Regulatory Toxicology and Pharmacology 63 (2012) 40–52

Contents lists available at SciVerse ScienceDirect

Regulatory Toxicology and Pharmacology

journal homepage: www.elsevier .com/locate /yr tph

Guiding principles for the implementation of non-animal safety assessmentapproaches for cosmetics: Skin sensitisation

Carsten Goebel a, Pierre Aeby b, Nadège Ade c, Nathalie Alépée c, Aynur Aptula d, Daisuke Araki e,Eric Dufour c, Nicola Gilmour d, Jalila Hibatallah f, Detlef Keller g, Petra Kern h, Annette Kirst i,Monique Marrec-Fairley j, Gavin Maxwell d, Joanna Rowland k, Bob Safford d, Florian Schellauf j,Andreas Schepky l, Chris Seaman m, Thomas Teichert l, Nicolas Tessier n, Silvia Teissier o,Hans Ulrich Weltzien p, Petra Winkler q, Julia Scheel g,⇑a Procter & Gamble, Berliner Allee 65, 64274 Darmstadt, Germanyb Route de Messidor 59, 1723 Marly, Switzerlandc L’Oreal, River Plaza 25-29, quai Aulagnier, 92600 Asnières-sur-Seine, Franced Unilever – SEAC, Colworth House, Sharnbrook, MK 44 1LQ Bedford, UKe Kanebo Cosmetics, 89/91 Rue du Faubourg St. Honoré, 75008 Paris, Francef Chanel, 135 Avenue Charles de Gaulle, 92521 Neuilly sur Seine, Franceg Henkel AG & Co. KGaA, Henkelstraße 67, 40191 Düsseldorf, Germanyh Procter & Gamble, Temselaan 100, 1853 Stroombeek Bever, Belgiumi KPSS – Kao Professional Salon Services GmbH, Pfungstaedter Strasse 92-100, 64297 Darmstadt, Germanyj Colipa, Avenue Herrmann Debroux 15A, 1160 Brussels, Belgiumk GlaxoSmithKline, St. George’s Avenue, Weybridge, KT13 0DE Surrey, UKl Beiersdorf, Unnastrasse 48, 20245 Hamburg, Germanym GlaxoSmithKline, 2FA12 Park Road, SG12 0DP Ware, Hertfordshire, UKn Spincontrol (representing FEBEA), 238 Rue Giraudeau, 37000 Tours, Franceo L’Oreal, 1 Avenue Eugène Schueller, 93600 Aulnais Sous Bois, Francep Schillhof 5, 79110 Freiburg, Germanyq Johnson & Johnson GmbH, Johnson & Johnson Platz 2, 41470 Neuss, Germany

a r t i c l e i n f o a b s t r a c t

Article history:Received 25 February 2011Available online 21 February 2012

Keywords:Skin sensitisationSafety assessmentCosmeticsAlternative methods

0273-2300/$ - see front matter � 2012 Elsevier Inc. Adoi:10.1016/j.yrtph.2012.02.007

Abbreviations: ACD, allergic contact dermatitis; Areview; COLIPA, the European Cosmetics Associatioconcentration of the test substance required to prodactivation test; HERA, human and environmental ristest; KC, keratinocytes; LC, Langerhans cells; LLNA,NESIL, no expected sensitisation induction level;assessment; (Q)SAR, (quantitative) structure activitySMARTS, SMiles ARbitrary Target Specificaton; TG,evidence.⇑ Corresponding author.

E-mail address: julia.scheel@henkel.com (J. Scheel

Characterisation of skin sensitisation potential is a key endpoint for the safety assessment of cosmeticingredients especially when significant dermal exposure to an ingredient is expected. At present themouse local lymph node assay (LLNA) remains the ‘gold standard’ test method for this purpose how-ever non-animal test methods are under development that aim to replace the need for new animaltest data. COLIPA (the European Cosmetics Association) funds an extensive programme of skin sensi-tisation research, method development and method evaluation and helped coordinate the early eval-uation of the three test methods currently undergoing pre-validation. In May 2010, a COLIPA scientificmeeting was held to analyse to what extent skin sensitisation safety assessments for cosmetic ingre-dients can be made in the absence of animal data. In order to propose guiding principles for theapplication and further development of non-animal safety assessment strategies it was evaluatedhow and when non-animal test methods, predictions based on physico-chemical properties (includingin silico tools), threshold concepts and weight-of-evidence based hazard characterisation could beused to enable safety decisions. Generation and assessment of potency information from alternative

ll rights reserved.

EL, acceptable exposure level; APC, antigen presenting cells; CEL, consumer exposure level; CIR, cosmetic ingredientn; DC, dendritic cells; DPRA, direct peptide reactivity assay; DST, dermal sensitisation threshold; EC3, effectiveuce a three-fold increase in the stimulation index compared to vehicle-treated controls; h-CLAT, human cell line

k assessment; HMT, human maximisation test; HPV, high production volume; HRIPT, human repeated insult patchlocal lymph node assay; LogPow, octanol–water partition coefficient; MUSST, myeloid U937 skin sensitisation test;NS, non-sensitisers; OECD, organisation for economic co-operation and development; QRA, quantitative riskrelationship; S, sensitisers; SAF, sensitisation assessment factor; SCCS, scientific committee for consumer safety;

test guideline; TSC, threshold of sensitisation concern; TTC, threshold of toxicological concern; WoE, weight-of-

).

C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52 41

tools which at present is predominantly derived from the LLNA is considered the future key researcharea.

� 2012 Elsevier Inc. All rights reserved.

1. Introduction hypothesis being that integration of data from a ‘toolbox’ of non-

A standard requirement within the safety assessment of cosmeticingredients is to characterise their potential to induce skin sensitisat-ion under product use conditions that may lead to allergic contactdermatitis (ACD) in humans. Despite extensive efforts to developalternative methods, the sensitising potential of an ingredient cur-rently needs to be identified on the basis of animal studies in manycases, i.e., the murine local lymph node assay (LLNA, OECD (organi-sation for economic co-operation and development) TG 429)(Basketter et al., 2007; OECD, 2010) and Guinea pig assays (OECDTG 406) (OECD, 1992), namely the Maximisation Test (Magnussonand Kligman, 1969) and the Buehler test (Buehler, 1965).

The LLNA is based on quantification of cell proliferation in thedraining auricular lymph nodes after repeated topical applicationsof the chemical. By testing multiple concentrations, the assay notonly identifies potential skin sensitisers, but also evaluates theirsensitising potency. Guinea pig tests are based on a visual scoringof skin reactions after topical application (Buehler test) or intrader-mal and topical application (Maximisation test) of the chemical ata dose which induces modest irritation. Approximately threeweeks later the potential of a chemical to elicit an immune re-sponse is analysed by virtue of a challenge exposure.

Immunologically, skin sensitisation can be described as a de-layed-type hypersensitivity reaction induced by low molecularweight reactive chemicals (haptens). It comprises two phases,induction and elicitation (Karlberg et al., 2008). Fig. 1 schemati-cally depicts the corresponding steps (1–7) as described in the fol-lowing: as a first step in the induction phase the (possibly oxidised)chemical must penetrate the skin (steps 1 and 2) to chemically re-act with endogenous proteins (step 5). Some chemicals requireactivation through enzymatic (pro-haptens, step 4) or oxidative(pre-haptens, step 1) processes in order to become haptens capableof binding to skin proteins (step 5). The first cells to be exposed tosensitisers are epidermal keratinocytes, which respond to chemicalstress with a cocktail of proinflammatory cytokines (step 3) (Cor-sini et al., 2009). Activated by these mediators as well as in somecases by direct hapten contact, epidermal Langerhans cells (LC)and immature dendritic cells (DC) take up and process haptenatedproteins. In parallel they mature into highly effective antigen pre-senting cells (APC) (Toebak et al., 2009). This maturation includesthe secretion of mediators like IL-8, as well as the expression ofsurface markers such as CD86, CD54, or chemokine receptors (step6) (Kroeze et al., 2009). The latter facilitate the migration of LC outof the epidermis and guide them to the nearest (local) lymph nodewhere they present haptenated protein fragments (antigens) to Tcells (step 7) (Ortmann et al., 1992). This step is the link betweenthe innate and the adaptive immune systems, i.e., recognition ofthe antigen by specific T cell receptors and subsequent specific Tcell activation. The activated (effector) T cells home to the skinwhere upon repeated contact with the same allergen (elicitationphase) they orchestrate an inflammatory response that can leadto dermal injury. Hence, they are representing the immunological‘‘memory’’ responsible for the specificity of the ACD (Martin andWeltzien, 1994).

COLIPA’s skin sensitisation research programme aims to furtherrefine our fundamental understanding of how each of these keypathways contribute to the induction of skin sensitisation and de-velop in vitro test methods that can predict the effect of a novelchemical on each of these key pathways (Aeby et al., 2010) the

animal test methods, each developed to model a different keypathway in vitro, will allow the precise characterisation of a chem-ical regarding its skin sensitising potency. Which (set) of thesetools will turn out to provide an appropriate prediction is notyet determined and is under investigation. On May 26–27, 2010COLIPA organised an expert workshop to propose how in vitro testmethods and combinations thereof may be applied to risk assess-ment decision-making, in combination with other non-animal riskassessment elements.

2. Review of existing tools for evaluation of skin sensitisationrisk assessment without animals

Risk assessment of cosmetic ingredients is not a standardisedprocedure, but a case-by-case consideration using best science.Usually a stepwise approach is employed, utilising the entire scopeof information available to reach science-based decisions in aweight-of-evidence (WoE) assessment. WoE is considered as thebasic principle to avoid unnecessary animal testing, since all rele-vant existing information is thoroughly evaluated before any newtesting is undertaken, and it is iteratively applied in each step ofthe assessment. An expert assessment of the relevance, i.e., scien-tific validity or suitability of the purpose of a method or approachneeds to be performed to decide how to weight individual pieces ofinformation. New testing may be required when the existing infor-mation is not adequate to support the safety of an ingredient orwhen safety issues arise. In general, information can be qualitative(used for hazard identification) and/or quantitative (used for haz-ard characterisation and risk assessment).

