kolle sullivan 2012

Regulatory Toxicology and Pharmacology 64 (2012) 402–414

Contents lists available at SciVerse ScienceDirect

Regulatory Toxicology and Pharmacology

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

Applicability of in vitro tests for skin irritation and corrosion to regulatoryclassification schemes: Substantiating test strategies with data from routine studies

Susanne N. Kolle a,⇑, Kristie M. Sullivan b, Annette Mehling c, Bennard van Ravenzwaay a, Robert Landsiedel a

a BASF SE, Experimental Toxicology and Ecology, Carl-Bosch-Straße 38, 67056 Ludwigshafen, Germanyb Physicians Committee for Responsible Medicine, 5100 Wisconsin Ave. NW, Washington, DC 20016, USAc BASF Personal Care and Nutrition GmbH, Henkelstraße 67, 40589 Düsseldorf, Germany

a r t i c l e i n f o

Article history:Received 14 July 2012Available online 30 August 2012

Keywords:Skin irritationSkin corrosionRoutine testingIn vitroRhEEpiDermOECD TG 431OECD TG 439Predictivity

0273-2300/$ - see front matter � 2012 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.yrtph.2012.08.015

⇑ Corresponding author. Fax: +49 621 60 58043.E-mail address: [email protected] (S.N. Koll

1 AAALAC, Association for Assessment and AccredCare; ANVISA, Agência Nacional de Vigilância sanitárSurveillance Agency); C, corrosive; Cat, category; CADSD, EU Dangerous Substance Directive ClassificatECVAM, European Centre for the Validation of AlternScientific Advisory Committee; EU, European Union; EUand Packaging, European GHS Regulation (EC) No. 12negative; FNR, false negative rate; FP, false positive; FPRGHS, Globally Harmonized System of Classification an2011); GLP, Good Laboratory Practice; HET-CAM, hebrane; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltblue, CAS No. 298-93-1; n/a, not applicable; NPV, negaticlassified; NC, not corrosive; nd, not determined; NI, nofor Economic Co-operation and Development; PBS, phpositive predictive value; REACH, EU Regulation 19Evaluation, Authorisation, and Restriction of ChemicalsRP, real positive; RhE, reconstructed human epidermis;sodium dodecyl sulfate; SIT, skin irritation test; TGNations; US EPA, United States Environmental Protecti

a b s t r a c t

Skin corrosion or irritation refers to the production of irreversible or reversible damage to the skin follow-ing the application of a test substance, respectively. Traditionally, hazard assessments are conductedusing the in vivo Draize skin test, but recently in vitro tests using reconstructed human epidermis(RhE) models have gained regulatory acceptance. In this study, skin corrosion (SCT) and irritation tests(SIT) using a RhE model were implemented to reduce the number of in vivo tests required by regulatorybodies. One hundred and thirty-four materials were tested from a wide range of substance classesincluded 46 agrochemical formulations. Results were assessed according to UN GHS, EU-CLP, ANVISAand US EPA classification schemes. There was high correlation between the two in vitro tests. Assessmentof the SCT sensitivity was not possible due to the limited number of corrosives in the data set; SCT spec-ificity and accuracy were 89% for all classification systems. Accuracy (63–76%) and sensitivity (53–67%)were low in the SIT. Specificity and concordance for agrochemical formulations alone in both the SCT andSIT were comparable to the values for the complete data set (SCT: 91% vs. 89% specificity, 91% vs. 89%accuracy and SIT: 64–88% vs. 70–85% specificity, 56–75% vs. 63–76% accuracy).

� 2012 Elsevier Inc. All rights reserved.

1. Introduction

Assessment of the potential of a substance to cause damage tothe skin is a basic endpoint evaluated in regulatory toxicology. Thisendpoint is used to predict the hazard, i.e. the intrinsic properties,of a substance upon accidental or intentional contact with the skin.

ll rights reserved.

e).itation of Laboratory Animalia (Brazilian National HealthS, Chemical Abstract Service;ion 67/548/EEC (EU, 1967);ative Methods; ESAC, ECVAM-CLP, Classification, Labelling,

72/2008 (EU, 2008); FN, false, false positive rate; I, irritant;d Labeling of Chemicals (UN,

n’s egg chorioallantoic mem-etrazolium bromide, thiazolylve predictive value; not cl, nott irritant; OECD, Organisationosphate buffered saline; PPV,0/2006 on the Registration,(EU, 2006); RN, real negative;SCT, skin corrosion test; SDS,

, test guideline; UN, Unitedon Agency.

Traditionally, the Draize skin irritation test has been used for manydecades to predict skin irritation hazard (OECD, 2002; Draize et al.,1944). In this test, the test material is applied topically onto theshaved skin of rabbits. Skin corrosion or irritation refers to the pro-duction of irreversible or reversible damage to the skin over timefollowing the application of a test material, respectively.

Within the EU, a harmonized approach to the classification andlabeling of chemicals was implemented via the Dangerous Sub-stances Directive (DSD,1 67/548/EEC (EU, 1967)) in 1967. The goalof this directive was to provide better protection for public healthand the environment. This directive has now been replaced by thenew European Chemical Regulation REACH (Registration, Evaluation,Authorisation and Restriction of Chemicals (EU, 2006)). Worldwide,however, legislation found in various countries differs in the require-ments instated for the purposes of the classification and labeling ofsubstances. Labeling is also used to convey information on the hazardsof a substance to users via the material safety data sheets. This non-harmonized approach is problematic not only in terms of transportand trade but can also hinder efforts to protect consumers and work-ers. An important step in worldwide harmonization was the adoptionof the Globally Harmonized System (GHS) of Classification and Label-ing of Chemicals (UN, 2011). This act has been/will be integrated intothe national legislation of numerous countries around the world with-in the near future. Enactment of the full GHS has/will take place in astep-wise fashion, starting with chemical substances then moving

http://dx.doi.org/10.1016/j.yrtph.2012.08.015

mailto:[email protected]

http://dx.doi.org/10.1016/j.yrtph.2012.08.015

http://www.sciencedirect.com/science/journal/02732300

http://www.elsevier.com/locate/yrtph

Table 1In vivo categorization cut-offs for the four major classification systems discussed.

US EPA/ANVISA Category/Class IV Category/Class III Category/Class II Category/Class IReaction Scores < 2a > 2 < 3a > 3a

> 0 < 4c

UN GHS Not Classified Category 3 Category 2 Category 1 A/B/CReaction Scores < 1.5b

> 1.5 < 2.3b > 2.3 < 4.0b > 0 < 4c

EU-CLP Not Classified Category 2 Category 1 A/B/CReaction Scores <2.3b

> 2.3 < 4.0b > 0 < 4c

a For US EPA categories II, III, and IV: reactions observed in more than 1 animal from gradings at 72 h. For ANVISA toxicity classes II, III, and IV: reactions observed in morethan 1 animal from gradings at any observed timepoint.

b For UN GHS and EU-CLP categories 3 and 2: reactions in at least 2 of 3 tested animals from gradings at 24, 48 and 72 hours.c For all classification systems discussed here, a corrosive substance is a test material that produces destruction of skin tissue, namely, visible necrosis through the

epidermis and into the dermis, in at least 1 tested anima after exposure up to a 4 hour duration.

