[Frontiers in Bioscience E5, 418-434, January 1, 2013]

A step change towards risk assessment in the 21st century

Grace Y Patlewicz1, Deborah R Lander1

1DuPont Haskell Global Centers for Health and Environmental Sciences, 1090 Elkton Road, Newark, DE 19711, USA

TABLE OF CONTENTS

1. Abstract
2. Introduction
2.1. Regulatory background
2.2. REACH
2.3. Integrated testing strategies
2.4. Non-testing approaches
2.5. Exposure assessment
3. Case study on selected methacrylates
3.1. Physicochemical properties
3.1.1. LogKow (octanol/water partition coefficient)
3.2. Mammalian toxicity
3.2.1. Skin sensitisation
3.2.2. Mutagenicity (in vitro): Ames
3.3. Environmental fate
3.3.1. Biodegradation
3.4. Ecotoxicity
3.4.1. Acute aquatic toxicity in fish
3.4.2. Acute aquatic toxicity in D. Magna
3.5 Summary remarks
4. Exposure assessment on methyl methacrylate
4.1. Dermal sensitisation exposure assessment
4.2. Determination of the local effects dermal DNEL
4.3. Methods for exposure assessment
4.4. Discussion
5. Summary and Perspectives
6. References

1. ABSTRACT

Chemical Regulation and the means by which data is generated for the purposes of risk assessment is undergoing a tremendous shift. There is a strong impetus in Europe, in particular, to move towards non-animal approaches to address data gaps for specific endpoints either in lieu of testing or as part of weight of evidence approaches within integrated testing strategies (ITS). An Exposure assessment considering workers and/or consumers is a critical component of a robust risk assessment. The EU chemicals legislation REACH, for example, provides considerable flexibility in the application of non-testing approaches such as (Q)SARs, chemical categories and read-across for data gap filling. There have been a number of efforts aimed at developing technical guidance, tools, and techniques for non-testing and tiered exposure approaches. Despite these efforts, there remains limited practical insight about how these approaches can be applied in the assessment of substances. Here, we first provide a background of the available approaches and how they can and should be practically utilised to address REACH requirements.

2. INTRODUCTION

2.1. Regulatory background

Chemical Regulation and the means by which data are generated and translated into information for the purposes of risk assessment is undergoing a massive shift. One of the major drivers for this change comes from the move towards non-animal alternatives. Animal welfare concerns within Europe, in particular, have provided significant momentum to investigate potential alternatives to animal testing which encompass the 3 "R"s (refine, reduce, and replace animal testing). For example; the 7th Amendment to the Cosmetics Directive has placed a ban on the in vivo testing of cosmetics ingredients. The ban for acute testing came into effect in March 2009, the ban for repeat dose testing will commence in 2013 (1). The new chemicals legislation within the EU, REACH (Registration, Evaluation, Authorisation and Restriction of Chemical substances) specifically calls for the use of non-animal alternatives (2). Vertebrate testing should only be used as a last resort. REACH advocates the use of integrated testing strategies (ITS) as an efficient means of addressing the information requirements for a given endpoint. ITS approaches comprise multiple elements such as optimised in vivo tests, in vitro tests, (Q)SARs and chemical categories (3). The elements themselves are not new but the integration in a framework where the generated information is non-standard provides a challenge for interpretation and decision making.

Another significant driver is the general public who play a role in demanding that chemicals are evaluated and determined to be safe for their intended applications. This places a burden on both Regulators and Industry to demonstrate that relevant and sufficient information is available to enable robust risk assessments to be undertaken and reviewed in a timely manner.

Thus, the regulatory landscape is dramatically evolving, opportunities and challenges remain in determining what the critical aspects in the evaluation of adverse outcomes are and how to interpret them. We will discuss the REACH regulatory programme where the new framework exists but the practical implementation is still evolving.

2.2. REACH

REACH (Registration, Evaluation, Authorisation and restriction of CHemicals) is the new EU legislation that came into force in June 2007 (2) and which superseded Directive 79/831/EEC. Under Directive 79/831/EEC, Existing chemicals comprised substances introduced between January 1971 and September 1981 and were listed on EINECS (European INventory of Existing commercial Chemical Substances) whereas New Chemicals were those substances introduced subsequently and listed on ELINCS (European LIst of Notified Chemical Substances). Under Directive 79/831/EEC, information requirements under the Notification of New Substances (NONS) procedure were mandated only for new substances with tonnages in excess of 10 kg. Existing substances were not subjected to the same information requirements hence the complement of data supporting their use has been extremely variable. For the c.a. 100,000 existing chemicals, supporting information rarely exists. Under REACH, this inconsistency should eventually become harmonised as REACH calls for equivalent information requirements for all new and existing chemicals manufactured or imported at quantities of 1 tonne or greater per annum. The specific information requirements for REACH depend on tonnage bands which are described in Annexes VII-X of the REACH legal text. The original estimates were that some 30,000 chemicals would need to be re-evaluated leading to additional information requirements, but given the number of pre-registrations (c.a.143,000) (4), this estimate appears conservative. In any case, the number of chemicals that are likely to require re-evaluation is vast and this does represent a significant challenge in terms of cost, animal numbers and resources in order to address the necessary information requirements. The REACH legal text affords considerable scope to utilise and exploit alternative (non-animal) approaches. Annex XI provides the framework for fulfilling these requirements by other means (examples include in vitro, Weight of Evidence (WOE), (Q)SARs, chemical grouping etc) thus limiting vertebrate testing to the fullest extent possible. There is an obligation to carry out vertebrate testing only as a last resort and to consider all other options before performing or requiring testing as described by Articles 13(1) and 25(1) of the REACH legal text.

2.3. Integrated testing strategies

Integrated testing strategies (ITS) are structured workflows of the different elements required to conduct a risk assessment. They conceptually describe how exposure information and effects information obtained from (Q)SARs, read-across methods, and in vitro tests prior to in vivo testing can afford a more rapid, efficient, and cost-effective way to perform risk assessment of chemicals (3). The REACH technical guidance describes endpoint specific ITS, to illustrate the hazard characterisation requirements. The technical guidance identifies the available approaches e.g. in vivo test methods, in vitro, (Q)SARs etc in brief for that endpoint and how to evaluate their respective outputs. The workflow (the ITS) provides the framework for how the information available/generated should be integrated together to arrive at an overall conclusion for hazard characterisation purposes (e.g. Classification & Labelling (C&L) and/or Risk Assessment) (5). The ITS does not address the practical aspects of how to perform an integrated assessment, it merely provides a conceptual framework to encourage a step change in thinking - instead of a checkbox approach of gathering data, a hypothesis based approach is advocated that considers what critical elements are really pertinent for risk assessment/management decisions. This ideally should take into account a weight of evidence approach; exploiting known synergies between endpoints as well as knowledge of mechanisms/modes of action.

2.4. Non-testing approaches

Under REACH, a registrant may adapt standard testing requirements for its chemicals based on scientific, technical or exposure informed considerations. This assumes that the same level of information can be potentially obtained by means other than vertebrate testing. Non-testing approaches comprising (Q)SARs and chemical categories are outlined in Annex XI as strategies to adapt standard testing requirements (2). Annex XI provides specific wording for (Q)SAR use and the conditions that must be satisfied:

"Results obtained from valid qualitative or quantitative structure-activity relationship models (Q)SARs may indicate the presence or absence of a certain dangerous properties. Results of (Q)SARs may be used instead of testing when the following conditions are met:

  • Results are derived from a (Q)SAR model whose scientific validity has been established
  • The substance falls within the applicability domain of the QSAR model
  • Results are adequate for the purpose of classification and labelling and/or risk assessment and
  • Adequate and reliable documentation of the applied method is provided.

The Agency in collaboration with the Commission, Member States and interested parties shall develop and provide guidance in assessing which (Q)SARs will meet these conditions and provide examples."

The wording emphasises how information provided by (Q)SARs may be used in lieu of experimental data provided certain conditions are met. In practice, (Q)SAR information can be used either as replacements for experimental testing or more likely as part of a weight of evidence evaluation. Detailed technical guidance for the use of (Q)SARs is available from ECHA (5).