WoE is broadly accepted by legislation and safety assessors as abasic principle in risk assessment and is explicitly mentioned inEuropean chemicals and cosmetics legislation, e.g., in the recastof the European Cosmetics Directive (EU, 2009), REACH (EU,2006), and the regulation on classification, labelling and packaging(‘‘CLP’’) of substances and mixtures (EU, 2008).

Elements that are available for skin sensitisation risk assess-ment to gain information that can be used within WoE-basedsafety assessment include:

(1) Prediction based on physico-chemical properties (withoutexperimental testing; expert judgment/in silico): presenceor absence of structural alerts ((quantitative) structureactivity relationships = (Q)SAR), indications for chemicalreactivity with nucleophiles, mechanistic assignment toreactivity domains including computer-based searches forstructural and functional similarities in data bases such asDEREK, TIMES, MULTICASE, OECD Toolbox, ToxTree etc.

(2) Read-across based on similar chemicals with availableexperimental data: this is usually done by making use ofWoE expert judgement, potentially assisted by in silico toolssuch as OECD toolbox, Toxtree etc.

(3) In vitro methods: includes binding capacity towards pro-teins; responses of human cell types, i.e., primary keratino-cytes (KC), dendritic cells (DC), and T cells or relevantimmortalised cell lines in terms of bio-markers, cytokinesecretion or gene expression (gene signature).

(4) Historical data: (i) animal studies reliably reporting on skinsensitisation effects. (ii) human experience with exposureto substances and preparations regarding cutaneous

NH2

NH2

T

DC

NH2

NH2

GP Tests, HRIPT

LLNA

1. Oxidation

4. Metabolism

6. DC activation

7. T cell reaction

2. Penetration

Induction Elicitation

Haptenisation

5. Protein binding

3. KC activation

LC

Fig. 1. Key events in skin sensitisation. Schematic representation of processes involved in the induction and elicitation phases of skin sensitisation by allergenic chemicals.KC: keratinocyte, DC: dendritic cell, StrC: stratum corneum, T: T cell, LLNA: local lymphnode assay, GP: guinea pig, HRIPT: human repeat insult patch test.

42 C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52

reactions, such as diagnostic clinical studies and experimen-tal human studies with volunteers (e.g., human repeatedinsult patch test, HRIPT).

(5) Exposure estimates: standard skin exposure scenarios asbased on guidance of the EU commission Scientific Commit-tee for Consumer Safety (SCCS) and/or information on der-mal bioavailability

(6) Risk assessment methods: quantitative risk assessment(QRA) for skin sensitisation for which WoE is considered abasic element, and threshold concepts.

2.1. Prediction based on physico-chemical properties

As the majority of organic skin sensitisers elicit their effect viacovalent bond formation with skin proteins, these reactions canbe described in terms of nucleophilic–electrophilic reaction chem-istry. The assessment of a molecular structure with regard to itspotential reactivity can be based on expert judgment, in silico toolsand in chemico reactivity determinations. This provides a frame-work and builds the basis for structure–activity relationship ap-proaches (SARs) intended to identify chemicals that may interactwith biological systems and have a potential to cause skinsensitisation.

In order to predict toxicity from structure, it is beneficial toestablish initially whether a compound is electrophilic in natureand secondly the type of electrophile, especially if this can be asso-ciated with a specific mechanism of action. Skin sensitisers can of-ten be grouped into several reaction mechanistic domains, themajor ones of which are Michael acceptor, Schiff base, SN2, SNAr,acyl transfer agents, or non reactive and non-pro reactive agents(Fig. 2). Recent publications summarised the chemistry of thesereaction mechanistic domains (Aptula and Roberts, 2006) andshowed how this concept can be successfully applied to classifyand chemically rationalise published skin sensitisation data sets(Roberts et al., 2007a, b) (Table 1).

Inclusion of quantitative considerations leads to quantitativestructure–activity relationship approaches (QSARs) able to approx-imate a chemical’s relative sensitising potency. An overview ofpublished reactivity-based QSAR models is provided in Table 2.More recent QSAR models (Roberts et al., 2011, 2006; Roberts

and Natsch, 2009) have been evaluated against OECD guidelines(OECD Quantitative Structure–Activity Relationships Project)(OECD) and were shown to meet the OECD principles for (Q)SARvalidation. Each one covers a diverse range of structures within asingle reaction mechanistic domain. Earlier QSARs were usually re-stricted to sets of structurally similar (congeneric series) of com-pounds. Relative alkylation index (RAI)-based QSARs (Franotet al., 1994; Roberts et al., 2007a, 1983; Roberts and Basketter,2000, 1997; Roberts and Williams, 1982) use a composite reactiv-ity + hydrophobicity parameter, calculated from reaction rate con-stants and partition coefficients. Those which are RAI-based arehybrid dose–response/QSARs, in which the biological response(stimulation index, erythema score etc.) is correlated with a com-bination of dose/reactivity/hydrophobicity.

2.1.1. In chemico reactivity determinationThe general concept that the rate determining step in the skin

sensitisation process is likely to be the reaction of the sensitiserwith skin nucleophiles has led to initiatives to develop methodsand to generate data on the reactivity of chemicals towards modelnucleophiles representing peptide and protein nucleophiles in theskin. These methods focus on the qualitative and quantitative mea-surement of reactivity.

Generally, empirical measures of the reactivity of chemicalswith model nucleophiles such as thiol can be used to simulatethe relative rates at which a reactive chemical is likely to bind tonucleophiles in cellular targets (Gerberick et al., 2008). To empha-sise the nature of the reaction, it has been termed it in chemicoreactivity (based only on organic chemistry). With this approachthe toxicity (e.g., skin sensitisation) of a new chemical were esti-mated from measured chemical data for certain subsets of chemi-cals (Schultz et al., 2009).

Practically, the activity of an electrophile can be quantified interms of the rate constant k, or its logarithm log k, for its reactionwith a nucleophile. They should be determined for a range of elec-trophiles under the same conditions (same nucleophile, solvent,temperature). Rate constants (absolute or relative) or their loga-rithms are often used in physical organic chemistry to express reac-tivity quantitatively. There are several approaches for determiningrate constants. They are listed in order of decreasing accuracy.

Mechanistic domain Protein binding reaction

Michael acceptors X Nu Protein

Modified protein

XNu

Protein

SNAr electrophiles

X NuNu X

Nu Protein

ProteinProtein

Y1, Y2… Y1, Y2… Y1, Y2…

Schiff base formers O

Protein

N Protein

X Nu Protein NuProtein

SN2 electrophiles

Acylating agents XO

Protein

O

ProteinNHNH2

NH2

Identification characteristics. Double or triple bond with electron-withdrawingsubstituent X, such as -CHO, -COR, -CO2R , -CN, -SO2R, -NO2...Includes para quinones and ortho quinones, often formed by oxidation of para and ortho di-hydroxy aromatics acting as pro-Michael acceptors. X can also be aheterocyclic group such as 2-pyridino or 4-pyridino.

Identification characteristics. X = halogen or pseudohalogen, Y's are electron withdrawing groups (at least two) such as -NO2, -CN, -CHO, -CF3, -SOMe,-SO2Me, ring fused nitrogen...One halogen is too weak to act as an X, butseveral halogens together can activate.

Identification characteristics. X = halogen or other leaving group, e.g.OSO2(R or Ar), OSO2O(R or Ar) bonded to primary alkyl, benzylic, or allylic carbon. OR and NHR or NR2 do not usually act as leaving groups, but can doso if part of a strained 3-membered ring (e.g. epoxides, ethylenimine andsubstituted derivatives).

Identification characteristics. X = halogen, or other group (e.g. -OC6H5)such that XH is acidic enough for X- to act as a good leaving group. Includesanhydrides, cyclic or non-cyclic. X = -Oalkyl does not qualify, except whenpart of a strained lactone ring, e.g. β-propiolactone (but not γ-butyrolactone). Analogous reactions can occur with attack at sulfonyl S, phosforyl Pand thioacyl C.

Identification characteristics. Reactive carbonyl compounds such as aliphatic aldehydes, some α,β- and α,γ-diketones, α-ketoesters. Not simplemonoketones and aromatic aldehydes. Other hetero-unsaturated systems canbehave analogously, e.g. C-nitroso compounds, thiocarbonyl compounds(C=S), cyanates and isocyanates, thiocyanates and isothiocyanates.

Fig. 2. Reaction mechanistic applicability domains (Aptula and Roberts, 2006).

Table 1Assignment of LLNA-typed chemicals to different reactivity domains. A total of 344 chemicals, which had previously been classified by LLNA into sensitisers and non-sensitisers,were assigned to different reactivity domains defined in Fig. 2 (Roberts et al., 2007a, b).

Chemical domain LLNA data distribution

Total Sensitiser Non-sensitiser Sensitisers [% of total]

Michael acceptors 97 89 8 92SNAr electrophiles 3 3 0 100SN2 electrophiles 61 49 12 80Schiff base formers 56 46 10 82Acylating agents 30 28 2 93Non-reactive 97 38 59 39

C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52 43

2.1.1.1. Full kinetics. In a typical kinetics experiment the extent ofreaction, either in terms of depletion of one of the reactants or interms of formation of a reaction product, is measured at several

recorded time intervals, and from these the rate constant k is cal-culated (for more details see (Chipinda et al., 2010; Roberts et al.,2008; Roberts and Natsch, 2009).

Table 2Published reactivity-based QSAR models.

