances for which each model was used to estimate a ‘predicted’ OEL in UPERCUT is in parenthesis):

• 1.Acute-intravenous (185)
• 2.Acute-subcutaneous (57)
• 3.Acute-cutaneous (181)
• 4.Acute-intra-peritoneal (178)
• 5.Acute-oral (321)
• 6.Chronic-oral (40)

• 7. Chronic-cutaneous (3)

DHR estimation

DHR was calculable for 1215 liquid and solid agents and 59 gasses. It was not possible to calculate DHRs for agents for which OELs or toxicological data were not available. For the purpose of reporting descriptive statistics of the DHR (Supplementary Appendix 3 is available at Annals of Occupational Hygiene online) a scenario of two hands exposed for eight hours was used for solids and liquids, and a scenario of the entire body exposed for eight hours was used for gasses. The estimated DHRs for the two hand, 8-h scenario are presented, along with 80% confidence intervals, in Fig. 3.

Figure 3.

Distribution of DHR values and associated 80% confidence interval stratified by type of occupational exposure limit. DHR are presented sorted by increasing value. They were calculated for a scenario of 2 hands exposed for 8h. Line DHR = 1 (1 hand, 30min) indicate a ratio of the internal to safe dose that would correspond to a DHR equal to 1 if this exposure scenario was selected. Values above this line suggest a health risk even for a minimal exposure scenario. Line DHR = 0.1 (whole body, 8h) indicate a value of the internal to safe dose that would correspond to a DHR equal to 0.1 if this exposure scenario was selected. Values below this line suggest the absence of a health risk even for a maximal exposure scenario.

Under the NIOSH criteria for assigning skin notations, dermal exposure should be considered significant and a notation assigned if the dose from the dermal route corresponds to 10% of the dose corresponding to inhalation at the OEL (DHR = 0.1) (NIOSH, 2009). Three hundred and six (24%) of the liquid and solid substances had a DHR>0.1 (50 or 85% for gasses), and 909 (76%) had a DHR>0.1 Y>1 (9 or 15% for gasses).

Uncertainty analysis

For the 618 substances that had at least one OEL available in the database, the uncertainty of the ‘calculated’ OELs were defined by a GSD of 2. The uncertainty for the ‘predicted’ OELs was defined with a GSD equal to the exponent of the model σ value in Table 2, ranging from 8 to 11 across the different regression models (Table 3). The variability between the estimated OELS across substances is also reported in Table 3. For ‘calculated’ OELs the variability was defined by a GSD of 20. The GSDs defining variability across the ‘predicted’ OELs ranged from 2 to 4.

Table 3.

Uncertainty and variability across substances of OELs depending on estimation method

Method	n (A)	Variability across substances (GSD)	Uncertainty (GSD)
Calculated OEL	618	20	2
‘Predicted’ OEL predicted from:
Acute-intravenous	185	4	10
Acute-subcutaneous	57	3	9
Acute-dermal	181	3	8
Acute-intra-peritoneal	178	3	9
Acute-oral	321	2	11
Chronic-oral	40	2	11
Chronic-dermal	3	2	8

(A) = Number of substances for which the method was used to obtain an OEL.

The estimated GSDs for K _p, J _max, and DHRs are presented in Table 4 for all of the substances for which calculation of these parameters were possible.

Table 4.

GSDs from uncertainty analysis and variability cross chemicals for K _p, J _max, and DHR

Parameter		N	Uncertainty (GSD)			Variability of metric across chemicals (GSD)
			Minimum	Maximum	Median
K p		1686	6.7	15.4	7.4	5.0
J max		1686	6.4	11.9	6.7	19.8
DHR	Calculated OELs	551	7.4	14.0	8.2	66.3
	Predicted OELs	664	7.4	31.5	20.5	20.2

For the two hands, 8- h scenario 909 of the 1215 (75%) liquids and gasses had an estimated DHR >0.1. However, if we require that the lower bound of the 80% confidence interval is above 0.1 in order to accept that the DHR is really >0.1 then 445 would be considered >0.1. Similarly, we could only conclude that 85 substances are <0.1 with 80% certainty instead of 306 for which the point estimate is<0.1.