It is perhaps useful to briefly clarify the terminology of these conditions. Scientific validity refers to the internationally agreed OECD principles for the validation of (Q)SARs. These were adopted by the 37th Joint Meeting of the Chemicals Committee and Working Party on Chemicals, Pesticides and Biotechnology in November 2004. Guidance to describe each of the principles in turn is provided in the following references (6-7). Whilst the principles are helpful to characterise a given (Q)SAR model, the critical factor to consider in the application of a model is whether the substance of interest lies within the domain of applicability, i.e. is the (Q)SAR model relevant for the chemical under evaluation?

The concept of the domain of applicability was first defined as part of a series of background papers to the ICCA-LRI Setubal workshop (8). ECVAM hosted a workshop 2 years later (9) and proposed the following definition:

"The applicability domain of a (Q)SAR model is the response and chemical structure space in which the model makes predictions with a given reliability."

There is no single way to characterise a domain for a QSAR model. Approaches depend on the type of QSAR, the endpoint of interest and the underlying training set - i.e. domain characterisation has to be context dependent. One approach could be to characterise a domain on the basis of structural fragments if these are indeed the factors pertinent to the endpoint of interest, another may be to use the descriptors from the training set to compute ranges or probability densities etc Ultimately whatever domain approach is derived for a given model, this is merely a first step in the evaluation. Careful examination of the extent to which the model being used is able to provide reasonable predictions for other related substances (analogues) is still required to provide the confidence that a robust and reliable prediction is feasible for the substance of interest. Software programs have been developed to implement various approaches for domain evaluation - notable examples include AMBIT Discovery v0.04 (released May 2006 by Ideaconsult Ltd, Bulgaria) freely accessible from http://ambit.acad.bg/downloads/AmbitDiscovery/ and Domain Manager which is developed and commercialised by the Laboratory of Mathematical Chemistry (LMC) (University ''Prof. As. Zlatarov,'' Bourgas, Bulgaria).

The technical guidance on (Q)SARs addresses the principles, the need to rationalise the reliability and adequacy of a (Q)SAR result as well as some of the available software tools in some detail in Chapter R6 (5). Work under the former QSAR Working Group, a subgroup under the EU's Technical Committee for New and Existing Substances (TCNES) agreed on reporting formats (templates) to capture the key pieces of information needed for REACH for both a (Q)SAR model and its prediction. Two formats were proposed and these are known as the QSAR Model Reporting Format and QSAR Prediction Reporting Format, abbreviated as QMRF and QPRFs. Both of these formats are structured on the OECD Validation Principles. Illustrative examples of these formats are provided on the website of the former ECB (see http://ihcp.jrc.ec.europa.eu/our_labs/computational_toxicology/qsar_tools/QRF).

Annex XI also contains the specific wording for the use of grouping methods (read-across and chemical categories). Specifically:

''Substances whose physicochemical, toxicological and ecotoxicological properties are likely to be similar or follow a regular pattern as a result of structural similarity may be considered as a group or 'category' of substances. Application of the group concept requires that physicochemical properties, human health effects and environmental effects or environmental fate may be predicted from data for a reference substance within the group by interpolation to other substances in the group (read-across approach). This avoids the need to test every substance for every endpoint. The similarities may be based on:

(1) a common functional group,

(2) the common precursors and/or the likelihood of common breakdown products via physical and biological processes, which result in structurally similar chemicals, or

(3) a constant pattern in the changing of the potency of the properties across the category...."

Whilst the development of chemical categories and (Q)SARs are underpinned by the same principles of chemical similarity, there is no specific requirement to validate a category. Most likely this is because adhoc categories have been routinely used under the High Production Volume (HPV) programmes within the US and under the OECD. Under REACH, the adequacy and reliability of the category approach must be substantiated and documented in a format known as the Category (Analogue) Reporting Format (CRF/ARF). An ARF is used when a read-across is carried out for one substance to another. A CRF considers a group of three of more substances. The technical guidance for chemical grouping is described in more detail in Chapter R6 (5). Case studies (10) developed by the drafting group authors of Chapter R6 are available at the former ECB website (http://ihcp.jrc.ec.europa.eu/our_labs/computational_toxicology/doc/EUR_22481_EN.pdf).

Read-across is the data gap filling mechanism used to interpolate predictions as part of the category. The approach can be qualitative or quantitative depending on the type of data available. Quantitative approaches might make use of a trend analysis (derivation of a local QSAR) using the category members themselves or rely on external QSARs or expert systems to derive the necessary predictions. It is beyond the scope to discuss the expert systems and other (Q)SARs here. There have been a several reviews that describe the state of the art of (Q)SARs for a number of REACH endpoints. Examples include references 11-18. Moreover the technical guidance for the different endpoints under REACH describe the availability of different (Q)SARs for each of the endpoints in turn. These (Q)SARs are not accepted for regulatory use, they are simply provided as a examples of what is described in the peer reviewed literature or available as software tools (both commercial and public).

For categories, the OECD Toolbox (OECD TB) is probably the best known tool since its prototype release in early 2008. Phase 2, a 4 year project funded by the European Chemicals Agency, ECHA commenced late 2008. The Toolbox aids in the development, evaluation, justification and documentation of chemical categories. It can verify whether a substance is part of an existing established category e.g. US EPA, OECD HPV category. It also possesses the functionality to develop endpoint specific categories making use of mode/mechanistic/empirical/structural "profilers". It is envisaged that the OECD Toolbox will be extensively used in evaluating the categories and (Q)SARs submitted for REACH and potentially for other regulatory programmes. The current version of the OECD Toolbox is v2.3 and is freely available from the following websites: www.oecd.org/existingchemicals/qsar and www.qsartoolbox.org. At the time of writing, the version of the OECD Toolbox was v2.0.

Other tools that can play a role in evaluating categories include the Industry funded AMBIT, and the Toxmatch chemical similarity tool - both of which were developed by Ideaconsult Ltd, the latter as part of a JRC contract and are freely available from http://ambit.sourceforge.net/ and http://ihcp.jrc.ec.europa.eu/our_labs/computational_toxicology/qsar_tools/toxmatch websites.

Whilst there is extensive technical guidance for the development of chemical categories and the use of (Q)SARs for regulatory purposes such as REACH, there remains little practical guidance or example case studies for how some of these tools can or should be applied in either the evaluation of existing (Q)SARs or the formation of robust categories. No doubt further guidance will be published by ECHA in due course for REACH, we aim in section 3 to provide some representative examples to illustrate how the principles and approaches can be potentially used in practice.

2.5. Exposure assessment

REACH requires evaluation throughout the entire life-cycle of a substance for exposure to workers, to consumers, and to all spheres of the environment. An exposure scenario is developed for each use of the substance. Risk characterisation is achieved by comparing the final hazard benchmark to the estimated exposure for a particular use. If the exposure is below the hazard benchmark, the use is determined to be safe. Since measured data for all uses is seldom available, models are used to estimate exposures. There are several models recommended by ECHA (19) for exposure estimations and each model has different inputs and assumptions. In the example discussed in section 4, the results from the different models will be compared for a professional worker dermal exposure scenario and compared to available measured data.

3. CASE STUDY ON SELECTED METHACRYLATES

Butyl methacrylate (97-88-1) (BMA) and methyl methacrylate (80-62-6) (MMA) were chosen as case study substances simply because they have been previously assessed under the OECD HPV programme (20) and hence were likely to be associated with a reasonable complement of (eco)-toxicity data. A full EU risk assessment report has also been completed for methyl methacrylate (80-62-6) and can be found at the former ECB website under http://esis.jrc.ec.europa.eu/index.php?PGM=ora (21). There is no intent to modify or change any of the conclusions/recommendations from these assessments, merely we wish to illustrate how non-testing approaches can play a role to either complement existing data in a weight of evidence approach by providing consistent and corroborating estimates.

For a selection of endpoints, we will demonstrate how external (Q)SARs can be utilised to fulfil datagaps and the extent to which these estimates are concordant with known experimental values. A brief data matrix shown in Table 1 has been constructed for BMA and MMA using the summary information as reported in the respective SIDS documents (20). Selected endpoints from the four main domains: Physicochemical properties, mammalian toxicity, environmental fate and ecotoxicity will be illustrated.

3.1. Physicochemical properties

3.1.1. LogKow (octanol/water partition coefficient)

LogKow, the octanol/water partition coefficient is a parameter that is relied upon to give insight towards an array of properties cross cutting all these endpoint domains. LogKow serves as a good model for hydrophobicity. The hydrophobicity of a compound can provide an indication of how easily a substance might transverse across a cell membrane, whether it will be absorbed readily through the skin (22) or gut (23); whether it might be taken up in groundwater to pollute waterways (24). Many QSAR models for aquatic acute toxicity rely on LogKow as their driving factor due to a narcosis mechanism (25-26). Hydrophobicity is also useful in evaluating likely bioaccumulation potential (16).