Ref. Reactivity parameter Other parameters Applicability domain Test system RAIa-based

Roberts et al. (2006) Substituent constants Hydrophobicity Schiff base LLNA NoRoberts and Natsch (2009) Peptide kinetics (logk) None Michael acceptors LLNA NoRoberts et al. (2011) Substituent constants None SNAr LLNA NoRoberts et al. (2007a) BuNH2

b kinetics/LFERc Hydrophobicity H-polar SN2 LLNA YesFranot et al. (1994) BuNH2 kinetics Hydrophobicity H-polar SN2 (single set of structural congeners) GPd (MSIATe) YesRoberts and Basketter (2000) BuNH2 kinetics Hydrophobicity H-polar SN2 (sulphonate esters) LLNA YesRoberts and Basketter (1997) BuNH2 kinetics Hydrophobicity H-polar SN2 (sulphonate esters) GP (MSIAT) YesRoberts et al. (1983) BuNH2 kinetics Hydrophobicity H-polar SN2 (single set of structural congeners) GP (MSIAT) YesRoberts and Williams (1982) BuNH2 kinetics Hydrophobicity Sultones (SN2 and Michael acceptors separately) GP (SIAT) Yes

a RAI – relative alkylation index.b BuNH2 – butylamine.c LFER – linear free energy relationship.d GP – guinea pig.e MSIAT – modified single injection adjuvant test.

44 C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52

2.1.1.2. Concentration giving 50% reaction in a fixed time(RC50). Methods based on this approach have been developed,which is designed to generate a quantitative reactivity databasewhich can be used for in chemico modelling for a variety of toxico-logical end points, including skin sensitisation. A concentration re-sponse method is used to quantify reactivity in aqueous solution tothe model nucleophile glutathione (GSH) in terms of the RC50 va-lue, with RC50 being the concentration of electrophile that de-pletes the thiol group of GSH by 50% (Roberts et al., 2008;Schultz et al., 2005).

2.1.1.3. Extent of reaction after a fixed time. Another method is todetermine the extent of reaction (which can be expressed as%depletion, DP of one of the reactants) after a fixed time. Althoughnot as accurate as a kinetics experiment, since only two data pointsare used, rough estimates for second order rate constants can beestimated (Roberts et al., 2008). This method has been used to de-velop a peptide depletion assay approach, namely the Direct Pro-tein Reactivity Assay (DPRA) (Gerberick et al., 2007, 2004) whichis described in detail below (see Section 2.3). Especially since thisassay is already well advanced (i.e., under prevalidation) it willbe used in the following (see Section 2.7) to represent chemicalreactivity determination. In a similar approach, quantitative LC-MS/MS technology was used to determine whether a chemical isreactive towards different protein nucleophiles by observing boththe disappearance of target peptide and formation of adduct(s)(Aleksik et al., 2009). Furthermore, an assay based on a heptapep-tide and quantitative LC-MS to measure peptide depletion and ad-duct formation was introduced (Natsch and Gfeller, 2008). Thismodification of the original HPLC assay has the advantage that alsoquantification of peptide oxidation (dimerisation) can bemeasured.

2.1.2. In silico toolsThe utility of two commonly used in silico SAR tools, namely the

database-assisted expert system DEREK (Judson, 2002; Rodfordet al., 2003) and the OECD Toolbox (OECD) were discussed with re-gard to their predictive capacity for skin sensitisation. It could beshown that the predictive capacity of these models is largely influ-enced by the data sources and prediction systems used. These toolsare not designed as a stand alone replacement, but may be usefulwithin a structured decision support system as part of a safetyassessment strategy.

Data was shared during the workshop from a company studythat was conducted in 2009, which compared the predictive capac-ity of the DEREK (version 10) and the OECD Toolbox (version 1), byconcordance analysis and WoE approach (Quayson et al., 2010).The output of each model (see Fig. 3) was compared to LLNA results

for 249 chemicals from an in-house database representing a broadspectrum of chemical classes known to cause skin sensitisation inanimals and/or man, including acrylates, ketones and diketones,aldehydes, aromatic amines, sulphonates, sulphates and quinines.The overall concordance with the LLNA was low in both systems,although DEREK performed slightly better than the OECD Toolbox(63% vs. 53% overall concordance). Both computer programmesgave many false positive predictions, which may, as indicated, bepartly due the limited dataset that was used to train the pro-grammes. Predictions for some chemical classes performed sub-stantially better than the systems in general, particularly forhaloalkanes, with correct predictions of up to ca 70%. Negative pre-dictivity reached up to 70%, indicating that (Q)SAR may be used aspart of a combination of approaches to obtain information for pri-ority setting. Importantly since these analyses were conducted in2009, data from these 249 chemicals, in addition data from othersources and systems, were uploaded to DEREK and made availablein version 12 to add to the existing dataset. Version 2 of the OECDToolbox has also been made available recently and might be fur-ther assessed. The OECD Toolbox is referred to as SAR tool andcan be used to identify structural alerts for contact hypersensitiv-ity. However, predictions based on structural alerts alone show apoor overall accuracy and use only one of the functionalities cur-rently implemented in the OECD Toolbox. Therefore, a more prom-ising evaluation strategy would be to build a category of relatedmolecules around the target chemical by sequentially applyingthe mode of action profilers, structural classification rules, and fi-nally the bioavailability prediction functionality.

The EU JRC commissioned the development of the open sourcecomputer programme ‘‘Toxtree’’ (2010, http://toxtree.source-forge.net/), which is able to estimate toxic hazard by applyingstructural rules. The skin sensitisation alerts were those developedby Aptula and Roberts (2006) and subsequently encoded asSMARTS patterns by Enoch et al. (2008).

2.1.2.1. Recommendations for further research. To support expertjudgement, in silico approaches are considered to have a potentialvalue when used as part of a WoE strategy for the assessment ofskin sensitisation.

Mechanistic chemistry classification and assignment to mecha-nistic domains is already possible for many chemicals. This can bedone through inspection of the structure and can be supported andrefined by reactivity determinations. Based on the current knowl-edge, quantitative measurement of reactivity is considered a verypromising approach helping to link protein reactivity to skin sensi-tisation potency. However, the available approaches are mainlybased on reactivity towards thiol groups and therefore have limita-tions in capturing all reactivity domains. Therefore, further

70

80

90

100

0

10

20

30

40

50

60

sensitivity specificity positive predictivity negative predictivity accuracy

Perc

enta

ge (%

)

DEREK

OECD Toolbox

Fig. 3. Comparison of the performance of DEREK and the OECD toolbox with LLNA data in predicting skin sensitisation.

C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52 45

approaches are needed to capture chemistry domains which aremore difficult to assess by e.g., amine-based, non-water assays. An-other limitation is that the assays detect only directly active elec-trophiles, while a number of skin allergens requires activation(either by abiotic oxidation or metabolic activation) to form anelectrophilic reactive molecule. Therefore, future in chemico assaysshould target to also capture not directly active chemicals. Promis-ing progress has already been made in this area by integrating anactivation step into the Direct Protein Reactivity Assay (Gerbericket al., 2009; Troutman et al., 2010).

The accuracy of the assessment (both negative and positive pre-dictivity) may be approved by additional considerations of phys-ico-chemical information such as molecular weight, LogPOW,solubility, predicted maximum flux/dermal penetration. Oneexample is high molecular weight chemicals such as polymerswhich lack structural alerts and fail to penetrate the stratum cor-neum. Exposure through the dermal route may thus be consideredas negligible and will not require further testing, as long as impu-rities or unconverted monomers have been evaluated and are notof concern. For other molecular structures these properties mustbe assessed on an individual basis possibly requiring more detailedinvestigation.

The principle of uploading available data to the DEREK andOECD toolbox databases should be actively supported. By extend-ing these databases the capacity for predicting skin sensitisingproperties of cosmetic ingredients may be improved.

2.2. Read-across based on similar chemicals with availableexperimental data

2.2.1. The OECD toolboxOne of the currently developed and officially promoted tools for

read-across considerations based on the chemical structure or do-main is the OECD Toolbox that has already been mentioned above.In some cases, sufficient read-across data on structurally similarmaterials might be available and provide information on the skinsensitising potential. The findings from the analysis described inthe previous (Section 2.1), reported an overall concordance of theOECD Toolbox with the LLNA of 53% based on 249 chemicals,demonstrating that the direct correlation to LLNA data is still

insufficient. There were many false positive predictions. Negativepredictivity was more accurate (70%), which still underlines theutility of the toolbox as part of a package of approaches for prioritysetting.

Using the OECD Toolbox with the data gap filling function didimprove the overall concordance of the predictions with the LLNAto 77%. Positive predictivity of both, DEREK and the OECD Toolbox,could be improved by considering physico-chemical propertiesused to predict dermal absorption, and bacterial mutagenicity dataas surrogate for information on reactivity (Quayson et al., in press).Considering this type of information may generally improvethe confidence in the output of (Q)SAR and/or read-acrossassessments.

2.2.2. Read-across and weight-of-evidence (WoE) in regulatory riskassessment. Examples from public databases

A meta-analysis of published literature and databases wasperformed to identify examples how read-across is applied in thecontext of regulatory decision-making. As explained above, read-across and WoE are considered basic principles in current riskassessment. However, the application needs thorough justificationand is often limited, especially for highly complex toxicological ef-fects. In general, data from SCCS opinions, CIR (cosmetic ingredientreview) reports, OECD HPV (high production volume) dossiers,HERA (human and environmental risk assessment) risk assess-ments, and PubMed were used. After a thorough review of thatliterature, it turned out that the identified examples can mainlybe grouped into several major categories which are listed in Table3 together with some examples identified for each category.

It was also identified that read-across is usually combined withWoE. An explicit mention of the use of these principles and a clearjustification, however, was found only in a limited number ofexamples. Assessments related almost exclusively to individualsubstances rather than categories. Major fields of application were(a) hazard identification to derive appropriate classification, or (b)risk assessment to define concentration levels and/or thresholdsfor safe use. Extra- or interpolation in case of data-poor substanceswas rarely done. In those cases where a category justification wasmade, a significant variation of quality and level of detail could beobserved. With regard to products, assessment is mostly based on

Table 3Read-across/WoE from public databases.