Validation

Figure 4 presents the K _p values estimated by equation (3) (along with their 80% uncertainty bands), and the experimentally derived K _p values in the EDETOX database for the same substances. Based on their experimental results, the authors of the EDETOX database selected the K _p value that they considered most appropriate for each substance. These are also plotted in Fig. 4.

Figure 4.

Comparison of K _p values estimated by UPERCUT with experimentally derived K _p values from the EDETOX database.

DISCUSSION

Assessment of risk from dermal exposure is challenging due to the limited availability of tools for measuring and evaluating exposure by this route. Predictive models for estimating dermal exposure have been developed including ECETOC TRA (ECETOC, 2009), DREAM (van Wendel de Joode et al., 2003), and RISKOFDERM (van Hemmen et al., 2003) but these models estimate only the mass on the skin and provide no information on dermal penetration. The RISKOFDERM toolkit includes an algorithm for assessing the risk by combining the exposure estimate from the RISKOFDERM model with a hazard score for the substance in question, but these hazard scores are based on the general toxicity of the substance and do not take dermal permeability into account (Oppl et al., 2003).

There are also tools available that estimate dermal penetration. NIOSH has developed two such tools that are both available online (NIOSH, 2013). The NIOSH Skin Permeation Calculator estimates the coefficient of permeability (K _p) for a substance of interest and the dermal flux (J) using three different methods; Frasch (2002), Potts and Guy (1992), and modified Robinson (Wilschut et al., 1995). NIOSH has also developed the Finite Dose Skin Permeation Calculator, based on the work of Kasting and Miller (2006) which goes a step further. Users enter the mass per unit area of dermal loading along with substance characteristics, and the calculator estimates the mass that will be absorbed, assuming a finite loading of the substance on the skin, and taking into account losses to evaporation. IH SkinPerm is another available tool that estimates dermal absorption using the dermal loading, the exposure duration, and substance’s physico-chemical properties. It was developed by the American Industrial Hygiene Association (AIHA) and is based on the work of ten Berge (2009). Like the Finite Dose Skin Permeation Calculator it takes into account losses to evaporation and estimates the mass that is absorbed (Tibaldi et al. 2014).

The NIOSH and IH SkinPerm dermal permeation tools represent great progress in dermal exposure assessment. These tools are focussed on estimating the internal dose following dermal exposure. UPERCUT fulfils a different role to these models by providing an indication of the risk posed by the dermal dose. The inputs required for UPERCUT are also less specific to the dermal route of exposure so it is easier for practitioners with limited dermal exposure assessment experience to use. UPERCUT also includes a large database (>1500 chemicals) that contributes to its ease of use. Another unique feature of UPERCUT, among dermal penetration tools, is the uncertainty assessment. This is in keeping with its role as a risk assessment screening tool. UPERCUT can be used at the first stage in a dermal risk assessment to identify substances and scenarios that may pose a dermal risk. The Finite Dose Skin Permeation Calculator and IH SkinPerm could be useful during the second stage when occupational hygienists perform a dermal exposure assessment for the substances and scenarios identified by UPERCUT as potentially harmful.

UPERCUT estimates DHRs using a method similar to that currently used by NIOSH to develop skin notations. UPERCUT can estimate DHRs for 1215 different substances, enabling dermal hazard assessment for substances that have not yet been reviewed by NIOSH. Furthermore, unlike the skin notation system, UPERCUT estimates the uncertainty surrounding the DHR. There is uncertainty associated with both the estimated and calculated OELs due to the use of multiple OELs and toxicological values in their derivation. Table 3 clearly shows the difference between calculated OELs (low uncertainty and high variability across agents) and predicted OELs (high uncertainty versus relatively moderate contrasts across agents). There is also considerable uncertainty associated with the K _p and the J _max due to uncertainty surrounding the physico-chemical properties and the QSAR models that are used to estimate them. This uncertainty is transmitted to the final uncertainty for DHR, with a median GSD of 8.2 for calculated OELs (~200 ratio between the upper and lower limits of a 80% confidence interval), and 20.5 for predicted OELs (~2000 ratio between the upper and lower limits of a 80% confidence interval). Such uncertainty might appear very high, but it is a transparent method of communicating the propagation of error along the path of estimating the DHR. Furthermore, the DHR values vary across substances and exposure scenarios with GSDs much greater than the GSD for uncertainty even for predicted OELs, showing that it is possible to draw useful conclusions despite wide confidence intervals (Fig. 3).