S.N. Kolle et al. / Regulatory Toxicology and Pharmacology 64 (2012) 402–414 403

to mixtures. Within the EU, GHS came into force in 2009 via the legis-lation referred to as Classification, Labelling and Packaging System(CLP; European GHS Regulation (EC) No. 1272/2008 (EU, 2008)) andis an integral part of REACH. EU-CLP has been in effect for substancessince December 2010, and will apply to mixtures as of June 2015.

Many of the evaluation schemes used for the identification of ahealth hazard were developed prior to GHS and some variations inclassification can take place. In the case of skin irritation and cor-rosion, EU-CLP differentiates between nonirritating or slightly irri-tating substances (no classification needed), skin irritatingsubstances (Category 2) and corrosive substances (Category 1 withthree subcategories: 1A, 1B and 1C). GHS Category 3 (mild irritantclassification) is optional (i.e. not classified according to EU-CLP).The cut-off scores used to differentiate between irritants and non-irritants have changed from an in vivo score of 2.0 to 2.3, which inturn has led to changes in classification. Substances with an in vivoscore between 2.0 and 2.3, which were classified as being irritantsunder DSD, are nonirritants according to EU-CLP. A comparison ofthe cut-off values for all four systems used here is provided in Ta-ble 1. Further, while the UN GHS and EU-CLP systems calculatein vivo scores by averaging scores at observation time points, theUS EPA and ANVISA (Brazilian National Health SurveillanceAgency) systems use the highest single score. Finally, while theUN GHS, EU-CLP, and US EPA systems allow a period of time forrecovery, the ANVISA system does not.

Over the past several years considerable progress has beenmade in the development of non-animal test methods for hazardidentification. Increasing concerns for animal welfare and the eth-ics of animal testing has been taken into consideration in REACH(EU, 2006) and even more so in the amendments of the EuropeanCosmetics Directive/Regulation (76/768/EEC (EU, 1976) and1223/2009 (EU, 2009)). A number of methods for the identificationof skin corrosion and skin irritation have gained a certain degree ofregulatory acceptance. The methods described in OECD TG 431(skin corrosion (OECD, 2004)) and OECD TG 439 (skin irritation(OECD, 2010)) utilize reconstructed human epidermis (RhE) mod-els to which the test material is applied. In this study, skin corro-sion and/or irritation test protocols using a RhE model wereintegrated into testing strategies to reduce the number of in vivoDraize skin irritation tests to be performed due to the require-ments of regulatory bodies. The 134 materials tested included awide range of substances from different chemical classes as wellas 46 agrochemical formulations. The results were assessed in aregulatory context according to the GHS (UN, 2011), EU-CLP (EU,2008), ANVISA (AENDA, 1992) and US EPA guidelines (EPA, 2007).

2. Materials and methods

2.1. Materials

Over the several years, 134 materials were tested in the in vivorabbit skin irritation tests (OECD TG 404) for registration purposes.

Using a tiered approach the materials were tested in the in vitroskin corrosion test (SCT; OECD TG 431) before subjected toin vivo testing. Thirty-eight materials were also tested in thein vitro skin irritation test (SIT; OECD TG 439). Identical batchesof all tested materials were used in both tests. Eighty-seven mate-rials were liquids (three were viscous) and 47 solids (one a waxysolid). Agrochemical formulations (n = 46) made up the largestproportion of the tested materials; other materials included acry-lates (n = 8), agrochemicals (n = 5), amines (n = 3), boron com-pounds (n = 6), emollients (n = 3), ionic liquids (n = 4),pharmaceutical compounds (n = 5), pigment/dyes (n = 8), polymers(n = 7), surfactants (n = 5), and a range of other materials (Fig. 1and Table 2).

2.2. In vivo acute dermal irritation/corrosion

The in vivo skin irritation test, initially described by Draize et al.(1944), was performed according to OECD TG 404 (OECD, 2002) inan AAALAC certified BASF SE laboratory under Good LaboratoryPractice (GLP) conditions and according to the provisions of theGerman animal welfare regulations. The potential of the test mate-rials to cause acute dermal irritation or corrosion was assessed by asingle topical application of 0.5 mL of the liquid test materials tothe intact skin of three White New Zealand rabbits (Centre LagoS.A., Vonnas, France) for 4 h. A stepwise procedure was used inwhich the test material was initially applied to one animal usinga patch with an area of 2.5 cm � 2.5 cm and covered with (semi)-occlusive dressing and, depending on the severity of the reactions,then to an additional two animals. After removal of the patch,residual material was removed from the application area viarinsing. The cutaneous reactions were assessed immediately afterremoval of the patch, approximately 1, 24, 48 and 72 h after re-moval of the patch, and then in weekly intervals until day 14. Skinreactions were evaluated by grading erythema, eschar formationand edema formation. Classifications presented here are based onthe results of the Draize skin irritation test and the criteria of thedifferent classification systems (Tables 3a and 3b). Here the op-tional Category 3 in the UN GHS system is used. All in vivo studieswere performed as regulatory requirements. No additional animaltesting was performed for the purpose of this study.

2.3. In vitro skin corrosion test (SCT)

The skin corrosion test using RhE was conducted in accordancewith OECD TG 431 (OECD, 2004). Briefly, the potential of the testmaterials to cause dermal corrosion was assessed following a sin-gle topical application of 50 lL for liquids or 25 lL bulk volume forsolids of the neat test material to a reconstructed three-dimen-sional human epidermis model (EpiDerm™, MatTek Corporation,Ashland, MA, USA). For this purpose, two EpiDerm™ tissues pertreatment were incubated with the test material for 3 min and1 h each. Tissue destruction was determined by measuring the

Fig. 1. Substance classes assessed in the in vitro skin corrosion (A) and skin irritation tests (B). The data sets for both the skin corrosion and skin irritation tests consist ofdiverse sets of substance classes; the largest is agrochemical formulations.

404 S.N. Kolle et al. / Regulatory Toxicology and Pharmacology 64 (2012) 402–414

metabolic activity of the tissue after exposure using a colorimetrictest. To this end, the reduction of mitochondrial dehydrogenaseactivity, assessed via formazan production after incubation witha tetrazolium salt (MTT), was measured. The formazan productionof the test-material treated tissues was compared to that of nega-tive control tissues. Sterile deionized water served as negative con-trol and 8 N potassium hydroxide solution (Sigma–Aldrich,Munich, Germany) as a positive control. A pretest was performedin order to assess the ability of the test material to directly reduceMTT. The test material was added to 0.9 mL of the MTT solutionand incubated in the dark at about 37 �C for 55–65 min. If thematerial was able to reduce the MTT directly or if visible residuesof the test material remained on the tissues after washing, freeze-killed control tissues were used.

The corrosive potential of the test materials was predicted fromthe mean relative tissue viabilities obtained after a 3-min treat-ment compared to the negative control tissues. A material wasconsidered to be ‘‘corrosive’’ if the mean relative tissue viabilityafter a 3-min treatment with a test material was less than to50%. In addition, those materials with a viability of P50% after a3-min treatment were considered to be ‘‘corrosive’’ if the mean rel-ative tissue viability after a 1-h treatment with a test material wasless than or equal to 15%.