There are numerous (Q)SAR models available both publically and commercially for the estimation of LogKow. Many of these are referenced in Dearden and Worth (27) as well as the Technical Guidance for REACH (5). One common model that is well known and commonly used is that of KOWWIN v1.68, developed by SRC Inc (formally Syracuse Research Corporation) and the US EPA (28). This model is routinely used as part of the US EPA's Premanufacture Notice (PMN) process (29) as well as the HPV programme under the OECD. Under REACH, a (Q)SAR model needs to be both characterised in accordance with its OECD principles and an assessment needs to be undertaken to substantiate that a substance of interest lies within the domain of applicability as defined for the (Q)SAR model (2). In an effort to illustrate this, we extracted the training set of compounds for KOWWIN v1.67 that is made available by the developers SRC, Inc. and the US EPA at http://esc.syrres.com/interkow/KowwinData.htm. At the time of writing, KOWWIN v1.67 was the most recent version. Since KOWWIN does not provide any explicit applicability domain aside from some guidance in terms of considering molecular weight and fragment descriptors as listed in the user manual; a domain was independently extracted using the commercial program, Domain Manager v1.02 software as cited previously. A structural domain on the basis of atom centred fragments was extracted owing to the fact that the (Q)SAR model itself as encoded in KOWWIN was based on structural fragments. The atom centred fragment approach is described in more detail in Dimitrov S et al (30). A set of rules are used to reflect the effect of different neighbours on a specified atom. The application of these rules allows the extraction of a set of atom centred fragments that can be used to characterise the structural domain of the atoms presented in a certain set of chemicals. The maximum and minimum values of molecular weight (MW) were also determined as a second criterion for assessing the KOWWIN domain. The MW range was as follows: Minimum MW 18.02, Maximum MW 719.92. In other words, for a substance to be considered within the domain of KOWWIN for our purposes, it had to be 100% within the structural domain as determined by atom centred fragments and have a MW value between 18 and 719. Butyl methacrylate and methyl methacrylate were introduced into the Domain Manager v1.02 software as a "test set" vs. the KOWWIN dataset as the "training set". The MW values of these substances are 142.18 and 100.12 which are well within the MW range of the KOWWIN training set. Both are also within 100% of the structural domain as defined by the atom centred fragments. On the basis of our domain assessment of KOWWIN, butyl methacrylate and methyl would be considered "preliminary valid" since they satisfy these two conditions we have nominally set for being "in domain". Accordingly, we would expect LogKow estimates for both substances to be reasonable. Belonging to an applicability domain of a model increases the likelihood of a reasonable prediction but is not a guarantee. Some evaluation of related analogues and their predictions relative to their experimental values needs to be conducted to substantiate that any (Q)SAR estimate is indeed both reasonable and predictive. These analogues may be taken from the training set of the (Q)SAR model itself, i.e. are there any related substances within KOWWIN or from other sources and how do their predicted/experimental LogKow values compare. In this case, Toxmatch v1.06 (Ideaconsult Ltd) was used to search and retrieve analogues from the KOWWIN training set to enable a comparison of experimental and estimated LogKow values for butyl methacrylate. The training set of KOWWIN model was imported into the Toxmatch v1.06 software. The Tanimoto distance (fingerprints,kNN) was chosen as the similarity index approach. The "Tanimoto distance (fingerprints,kNN)" method calculates the average Tanimoto index between the fingerprints for each query chemical within the KOWWIN dataset and the fingerprints for the k most similar chemicals from the set (where k = 10 as default). The most similar chemicals are those with the highest Tanimoto index values. Butyl methacrylate, our substance of interest was then imported as the test set chemical and a pairwise similarity performed using the Tanimoto distance relative to the training set. The most similar analogues according to this Tanimoto distance index were then reviewed, where an index of 0.7 was arbitrarily taken as a quantitative measure of "most similar". Five other analogues were identified which are shown in Table 2 together with their respective similarity index values, experimental LogKow values and estimated LogKow values (by KOWWIN). All the methacrylates were reasonably predicted by KOWWIN v1.67 compared with their reported experimental LogKow values. Accordingly the estimated LogKow of butyl methacrylate, is considered to be a reasonable one. KOWWIN actually has butyl methacrylate as part of its training set of compounds. The estimated and experimental values are in good agreement with each other. The same procedure was conducted for methyl methacrylate, results not shown.

LogKow is well recognised as an important parameter for a range of different endpoints. KOWWIN v1.67 is an example of one (Q)SAR model that enables the prediction of LogKow. A evaluation of KOWWIN v1.67 with respect to the OECD principles is required to meet the conditions for (Q)SARs as laid out in Annex XI. Here we have described one of the principles - domain of applicability and illustrated how this can be extracted using the available training set of data from KOWWIN v1.67. We have applied this approach to butyl methacrylate and methyl methacrylate and found both to satisfy the two criteria specified for domain inclusion. To substantiate that the predicted value of butyl methacrylate was reasonable, an evaluation of related analogues within the training set using Toxmatch v1.06 identified 5 other analogues with predicted and experimental values in good agreement. This demonstrated that methacrylates are indeed well represented as a chemical class in the training set with robust predictions. Subject to the appropriate documentation being drafted (i.e. QMRF, QPRF) to substantiate this evaluation made, the conditions of use for REACH have been met.

3.2. Mammalian toxicity

3.2.1. Skin sensitisation

Much is understood about the mechanisms underpinning skin sensitisation. Further information can be found in the following references (31-32). A substance must negotiate a number of steps before sensitisation is induced. A sensitising chemical must penetrate through the stratum corneum and form a stable association with a carrier protein in order to deliver dermal trauma sufficient to induce and upregulate epidermal cytokines. These processes are necessary for the mobilisation, migration and maturation of LC, and for the chemical to be inherently immunogenic in such a way that a T lymphocyte response of sufficient magnitude is stimulated (32). All these steps are not considered to be equally important. The step which is dependent on the sensitising chemical itself, the rate determining step, provides insight on how to evaluate the skin sensitisation potential of a given chemical and has been discussed at some length in references (33-34). Efforts to predict skin sensitisers typically focus on identifying electrophilic features in chemicals (structural alerts) and relating these back to skin sensitisation potential.

There have been many efforts to develop models that relate the electrophilic features of chemicals to the skin sensitisation potential as shown in references (35-36). Most recent efforts have been described by Roberts and Aptula (37) who described a set of principles that characterised sensitisers by their reaction mechanisms which have since been implemented into Smiles ARbitrary Target Specification (SMARTS) codes for easy re-use (38) (available as a module within Toxtree, from the former ECB website, see http://ihcp.jrc.ec.europa.eu/our_labs/computational_toxicology/qsar_tools/toxtree). Other initiatives have focused on the development of expert systems. Derek for Windows (DfW) is a well known knowledge based system that encodes chemistry and toxicity information in the form of toxicophores (39-40). The hybrid expert system TIssue MEtabolism Simulator for Skin Sensitisation (TIMES-SS) encodes structure toxicity and structure metabolism relationships through a number of transformations simulating skin metabolism and interaction of the generated reactive metabolites with skin proteins. The skin metabolism simulator mimics metabolism using 2D structural information. Metabolic pathways are generated based on a set of 236 hierarchically ordered principal transformations including spontaneous reactions and enzyme catalysed reactions (phase I and II). The covalent reactions with proteins are described by 47 alerting groups (structural alerts). Some of these alerts are additionally underpinned by mechanistically based 3D-QSARs to refine the predictions. These 3D-QSAR models depend on both the structural alert and factors that influence its reactivity - steric effects, molecular size, shape, solubility, lipophilicity and electronic properties (41-43). The rules implemented into TIMES-SS have also since been incorporated into the OECD Toolbox.

Butyl methacrylate and methyl methacrylate were processed within TIMES, the OECD Toolbox and the SMARTS implementation within Toxtree to gain some insights about likely skin sensitisation potential.