Chemical group characteristics Examples Source

1 Salts where the respective counter-cations or anions are ‘‘neutral’’; salts andtheir respective acids

Nickel salts OECDThioglycolic acid and ammonium or sodium salts OECDBenzoic acid SCCP

2 Close homologues with varying carbon chain length Anionic surfactants (alkyl sulphates, alkane sulphonates, a-olefinesulphonates)

OECD

a-Olefines OECDAlkylamidopropyl betaines (amphoteric surfactants, carboxylic structure,varying chain length)

OECD

Linear alkylbenzene sulphonates HERA3 Ethers from the same alcohol, or alkoxylates from different alcohols Ceteareth-n, various polyoxyethylene ethers of cetearyl/stearyl/

cetylacloholCIR

Ceteth-n, various PEG ethers of cetyl alcohol CIRAlcohol ethoxylates HERAAlcohol ethoxysulphates HERAPropylene glycol ethers (Vink et al., 2010) PubMed

4 Related esters Ester quats HERA2 Alkyl ester quarternary ammonium cpds. (Jowsey et al., 2007) PubMedGlyceryl myristate and related esters CIRPEGs sorbitan fatty acid esters CIR

5 Similar polymers Polyacrylic acid homopolymers, poly-(acrylic/maleic) acid copolymersand their sodium salts

HERA

Acrylates, polyacrylates, various acrylate copolymers CIRRosin-based substances (Illing et al., 2009) PubMed

6 Plant extracts Storax (Liquidambar orientalis balsam extract + oil; Liquidmbar styracifluabalsam extract + oil)

SCCP

Opoponax (Commiphora erythrea glabrescens gum extract + oil) SCCP

Table 4Alternative assays available within COLIPA and other organisations.

Method Developed by Assay type Reference Mirroring step inFig. 1

1. Toxicokinetic J. Kasting, Uni ofCincinnati, USA

In silico Wang et al. (2006) 2

2. DPRA P & G In chemico Gerberick et al. (2004, 2007) 53. MUSST L’Oreal DC in vitro Ade et al. (2006) 64. h-CLAT KAO&Shiseido DC in vitro Ashikaga et al. (2006), Sakaguchi et al. (2006) 65. NCTC IL-18 test Sens-it-iv KC in vitro Corsini et al. (2009) 36. Tiered approach Sens-it-iv KC and epidermal equivalent Corsini et al. (2009), Spiekstra et al. (2009) 2, 37. AREc32 CXR Biosciences AREc32 cells in vitro reporter gene

activityNatsch and Emter (2008) 3, 6

8. MUTZ-3 Genesignature

Sens-it-iv DC in vitro gene expression Johansson et al. (2011) 6

9. Peptide reactivity Unilever In chemico Aleksik et al. (2009) 510. DPRA next

generationP & G Univ. Strasbourg In chemico Gerberick et al. (2009) 1, 4, 5

11. KeratinoSens Givaudan KC in vitro, reporter gene activation Emter et al. (2010) 3, 612. VITOSens™ CARDAM-VITO DC in vitro Gene expression Hooyberghs et al. (2008) 613. PBMDC Beiersdorf moDC in vitro Reuter et al. (2011) 614. SenCeeTox™ CeeTox GSH depletion & KC in vitro gene

expressionMcKim et al. (2010) 3, 5

15. Multi parameterbiosensor

Univ. of Toledo, Ohio, USA DC in vitro, reporter gene activation Mizumoto et al. (2002, 2003, 2005) 3, 6

16. DC migration Imperial College London DC in vitro, reporter gene activation Pease and Williams (2006) 617. T cell proliferation Univ. Freiburg PBMC in vitro Martin et al. (2010), Kotturi et al. (2008), Moon

et al. (2007)7

18. T cell proliferation Univ. Lyon PBMC in vitro Vocanson et al. (2006) 7

46 C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52

ingredient data, but also experience with the product itself or sim-ilar products could be used. In some cases the WoE approach is alsoused for the quality evaluation of specific test data.

As a general recommendation, a sound rationale for the ap-proach in each individual case should be provided to demonstratethe scientific validity. The rationale needs to take into consider-ation the applicability for the class of chemical under investigation.This can be done by providing in vitro data demonstrating that thechemistry-based prediction is supported by real data for directcomparison to chemicals of the identical applicability domain.

2.3. In vitro methods

The non-animal test methods currently under development orevaluation (within the COLIPA research programme or by otherorganisations (Table 4) aim to replace the need for new animal testdata through predicting how cosmetic ingredients will affect thevarious chemical or biological pathways identified as relevant tothe induction of skin sensitisation in humans so far. At present, noneof these in vitro test methods has been accepted by national or inter-national regulatory authorities as a partial or full replacement for

C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52 47

animal tests. However, according to a survey among COLIPA mem-bers, the following three assays are being evaluated for inclusion intesting strategies for hazard and risk assessment in severalcompanies:

� The direct peptide reactivity assay (DPRA) discriminates sensi-tisers (S) from non-sensitisers (NS) on the basis of their chemi-cal reactivity towards two synthetic peptides (containing eithera single cysteine or lysine as a reaction target) and by measur-ing their depletion using High Performance Liquid Chromatog-raphy. As an improvement to this assay an oxidation step hasbeen included (No. 10 in Table 4) using hydrogen peroxide inthe presence or absence of horseradish peroxidase to oxidisethe chemical under investigation to enable detection of pre-and pro-haptens (Gerberick et al., 2007).� The myeloid U937 skin sensitisation test (MUSST) uses flow

cytometry to determine induction of the activation markerCD86 on the DC line U937 for S/NS discrimination (Ade et al.,2006; Rousset et al., 2002).� The human cell line activation test (h-CLAT) employs the DC

line THP-1 and bases S/NS decision on CD86 plus CD54 induc-tion (Ashikaga et al., 2006; Sakaguchi et al., 2006).

DPRA, MUSST and h-CLAT are part of an ECVAM Phase III pre-validation study (Hartung et al., 2004) and some of the other tests,such as the tired approach (using keratinocytes and epidermalequivalents; see No. 6 in Table 4) and the KeratinoSens (using re-porter gene activation in keratinocytes; see No. 11 in Table 4),are short of entering pre-validation experiments.

2.3.1. Recommendations for further researchThe improvement of in vitro assays is considered as a major to-

pic for further development to allow for hazard identification andpotency prediction in a future WoE approach. Currently, the DPRA,h-CLAT and MUSST have been the subject of substantial evaluationincluding inter-laboratory assessments and have been accepted byECVAM for inclusion in a pre validation study. Assuming a success-ful outcome, they may soon be included in a hazard identificationtoolbox. However, to generate potency information, there is cur-rently only a limited understanding which assays or which combi-nation thereof may be best suited. Thus, further research is neededto identify the most promising assays and to increase the under-standing how in vitro results can be interpreted to provide potencyinformation to finally feed into practical decision-making pro-cesses as applied under QRA approaches.

2.4. Historical data

The evaluation of the quality of data and its subsequent use inthe hazard and risk assessment is an important aspect. In general,information regarding the test substance, the test report/protocolis critical and their relevance and reliability should be consideredbased upon expert judgement. This also applies to studies pub-lished in the peer-reviewed literature and justification of theacceptability of these studies might be required. Decisions to in-clude or reject a study for risk assessment purposes must be madeon case-by-case bases taking into account factors such as theimportance of the study in the context of the risk assessmente.g., the extent of the deficiencies of the existing study versus cur-rent testing methodologies and the applicability/extent of othersupporting data.

Evaluation of standardised animal studies should follow thegeneral requirements for the LLNA or the Guinea pig tests as pro-vided in the respective OECD guidelines and in the ECETOC techni-cal report on skin sensitisation (ECETOC, 2003).

Evaluation of historical human experience with exposure to sub-stances and preparations that report on cutaneous reactions (aller-gic contact dermatitis, eczema) may come from a variety of sources:

� Consumer experience and comments, possibly followed up byprofessionals (e.g., diagnostic patch tests, bronchial provocationtests, skin prick tests and measurements of specific IgE serumlevels).� Diagnostic clinical studies.� Records of workers’ experience, accidents, and exposure studies

including medical surveillance.� Case reports in the general scientific and medical literature.� Consumer tests (monitoring by questionnaire and/or medical

surveillance).

For the above listed data frequency and severity of the reportedeffects needs to be evaluated on a case by case basis and requiresexpert judgement.

More precise information can be obtained if human experimen-tal data are available. For skin sensitisation there are volunteerstudies such as the HRIPT and the human maximisation test(HMT). Respective outlines of these clinical test protocols and forthe necessary ethical considerations are available in the ECETOCmonograph No. 32 (ECETOC, 2002). The robustness of a given pro-tocol depends among others on the number of dose groups, theapplication frequency, the amount applied and the skin site usedfor application. Depending on the quality of the data, potencyinformation can be obtained.

2.5. Exposure assessment

The objective of exposure assessment is an estimate of the dose/concentration of the substance to which humans may be exposed,considering duration and frequency of their exposure. Since cos-metic products cover a wide range of product types, various expo-sure scenarios exist. Regarding skin sensitisation, every specificexposure scenario will be linked to a certain amount of substancethat may be applied to the skin surface and might have access tothe living epidermis through dermal absorption.

Generally, in use exposure levels can only be obtained on acase-by-case basis taking into consideration at least the followingfactors (SCCP, 2006):

- class of cosmetic product(s) in which the ingredient may beused,

- method of application,- concentration of the ingredient in the finished cosmetic

product,- quantity of product used at each application,- frequency of application,- total area of skin contact,- site of skin contact,- duration of contact (e.g., rinse-off products),- foreseeable misuse which may increase exposure,- consumer target group (e.g., children, people with ‘‘sensitive

skin’’).

Furthermore, usage of cosmetic products may depend on habitsand practices of usage including factors such as age group, seasonalvariations, local habits, fashion, trends, disposable income, andproduct innovation. Typical exposure scenarios for cosmetic prod-ucts are provided in the SCCP’s ‘‘Notes of guidance’’ (SCCP, 2006).

For a skin sensitisation exposure assessment the skin surfacearea exposed with the cosmetic product (dose/unit area) contain-ing the ingredient under study, has to be known, as well as the fre-quency of application of the product.

Identify sensitisation potential, e.g. LLNA

Define no expected skin sensitisation induction level (NESIL)

Apply sensitisation assessment factors (SAFs): •inter-individual variability (x 10) •vehicle/product matrix effects (x 1 – x 10) •use considerations (x 1 – x 10) Acceptable exposure level (AEL)

Compare AEL with consumer exposure level (CEL)

Market product and monitor

consumer and clinical feedback

?