2.4. In vitro skin irritation test (SIT)

The skin irritation test using RhE was conducted in accordancewith OECD TG 439 (OECD, 2010). Briefly, the potential of test mate-rials to cause dermal irritation was assessed by a single topicalapplication of 30 lL for liquids or 25 lL bulk volume for solids ofthe neat test material to the EpiDerm™ tissues (MatTek Corpora-tion, Ashland, MA, USA). Three EpiDerm™ tissues were incubatedwith the test material for 1 h followed by a 42-h post-incubationperiod. Tissue destruction was determined by measuring the met-abolic activity of the tissue after exposure and post-incubationusing a colorimetric test. The reduction of mitochondrial dehydro-genase activity, assessed via decreased formazan production fol-lowing incubation with MTT, was measured. The formazanproduction of the test-material treated RhE tissues was comparedto that of negative control tissues. Sterile phosphate buffered saline(PBS) served as a negative control, 5% (w/v) sodium dodecyl sulfate(SDS, Sigma, Germany) in sterile deionized water as a positive con-trol. To assess the ability of the test material to directly reduce MTTthe pretest described above (SCT) was performed.

The irritation potential of the test material was predicted bycomparing the mean relative tissue viabilities of the test-material

treated tissues to those of the negative control tissues. A test mate-rial was considered an ‘‘irritant’’ if the mean relative tissue viabilitywith a test material was less than or equal to 50%.

2.5. Classification

Table 2 shows the in vivo classification for each material accord-ing to the UN GHS, EU-CLP, US EPA, and ANVISA systems. The com-parision of the classifications according to the different systems issummarized in Table 3. As described above, minor differences inclassification scoring guidelines lead to different classificationsfor the same material.

2.6. Statistics

All calculations were performed with Microsoft Office Excel2010. After assessment of classification for skin irritating potentialaccording to the criteria mentioned above, the number of materialscorrectly or incorrectly identified by the in vitro SIT and SCT meth-ods were determined for the different classification systems.Thereby, the concordance (=(RP + RN)/all � 100%), the false nega-tive rate (FNR = FN/(RP + FN) � 100%) and the false positive rate(FPR = FP/(RN + FP) � 100%) were calculated; the latter resultingin sensitivity (=100% � FNR) and specificity (=100% � FPR). Thepositive predictive value (PPV) was determined as the number ofcorrect positive predictions amongst all positive predictions bythe in vitro tests (RP/(RP + FP) � 100%) and the negative predictivevalue (NPV) as the number of correct negative predictions amongstall negative predictions by the in vitro tests (RN/(RN + FN) � 100%).In addition, to account for the unbalanced data set (e.g. underrep-resentation of corrosives) the balanced accuracy was determinedas the average of sensitivity and specificity.

3. Results

Since the implementation of the SCT in 1999 and SIT in 2007 inthe BASF laboratory, historical control data for more than 90 SCTsand 30 SITs has been generated. The acceptance criteria as speci-fied in OECD TGs 431 and 439 were always met. During routinetesting, 134 materials (including 46 agrochemical formulations)were tested in both the in vitro (SCT) and in vivo skin corrosiontests. For a subset of 38 materials (including 16 agrochemical for-mulations), in vitro skin irritation test (SIT) data was also available.For any materials able to reduce MTT directly, comparisons to thefreeze-killed control tissues were made. Contingency tables for theSCT and SIT in comparison to the in vivo data for all materials

Table 2Materials assessed and experimental data.a,b

Test material information In vitro results In vivo classification

SCT SIT

Name Substance /use class Physical state pHb Viability 3 min [%] Viability1 h [%]

Result Viability1 h [%]

Result UNGHS

EU-CLP EPA Brazil

1-Aminotetralin Unspecified Liquid 8.5 56 11 C Cat 1 Cat 1 Cat I Class IAlkoxylated aliphatic alcohol 08/0313 Alkoxylated aliphatic alcohol liquid 5 106 106 NC 95 NI Cat 1 Cat 1 Cat I Class IIonic liquid 09/0478 ionic liquid liquid 5 92 14 C 6 I Cat 1 Cat 1 Cat I Class IIonic liquid 10/0031 ionic liquid liquid 6 83 9 C 10 I Cat 1 Cat 1 Cat I Class IPharma 02/0151 pharma liquid nd 43 12 C Cat 1 Cat 1 Cat I Class IN,N-Dimethylisopropylamine amine liquid 5 30 7 C Cat 1 Cat 1 Cat I Class IRhenium(VII)-oxid unspecified solid ca 0 (10%

aqueouspreparation)

34 12 C Cat 1 Cat 1 Cat I Class I

Amine 03/0085 amine liquid 8 32 17 C Cat 1 Cat 1 Cat I Class IAcrylate 04/0651 acrylate liquid 5 88 7 C Cat 2 Cat 2 Cat IV Class IIIAcrylate 05/0006 acrylate liquid 5.5 91 20 NC 8 I Cat 2 Cat 2 Cat IV Class IIIAgrochemical formulation 02/0550 agr. form. liquid 6 103 113 NC 86 NI Cat 2 Cat 2 Cat III Class IIIAgrochemical formulation 03/0020 agr. form. liquid 7 101 111 NC 7 I Cat 2 Cat 2 Cat III Class IIIAgrochemical formulation 04/0283 agr. form. liquid 7 98 99 NC Cat 2 Cat 2 Cat III Class IIIAgrochemical formulation 05/0278 agr. form. liquid 7 105 125 NC Cat 2 Cat 2 Cat III Class IIIAgrochemical formulation 05/0293 agr. form. liquid nd 90 91 NC 13 I Cat 2 Cat 2 Cat III Class IIIAgrochemical formulation 05/0299 agr. form. liquid 7 101 14 C Cat 2 Cat 2 Cat II Class IIIAgrochemical formulation 05/0384 agr. form. liquid 8 87 106 NC Cat 2 Cat 2 Cat IV Class IVAgrochemical formulation 05/0911 agr. form. liquid 5 100 102 NC 72 NI Cat 2 Cat 2 Cat II Class IIAgrochemical formulation 06/0419 agr. form. liquid 6 112 124 NC 88 NI Cat 2 Cat 2 Cat II Class IIBorane-tetrahydrofuran complex in tetrahydrofuran 1M boron compound liquid 5 20 8 C Cat 2 Cat 2 Cat II Class IICatecholborane boron compound liquid 6 7 -2 C Cat 2 Cat 2 Cat II Class IICyclohexanone unspecified liquid nd 85 7 C 7 I Cat 2 Not cl Cat III Class IIIEmollient 06/0177 emollient liquid 5 99 103 NC 87 NI Cat 2 Cat 2 Cat III Class IIIEmulsifier 11/0445 emulsifier solid 6 103 107 NC 117 NI Cat 2 Cat 2 Cat III Class IIIFatty alcohol 03/0197 unspecified liquid 5-6 111 111 NC Cat 2 Cat 2 Cat III Class IIIIonic liquid 04/0204 ionic liquid solid (at RT)

viscous (heatedto 80 �C)