The uncertainty analysis carried out by UPERCUT was included to ensure that the tool’s limitations were transparent to enable users to make decisions while taking into account uncertainty. The information provided on the probability that the dermal dose is either equal to, between 10 and 100% of, or less than 10% of the inhalation dose can help users to make a decision despite the range of possible DHRs that are presented. A partial evaluation of the model performance was presented in Fig. 4. This demonstrated the relationship between K _p values estimated by the QSAR used by Upercut, and experimentally derived K _p values. All but one of the experimentally derived K _p values fell within the 80% uncertainty band generated by the QSAR equation. This emphasizes the importance of considering the uncertainty surrounding UPERCUT estimates.

Although UPERCUT can be used as part of a dermal risk assessment, it is not a complete risk assessment tool. It is a hazard assessment tool that enables users to identify substances and scenarios that may be of concern and that require a more complete risk assessment. As such, it has a number of limitations. As discussed above, it does not take into account all of the processes that can affect dermal absorption, such as losses to evaporation or hand-washing, differences in absorption between different areas of the body, metabolization within the skin, dermal integrity, penetration enhancers, and occlusion (Semple, 2004; Sartorelli et al., 2007). Furthermore, the QSAR used to estimate dermal absorption was based on analysis of data from only a few hundred substances, so it may not be applicable to all of the chemicals available in the UPERCUT database (this may be particularly the case for very lipophilic substances). Also, the exposure scenarios used to estimate the dermal dose assume exposure to an infinite quantity of a saturated aqueous solution. This will rarely be the case in an occupational setting and a finite dose model would be more applicable (but also more complex). The tool is designed to assess the risk of systemic health effects following dermal absorption. It does not allow assessment of dermal sensitization. When substances that are dermal sensitizers are selected, a warning message is displayed. A similar warning message is displayed for possible carcinogens and mutagens to advise users that safe levels of exposure may not exist and that minimizing exposure should be a priority.

There are also limitations associated with the ‘OELs’ used in the DHR to assess the hazard posed by the estimated dose. Firstly, the ‘calculated’ OELs are based on OELs that were developed for exposure by the inhalation pathway. This includes OELs that may be relevant mainly to respiratory irritation, and the dose associated with such OELs may not pose a similar health risk when received by the dermal route. Also, some of the OELs had to be predicted based on mostly acute animal toxicology data, and the OEL prediction models, similar to other efforts (Suda et al., 1999; Whaley et al., 2000), had only moderate predictive power (R ² ranged from 0.12 to 0.45), also reflected in the much larger uncertainty associated with ‘predicted’ OELs.

Despite its limitations, UPERCUT provides an integrated picture of a large amount of available data. It includes a large database with information on over 1500 chemicals. It uses both QSAR modelling techniques and empirical toxicological data to present a portrait of the maximum amount of available data. It integrates information from several sources into a single metric, the DHR. This allows users to easily use and interpret the available information. In the absence of dermal exposure limits, and given the limited availability of skin notations UPERCUT answers a current need for a systematic approach for screening dermal exposure scenarios. It is a useful hazard assessment tool with quite simple input requirements for users that can be used by occupational hygiene generalists with little experience with dermal exposure assessment. It could also be a useful education tool to help occupational health, exposure and risk assessment students and professionals to understand the systematic health effects that can arise from dermal exposure, and how the associated risks can be evaluated.

Now that the UPERCUT framework has been developed, there is the potential for future development to create a more complete risk assessment tool. As more toxicological data on chronic-cutaneous exposure becomes available the ‘predicted’ OEL models could be updated. A more detailed dermal penetration model could also be incorporated to model doses from non-steady state and finite dose scenarios, and from exposure to mixtures, as well as losses to evaporation and washing. As it currently exists the UPERCUT tool fills a gap in dermal hazard assessment by estimating both the dose from the dermal route of exposure, and the potential health risk posed by that dose.

SUPPLEMENTARY DATA

Supplementary data can be found at http://annhyg.oxfordjournals.org/.

FUNDING AND DECLARATION

The project was funded by Agence nationale de sécurité sanitaire de l’alimentation, de l’environnement et du travail (ANSES, grant number EST-2007-003). M.G.N. was supported through personal research funds from JL for the writing of the manuscript.