The SMARTS implementation within Toxtree identified both methacrylates as potential Michael acceptor. This is commensurate with the reaction mechanistic domains outlined in Roberts and Aptula (37). The same alerting group is flagged by the protein binding alerts as encoded in the OECD Toolbox v2.0. The rationale provided is as follows:

"Michael acceptors are double or triple bonds with neighbouring electron-withdrawing group. They can clearly be seen to have an electron-deficient double bond that is susceptible to nucleophilic attack. Michael-type addition provides a means of covalent adduct formation at an electrophilic centre, without any leaving group. Direct addition of a nucleophile can take place across a double or triple carbon-carbon bond if it is attached to a highly polarised substituent that permits the resultant negatively charged transition state to be stabilised, as for example in acrolein, acrylamide. Compounds with double or triple bounds adjacent to a C=O group are known as alpha,beta-unsaturated carbonyl compounds. Nucleophiles will undergo conjugate additions with them. C=O group profoundly affects the reactivity of the double or triple bond. Isolated C=C bonds are nucleophilic but conjugated C=C bonds are electrophilic. Proteins are good nucleophiles for conjugate addition reactions with these compounds. In the LLNA, most of acrylates and methacrylates are much less potent than expected from their reactivity. This is believed to be partly due to their volatility (for lower homologues such as ethyl acrylate, theoretical EC3 = 1.3 %, experimental EC3 = 28 %) and partly due to their rapid polymerisation (with loss of electrophilic reactivity) when exposed to air. Acrylates and methacrylates usually fail to show their true potential to sensitise because of polymerisation and in some cases evaporation under normal LLNA conditions and under many occupational and domestic exposure conditions".

Within TIMES, both methacrylates form part of the training set, in that there is a flag that experimental results exist. For butyl methacrylate, this value is an assignment made by the BfR (44) which categorises butyl methacrylate as a questionable or unlikely sensitiser and methyl methacrylate is cited as a very weak - non-sensitising. The respective predictions from TIMES were non-sensitiser for butyl methacrylate and "unable to predict" for methyl methacrylate owing to TIMES being unable to discriminate between a weak and non sensitising classification. An assessment of domain is automatically conducted for butyl methacrylate within the TIMES system. TIMES uses a modular approach to domain assessment starting with global requirements, structural domain, metabolic domain and interpolation domain. More explanation is provided in the following references that describe how TIMES processes its structures and assesses the relevance of the predictions made. Reference (42) in particular describes the overall characterisation of the TIMES sensitisation model with respect to all the OECD principles. Data from the OECD Toolbox for butyl methacrylate cites several data points, 3 from guinea pig assays, as Freund's complete adjuvant test and a final data point which is cited within the TIMES training set. In two out of the three guinea pig cases, the overall call was negative. Overall, the weight of evidence would support a negative call for butyl methacrylate. For methyl methacrylate, the calls are conflicting with a sensitising call in guinea pigs and a negative call in the LLNA owing to a reported EC3 of 60% and 90% in Acetone and AOO respectively where AOO is an acetone/olive oil mix (45). The OECD Toolbox was used to collect data on related methacrylates; on the basis of a qualitative read-across, based on 5 nearest neighbours (Figure 1), an overall call of positive was inferred for methyl methacrylate. In Figure 1, the nearest neighbours are marked in grey circles, the target substance MMA is a black triangle (note data exists for MMA hence there is more than one triangle reflected in the figure) and other related analogues as black squares.

Butyl and methyl methacrylate have the potential to be sensitisers due to their ability to react via a Michael addition route. Experimental evidence available within TIMES and the OECD Toolbox suggest that sensitisation is unlikely to be observed for butyl methacrylate. Equivocal data exists for methyl methacrylate and on the basis of read-across within the Toolbox, a conservative estimate of sensitising would be made.

3.2.2. Mutagenicity (in vitro): Ames

Covalent bond formation as a rate determining step is not unique to skin sensitisation. Schultz et al. (46) described a conceptual framework for predicting the toxicity of reactive chemicals where plausible molecular initiating events were based on covalent reactions with nucleophiles in proteins or DNA, and would ultimately lead to a variety of different adverse outcomes such as aquatic fish toxicity, mutagenicity, hepatocyte cytoxicity or respiratory toxicity. Aptula and Roberts (37) illustrated this concept using aquatic toxicity and sensitisation as example endpoints. Electrophilicity is well known to be an important factor in driving mutagenicity and carcinogenicity (47). Seminal work by Ashby and Tennant (48-50) described structural alerts for carcinogenicity. They found that the electrophilicity of chemicals correlated very well with mutagenicity in the Ames test. Similar findings have been published by Benigni and co-workers (51). His extensive evaluations on structural alerts developed by Ashby and Tennant (48), Kazius et al. (52) and Woo, (53) have been additionally implemented into the software program Toxtree (54). Mekenyan et al. have developed the TIMES system for the prediction of Ames mutagenicity (55) and in vitro chromosomal aberrations (56). The structure and implementation is similar to that for skin sensitisation. Structure toxicity and structure metabolism relationships are encoded through a number of transformations simulating liver metabolism and interaction of the generated reactive metabolites with DNA.

Butyl methacrylate was processed through the OECD Toolbox and TIMES to determine what alerts if any were flagged and what the likely prediction in Ames was. The OECD Toolbox had experimental Ames data with/without metabolic activation which showed butyl methacrylate to be negative. A review of alerts identified no alerts from the DNA OASIS binding scheme but a polarised alkene/Michael addition flag from the DNA OECD binding profiler. The following rationale was provided: "An initial Michael addition mechanism has been suggested to be primarily responsible for the ability of these chemicals to alkylate DNA" (57). TIMES reported an overall negative call for butyl methacrylate since it formed part of the training set and was associated with a negative Ames result. It did flag an alerting group on account of the Michael addition mechanism and this actually drove a positive prediction. Butyl methacrylate was within 90% of the structural domain and 100% of the alert performance. It was categorised as being outside the domain on account of the 100% criterion for structural domain not being met.

Within the TIMES chromosomal aberration model, butyl methacrylate is predicted positive, the same alerting group for Michael addition being triggered which drives the potential response.

Thus the OECD Toolbox highlighted the same Michael addition route as potentially contributing to positive outcomes without activation in Ames and in vitro chromosomal aberration assays. The TIMES system predicted butyl methacrylate as being positive on account of this alerting group, however for Ames, it refuted the prediction in favour of the available experimental data.

A category approach was attempted within the OECD Toolbox to make complementary predictions of Ames and chromosomal aberration results. In this case, analogues with available data and similar with respect to this reaction mechanism were identified. The Oncologic profiler within the OECD Toolbox categorised butyl methacrylate as a substance containing a reactive acrylate functionality. This resulted in 1079 potential analogues including butyl methacrylate. Of these analogues, only 36 were associated with Ames experimental data (Ames without S9 since it was assumed that Butyl methacrylate was a direct acting electrophile). A subcategorisation was performed to ensure that the analogues used in the read-across were similar to butyl methacrylate in that they did not contain any other functional groups that would enable an alternative reaction pathway to occur. On the basis of the 5 nearest neighbours, butyl methacrylate is inferred to be negative in Ames. The category members substantiate the hypothesis that despite the potential for a reaction to occur with DNA, in practice, methacrylates and acrylates are not found to be experimentally mutagenic in Ames. Table 3 lists the category members together with their CAS numbers, names and SMILES codes.

A similar read-across exercise was conducted to elicit an estimated call for in vitro chromosomal aberration. In this case, from the starting set of analogues categorised by the Oncologic profiler, only a handful of analogues remained which inferred a likely positive outcome for butyl methacrylate. However in evaluating the analogues (Table 4) more closely, we see that the LogKow value of butyl methacrylate far exceeds that of the remaining 3 analogues, suggesting that this result is subject to considerable uncertainty since this would be an extrapolation for butyl methacrylate which lies outside of the domain of the read-across. The 3 analogues with available experimental data all are significantly less hydrophobic than butyl methacrylate.

A second search was conducted in Leadscope (http://www.leadscope.com), a data mining tool that contains structured data for a number of mammalian endpoints. A similarity search using butyl methacrylate as a key together with a flag to extract in vitro chromosomal aberration study results was undertaken. Table 5 identifies the substances and their associated data.

Again the set of data is limited but it reveals that short chain acrylates and methyl methacrylate have been shown to be positive. This provides an indication that butyl methacrylate may be a potential clastogen as part of a weight of evidence approach, even though a qualitative read-across for butyl methacrylate could not be reliably performed.