6 92 50 NC Cat 2 Cat 2 Cat III Class III

Isocyanate 02/0045 isocyanate liquid 5 99 108 NC Cat 2 Cat 2 Cat III Class IIIN-ButylethanolAmine amino alcohol liquid 6 56 18 NC Cat 2 Cat 2 Cat III Class IIIPolymer 03/0049 polymer solid 8 (10% aqueous

preparation)100 107 NC Cat 2 Cat 2 Cat III Class III

Polymer 03/0421 polymer solid nd SCT 108 NC 112 NI Cat 2 Cat 2 Cat IV Class IVSurfactant 05/0511 surfactant liquid 6 92 100 NC 9 I Cat 2 Cat 2 Cat II Class IISurfactant 11/0560 surfactant liquid 6 82 95 NC 6 I Cat 2 Cat 2 Cat II Class IIAgrochemical formulation 02/0325 agr. form. liquid 6-8 105 103 NC Cat 3 Not cl Cat IV Class IVAgrochemical formulation 02/0638 agr. form. liquid 4.5 103 11 C Cat 3 Not cl Cat IV Class IVAgrochemical formulation 03/0026 agr. form. liquid 4 110 20 NC Cat 3 Not cl Cat IV Class IIIAgrochemical formulation 03/0180 agr. form. liquid 6-8 91 6 C Cat 3 Not cl Cat III Class IIIAgrochemical formulation 04/0349 agr. form. liquid 4 103 17 NC 9,6 I Cat 3 Not cl Cat IV Class IIIAgrochemical formulation 05/0595 agr. form. liquid 7 113 42 NC Cat 3 Not cl Cat III Class IIIAgrochemical formulation 10/0080 agr. form. liquid 5.5 96 95 NC 5,3 I Cat 3 Not cl Cat IV Class IIIEmollient 11/0434 emollient liquid 5 104 105 NC 97 NI Cat 3 Not cl Cat IV Class IVMulticomponent mixture 04/0025 multicomponent mixture liquid 5.5 79 -1 C Cat 3 Not cl Cat III Class IIISurfactant 09/0702 surfactant liquid 12 88 94 NC 15 I Cat 3 Not cl Cat III Class IIISurfactant 11/0209 surfactant liquid 11 101 106 NC 12 I Cat 3 Not cl Cat IV Class IVSurfactant 11/0559 surfactant liquid 7 96 81 NC 53d NI Cat 3 Not cl Cat IV Class IVTrimethoxyboroxine boron compound liquid 5 89 46 NC Cat 3 Not cl Cat IV Class IIITrimethylboroxine in tetrahydrofuran 50% solution boron compound liquid 5 92 28 NC Cat 3 Not cl Cat IV Class IV

(continued on next page)

S.N.K

olleet

al./Regulatory

Toxicologyand

Pharmacology

64(2012)

402–414

405

Table 2 (continued)

Test material information In vitro results In vivo classification

SCT SIT

Name Substance /use class Physical state pHb Viability 3 min [%] Viability1 h [%]

Result Viability1 h [%]

Result UNGHS

EU-CLP EPA Brazil

(R)-2-Chloromandelic acid unspecified solid nd 94 9 C Not cl Not cl Cat IV Class IV1-Vinylimidazol unspecified liquid 7 56 12 C Not cl Not cl Cat IV Class IV2-Cyano-2-methyl–propanoic acid-methyl ester unspecified liquid 5 99 94 NC Not cl Not cl Cat IV Class IV4-(4-Acryloyloxy-butoxycarbonyloxy)-benzoic acid unspecified solid 4 (10% aqueous

preparation)86 109 NC Not cl Not cl Cat IV Class IV

4-Amino-2-chlor-6,7-dimethoxychinazolin unspecified solid nd 105 96 NC Not cl Not cl Cat IV Class IV4-tert-Butyl-3-hydroxy-2,6-xylylacetonitrile unspecified solid 4.5 107 112 NC Not cl Not cl Cat IV Class IV4-Vinyloxybenzophenone polymer solid (at RT)

liquid (aftermelting at about60 �C)

4 94 113 NC Not cl Not cl Cat IV Class III

Acrylate 02/0178 acrylate liquid 8 106 106 NC Not cl Not cl Cat IV Class IVAcrylate 03/0410 acrylate liquid 6 76 72 NC Not cl Not cl Cat IV Class IVAcrylate 03/0423 acrylate viscous 3 94 77 NC Not cl Not cl Cat IV Class IVAcrylate 04/0114 acrylate liquid 5 95 75 NC 49.5c I Not cl Not cl Cat IV Class IVAcrylate 08/0289 acrylate liquid 7 99 113 NC Not cl Not cl Cat IV Class IVAcrylate 08/0322 acrylate liquid 4 97 93 NC Not cl Not cl Cat IV Class IVAgrochemical 02/0134 agrochemical liquid nd 103 98 NC Not cl Not cl Cat IV Class IVAgrochemical 04/0046 agrochemical liquid 4 100 90 NC Not cl Not cl Cat IV Class IVAgrochemical 05/0681 agrochemical solid nd 94 104 NC Not cl Not cl Cat IV Class IVAgrochemical 05/0682 agrochemical solid nd 98 87 NC Not cl Not cl Cat IV Class IVAgrochemical 05/0683 agrochemical solid nd 91 85 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 01/0414 agr. form. solid 4 (10% aqueous


Agrochemical formulation 02/0163 agr. form. liquid nd 104 101 NC 90 NI Not cl Not cl Cat IV Class IVAgrochemical formulation 02/0244 agr. form. liquid 6-7 98 75 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 02/0419 agr. form. liquid 6-8 95 108 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 02/0552 agr. form. liquid 6-7 83 90 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 02/0582 agr. form. liquid 5 108 100 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 02/0653 agr. form. liquid 8 76 10 C Not cl Not cl Cat IV Class IVAgrochemical formulation 03/0069 agr. form. liquid 5 101 93 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 03/0076 agr. form. solid 6-7 (10%

aqueouspreparation)

104 102 NC Not cl Not cl Cat IV Class IV

Agrochemical formulation 03/0298 agr. form. solid 12 (10% aqueouspreparation)


Agrochemical formulation 03/0347 agr. form. liquid nd 90 47 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 03/0443 agr. form. solid nd 81 76 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 03/0500 agr. form. solid 4 (10% aqueous





102 94 NC 102 NI Not cl Not cl Cat IV Class IV

Agrochemical formulation 04/0100 agr. form. liquid 5 83 72 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 04/0110 agr. form. liquid 6.16 94 78 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 04/0224 agr. form. liquid 7.4 95 99 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 04/0724 agr. form. liquid 6 94 85 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 05/0041 agr. form. liquid 7 95 87 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 05/0295 agr. form. liquid 6 139 118 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 05/0302 agr. form. liquid 6 87 102 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 06/0165 agr. form. liquid 7 105 101 NC 103 NI Not cl Not cl Cat IV Class IV