Fig. 4. Quantitative risk assessment (QRA). Simplified flow diagram for the conductof a quantitative risk assessment (QRA) using animal-derived skin sensitisationpotency data (LLNA = local lymph node assay in mice); modified from (Basketteret al., 2008).

48 C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52

Specific information on exposure under use conditions of cos-metics can be obtained from dermal absorption studies followingthe requirements of the corresponding OECD guidelines 427 and428 (OECD, 2004a,b).

2.6. Risk assessment approaches

2.6.1. Quantitative risk assessmentThe risk of induction of skin allergy to a particular ingredient in a

product under specific exposure conditions can be determinedusing the quantitative risk assessment approach (QRA; see Fig. 4).This is a predictive model based upon evaluation of two parameters.

Firstly, the identification of the intrinsic sensitisation potency ofan ingredient (defined in a WoE approach) is undertaken to obtaina no effect level (No Expected Sensitisation Induction Level, NESIL).To this, appropriate Sensitisation Assessment Factors (SAF) have tobe applied in order to provide an estimate of acceptable humanexposure levels below which the adverse effect, i.e., the inductionof skin sensitisation, should not occur in a population of consumersusing this product.

The second component of the QRA is the quantification of theactual consumer exposure (Consumer Exposure Level, CEL, ex-pressed as dose per unit area) to the ingredient during use. Onlyboth components together provide a scientifically relevant quanti-fication of the risk of skin sensitisation associated with productusage. Comprehensive details of this approach have been delin-eated in a series of publications (e.g., Felter et al., 2002, 2003).The major elements of the QRA concepts will be briefly describedas follows:

2.6.1.1. Determination of the NESIL. The WoE NESIL is a derivedthreshold taking into account all relevant skin sensitisation data,mainly from LLNA and if available from HRIPT data and/or Guineapig studies used as supportive evidence (Api and Vey, 2008;Basketter et al., 2000, 2005; Gerberick et al., 2001; Griem et al.,2003; Schneider and Akkan, 2004). The NESIL is always definedas a dose per unit area (e.g., lg/cm2).

Based on potency values obtained in the LLNA (EC3, effectiveconcentration of the test substance required to produce a threefoldincrease in the stimulation index compared to vehicle-treated con-trols), sensitisers are grouped into four categories, each coveringone order of magnitude (i.e., extreme (EC3 <0.1%), strong (EC3P0.1–<1%), moderate (EC3 P1–<10%) and weak (EC3 P10–6100%)) (ECETOC, 2003; Kimber et al., 2003; Loveless et al.,2010). In cases where LLNA data exist for structural analogueswithin the same chemical class, a surrogate NESIL (based uponlowest EC3 in potency class) might be assigned based upon theabove for use in a QRA. This procedure includes the conservativeassumption to use the lowest EC3 value of the category assigned.

2.6.1.2. Derivation of an SAF. In dermal sensitisation risk assess-ment in general, it is necessary to extrapolate from the definedand controlled experimental conditions within the hazard charac-terisation tools used to obtain the data for defining the NESIL to theconditions that exist during consumer exposure to the product.

This is achieved by the application of SAFs, which accounts forthree parameters: (a) inter-individual variability, (b) vehicle/prod-uct matrix differences and (c) use considerations. The total SAF ap-plied will be product specific and in practice could range from 10(for products not intended to come into contact with skin) to1000 (for products applied to sensitive areas of the body). The con-cept and the parameters affecting the SAF are well described in theliterature (reviewed in Felter et al., 2002, 2003).

2.6.1.3. Human exposure. The CEL is a measure of exposure underintended and foreseeable conditions of use and takes into account,

habits and practices, frequency and duration of use and amount ofproduct used per application.

The two key elements for risk characterisation are the Accept-able Exposure Level (AEL) and the comparison of the AEL to theCEL. The AEL is determined by dividing the WoE NESIL by the SAFand is expressed as dose per unit area. If for a particular ingredientthe relationship is AEL > CEL, it is considered that the proposedexposure to this ingredient poses negligible risk of induction of skinsensitisation (see Fig. 4).

2.6.2. Threshold conceptsInspired by the TTC (Threshold of Toxicological Concern) con-

cept for systemic toxicity, originally developed in the food sector(Rulis, 1986) and accepted in that area for certain applicationse.g., for food flavours, and adaptation of the TTC concept has re-cently been suggested for skin sensitisation for ingredients usedin cosmetic and personal care products. In general, the TTC conceptdefines chemical exposures below which the general risks of ad-verse toxicological effects are considered to be negligible (Kroeset al., 2004).

Two similar TTC concepts for the potential to induce skin sensi-tisation have been recently suggested based on animal and humandata sets: the Dermal Sensitisation Threshold (DST) (Safford, 2008)and the Threshold of Sensitisation Concern (TSC) (Keller et al.,2009). While DST and TSC both address the same human sensiti-sation threshold expressed in the appropriate dose metric (lg/cm2), the calculations are based on different data sets: the DST re-lies on available animal (LLNA) data, the TSC employs a databasewith human experience. The two analyses therefore can be consid-ered as complementary exercises. The combined results indicatethat human data-derived findings largely support conclusionsdrawn from the LLNA data.

In addition to the published data (Safford, 2008), a proposal hasbeen made to further refine the DST by classifying chemicalsaccording to their protein binding potential using data of (Aptulaand Roberts, 2006; Safford et al., 2011). In this way a higher thresh-old value can be assigned to chemicals which are classified into thenonreactive domain. Such an analysis of structural alerts could alsobe used to derive more differentiated TSC values as indicated byKeller et al. (2009). Table 5 shows maximum concentrations of

C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52 49

ingredients for some typical cosmetic applications based on theapplication of the DST, refined DST and TSC approaches.

2.6.2.1. Recommendations for further research. Identification of thepresence/absence and the quality of structural alerts (as indicatedby chemistry prediction) might be used to refine the thresholddetermination and lead to a more targeted and accurate applica-tion of threshold concepts. The same may in the future hold truefor testing information from in vitro methods in a similar manneras it was suggested for the TTC concept (Cheeseman et al., 1999).In addition, expanding the underlying datasets which define thethreshold would further substantiate the applicability domain ofthese approaches and possibly lead to a more comprehensivesystem.

2.7. Integration of the available tools

The integration of data/information from the available tools isconsidered the way forward for future safety assessment, but indi-vidual tests and combinations thereof are not yet at a final stage ofdevelopment. To further optimise the approach, generation ofcomprehensive databases including the results obtained with a lar-ger set of chemicals is needed to increase its value for safetyassessment purposes. This is already the case for tools in an ad-vanced development stage, which in many instances are alreadyapplied for decision-making within industry. For example, compar-ison of in vivo benchmark data (e.g., LLNA data) with results fromthe same chemical analysed in the three in vitro assays currentlyunder pre-validation (e.g., DPRA, h-CLAT and MUSST) could be fur-ther expanded by performing in silico predictions and applicationof threshold concepts. Table 6 illustrates which mechanistic as-pects are included in the three in vitro assays currently underpre-validation compared to in vivo and T-cell assays. Since noneof the in vitro assays (with the DPRA being also representative ofin chemico reactivity prediction) covers all mechanistic aspects ofthe in vivo immune response, it needs to be further investigatedwhich (set of) tools is needed in order to achieve an appropriateprediction. It should be noted that DPRA/chemical reactivity deter-mination provides a high correlation with LLNA results (Gerbericket al., 2008) but does not address known chemical reactivity inde-pendent mechanisms as e.g., described for metal ions such as nick-el (Martin et al., 2010).

Gaining and sharing experience of data integration for riskassessment should allow to identify what data/information willbe most valuable for certain chemicals and exposure scenarios.The aim will be to create guidelines on how to make use of genericdata types (e.g., peptide reactivity data) rather than specific testmethods. It may turn out that an appropriate prediction can al-ready be derived from a quite limited dataset without the necessityto reflect all key mechanistic in vivo processes.

Table 5Examples of the application of threshold concepts (DST and TSC) to suggest maximum us

Product type DSTa derived AEL (lg/cm2)/use conc. (%) Refined D

Rinse-offShampood (rinse-off, SAF 100) 1.64/2.1 9/11.7Conditioner (rinse-off, SAF 100) 1.64/1.6 9/9.3

Leave-ond

Face cream (leave-on, SAF 100) 1.64/0.06 9/0.33Body lotion (leave-on, SAF 300) 0.55/0.11 3/0.6Deo/AP (leave-on, SAF 300) 0.55/0.007 3/0.04

a Derived by probabilisitc analysis, gamma distribution, 75 percentile from distributiob Chemicals classified in the non-reactive domain (Safford et al., 2011).c Derived by probabilisitc analysis, skew normal distribution, 95 percentile from distrd Product exposure estimates were performed according to SCCS Notes of Guidanc

publications cited under.a–c

3. Guiding principles for the use and further development ofavailable tools in the context of skin sensitisation safetyassessment

Based on the thorough analysis of the currently available toolsdescribed in the previous chapter, it was evaluated as to how a pro-cess to assess the risk to induce skin sensitisation for a new chem-ical could be structured without new in vivo (LLNA) data. The term‘guiding principles’ was considered appropriate to express that thesuggested process is a pragmatic starting point to combine thetools available for a risk assessment in a WoE approach givingspace for further refinement/development.

In addition, there is the need to emphasise that expert knowl-edge is a key factor to decide if the available data are sufficientto perform a risk assessment on a case by case basis when follow-ing the suggested guiding principles or if for the chemical class un-der review additional information is needed that might not beobtained with the currently available tools.

As an example, Fig. 5 suggests a process for data combinationand interpretation. It allows in a stepwise manner a considerationfor both hazard classification and/or risk assessment or if addi-tional data would be needed to allow for a decision regarding therisk of skin sensitisation under use conditions.

An initial need in the safety assessment process is to answer thequestion whether a certain ingredient may present a hazard at alland is usually considered in parallel with the expected exposurefrom the product type(s) of interest.