3.3. Environmental fate

3.3.1. Biodegradation

Catalogic (58), an expert system built on a same platform to that of TIMES, was used to make an estimate of ready biodegradation on the basis of the OECD MITI 301C study protocol. The Biochemical Oxygen Demand (BOD) value generated was 0.853, i.e. butyl methacrylate is likely to be readily biodegradable since the 60% threshold for 28 days is met on the basis of this 301C test guideline. In this case, within 28 days, 85% of BMA is expected to degrade, thus satisfying the criteria of ready biodegradability.

3.4. Ecotoxicity

3.4.1. Acute aquatic toxicity in fish

A reported value of 5.6 mg/L 96 hr LC50 for O latipes from the OECD Toolbox was actually available for butyl methacrylate. The profilers within the OECD TB assign BMA to a category of unspecific toxicity due to the methacrylate functionality. Verhaar et al (59) as encoded in the TB, places it in Class 3 (unspecific toxicity), OASIS MOA places it in unspecific reactivity and ECOSAR categorises it as a methacrylate vs. neutral organic (whereupon the mode of action would be simply narcosis and hence driven by hydrophobicity (25-26)).

ECOSAR, one of a suite of tools developed between SRC Inc and US EPA, was used to derive an acute fish LC50 prediction (29). A 96 hr LC50 of 5.55 mg/L was predicted based on the methacrylates QSAR and relying on the experimental LogKow value of 2.99 as reported in the SIDS document (20). The fish 96 hr LC50 values used to develop the SAR were measured and the octanol water partition coefficients (Kow) were calculated using KOWWIN v 1.67. The actual reported SAR equation is: Log 96 hr LC50 (mmol/L) = -0.6845 (LogKow) + 0.6381 The LC50 is in millimoles per litre (mM/L); N = 12 + 4; and the coefficient of determination (R2) = 0.7139. The dataset underpinning this QSAR is actually based on results from 9 different substances, one of these is subject to confidential business information (CBI), the remaining substances include methyl methacrylate amongst others. To evaluate the QSAR model further, the training set as reported in the ECOSAR Help manual was extracted as shown in Table 6.

This training set provided a basis for defining an applicability domain. The ranges of estimated LogKow and MW values together with the mode of action as being methacrylate as per ECOSAR's classes was proposed as appropriate criteria for a domain. The LogKow and MW ranges were as follows: 0.3-2.98 and 100-248. Evaluating the ranges with respect to butyl methacrylate which has an estimated LogKow of 2.75 and a MW of 142.20 shows that the model should be appropriate for use. However as a second step, an attempt was made to re-derive the QSAR based on the training set to verify that the regression equation and corresponding reported statistics were reasonable. In fact, a simple linear regression performed within Minitab v15 statistical package resulted in a different regression equation with substantially poorer statistics than those reported by the ECOSAR user manual. The equation was as follows: Log 96 hr LC50 (mmol/L) = -0.7661 (LogKow) + 0.7578. The coefficient of determination in this case was only 0.554.

From the line plot depicted in Figure 2, it is evident that there is variability in the experimental values and a great deal of scatter with some substances notably allyl methacrylate (96-05-9) being poorly predicted. This raised some uncertainties in relying on the ECOSAR estimate alone and a trend analysis was attempted in the Toolbox to establish a more reliable estimate for the 96 hr LC50 in fish.

A category was defined based on the ECOSAR classes. This resulted in 116 substances. Of those, there were 13 substances with LC50 data in fish (a variety of species). Removing substances with very low water solubilities and subcategorising to remove additional functional groups resulted in a set of 9 analogues including butyl methacrylate from which an interpolated LC50 could be derived. A LC50 mol/L of 2.68E-04 mol/L was estimated for butyl methacrylate which equated to 38.1 mg/L. This appears to be a reasonable estimate of the potential toxicity of BMA and within an order of magnitude of that observed in O. latipes as referenced in the OECD Toolbox.

The ECOSAR estimate for methyl methacrylate gave rise to a conservative estimate of 49.43 mg/L. Methyl methacrylate lies within the domain of the model, on account of forming a part of the training set. However as evidenced from the evaluation already conducted, the ECOSAR model is not very robust and fails to provide reliable realistic predictions of methacrylates. Using the OECD TB, an estimated value was derived which provided a comparable yet conservative value of 82.8 mg/L. The OECD SIDS dossier (20) quoted values ranging from >79 mg/L in O. mykiss to 191 mg/L in L. macrochirus.

3.4.2. Acute aquatic toxicity in D. Magna

Methyl methacrylate was assessed for its toxicity to Daphnia magna. An ECOSAR prediction was derived which gave rise to an estimated 48 hr LC50 of 33.112 mg/L. A closer evaluation of the SAR underpinning this prediction was performed and ascertained that the SAR within ECOSAR was based on a sole data point from a substance subject to CBI. The SAR quoted is given as 48 hr LogLC50 (mmol/L) = 0.377 - 0.6214 (LogKow). It is unclear how a regression equation was arrived from this datapoint. For REACH, it is unclear how the conditions of use could be fulfilled since the training set of 1 chemical is insufficient. Accordingly a trend analysis on the basis of deriving a category within the OECD TB was conducted. Initially a category was created on the basis of the OASIS Mode of Action to extract out a starting category of 29,309 substances. From that starting list, a trend analysis was undertaken to extract potential analogues that had associated D. magna 48 hr EC50 data. 135 analogues were identified with associated experimental data. These were filtered to remove substances with low predicted water solubility. The remaining analogues were subcategorised to remove inorganics, mixtures and dissociating substances. This resulted in 116 analogues, from which further subcategorisations were conducted to remove substances not acting by the same reaction mechanism. The remaining data points resulted in a fair correlation from which an estimated EC50 of 1.38E-03 mol/L (138 mg/L) was derived (results not shown). This is in reasonable concordance with the reported EC50 in the Toolbox (84 mg/L) as well as the SIDS experimental value of 69 mg/L (20).

3.5. Summary remarks

In the previous sections, we have outlined the REACH framework and the scope and flexibility of using non-testing approaches to address information requirements. For a selection of endpoints for 2 compounds, we have shown the practical possibilities of using different tools to evaluate existing (Q)SAR models, substantiate their estimates with analogues as well as derive new (Q)SAR models. The results were promising with estimates being derived that corroborated the experimental data available. In spite of the apparent stringent conditions of use required for non-testing approaches, we have demonstrated with practical examples how compliance for hazard characterisation may be adequately addressed for REACH.

4. EXPOSURE ASSESSMENT ON METHYL METHACRYLATE

Exposure assessment is a critical aspect of any risk assessment. Here we illustrate an exposure assessment for methyl methacrylate focusing in on dermal exposure given the hazard profile of this compound and its sensitising potential. Dermal exposures can be estimated under REACH using either measured data or recommended models. A toolbox of tiered models are described in the ECHA guidance to address both occupational and consumer exposures. The simplest models, Tier 1 models require few inputs and give rise to conservative outputs. Acceptable risk values obtained from Tier 1 models as quantified by risk characterisation ratios (RCR) less than 1 usually merit no further action. Tier 2 models require more detailed inputs. Occupational exposure estimation is outlined in Chapter R14 of the ECHA guidance (19). The ECETOC Targeted Risk Assessment Model (TRAM) is described as an acceptable Tier 1 model (60). The RCR is the estimated exposure divided by the derived no effects level (DNEL). The other dermal exposure model recommended for REACH is RISKOFDERM which is considered a Tier 2 model as it requires more detailed input (61).

4.1. Dermal sensitisation exposure assessment

A dermal sensitisation exposure assessment was undertaken for methyl methacrylate (MMA) using the ECHA guidance. Models available for estimating dermal exposure were compared to published dermal exposure data in order to determine whether there was control of risk for REACH. The exposure scenario considered was a dental technician working in a dental laboratory for 8 hrs a day. The scenario considered was a dental technician handling monomer MMA liquid whilst manufacturing or repairing orthodontic splints and dentures. Dermal exposure and inhalation exposure could therefore occur as the technician poured liquid MMA into moulds and added the polymerisation initiator.