406S.N

.Kolle

etal./R

egulatoryToxicology

andPharm

acology64

(2012)402–

414

Agrochemical formulation 06/0166 agr. form. liquid 8 106 102 NC 103 NI Not cl Not cl Cat IV Class IVAgrochemical formulation 06/0352 agr. form. solid 8 104 93 NC Not cl Not cl Cat IV Class IVAgrochemical formulation 07/0075 agr. form. liquid 5 111 102 NC 10 I Not cl Not cl Cat IV Class IIIAgrochemical formulation 07/0081 agr. form. solid 4 119 117 NC 91 NI Not cl Not cl Cat IV Class IVAgrochemical formulation 07/0487 agr. form. liquid 5 93 27 NC 17 I Not cl Not cl Cat IV Class IVAgrochemical formulation 08/0124 agr. form. liquid 5 93 104 NC 76 NI Not cl Not cl Cat IV Class IVAgrochemical formulation 11/0588 agr. form. liquid 7 104 111 NC 107 NI Not cl Not cl Cat IV Class IVAmine 04/0280 amine solid about 7 (10%

aqueouspreparation)


Bis(pinacolato)diboron boron compound solid 5 (10% aqueouspreparation


Bisoxethylanilin unspecified liquid 5 110 116 NC Not cl Not cl Cat IV Class IVCatalyst 03/0195 catalyst solid 9.6 (10%

aqueouspreparation)


Chromium trichloride unspecified solid 4.5 112 107 NC 101 NI Not cl Not cl Cat IV Class IVCinchocaineHCl pharma solid 5 85 23 NC Not cl Not cl Cat IV Class IVCs-Dodecahydrododecaborate boron compound solid 8 (10% aqueous


DKDS-Reinkristallisat intermediate solid 4.5 (10%aqueouspreparation)


Emollient 11/0444 emollient liquid 5 91 105 NC 109 NI Not cl Not cl Cat IV Class IVEther 05/0701 ether liquid 5 94 50 NC Not cl Not cl Cat IV Class IVFatty acid/amine condensate 06/0549 fatty acid/amine condensate solid (waxy) 5 105 119 NC Not cl Not cl Cat IV Class IVFluoxetine HCl pharma solid nd 89 1 C Not cl Not cl Cat IV Class IVHydrated Bisphenol-A bisglycidylether ether liquid, viscous 5 100 107 NC Not cl Not cl Cat IV Class IVIonic liquid 05/0233 ionic liquid liquid 5 86 25 NC Not cl Not cl Cat IV Class IVlubricant 11/0567 lubricant liquid 5 92 100 NC 112 NI Not cl Not cl Cat IV Class IVMulticomponent mixture 04/0233 multicomponent mixture liquid 4 46 9 C Not cl Not cl Cat IV Class IVN-(2-Hydroxyethyl)-oxazolidin-2-one unspecified liquid (slight

viscous)6 97 79 NC Not cl Not cl Cat IV Class IV

Pearlizer 11/0562 pearlizer solid 4 99 102 NC 108 NI Not cl Not cl Cat IV Class IVPharma compound pharma solid 5 105 119 NC 105 NI Not cl Not cl Cat IV Class IVPharma compound 03/0371 pharma solid 6 (10% aqueous


Phenyl Acetate unspecified liquid 5 106 11 C Not cl Not cl Cat IV Class IVPigment/dye 00/0180 pigment/dye solid nd 101 72 NC Not cl Not cl Cat IV Class IVPigment/dye 02/0576 pigment/dye solid nd 94 101 NC Not cl Not cl Cat IV Class IVPigment/dye 02/0577 pigment/dye solid 5 (10% aqueous


Pigment/dye 03/0013 pigment/dye solid 4 (10% aqueouspreparation)


Pigment/dye 03/0240 pigment/dye solid nd 101 106 NC Not cl Not cl Cat IV Class IVPigment/dye 03/0450 pigment/dye solid 5 (10% aqueous


Pigment/dye 04/0399 pigment/dye solid 5.2 (10%aqueouspreparation)


Pigment/dye 07/0285 pigment/dye solid nd 92 119 NC 100 NI Not cl Not cl Cat IV Class IVPolymer 02/0335 polymer liquid 8 92 93 NC Not cl Not cl Cat IV Class IVPolymer 02/0358 polymer liquid 4 108 104 NC Not cl Not cl Cat IV Class IVPolymer 05/0655 polymer solid 4 (10% aqueous


Polymer 05/0927 polymer liquid 5 96 93 NC Not cl Not cl Cat IV Class IVPolyol 03/0183 polyol solid (at RT)

liquid5 (10% aqueouspreparation)


Postassium-N-methylalaninate in water solvent liquid 12.5 99 60 NC Not cl Not cl Cat IV Class IV

(continued on next page)

S.N.K

olleet

al./Regulatory

Toxicologyand

Pharmacology

64(2012)

402–414

407

Tabl

e2

(con

tinu

ed)

Test

mat

eria

lin

form

atio

nIn

vitr

ore

sult

sIn

vivo

clas

sifi

cati

on

SCT

SIT

Nam

eSu

bsta

nce

/use

clas

sPh

ysic

alst

ate

pHb

Via

bili

ty3

min

[%]

Via

bili

ty1

h[%

]R

esu

ltV

iabi

lity

1h

[%]

Res

ult

UN

GH

SEU

-CLP

EPA

Bra

zil

Pyri

dazi

ne

un

spec

ified

liqu

id5

(un

dilu

ted)

4(1

0%aq

ueo

us

prep

arat

ion

)

8532

NC

Not

clN

otcl

Cat

IVC

lass

IV

Silo

xan

es/

sili

con

es08

/016

8si

loxa

nes

/si

lico

nes

liqu

id5

113

104

NC

Not

clN

otcl

Cat

IVC

lass

IVSy

ner

gist

10/0

463

syn

ergi

stso

lid

7.61

(1%

susp

ensi

on)

100

106

NC

85,9

NI

Not

clN

otcl

Cat

IVC

lass

IV

Tran

s-1,

2-D

ith

ian

e-4,

5-D

iol,

98%

diol

soli

d5

(10%

aqu

eou

spr

epar

atio

n)

9395

NC

Not

clN

otcl

Cat

IVC

lass

IV

Vit

amin

02/0

457

vita

min

eso

lid

nd

8281

NC

Not

clN

otcl

Cat

IVC

lass

IV

aN

C,n

otco

rros

ive;

C,c

orro

sive

;C

at,c

ateg

ory;

NI,

not

irri

tan

t;n

otcl

,not

clas

sifi

ed;

I,ir

rita

nt;

SCT,

skin

corr

osio

nte

st;

SIT,

skin

irri

tati

onte

st;

nd,

Not

dete

rmin

ed(a

gr.f

orm

.,ag

roch

emic

alfo

rmu

lati

on).

bO

ECD

Gu

idel

ine

for

the

test

ing

ofch

emic

als.