First, an assessment of all relevant existing data including thosefrom humans and animals should be performed. If sufficient dataare available to identify a material as a non-sensitiser, then thisstatus can be assigned. If this is not the case, a first step in evalu-ating the safety of a material would be to consider the degree ofhuman exposure and consider whether it is appropriate to applythe concept of the TTC.

If not, information on the physico-chemical properties, verifica-tion of structural alerts, and the mechanism of action may be suf-ficient to allow for hazard assessment and/or for priority setting.

If there is still insufficient information, data from in vitro assayscould be considered. It should be recognised that data from one ormultiple assays may be required. One criterion for the assay selec-tion could be its performance during interlaboratory evaluation(ring trials) and the number of chemicals tested. Currently, theseare the DPRA for information on chemical reactivity and theh-CLAT and MUSST for dendritic cell responses.

The other initial key consideration apart from the yes/no ques-tion regarding the hazard is the exposure assessment, as the quan-tification of exposure is a prerequisite for the quantification of risk.For cosmetics, this information can be taken from generally appliedstandard data on habits and practices of use (e.g., SCCS, 2010) andshould include information on intensity, frequency and duration of

e levels in cosmetic rinse-off and leave-on products.

STb derived AEL(lg/cm2)/use conc. (%) TSCc derived AEL(lg/cm2)/use conc. (%)

0.9/1.20.9/0.93

0.9/0.0330.3/0.060.3/0.004

n (Safford, 2008).

ibution (Keller et al., 2009).e (SCCS, 2010) and values were modified accordingly compared to the original

Table 6Mechanistic skin sensitisation parameters addressed in individual test systems.

Skinpenetration

Skinmetabolism

Chemicalreactivity

Innate immune activation (dendritic cells) Adaptive immuneactivation(T cells)

Contact dermatitis (patch test) + + + + +

LLNA +a +a + + ++

DPRA – –b ++ – –MUSST – ? + ++ –h-CLAT – ? + ++ ––T cell assay – ? + + ++

+ Parameter indirectly assessed.++ Parameter directly assessed.

? Ability to address parameter unclear.a May be different to human skin.b Metabolic activation has been included in next generation of assay (Gerberick et al., 2009).

hazard characterisation/ potency

hazard identification exposure assessment

chemistry prediction:(non-testing)

• phys-chem properties• structural alerts• mechanistic domain

threshold concepts(DST/TSC)

WoE consideration for potency/ read across

• based on chemistry predictions• based on historical data

skin metabolism (pro-haptens) and/or

auto-oxidation (pre-haptens)

dermal bioavailabilitysuch as percutaneous

absorption in vitro

potency (W, M, S, E)„surrogate NESIL“

QRA

information for hazardclassification

expected human exposurescenario

(standard procedure: SCCS guidance)

in vitro assays selected fromtoolbox for induction of skin

sensitisation

Fig. 5. Elements for non-animal skin sensitisation safety assessment approaches of cosmetics.

50 C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52

the skin exposure. Exposure assessments could be refined byassessing the dermal bioavailability, i.e., the amount that can reachthe living parts of the epidermis where skin sensitisation is initi-ated provided the knowledge base for interpretation is sufficientlydeveloped.

In vitro assays, such as the DPRA, h-CLAT and MUSST (Nos. 2, 3and 4 in Table 4) have been discussed in Section 2.3. Although cur-rently still under evaluation for the purpose of hazard identifica-tion, they might be used to support the decision making processin a WoE risk assessment. There are initial indications that withincertain applicability domains the results can also be used forobtaining relative potency estimations. In order to test this hypoth-esis, the in vitro derived potency information should be comparedto the corresponding in vivo data (within their respective applica-bility domains). If such an approach is considered valid no newin vivo data might be needed.

Furthermore, possible activation of the chemical through skinmetabolism (pro-hapten) and/or auto oxidation (pre-hapten)should also be considered during the WoE evaluation. For pre-hap-tens, the oxidised form is relevant for the induction of skin sensiti-sation but depending on the exposure conditions might not begenerated during product use.

Already today there can be cases where all above considerationsmay justify the assignment of a potency class (i.e., weak, moderate,

strong, extreme; for reference see (ECETOC, 2003) and derivation ofa corresponding surrogate NESIL to be finally used in a QRA asdescribed in Section 2.5. In most instances however the currenttoolbox is not yet sufficiently developed to assign a potency classwithout animal data.

4. Conclusions

Current hazard and risk assessment of skin sensitisation is com-monly based on WoE assessments, which in most cases make useof animal (LLNA) data. Future WoE-based approaches will dependon the use, optimisation and new development of non-animal toolsand their integration into appropriate assessment/testing strate-gies. The COLIPA research programme has been designed to devel-op in vitro assays for several key mechanistic processes involved inskin sensitisation. Several protocols for the detection of skin sensi-tising potential of chemicals (hazard) are currently at the pre-validation stage (i.e., DPRA, h-CLAT and MUSST). However, thegeneration of in vitro data that can be translated into quantifiablepotency information for practical risk assessment remains a keychallenge.

Equally important is the optimisation of physico-chemical prop-erty-based prediction models (including in silico tools) and toexplore how the prediction from these models can be appropriately

C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52 51

combined with in vitro data, within integrated testing and assess-ment strategies. To achieve this goal, more chemicals need to beevaluated within the different available prediction models in orderto investigate their predictivity. Similarly, threshold concepts arelikely to be refined when more data from different chemical classesare integrated, including the consideration of data from in vitro andin silico methods to refine for the absence or presence of alerts.

While data gaps and research needs are still to be worked on, aconceptual approach for the integration of data from varioussources has been outlined. These guiding principles suggestinghow to apply and combine the available tools in a WoE approachare considered a pragmatic starting point in order to instruct fur-ther research towards the non-animal quantitative risk assessmentof skin sensitisation. In addition to perspectives provided in recentreviews (Basketter and Kimber, 2009, 2010; Vandebriel and vanLoveren, 2010), this approach highlights that identification of therelevant mechanistic steps and the optimum set of tools suitedto reliably predict the skin sensitisation potency is a decisive futureneed.

Conflict of interest statement

The authors declare they have no conflict of interest. The workon which this publication is based was conducted under the um-brella of research for alternatives to animal testing of COLIPA, theEuropean Cosmetics Association, a research programme assistingindustry with regulatory compliance and without intention ofmaking profit from research results.

References

Ade, N., Martinozzi-Teissier, S., Pallardy, M., Rousset, F., 2006. Activation of U937cells by contact sensitizers: CD86 expression is independent of apoptosis. JImmunotoxicol. 3, 189–197.

Aeby, P., Ashikaga, T., Bessou-Touya, S., Schepky, A., Gerberick, F., Kern, P., Marrec-Fairley, M., Maxwell, G., Ovigne, J.M., Sakaguchi, H., Reisinger, K., Tailhardat, M.,Martinozzi-Teissier, S., Winkler, P., 2010. Identifying and characterizingchemical skin sensitizers without animal testing: Colipa’s research andmethod development program. Toxicol. In Vitro 24, 1465–1473.

Aleksik, M., Thain, E., Roger, D., Saib, Q., Davies, M., Li, J., Aptula, A., Zazzeroni, R.,2009. Reactivity profiling: covalent modification of single nucleophile peptidesfor skin sensitization risk assessment. Toxicol. Sci. 108, 401–411.

Api, A.M., Vey, M., 2008. Implementation of the dermal sensitization quantitativerisk assessment (QRA) for fragrance ingredients. Regul. Toxicol. Pharmacol. 52,53–61.

Aptula, A.O., Roberts, D.W., 2006. Mechanistic applicability domains for nonanimal-based prediction of toxicological end points: general principles and applicationto reactive toxicity. Chem. Res. Toxicol. 19, 1097–1105.

Ashikaga, T., Yoshida, Y., Hirota, M., Yoneyama, K., Itagaki, H., Sakaguchi, H.,Miyazawa, M., Ito, Y., Suzuki, H., Toyoda, H., 2006. Development of an in vitroskin sensitization test using human cell lines: the human Cell Line ActivationTest (h-CLAT). I. Optimization of the h-CLAT protocol. Toxicol. In Vitro 20, 767–773.

Basketter, D., Cockshott, A., Corsini, E., Gerberick, G.F., Idehara, K., Kimber, I., VanLoveren, H., Matheson, J., Mehling, A., Omori, T., Rovida, C., Sozu, T., Takeyoshi,M., Casati, S., 2008. An evaluation of performance standards and non-radioactive endpoints for the local lymph node assay. The report andrecommendations of ECVAM Workshop 65. [Review] [60 refs]. ATLA 36, 243–257.

Basketter, D.A., Blaikie, L., Dearman, R.J., Kimber, I., Ryan, C.A., Gerberick, G.F.,Harvey, P., Evans, P., R., W.I., Rycroft, R.J.G., 2000. Use of the local lymph nodeassay for the estimation of relative contact allergenic potency. ContactDermatitis 42, 344-348.

Basketter, D.A., Clapp, C., Jefferies, D., Safford, B., Ryan, C.A., Gerberick, F., Dearman,R.J., Kimber, I., 2005. Predictive identification of human skin sensitizationthresholds. Contact Dermatitis 53, 260–267.

Basketter, D.A., Gerberick, G.F., Kimber, I., 2007. The local lymph node assay: currentposition in the regulatory classification of skin sensitizing chemicals. Cutan.Ocul. Toxicol. 26, 293–301.

Basketter, D.A., Kimber, I., 2009. Updating the skin sensitization in vitro dataassessment paradigm in 2009. J. Appl. Toxicol. 29, 545–550.

Basketter, D.A., Kimber, I., 2010. Updating the skin sensitization in vitro dataassessment paradigm in 2009 – a chemistry and QSAR perspective. J. Appl.Toxicol. 30, 289.

Buehler, E.V., 1965. Delayed contact hypersensitivity in the guinea pig. Arch.Dermatol. 91, 171–175.

Cheeseman, M.A., Machuga, E.J., Bailey, A.B., 1999. A tiered approach to threshold ofregulation. Food Chem. Toxicol. 37, 387–412.