4.2. Determination of the local effects dermal DNEL

MMA has been identified as a skin sensitiser. There is ample human evidence of skin sensitisation from workplace contact with MMA resulting in contact allergy (62). The former ECB published a risk assessment for MMA in 2002 (21). In this document, only systemic long term effects from dermal exposure based on the older Estimation and Assessment of Substance Exposure (EASE) model was addressed. The resulting daily dermal exposure estimate was 1000 micro g MMA/cm2. Under REACH, local effects such as sensitisation must also be considered. Data from the local lymph node assay (LLNA) on the skin sensitisation potency of MMA (45) was available. The ECHA guidance from R.8 Appendix 8-10 Skin Sensitisation and the EC3 value from LLNA was used to derive a DNEL for local dermal effects. The EC3 is the effective concentration for a stimulation index (SI) of 3 in proliferation of lymph node cells. A SI of 3 in one or more test concentration categorises that substance as a sensitiser. The EC3 value was reported as 60% (categorised as a weak sensitiser; (63)) and ECHA guidance indicates this is considered as a LOAEL for induction (45). Following the R.8 Appendix 8-10 guidelines, the EC3 value can be converted from % to micro g/cm2 by multiplying by 250 i.e. EC3 = 60% * 250 = 15,000 micro g/cm2. The assessment factors are given in Table 7. The resulting local effects dermal DNEL is determined by dividing the EC3 by the overall AF. Thus, the Local effects dermal DNEL is 15,000 micro g/cm2 / 250 = 60 micro g/cm2.

4.3. Methods for exposure assessment

Models can be used to estimate potential dermal exposure to the dental technicians based on defined exposure scenarios as follows. The dental technician (considered a professional worker under REACH, rather than an industrial worker) works 8 hrs a day. He/she mixes and pours small quantities of liquid MMA into moulds and after the cure process handles the finished product containing residual levels of MMA. The activities performed need to be matched with the ECHA R.12 Use descriptor system. Perfect matches do not always exist but in this case there were two possible matches. The most reasonable process categories (PROC) were PROC 15, use of substance at small laboratory scale (1 L or 1 kg) or PROC 13 treatment of articles by dipping and pouring. The first model considered was ECETOC-TRAM, the Tier 1 model. When a Tier 1 outcome indicates control of risk there is usually high confidence in the results. In practice, deciding on the appropriate PROC can be difficult which results in additional uncertainty as in this case. The inputs to the TRAM for worker exposure and are listed in Table 8 and the outputs from the TRAM are listed in Table 9.

The model reports potential systemic dermal exposure. This can be converted to a dose per surface area. The exposure occurs on the hands and for this scenario a typical male worker was used to match the defaults of the TRAM. The surface area exposed is 820 cm2 and the body weight is 65 kg which enables conversion to micro g/cm2 as follows:

(0.34 mg/kg/day x 65 kg)/ 820 cm2 = 0.027 mg/cm2 or 27 micro g/cm2 for PROC 15

(13.7 mg/kg/day x 65 kg)/820 cm2 = 1.06 mg/cm2 or 1060 micro g/cm2 for PROC 13

These results are then divided by the DNEL to obtain the risk characterisation ratio (RCR). As can be seen from Table 9, PROC 15 shows control of risk but PROC 13 does not. In addition, the ECB report (21) as written would not have demonstrated control of risk under REACH. At this point it is difficult to feel confident that PROC 15 is sufficiently conservative for this scenario since both PROC 13 and the ECB report are more than 30 times higher (21). A higher tier model such as RISKOFDERM can be utilised which incorporates more information into the scenario. In RISKOFDERM, the type of activity is selected from six choices. For this scenario, filling, mixing and loading was deemed the best choice. The selected activity had specific questions to answer. The inputs selected are listed in Table 10. The 0.005 kg/min application rate results in the use of 2.4 kg/day which was selected to be higher than PROC 15 to ensure conservatism in the inputs. The text inputs are selected from a pull down menu where more information is given on the options.

The 90th percentile result from RISKOFDERM was an exposure rate of 0.6 mg/min which when multiplied by 480 minutes results in an exposure to the hands of 0.36 mg/cm2 or 360 micro g/cm2. This results in an RCR of 6 which still does not show control of risk.

It must be remembered that these models evaluate potential dermal exposure. Neither of these models incorporate the fact that MMA is quite volatile and will evaporate off the skin while it is trying to absorb into the skin. ECHA R.14 Occupational Exposure Estimate details in Appendix 14.1 how to take the evaporation rate into account (19) with the following equation where the inputs are explained in Table 11:

t(s) = (mRT)K

(MbetapA)

The total potential mass from the RISKOFDERM model is input into the equation. The only question is the dermal exposure area. In this scenario, there is no expectation for contamination to spread all over the 2 hands back and front since work is done on a small scale in a small area. The conservative approach is to put in a small area which will result in a longer time on the skin such as a finger tip or 2 cm2. The resulting time on skin is 53 minutes rather than the entire workday of 480 minutes. When the 53 minutes rather than the 480 minutes are multiplied by the 0.6 mg/min exposure rate the refined result from RISKOFDERM is 39 micro g/cm2.

As can be seen, the RISKOFDERM model with the inclusion of the evaporation effect is able to show control of risk and reinforces the earlier selection of PROC 15 which did not include the evaporation effect. To put into context, it is important to compare how these model results concur with actual measured dermal exposures. A recent publication which measured the potential dermal exposure to MMA for dental technicians was identified (64). Potential dermal exposures to the hands were evaluated using patches made of activated charcoal, one sandwiched between two layers of cotton fabric, which was attached to gloves donned by the technicians. The use of activated charcoal did not allow for MMA evaporation once adsorbed into the charcoal. Consequently, the results showed very high retention of MMA with a maximum of 25 mg/cm2 and an average of 2.5 mg/cm2; the estimated dermal exposure (95th percentile) was 8 mg/cm2 based on the standard deviation of 3 mg/cm2. The total surface area of the charcoal patches was 60 cm2. The resulting potential dermal exposure was 8 mg/cm2 x 60 cm2 = 480 mg/day. Since the experimental data is based on 60 cm2, this value is used to estimate the time on the skin. The resulting time on skin is only 3 minutes. RISKOFDERM can be used to estimate the actual dermal exposure in these experiments by forcing the model to output 480 mg/day (based on 60 cm2) over 480 minutes and then using that resulting exposure rate with the actual time on skin. The estimated exposure rate was 1 mg/min, so the estimated actual dermal exposure was 3 mg/day over the measured area of 60 cm2 or 50 micro g/cm2. The final results are listed in Table 12 and indicate control of risk since the RCRs are less than 1.

4.4. Discussion

This exercise has shown how the tiered approach to risk assessment is performed when there is little to no experimental data available. Incorporating both details in the activity performed and physical properties of the substance can result in significant refinement and increases the confidence in the results. Based on this example, it probably would have been reasonable to use PROC 15 and stop there but the insights derived from the RISKOFDERM model and the evaporation rate increased overall confidence as part of a weight of evidence approach. Searching the literature and finding experimental data that can be compared to the models increases the level of confidence further. In addition, it demonstrates that experimental data is not always able to measure actual exposure and it can be necessary to refine both model and experimental data to arrive at meaningful dermal exposure estimates.

5. SUMMARY AND PERSPECTIVES

Regulation is changing and new approaches need to be exploited in an effort to identify hazards, evaluate exposure and perform robust risk assessments in a more timely and cost effective manner. REACH is one such example where a new framework has been laid out which advocates the use of alternative approaches for data gap filling and provides exposure assessment models to perform risk assessment requirements. We have aimed to show a practical perspective on this framework by providing illustrative examples of how non-testing approaches can be undertaken and how exposure assessments can be performed. This is merely one step in the change towards a new risk assessment paradigm for the 21st Century.