Dra

ftPr

opso

sal

for

anU

pdat

eof

the

test

guid

elin

e43

1.In

Vit

roSk

inC

orro

sion

:R

econ

stru

cted

Hu

man

Epid

erm

is(R

hE)

Test

Met

hod

pHu

ndi

lute

dor

moi

sten

edw

ith

wat

er.

cC

alcu

late

dm

ean

valu

eof

two

inde

pen

den

tru

ns

(48%

and

51%

).d

Cal

cula

ted

mea

nva

lue

ofth

ree

inde

pen

den

tru

ns

(48%

,55%

,an

d56

%).


(Table 4a), agrochemical formulations alone (Table 4b, subset ofTable 4a) and materials except agrochemical formulations(Table 4c, subset of Table 4a), are shown according to the EU-CLPclassification system. The resulting predictivity calculations forall classification systems are depicted in Tables 5a (all materials),5b (agrochemical formulations alone) and 5c (all materials exceptagrochemical formulations).

The accuracy and specificity of the SCT predictions compared toin vivo classifications for all materials, including agrochemical for-mulations, was above 89% across all classification systems. Thedataset included only eight materials (no agrochemical formula-tions) classified as ‘‘corrosive’’ (UN GHS/EU-CLP Category 1/USEPA Category I/ANVISA Toxicity Class I) by the in vivo test. Dueto this small number of UN GHS Category 1 materials, an assess-ment of sensitivity of the SCT for corrosive materials was not fea-sible. Of the eight corrosive materials, three were also tested in theSIT. Two yielded correct positive results in the SIT whereas thethird gave false negative results in the SIT. Of 126 in vivo noncorro-sive materials tested, 14 materials (including four agrochemicalformulations) were false positive in the SCT. These 18, were equallydistributed between in vivo irritants (UN GHS Categories 2 or 3/EU-CLP Category 2/US EPA Category II or III/Brazilian Toxicity Class IIor III) and nonirritants in vivo (UN GHS/EU CLP Category no classi-fication/US EPA Category IV/ANVISA Toxicity Class IV) (Tables 4aand 5a). Of 21 (of all 134 materials tested) materials positive inthe SCT, three were also tested in the SIT. All three yielded positiveresults and two of these were identified as being a skin irritantin vivo in at least three classification schemes. According to UNGHS, the number of overpredictions in the in vitro skin corrosiontest for agrochemical formulations was also low (4/46) (Tables4b and 5b). Three of these four exhibited some irritant potentialin vivo (two were classified as Category 3 irritants and one wasclassified as a Category 2 irritant in vivo according to the UNGHS); one was ‘‘not classified’’ according to UN GHS classificationin vivo. The number of corrosive materials in the data set was verylimited. One material tested false negative in both the SCT and SIT(i.e. one material classified as being corrosive in the in vivo test butnegative in the SCT and SIT, Tables 4a, 4c, and 5a, 5c).

For this data set of 134 test materials, pH or physical state (so-lid, liquid, or viscous; details are listed in Table 2) did not appear toaffect whether the material was concordantly classified in the SCT.For example, while no viscous material was overpredicted. Twosolids were overpredicted (2/42 in SCT and 0/7 in SIT), but manywere concordantly (45/47, overall concordance 96%) predicted.For solids, 100% sensitivity (although only one corrosive solidwas in the data set) and 96% specificity were obtained for theSCT. In the SIT, 7/7 nonirritant solids were correctly predictedwhile 2/2 UN GHS Category 2 solids were underpredicted to benonirritants. The false positive rates for liquids in both the SCTand SIT were higher than for solids (15% and 30% vs. 4% and 0%)and the false negative rate was high in the SIT (37%) resulting ina good overall concordance for liquids in the SCT (85%) and a weakoverall concordance for liquids in the SIT (66%).

Concerning borderline cases (cut-off values ± 0.1� cut-off) inthe SCT, one material (in vivo noncorrosive) resulted in a borderlineviability after the 3-min exposure in the SCT (i.e. 45% < relative via-bility < 55%), while the viability after the 1 h-exposure was clearlybelow the cut-off. Of two other materials (i.e. 2% of all materialstested in the SCT) resulting in borderline viability values after the1-h exposure (12% < relative viability < 17%), both were clearlynegative (relative viability > 55%) after the 1-h exposure. Of thosematerials with relative viability values between 12 and 15% (ratedpositive for all remaining evaluations), one in vivo UN GHS Cate-gory 1 material was correctly and one in vivo noncorrosive wasoverpredicted according to the 3-min exposure. In the SIT, 2/38materials resulted in borderline relative viability cut-offs

Table 3aComparison of UN GHS classification to US EPA classification schemes for the assessed dataset.a

a Not cl, not classified.

Table 3bComparison of UN GHS classification to ANVISA classification schemes for the assessed dataset.a

a Not cl, not classified.


(45% < relative viability < 55%). The borderline results were repro-duced in repetitions of the individual experiments and mean val-ues of all experiments performed were considered for allremaining evaluations.

Of the compound classes represented by five or more materialsin the SCT, all polymers (7/7), surfactants (5/5), pigments/dyes (8/8), and agrochemicals (5/5) were concordantly predicted. In theSIT, most surfactants were concordantly predicted, depending onthe classification system. All emollients were negative in vitro,but two were classified as in vivo irritants according to UN GHS(one was in vivo UN GHS Category 3 and one was in vivo UN GHSCategory 2); one of these materials is not considered irritantsaccording to the US EPA or ANVISA classification systems. For all

34 diverse materials (‘‘others’’), the SIT showed good specificityand fair sensitivity depending on the classification scheme(Table 5c). For this data set the SIT was most concordant withthe ANVISA and US EPA systems.

Amongst the 134 materials evaluated in this study, 46 wereagrochemical formulations and were tested in the SCT; 16 agro-chemical formulations were also tested in the SIT. Hence, a sepa-rate analysis was made for this particular substance class. Foragrochemical formulations, the SIT showed good specificity andfair sensitivity (Table 5b) depending on the classification scheme,with comparable performance compared to the entire SIT dataset. Noncorrosive agrochemical formulations were concordantlypredicted in the UN GHS system as irritants (UN GHS Categories

Table 4aAll test materials.a

a NC, not corrosive; C, corrosive; NI, not irritant; Not cl, not classified; I, irritant.

Table 4bAgrochemical formulations alone.a



2 and 3; 4/7) or nonirritants (no UN GHS Category, 7/9) (Table 4b).The data set did not contain any corrosive agrochemicalformulations.

If the data set is analyzed without including the agrochemicalformulations (Tables 4c and 5c), sensitivity, specificity and overallconcordance were lower, according to UN GHS classification, com-pared to the agrochemical formulations alone or to the completedata set. The predictive capacity (with the exception of the positivepredictive value for UN GHS and EU-CLP classification systems) ofthe SIT according to UN GHS, US EPA and ANVISA classification sys-tems was superior to that of the complete data set or the datasetwithout agrochemical formulations.