Chipinda, I., Ajibola, R.O., Morakinyo, M.K., Ruwona, T.B., Simoyi, R.H., Siegel, P.D.,2010. Rapid and simple kinetics screening assay for electrophilic dermalsensitizers using nitrobenzenethiol. Chem. Res. Toxicol. 23, 918–925.

Corsini, E., Mitjans, M., Galbiati, V., Lucchi, L., Galli, C.L., Marinovich, M., 2009. Use ofIL-18 production in a human keratinocyte cell line to discriminate contactsensitizers from irritants and low molecular weight respiratory allergens.Toxicol. In Vitro 23, 789–796.

ECETOC, 2002. Use of Human Data in Hazard Classification for Irritation andSensitisation. ECETOC Monograph No. 32. European Centre for Ecetoxicologyand Toxicology of Chemicals, Brussels, Belgium.

ECETOC, 2003. Contact sensitisation: Classification according to potency. Technicalreport No. 87.

Emter, R., Ellis, G., Natsch, A., 2010. Performance of a novel keratinocyte-basedreporter cell line to screen skin sensitizers in vitro. Toxicol. Appl. Pharmacol.245, 281–290.

Enoch, S.J., Madden, J.C., Cronin, M.T., 2008. Identification of mechanisms of toxicaction for skin sensitisation using a SMARTS pattern based approach. SAR QSAREnviron. Res. 19, 555–578.

EU, 2006. Regulation (EC) No. 1907/2006 of the European Parliament and of theCouncil of 18 December 2006 concerning the Registration, Evaluation,Authorisation and Restriction of Chemicals (REACh), establishing a EuropeanChemicals Agency, amending Directive 1999/45/EC and repealing CouncilRegulation (EEC) No. 793/93 and Commission Regulation (EC) No. 1488/94 aswell as Council Directive 76/769/EEC and Commission Directives 91/155/EEC,93/67/EEC, 93/105/EC and 2000/21/EC. Official Journal of the European Union. L396.

EU, 2008. Regulation (EC) No. 1272/2008 of the European Parliament and of theCouncil on classification, labelling and packaging of substances and mixtures.Official Journal of the European Union. 2008, 1-1354.

EU, 2009. Regulation (EC) No 1223/2009 of the European Parliament and of thecouncil of 30 November 2009 on cosmetic products (recast). Official Journal ofthe Eurpean Union. L 342/59-209.

Felter, S.P., Robinson, M.K., Basketter, D.A., Gerberick, G.F., 2002. A review of thescientific basis for uncertainty factors for use in quantitative risk assessment forthe induction of allergic contact dermatitis. Contact dermatitis 47, 257–266.

Felter, S.P., Ryan, C.A., Basketter, D.A., Gilmour, N.J., Gerberick, G.F., 2003.Application of the risk assessment paradigm to the induction of allergiccontact dermatitis. Regul. Toxicol. Pharmacol. 37, 1–10.

Franot, C., Roberts, D.W., Basketter, D.A., Benezra, C., Lepoittevin, J.P., 1994.Structure-activity relationships for contact allergenic potential of gamma,gamma-dimethyl-gamma-butyrolactone derivatives. 2. Quantitative structure-skin sensitization relationships for alpha-substituted-alpha-methyl-gamma,gamma-dimethyl-gamma-butyrolactone s. Chem. Res. Toxicol. 7, 307–312.

Gerberick, F., Aleksic, M., Basketter, D., Casati, S., Karlberg, A.T., Kern, P., Kimber, I.,Lepoittevin, J.P., Natsch, A., Ovigne, J.M., Rovida, C., Sakaguchi, H., Schultz, T.,2008. Chemical reactivity measurement and the predictive identification of skinsensitisers. Altern. Lab. Anim. 36, 215–242.

Gerberick, G.F., Robinson, M.K., Felter, S., 2001. Understanding fragrance allergyusing an exposure-based risk assessment approach. Contact Dermatitis 45,333–440.

Gerberick, G.F., Vassallo, J.D., Bailey, R.E., Chaney, J.G., Morrall, S.W., Lepoittevin, J.P.,2004. Development of a peptide reactivity assay for screening contact allergens.Toxicol. Sci. 81, 332–343.

Gerberick, G.F., Vassallo, J.D., Foertsch, L.M., Price, B.B., Chaney, J.G., Lepoittevin, J.P.,2007. Quantification of chemical peptide reactivity for screening contactallergens: a classification tree model approach. Toxicol. Sci. 97, 417–427.

Gerberick, G.F., Troutman, J.A., Foertsch, L.M., Vassallo, J.D., Quijano, M., Dobson,R.L., Goebel, C., Lepoittevin, J.P., 2009. Investigation of peptide reactivity of pro-hapten skin sensitizers using a peroxidase-peroxide oxidation system. Toxicol.Sci. 112, 164–174.

Griem, P., Goebel, C., H., S., 2003. Proposal for a risk assessment methodology forskin sensitization based on sensitization potency data. Regul Toxicol Pharmacol38, 269-290.

Hartung, T., Bremer, S., Casati, S., Coecke, S., Corvi, R., Fortaner, S., Gribaldo, L.,Halder, M., Hoffmann, S., Roi, A.J., Prieto, P., Sabbioni, E., Scott, L., Worth, A.,Zuang, V., 2004. A modular approach to the ECVAM principles on test validity.Altern. Lab. Anim. 32, 467–472.

Hooyberghs, J., Schoeters, E., Lambrechts, N., Nelissen, I., Witters, H., Schoeters, G.,Van Den Heuvel, R., 2008. A cell-based in vitro alternative to identify skinsensitizers by gene expression. Toxicol. Appl. Pharmacol. 231, 103–111.

Illing, H.P., Malmfors, T., Rodenburg, L., 2009. Skin sensitization and possiblegroupings for ‘read across’ for rosin based substances. Regul. Toxicol.Pharmacol. 54, 234–241.

Johansson, H., Lindstedt, M., Albrekt, A.S., Borrebaeck, C.A., 2011. A genomicbiomarker signature can predict skin sensitizers using a cell-based in vitroalternative to animal tests. BMC Genomics. 12, 399 ([Epub ahead of print]).

Jowsey, I.R., Kligman, A.M., White, I.R., Goossens, A., Basketter, D.A., 2007. Evidencethat two alkyl ester quaternary ammonium compounds lack substantial humanskin-sensitizing potential. Dermatitis 18, 32–39.

Judson, P.N., 2002. Prediction of toxicity using DEREK for windows. Toxicol. Sci. 27,278.

Karlberg, A.T., Bergstrom, M.A., Borje, A., Luthman, K., Nilsson, J.L., 2008. Allergiccontact dermatitis–formation, structural requirements, and reactivity of skinsensitizers. Chem. Res. Toxicol. 21, 53–69.

52 C. Goebel et al. / Regulatory Toxicology and Pharmacology 63 (2012) 40–52

Keller, D., Krauledat, M., Scheel, J., 2009. Feasibility study to support a Threshold ofSensitization Concern (TSC) concept in risk assessment based on human data.Arch. Toxicol. 8, 1049–1060.

Kimber, I., Basketter, D.A., Butler, M., Gamer, A., Garrigue, J.L., Gerberick, G.F.,Newsome, C., Steiling, W., Vohr, H.W., 2003. Classification of contact allergensaccording to potency: proposals. Food Chem. Toxicol. 12, 1799–1809.

Kotturi, M.F., Scott, I., Wolfe, T., Peters, B., Sidney, J., Cheroutre, H., von Herrath, M.G.,Buchmeier, M.J., Grey, H., Sette, A., 2008. Naive precursor frequencies and MHCbinding rather than the degree of epitope diversity shape CD8+ T cellimmunodominance. J. Immunol. 181, 2124–2133.

Kroes, R., Renwick, A.G., Cheeseman, M., Kleiner, J., Mangelsdorf, I., Piersma, A.,Schilter, B., Schlatter, J., van Schothorst, F., Vos, J.G., Wurtzen, G., 2004.Structure-based thresholds of toxicological concern (TTC): guidance forapplication to substances present at low levels in the diet. Food. Chem.Toxicol. 42, 65–83.

Kroeze, K.L., Jurgens, W.J., Doulabi, B.Z., van Milligen, F.J., Scheper, R.J., Gibbs, S.,2009. Chemokine-mediated migration of skin-derived stem cells: predominantrole for CCL5/RANTES. J. Invest. Dermatol. 129, 1569–1581.

Loveless, S.E., Api, A.M., Crevel, R.W., Debruyne, E., Gamer, A., Jowsey, I.R., Kern, P.,Kimber, I., Lea, L., Lloyd, P., Mehmood, Z., Steiling, W., Veenstra, G., Woolhiser,M., Hennes, C., 2010. Potency values from the local lymph node assay:application to classification, labelling and risk assessment. Regul. Toxicol.Pharmacol. 56, 54–66.

Magnusson, B., Kligman, A.M., 1969. The identification of contact allergens byanimal assay. The guinea pig maximization test. J. Invest. Dermatol. 52, 268–276.

Martin, S., Weltzien, H.U., 1994. T cell recognition of haptens, a molecular view. Int.Arch. Allergy Immunol. 104, 10–16.

Martin, S.F., Esser, P.R., Schmucker, S., Dietz, L., Naisbitt, D.J., Park, B.K., Vocanson,M., Nicolas, J.-F., Keller, M., Pichler, W.J., Peiser, M., Luch, A., Wanner, R., Maggi,E., Cavani, A., Rustemeyer, T., Richter, A., Thierse, H.-J., Sallusto, F., 2010. T cellrecognition of chemicals, protein allergens and drugs: towards thedevelopment of in vitro assays. Cell. Mol. Life Sci. 24, 4171–4184.

McKim, J.M.J., Keller III, D.J., Gorski, J.R., 2010. A new in vitro method for identifyingchemical sensitizers combining peptide binding with ARE/EpRE-mediated geneexpression in human skin cells. Cutan. Ocul. Toxicol. 29, 171–192.