6. REFERENCES

1. C Eskes, V Zuang: Alternative (Non-Animal) Methods for Cosmetics Testing: Current Status and Future Prospects. A report prepared in the context of the 7th Amendment to the Cosmetics Directive for establishing the timetable for phasing out animal testing. Altern Lab Anim 33, 1-18 (2005)
PMid:15804212

2. EC: Regulation (EC) No. 1907/2006 of the European Parliament and the Council of 18 December 2006 concerning the Registration, Authorisation and Restriction of Chemicals (REACH), establishing a European Chemicals Agency, amending Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/EEC, 93/105/EC and 2000/21/EC. Off J Eur L396, 1-849 (2006)

3. CJ van Leeuwen, GY Patlewicz, AP Worth. Intelligent Testing Strategies. In: Risk Assessment of Chemicals: An Introduction. Eds: CJ van Leeuwen, T Vermeire, Springer, New York (2007)
http://dx.doi.org/10.1007/978-1-4020-6102-8_11

4. http://apps.echa.europa.eu/preregistered/pre-registered-sub.phpx

5. http://guidance.echa.europa.eu/docs/guidance_document/information_requirements_en.htm (accessed 30 October 2010)

6. AP Worth, A Bassan, A Gallegos, TI Netzeva, G Patlewicz, M Pavan, I Tsakovska, M Vracko: The Characterisation of (Quantitative) Structure-Activity Relationships: Preliminary Guidance. EUR 21866 EN, (2005)

7. OECD: Guidance Document on the Validation of (Quantitative) Structure-Activity Relationships ((Q)SAR) Models. ENV/JM/MONO(2007)2 (2007) Available at: http://www.oecd.org/dataoecd/55/35/38130292.pdf (Accessed 7 July 2010)

8. L Eriksson, J Jaworska, AP Worth, MTD Cronin, RM McDowell, P Gramatica: Methods for reliability and uncertainty assessment and for applicability evaluations of classification- and regression-based QSARs. Environ Health Perspect 111, 1361-1375 (2003)
http://dx.doi.org/10.1289/ehp.5758
PMid:12896860    PMCid:1241620

9. TI Netzeva, A Worth, T Aldenberg, R Benigni, MT Cronin, P Gramatica, JS Jaworska, S Kahn, G Klopman, CA Marchant, G Myatt, N Nikolova-Jeliazkova, GY Patlewicz, R Perkins, D Roberts, T Schultz, DW Stanton, JJ van de Sandt, W Tong, G Veith, C Yang: Current status of methods for defining the applicability domain of (quantitative) structure-activity relationships. The report and recommendations of ECVAM Workshop 52, Altern Lab Anim 33, 155-173 (2005)
PMid:16180989

10. A Worth, G Patlewicz: A Compendium of Case Studies that helped to shape the REACH Guidance on Chemical Categories and Read Across. EUR 22481 EN (2007)

11. I Tsakovska, A Worth: The Use of Computational Methods for the Assessment of Chemicals in REACH. Bioautomation 13, 151-162 (2009)

12. MTD Cronin, AP Worth: (Q)SARs for Predicting Effects Relating to Reproductive Toxicity. QSAR Comb Sci 27, 91-100 (2008)
http://dx.doi.org/10.1002/qsar.200710118

13. A Gallegos Saliner, G Patlewicz, AP Worth: A Review of (Q)SAR Models for Skin and Eye Irritation and Corrosion. QSAR Comb Sci 27, 49-59 (2008)
http://dx.doi.org/10.1002/qsar.200710103

14. TI Netzeva, M Pavan, AP Worth: Review of (Quantitative) Structure-Activity Relationships for Acute Aquatic Toxicity. QSAR Comb Sci 27, 77-90 (2008)
http://dx.doi.org/10.1002/qsar.200710099

15. G Patlewicz, AO Aptula, DW Roberts, E Uriarte: A Minireview of Available Skin Sensitization (Q)SARs/Expert Systems. QSAR Comb Sci 27, 60-76 (2008)
http://dx.doi.org/10.1002/qsar.200710067

16. M Pavan, TI Netzeva, AP Worth: Review of Literature-Based Quantitative Structure-Activity Relationship Models for Bioconcentration. QSAR Comb Sci 27, 21-31 (2008)
http://dx.doi.org/10.1002/qsar.200710102

17. M Pavan, AP Worth: Review of Estimation Models for Biodegradation. QSAR Comb Sci 27, 32-40 (2008)
http://dx.doi.org/10.1002/qsar.200710117

18. I Tsakovska, I Lessigiarska, T Netzeva, AP Worth: A Mini Review of Mammalian Toxicity (Q)SAR Models. QSAR Comb Sci 27, 41-48 (2008)
http://dx.doi.org/10.1002/qsar.200710107

19. ECHA: European Chemicals Agency (ECHA) Guidance on Information Requirements and Chemical Safety Assessments, R.14 Occupational Exposure Estimation (2010)

20. http://www.chem.unep.ch/irptc/sids/oecdsids/sidspub.html

21. ECB: European Chemicals Bureau Risk Assessment Report for Methyl Methyacrylate. (2002) Downloaded from http://esis.jrc.ec.europa.eu/

22. GL Flynn. Physicochemical determinants of skin absorption. In: Principles of Route-to-Route Extrapolation for Risk Assessment. Eds: TR Gerrity, CJ Henry, Elsevier, New York (1990)

23. CA Lipinski, F Lombardo, BW Dominy, PJ Feeney: Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Del Rev 23, 3-25 (1997)
http://dx.doi.org/10.1016/S0169-409X(96)00423-1

24. A Sabljic: Predictions of the nature and strength of soil sorption of organic pollutants by molecular topology. J Agric Food Chem 32, 243-246 (1984)
http://dx.doi.org/10.1021/jf00122a016

25. P Donkin. Quantitative structure-activity relationships. In: Handbook of Ecotoxicology, Vol 2. Ed: P Calow, Blackwell Scientific Publications, London (1994)

26. H Konemann: Quantitative structure-activity relationships in fish toxicity studies 1. Relationship for 50 industrial pollutants. Toxicology 19, 209-221 (1981)
http://dx.doi.org/10.1016/0300-483X(81)90130-X

27. J Dearden, A Worth: In Silico Prediction of Physicochemical Properties. EUR 23051 EN (2007)

28. WM Meylan, PH Howard: Atom/fragment contribution method for estimating octanol-water partition coefficients. J Pharm Sci 84, 83-92 (1995)
http://dx.doi.org/10.1002/jps.2600840120
PMid:7714751

29. http://www.epa.gov/oppt/newchems/index.htm

30. S Dimitrov, G Dimitrova, T Pavlov, N Dimitrova, G Patlewicz, J Niemala, O Mekenyan: A Stepwise approach for defining the applicability domain of SAR and QSAR models. J Chem Inf Comput Sci 45, 839-849 (2005)
http://dx.doi.org/10.1021/ci0500381
PMid:16045276

31. CK Smith Pease: From xenobiotic chemistry and metabolism to better prediction and risk assessment of skin allergy. Toxicology 192, 1-22 (2003)
http://dx.doi.org/10.1016/S0300-483X(03)00246-4

32. I Kimber, RJ Dearman: What makes a chemical an allergen? Ann Allergy Asth Immunol 90, 28-31 (2003)
http://dx.doi.org/10.1016/S1081-1206(10)61645-6

33. DW Roberts, AO Aptula: Determinants of skin sensitisation potential. J Appl Toxicol 28, 377-387 (2008)
http://dx.doi.org/10.1002/jat.1289
PMid:17703504

34. DW Roberts, G Patlewicz. Chemistry based non-animal predictive modeling for skin sensitization. In: Ecotoxicology Modeling. Ed: J Devillers, Springer, New York (2009)
http://dx.doi.org/10.1007/978-1-4419-0197-2_3

35. G Patlewicz, AO Aptula, E Uriarte, DW Roberts, PS Kern, GF Gerberick, I Kimber, RJ Dearman, CA Ryan, DA Basketter: An evaluation of selected global (Q)SARs/expert systems for the prediction of skin sensitisation potential. SAR QSAR Environ Res 18, 515-541 (2007)
http://dx.doi.org/10.1080/10629360701427872
PMid:17654336

36. DW Roberts, DL Williams: The derivation of quantitative correlations between skin sensitisation and physico-chemical parameters for alkylating agents and their application to experimental data for sultones. J Theo Biol 99, 807-825 (1982)
http://dx.doi.org/10.1016/0022-5193(82)90199-0

37. AO Aptula, DW Roberts: Mechanistic applicability domains for nonanimal-based prediction of toxicological end points: general principles and application to reactive toxicity. Chem Res Toxicol 19, 1097-1105 (2006)
http://dx.doi.org/10.1021/tx0601004
PMid:16918251

38. SJ Enoch, JC Madden, MT Cronin: Identification of mechanisms of toxic action for skin sensitisation using a SMARTS pattern based approach. SAR QSAR Environ Res 19, 555-578 (2008)
http://dx.doi.org/10.1080/10629360802348985
PMid:18853302

39. MD Barratt, DA Basketter, M Chamberlain, GD Admans, JJ Langowski: An expert system rulebase for identifying contact allergens. Toxicol In Vitro 8, 1053-1060 (1994)
http://dx.doi.org/10.1016/0887-2333(94)90244-5