4. Discussion

In this study we provide data to contribute to the overall perfor-mance assessment of in vitro RhE methods for skin corrosion andirritation. During routine testing, 134 materials were testedin vitro for skin corrosion (SCT); 38 of these materials were also

tested in the in vitro skin irritation test (SIT). Any in vivo data usedfor this comparative study was obtained as a regulatory require-ment and not for the purposes of this study. The materials includeda diverse set of substance classes and physical forms, including 46agrochemical formulations. The data set includes several ‘‘chal-lenging’’ types of substances such as polymers, viscous fluids andagrochemical formulations. We find excellent concordance withthe classifications made using the in vivo tests for a wide rangeof chemical classes and physical states in the SCT, including foragrochemical formulations, solids, polymers, surfactants, and pig-ments. Of the 126 noncorrosive materials 14 were false positive;none of these have extreme pH. One of eight corrosive materialswas underpredicted in the SCT. This alkoxylated aliphatic alcoholwas however also incorrectly classified as negative in the SIT in thisstudy, and in the Bovine Corneal Opacity and Permeability assay(BCOP) (Schrage et al., 2011), the hen’s egg chorioallantoic mem-brane test (HET-CAM; unpublished BASF data), and the EpiOcularassay (Kolle et al., 2011). Whether this material is corrosive in hu-mans is not known. Based on data obtained in human irritationpatch tests, three materials underpredicted compared to the

Table 4cWithout agrochemical formulations.a


Table 5aPerformance results for all test materials.a

Test SCT SIT

Classificationsystem

Allclassificationsystems

UN GHS EU-CLP US EPA ANVISA

Prediction ModelDescription

No UN GHS Category vs. UNGHS Categories 3, 2, 1

No UN GHS Category, Category 3 vs.UN GHS Categories 2, 1

EPA Category IV vs.Categories III, II, I

ANVISA Toxicity Class IVvs. Classes III, II, I

a SIT, skin irritation test; SCT, skin corrosion test; PPV, positive predictive value; NPV, negative predictive value; FP, false positive; FN, false negative.

Table 5bPerformance results for agrochemical formulations alone.a

Test SCT SIT










Table 5cPerformance results without agrochemical formulations.a

Test SCT SIT











animal tests (the three emollients) have low skin irritation poten-tial in humans (unpublished BASF data). Overprediction of irrita-tion by animal tests compared to human tests has been reportedin a number of papers (e.g. Robinson et al., 2001; Basketter et al.,2004). In a more recent report by Jirova et al. (2010), studies con-

firmed that the animal tests tend to be overpredictive of skin irri-tation potentials in humans and that the RhEs had a betterconcordance to human data. This emphasizes the necessity to val-idate in vitro methods to reflect human and not animal hazardpotentials.


While the specificity for the SIT for most classification systemsis acceptable (70–85%) the sensitivity (53–67%) and thus overallconcordance (63–76%) is less robust (Tables 5). The sensitivityand specificity are comparable for agrochemical formulations(40–63% and 64–88%) when considered alone. This is in contrastto findings with 20 agrochemical formulations (cited in Eskeset al., 2012), which found more overpredictions than underpredic-tions. As formulations are multicomponent systems with differingingredients and actives, the differences in predictivity may beattributable to differences in the composition of the formulations.

In December 2008, ESAC stated the validity of RhE model testsfor skin irritation and the OECD guideline was adopted in 2010.One reason for the unbalanced data set with respect to corrosivesin this study is the sequential nature of the international accep-tance of the two in vitro methods. In vivo corrosives are underrep-resented because of the earlier acceptance of OECD TG 431 (SCT)compared to TG 439 (SIT). Even prior to the regulatory acceptanceof the SIT, a positive result in the SCT did not require further animaltesting. In contrast, if the SCT was negative, further testing in vivowas mandatory to exclude a skin irritation potential. Furthermore,when compiling this data for this analysis, only a few SIT data gapscould be filled at a later date due to test material expiration.

Upon availability of regulatory agency-accepted in vitro skincorrosion and irritation tests, and with the adoption of the OECDguidelines, a testing strategy deemed to be most efficient basedon the expected irritant potency of the test material was adopted(Fig. 2) in accordance with the strategy described by CosmeticsEurope (formerly COLIPA, Macfarlane et al., 2009). In this respectif a material was expected to be corrosive, the in vitro skin corro-sion test was performed first. Likewise if the material was expectedto be a nonirritant, the in vitro skin irritation test was performedfirst. If no prediction based on e.g. expert judgment, structureactivity relationship and/or pH could be made, both tests were per-formed in parallel (combination test). This is another reason for thesmall number of SCT positives tested in the SIT or in vivo, as furthertesting in vivo was avoided if a material was evaluated as beingcorrosive in the SCT.

The UN GHS and EU-CLP regulations also contain subclassifica-tions within the corrosive classification of 1A, 1B, and 1C, whichthen define packaging and transportation requirements. The OECDis currently revising TG 431 to specify that the RhE methods Epi-Derm SCT (used in this study), EpiSkin™ Standard Model andSkinEthic™ RHE can discriminate between 1A and 1B/1C (OECD,2012). Unfortunately there are too few corrosive materials evalu-

Fig. 2. In vitro testing strategy. Depending on the known characteristics

ated in this study to determine the performance of the SCT forsub-classification.

4.1. Classification considerations

As discussed above, differing regional classification system rulescan influence the classification of the same materials; using differ-ing classification schemes as ‘‘gold standards’’ in the validation ofin vitro tests can affect the performance of in vitro methods. For thisdata set, there were no differences between classification systemsfor the corrosive (UN GHS Category 1/EU-CLP Category 1/US EPACategory I/ANVISA Toxicity Class I) materials, so only unified datais presented. Although there were differences in the performanceof the SIT between classification systems, most of the time the pre-dictivity of the in vitro tests differed by a few percentage points.While some materials were underpredicted in only one system, afew others were underpredicted in only one other system. Forexample, surfactant 11/0559 was underpredicted as nonirritantsin vitro only according to UN GHS. Incidentally, this material wasUN GHS Category 3, which under some implementations of theEU-CLP would be considered not classified; furthermore the surfac-tant was not irritant in humans (unpublished BASF data).

The SIT performance was the least robust when comparing tothe in vivo data classified according to the EU-CLP classificationsystem, for all materials and for agrochemical formulations (Tables5). Better performance was obtained according to the ANVISA sys-tem for all materials, and according to the US EPA system for agro-chemical formulations. This may be because the cut-off valuebetween ‘‘irritant’’ and ‘‘no label’’ in the previous EU classificationsystem, the DSD, was 2.0, as it is still for the US EPA and ANVISAsystems. The cut-off for the EU-CLP systems is now 2.3 (mild irri-tants are not subject to classification), and is 1.5 for UN GHS (be-cause here ‘‘mild irritant’’ UN GHS Category 3 was included(Table 1)). So, very mild irritants (UN GHS Category 3/US EPA Cat-egory III, ANVISA Toxicity Class III) are picked up by the SIT andcorrectly identified, but because the cut-off value for the EU-CLPhas moved up, these are identified as false positives, with a result-ing higher false positive rate than the other classification systems(Tables 5). This illustrates the difficulties imposed on test methoddevelopers and laboratories faced with the need to maintain thevalidation status of in vitro methods, which are subject to very highscrutiny, in the face of diverse and ever-changing classificationrequirements.

of a test material, a testing strategy can help guide in vitro testing.