Mizumoto, N., Kumamoto, T., Robson, S.C., Sevigny, J., Matsue, H., Enjyoji, K.,Takashima, A., 2002. CD39 is the dominant Langerhans cell-associated ecto-NTPDase: modulatory roles in inflammation and immune responsiveness. Nat.Med. 8, 358–365.

Mizumoto, N., Mummert, M.E., Shalhevet, D., Takashima, A., 2003. Keratinocyte ATPrelease assay for testing skin-irritating potentials of structurally diversechemicals. J. Invest. Dermatol. 121, 1066–1072.

Mizumoto, N., Hui, F., Edelbaum, D., Weil, M.R., Wren, J.D., Shalhevet, D., Matsue, H.,Liu, L., Garner, H.R., Takashima, A., 2005. Differential activation profiles ofmultiple transcription factors during dendritic cell maturation. J. Invest.Dermatol. 124, 718–724.

Moon, J.J., Chu, H.H., Pepper, M., McSorley, S.J., Jameson, S.C., Kedl, R.M., Jenkins,M.K., 2007. Naive CD4(+) T cell frequency varies for different epitopes andpredicts repertoire diversity and response magnitude. Immunity 27, 203–213.

Natsch, A., Emter, R., 2008. Skin sensitizers induce antioxidant response elementdependent genes: application to the in vitro testing of the sensitizationpotential of chemicals. Toxicol. Sci. 102, 110–119.

Natsch, A., Gfeller, H., 2008. LC-MS-based characterization of the peptide reactivityof chemicals to improve the in vitro prediction of the skin sensitizationpotential. Toxicol. Sci. 106, 464–478.

OECD. Quantitative Structure–Activity Relationships Project [(Q)SARs] http://www.oecd.org/document/23/0,2340,en_2649_34365_33957015_1_1_1_1,00.htm.

OECD, 1992. OECD Guideline for the Testing of Chemicals 406: Skin Sensitisation,adopted: 17 July 1992.

OECD, 2004a. Guideline for the Testing of Chemicals 427: Skin absorption: in vivomethod, adopted: 13 April 2004.

OECD, 2004b. Guideline for the Testing of Chemicals 428: Skin absorption: in vitromethod, adopted: 13 April 2004.

OECD, 2010. OECD Guideline for the Testing of Chemicals 429: Skin Sensitisation:Local Lymphnode Assay, adopted: 22 July 2010. p. 7.

Ortmann, B., Martin, S., von Bonin, A., Schiltz, E., Hoschutzky, H., Weltzien, H.U.,1992. Synthetic peptides anchor T cell-specific TNP epitopes to MHC antigens. J.Immunol. 148, 1445–1450.

Pease, J.E., Williams, T.J., 2006. Chemokines and their receptors in allergic disease. J.Allergy Clin. Immunol. 118, 110–119.

Quayson, A., Seaman, C., Olson, M., Guerriero, F., 2010. Predictive performance ofsome structure activity tools for identification of dermal sensitizers inpharmaceutical operations. Toxicologist 114, 384.

Quayson, A., Seaman, C., Olson, M., Guerriero, F., in press. A weight of evidenceapproach to improving SAR prediction of allergic contact dermatitis.Toxicologist.

Reuter, H., Spieker, J., Gerlach, S., Engels, U., Pape, W., Kolbe, L., Schmucker, R.,Wenck, H., Diembeck, W., Wittern, K.P., Reisinger, K., Schepky, A.G., 2011. Invitro detection of contact allergens: development of an optimized protocolusing human peripheral blood monocyte-derived dendritic cells. Toxicol. InVitro 25, 315–323.

Roberts, D.W., Aptula, A.O., Patlewicz, G., 2007a. Electrophilic chemistry related toskin sensitization. Reaction mechanistic applicability domain classification for apublished dataset of 106 chemicals tested in the mouse local lymph node assay.Chem. Res. Toxicol. 20, 44–60.

Roberts, D.W., Patlewicz, G., Kern, P.S., Gerberick, F., Kimber, I., Dearman, R.J., Ryan,C.A., Basketter, D.A., Aptula, A.O., 2007b. Mechanistic applicability domainclassification of a local lymph node assay dataset for skin sensitization. Chem.Res. Toxicol. 20, 1019–1030.

Roberts, D.W., Aptula, A.O., Patlewicz, G., 2006. Mechanistic applicability domainsfor non-animal based prediction of toxicological endpoints. QSAR analysis of theschiff base applicability domain for skin sensitization. Chem. Res. Toxicol. 19,1228–1233.

Roberts, D.W., Natsch, A., 2009. High throughput kinetic profiling approach forcovalent binding to peptides: application to skin sensitization potency ofMichael acceptor electrophiles. Chem. Res. Toxicol. 22, 592–603.

Roberts, D.W., Aptula, A.O., Patlewicz, G.Y., 2011. Chemistry-based risk assessmentfor skin sensitization: quantitative mechanistic modeling for the S(N)Ardomain. Chem. Res. Toxicol. 24, 1003–1011.

Roberts, D.W., Basketter, D.A., 2000. Quantitative structure–activity relationships:sulfonate esters in the local lymph node assay. Contact Dermatitis 42, 154–161.

Roberts, D.W., Basketter, D.A., 1997. Further evaluation of the quantitativestructure-activity relationship for skin-sensitizing alkyl transfer agents.Contact Dermatitis 37, 107–112.

Roberts, D.W., Goodwin, B.F., Williams, D.L., Jones, K., Johnson, A.W., Alderson, J.C.,1983. Correlations between skin sensitization potential and chemical reactivityfor p-nitrobenzyl compounds. Food Chem. Toxicol. 21, 811–813.

Roberts, D.W., Williams, D.L., 1982. The derivation of quantitative correlationsbetween skin sensitisation and physio-chemical parameters for alkylatingagents, and their application to experimental data for sultones. J. Theor. Biol. 99,807–825.

Roberts, D.W., Aptula, A.O., Patlewicz, G., Pease, C., 2008. Chemical reactivity indicesand mechanism-based read-across for non-animal based assessment of skinsensitisation potential. J. Appl. Toxicol. 28, 443–454.

Rodford, R., Patlewicz, G., Walker, J.D., Payne, M.P., 2003. Quantitative structure–activity relationships for predicting skin and respiratory sensitization. Environ.Toxicol. Chem. 22, 1855–1861.

Rousset, F., Verda, D., Garrigue, J.L., Mariani, M., Leclaire, J., 2002. In vitro predictionof contact sensitivity with human cell lines. Contact Dermatitis 46, 6.

Rulis, A.M., 1986. De minimis and the Threshold of Regulation, vol. 1. LewisPublishers, Chelsea, MI.

Safford, R.J., 2008. The dermal sensitisation threshold-A TTC approach for allergiccontact dermatitis. Regul. Toxicol. Pharmacol. 51, 195–200.

Safford, R.J., Aptula, A.O., Gilmour, N., 2011. Refinement of the Dermal SensitisationThreshold (DST) approach using a larger dataset and incorporating mechanisticchemistry domains. Regul. Toxicol. Pharmacol. 60, 218–224.

Sakaguchi, H., Ashikaga, T., Miyazawa, M., Yoshida, Y., Ito, Y., Yoneyama, K., Hirota,M., Itagaki, H., Toyoda, H., Suzuki, H., 2006. Development of an in vitro skinsensitization test using human cell lines; human Cell Line Activation Test (h-CLAT). II. An inter-laboratory study of the h-CLAT. Toxicol. In Vitro 20, 774–784.

SCCS, 2010. (Scientific Committee on Consumer Products) The SCCS’s Notes ofGuidance for the Testing of Cosmetic Ingredients and their Safety Evaluation,7th revision. 14th December 2010.

Schneider, K., Akkan, Z., 2004. Quantitative relationship between the local lymphnode assay and human skin sensitization assays. Regul. Toxicol. Pharmacol. 39,245–255.

Schultz, T.W., Yarbrough, J.W., Johnson, E.L., 2005. Structure-activity relationshipsfor reactivity of carbonyl-containing compounds with glutathione. SAR QSAREnviron. Res. 16, 313–322.

Schultz, T.W., Rogers, K., Aptula, A.O., 2009. Read-across to rank skin sensitizationpotential: subcategories for the Michael acceptor domain. Contact Dermatitis.60, 21–31.

Spiekstra, S.W., Dos Santos, G.G., Scheper, R.J., Gibbs, S., 2009. Potential method todetermine irritant potency in vitro – comparison of two reconstructedepidermal culture models with different barrier competency. Toxicol. In Vitro23, 349–355.

Toebak, M.J., Gibbs, S., Bruynzeel, D.P., Scheper, R.J., Rustemeyer, T., 2009. Dendriticcells: biology of the skin. Contact Dermatitis 60, 2–20.

Troutman, J.A., Foertsch, L.M., Kern, P.S., Dai, H.J., Quijano, M., Dobson, R.L., Lalko,J.F., Lepoittevin, J.P., Gerberick, G.F., 2010. The incorporation of lysine into theperoxidase peptide reactivity assay for skin sensitization assessments. Toxicol.Sci.

Vandebriel, R.J., van Loveren, H., 2010. Non-animal sensitization testing: state-of-the-art. Crit. Rev. Toxicol. 40, 389–404.

Vink, S.R., Mikkers, J., Bouwman, T., Marquart, H., Kroese, E.D., 2010. Use of read-across and tiered exposure assessment in risk assessment under REACH – a casestudy on a phase-in substance. Regul. Toxicol. Pharmacol. 58, 64–71.

Vocanson, M., Hennino, A., Cluzel-Tailhardat, M., Saint-Mezard, P., Benetiere, J.,Chavagnac, C., Berard, F., Kaiserlian, D., Nicolas, J.F., 2006. CD8+ T cells areeffector cells of contact dermatitis to common skin allergens in mice. J. Invest.Dermatol. 126, 815–820.

Wang, T.F., Kasting, G.B., Nitsche, J.M., 2006. A multiphase microscopic diffusionmodel for stratum corneum permeability. I. Formulation, solution, andillustrative results for representative compounds. J. Pharm. Sci. 95, 620–648.