40. JE Ridings, MD Barratt, R Cary, CG Earnshaw, CE Eggington, MK Ellis, PN Judson, JJ Langowski, CA Marchant, MP Payne, WP Watson, TD Yih: Computer prediction of possible toxic action from chemical structure: an update on the DEREK system. Toxicology 106, 267-279 (1996)
http://dx.doi.org/10.1016/0300-483X(95)03190-Q

41. S Dimitrov, L Low, G Patlewicz, PS Kern, GD Dimitrova, MH Comber, RD Phillips, J Niemela, PT Bailey, OG Mekenyan: Skin sensitization: Modelling based on skin metabolism simulation. Int J Toxicol 24, 189-204 (2005)
http://dx.doi.org/10.1080/10915810591000631
PMid:16126613

42. G Patlewicz, S Dimitrov, LK Low, PS Kern, GD Dimitrova, MI Comber, AO Aptula, RD Phillips, J Niemelä, C Madsen, EB Wedebye, DW Roberts, PT Bailey, OG Mekenyan: TIMES-SS - A promising tool for the assessment of skin sensitization hazard. A characterization with respect to the OECD validation principles for (Q)SARs and an external evaluation for predictivity. Regul Toxicol Pharmacol 48, 225-239 (2007)
http://dx.doi.org/10.1016/j.yrtph.2007.03.003
PMid:17467128

43. DW Roberts, G Patlewicz, S Dimitrov, LK Low, AO Aptula, PS Kern, GD Dimitrova, MI Comber, RD Phillips, J Niemelä, C Madsen, EB Wedebye, PT Bailey P. T, OG Mekenyan: TIMES-SS - A mechanistic evaluation of an external validation study using reaction chemistry principles. Chem Res Toxicol 20, 1321-1330 (2007)
http://dx.doi.org/10.1021/tx700169w
PMid:17713962

44. E Schlede, W Aberer, T Fuchs, I Gerner, H Lessmann, T Maurer, R Rossbacher, G Stropp, E Wagner, D Kayser: Chemical substances and contact allergy-244 substances ranked according to allergenic potency. Toxicology 193, 219-259 (2003)
http://dx.doi.org/10.1016/S0300-483X(03)00266-X

45. C Betts, RJ Dearman, JR Heyling, I Kimber, DA Basketter: Skin sensitization potency of methyl methacrylate in the local lymph node assay: comparisons with guinea-pig data and human experience. Contact Dermatitis 55, 140-147 (2006)
http://dx.doi.org/10.1111/j.1600-0536.2006.00898.x
PMid:16918612

46. TW Schultz, RE Carlson, MT Cronin, JL Hermens, R Johnson, PJ O'Brien, DW Roberts, A Siraki, KB Wallace, GD Veith: A conceptual framework for predicting the toxicity of reactive chemicals: modeling soft electrophilicity. SAR QSAR Environ Res 17, 413-428 (2006)
http://dx.doi.org/10.1080/10629360600884371
PMid:16920662

47. JA Miller, EC Miller. Ultimate chemical carcinogens as reactive mutagenic electrophiles. In: Origins of Human Cancer. Eds: HH Hiatt, JD Watson, JA Winsten, Cold Spring Harbor Laboratory, Cold Spring Harbor (1977)

48. J Ashby, RW Tennant: Chemical structure, Salmonella mutagenicity and extent of carcinogenicity as indicators of genotoxic carcinogenesis among 222 chemicals tested in rodents by the U.S. NCI/NTP. Mutat Res 204, 17-115 (1988)
http://dx.doi.org/10.1016/0165-1218(88)90114-0

49. J Ashby, RW Tennant, E Zeiger, S Stasiewicz: Classification according to chemical structure, mutagenicity to Salmonella and level of carcinogenicity of a further 42 chemicals tested for carcinogenicity by the U.S. National Toxicology Program. Mutat Res 223, 73-103 (1989)
http://dx.doi.org/10.1016/0165-1218(89)90037-2

50. RW Tennant, J Ashby: Classification according to chemical structure, mutagenicity to Salmonella and level of carcinogenicity of a further 39 chemicals tested for carcinogenicity by the U.S. National Toxicology program. Mutat Res 257, 209-227 (1991)
PMid:2014033

51. R Benigni, C Bossa: Structure-activity models of chemical carcinogens: state of the art, and new directions. Ann Ist Super Sanita 42, 118-126 (2006)
PMid:17033131

52. J Kazius, R McGuire, R Bursi: Derivation and validation of toxicophores for mutagenicity prediction. J Med Chem 48, 312-320 (2005)
http://dx.doi.org/10.1021/jm040835a
PMid:15634026

53. YT Woo. Mechanisms of action of chemical carcinogens, and their role in structure-activity relationships (SAR) analysis and risk assessment. In: Quantitative Structure-Activity Relationship (QSAR) Models of Mutagens and Carcinogens. Ed: R. Benigni, CRC Press, Boca Raton (2003)

54. R Benigni, C Bossa, N Jeliazkova, T Netzeva, A Worth: The Benigni / Bossa rulebase for mutagenicity and carcinogenicity - a module of Toxtree. EUR 23241 EN (2008)

55. R Serafimova, M Todorov, T Pavlov, S Kotov, E Jacob, A Aptula, O Mekenyan: Identification of the structural requirements for mutagencitiy, by incorporating molecular flexibility and metabolic activation of chemicals. II. General Ames mutagenicity model. Chem Res Toxicol 20, 662-676 (2007)
http://dx.doi.org/10.1021/tx6003369
PMid:17381132

56. O Mekenyan, M Todorov, R Serafimova, S Stoeva, A Aptula, R Finking, E Jacob: Identifying the structural requirements for chromosomal aberration by incorporating molecular flexibility and metabolic activation of chemicals. Chem Res Toxicol 20, 1927-1941 (2007)
http://dx.doi.org/10.1021/tx700249q
PMid:18052113

57. R Benigni, C Bossa: Structure alerts for carcinogenicity, and the Salmonella assay system: A novel insight through the chemical relational databases technology. Mutat Res 659, 248-261 (2008)
http://dx.doi.org/10.1016/j.mrrev.2008.05.003
PMid:18621573

58. J Jaworska, S Dimitrov, N Nikolova, O Mekenyan: Probablistic assessment of biodegradability based on metabolic pathways. CATABOL system. SAR QSAR Environ Res 13, 445-455 (2002)
http://dx.doi.org/10.1080/10629360290002794
PMid:12071658

59. HJM Verhaar, CJ van Leeuwen, JLM Hermens: Classifying environmental pollutants. 1. Structure-activity relationships for prediction of aquatic toxicity. Chemosphere 25, 471-491 (1992)
http://dx.doi.org/10.1016/0045-6535(92)90280-5

60. ECETOC: Technical Report No. 107, Addendum to ECETOC Targeted Risk Assessment Technical Report No. 93 European Centre for Ecotoxicology and Toxicology of Chemicals, Brussels, Belgium (2009)

61. RISKOFDERM: The RISKOFDERM Dermal Exposure Model Version 2.0, TNO, Netherlands (2006) Downloaded from http://www.tno.nl

62. S Geukens, A Goossens: Occupational contact allergy to (meth)acrylates. Contact Dermatitis 44, 153-159 (2001)
http://dx.doi.org/10.1034/j.1600-0536.2001.044003153.x
PMid:11217987

63. SE Loveless, A-M Api, R Crevel, E Debruyne, A Gamer, I Jowsey, P Kern, I Kimber, L Lea, P Lloyd, Z Mehmood, W Steiling, G Veenstra, M Woolhiser, C Hennes: Potency values from the local lymph node assay: Application to classification, labeling and risk assessment. ECETOC Monograph 46 (2008)

64. IE Liljelind, A Hagenbjörk-Gustafsson, LO Nilsson: Potential dermal exposure to methyl methacrylate among dental technicians; variability and determinants in a field study. J Environ Monit 11, 160-165 (2009)
http://dx.doi.org/10.1039/b810355k
PMid:19137152

Key Words: REACH, Non-testing approaches, (Q)SARs, Chemical Categories, Read-Across, Exposure Assessment, Review

Send correspondence to: Grace Patlewicz, DuPont Haskell Global Centers for Health and Environmental Sciences, 1090 Elkton Road, Newark, DE 19711, Tel: 1-302-366-5090, Fax: 1-302-451-4531, E-mail: patlewig@hotmail.com