The SIT provides a dichotomous classification (irritant or nonir-ritant). In order to make the comparisons here for four-categorysystems, we considered materials in both irritant categories (UNGHS Category 2 or 3/US EPA Category II or III/ANVISA Toxicity ClassII or III) as ‘‘irritant.’’ Officially collapsing these middle categorieswould allow the direct application of in vitro methods to local skintoxicity assessment in regions that have not yet adopted UN GHS.Indeed, the text of the UN GHS explicitly acknowledges that inter-animal variability supports a single irritant category (UN, 2007;Chapter 3.2.2.5.1, sub point c).

Contingency tables (Tables 4) indicate correct and incorrectpredictions of the SCT and SIT, for all materials (Table 4a), agro-chemical formulations (Table 4b), and all materials withoutagrochemical formulations (Table 4c), under UN GHS. A similarexercise was performed for the US EPA and ANVISA systems andthe tables were found to have minimal differences. The absenceof drastic differences between systems, despite different in vivoclassifications for some materials, indicates the robust natureof the in vitro methods to predict skin irritation as an apicalendpoint.

5. Conclusion and outlook

The in vitro SCT and SIT methods can be easily integrated into adiverse test material assessment program and lead to major reduc-tions in the number of rabbits used for skin irritation testing. RhEmodels in general are commercially available in a number of differ-ent regions (Deshmukh et al., 2012; Kojima et al., 2012; Kandarovaet al., 2009a,b; Spielmann et al., 2007; Hoffmann et al., 2005), andeasily handled and used according to the testing protocol. For re-gions where RhE models are not (yet) commercially available ormay not be imported, ‘‘open-source’’ tissue projects are underwayto allow the in house creation of RhE models according to com-monly agreed production standards (Poumay et al., 2004; Poumayand Coquette, 2007). By air–liquid interface culture, RhE models al-low easier material application than cell culture models in aqueousmedia, or rabbits, depending on the physical state of the material.Furthermore RhE models closely mimic the biochemical and phys-iological properties of the human epidermis, and directly assess themechanism of action (cell and tissue damage resulting in localizedtrauma) that occurs during irritation in vivo (OECD, 2010). It is alsoclear that the Draize rabbit test is highly variable (Griesinger et al.,2010) and overpredictive (Jirova et al., 2010; Robinson et al., 2001;Basketter et al., 2004) of human irritation response in vivo, andclassification system rules were written based on in vivo rabbit re-sponses; nevertheless the current situation necessitates compara-ble sensitivity and specificity be obtained with the in vitro tests.

For this data set of 134 test materials, we obtained low sensitiv-ity in the SIT (53–67%), despite following a protocol designed toimprove this performance measure (Kandarova et al., 2009a). Fur-ther, while a sufficient performance assessment of the SCT sensi-tivity was not possible for agrochemical formulations alone orthe complete data set, specificity and overall concordance of bothSCT and SIT were comparable the values for the complete dataset (SCT: 91% vs. 89% specificity, 91% vs. 89% accuracy and SIT:64–88% vs. 70–85% specificity, 56–75% vs. 53–76% accuracy).Therefore, while generally some improvements of the RhE-basedassays for the different classification systems may be feasible, wehave shown that RhE-based methods seem to be applicable to as-sess the irritant potential of agrochemical formulations. In severalregions, the in vitro SIT is not accepted for regulatory use for agro-chemicals and agrochemical formulations. We encourage otherindustry laboratories to share data with the wider scientific com-munity in order to facilitate replacement of the rabbit test for thislarge sector of substances.

Conflict of interest statement

The authors declare no conflict of interest.

Acknowledgments

We would like to thank the Laboratory for Applied AlternativeMethods at BASF SE for their excellent technical assistance andall sponsors for the authorization to publish their data.

References

AENDA, 1992. Portaria N� 03/MS/SNVS, De 16 De Janeiro de 1992.Basketter, D.A., York, M., McFadden, J.P., Robinson, M.K., 2004. Determination of

skin irritation potential in the human 4-h patch test. Contact Dermatitis 51, 1–4.

Deshmukh, G.R., Kumar, K.H., Reddy, P.V., Rao, B.S., 2012. In vitro skin corrosion:human skin model test – a validation study. Toxicol. In Vitro 6, 1072–1074.

Draize, J.H., Woodard, G., Calvery, H.O., 1944. Methods for the study of irritation andtoxicity of substances applied topically to the skin an mucous membranes. J.Pharmacol. Exp. Ther. 82, 377–390.

EPA, 2007. Label Review Manual Chapter 7.Eskes, C., Detappe, V., Koeter, H., Kreysa, J., Liebsch, M., Zuang, V., Amcoff, P.,

Barroso, J., Cotovio, J., Guest, R., Hermann, M., Hoffmann, S., Masson, P., Alepee,N., Arce, L.A., Bruschweiler, B., Catone, T., Cihak, R., Clouzeau, J., D’Abrosca, F.,Delveaux, C., Derouette, J.P., Engelking, O., Facchini, D., Frohlicher, M., Hofmann,M., Hopf, N., Molinari, J., Oberli, A., Ott, M., Peter, R., Sa-Rocha, V.M., Schenk, D.,Tomicic, C., Vanparys, P., Verdon, B., Wallenhorst, T., Winkler, G.C., Depallens, O.,2012. Regulatory assessment of in vitro skin corrosion and irritation data withinthe European framework: Workshop recommendations. Regul. Toxicol.Pharmacol. 62, 393–403.

EU, 1967. Council Directive 67/548/EEC of 27 June 1967 on the approximation oflaws, regulations and administrative provisions relating to the classification,packaging and labelling of dangerous substances. Official J. Eur. Union 196.

EU, 1976. Council Directive 76/768/EEC of 27 July 1976 on the approximation of thelaws of the Member States relating to cosmetic products. Official J. Eur. UnionL262, 169–200.

EU, 2006. Commission Regulation (EU) No 1907/2006 of the European Parliamentand of the Council of 18 December 2006 concerning the Registration,Evaluation, Authorisation and Restriction of Chemicals (REACH), establishinga European Chemicals Agency, amending Directive 1999/45/EC and repealingCouncil Regulation (EEC) No. 793/93 and Commission Regulation (EC) No. 1488/94 as well as Council Directive 76/769/EEC and Commission Directives 91/155/EEC, 93/67/EEC, 93/105/EC and 2000/21/EC. Official J. Eur. Union 49, 1–849.

EU, 2008. Regulation (EC) No. 1272/2008 of the European Parliament and of theCouncil of 16 December 2008 on classification, labelling and packaging ofsubstances and mixtures, amending and repealing Directives 67/548/EEC and1999/45/EC, and amending Regulation (EC) No. 1907/2006. Official J. Eur. UnionL 353, 1–1355.

EU, 2009. Regulation (EC) No. 1223/2009 of the European Parliament and of theCouncil of 30 November 2009 on cosmetic products. Official J. Eur. Union L 342,59–209.

Griesinger, C., Barroso, J., Zuang, V., et al., 2010. Explanatory Background Documentto the OECD Draft Test Guideline on in vitro Skin Irritation Testing. Available at:http://www.oecd.org/chemicalsafety/testingofchemicals/43670220.pdf (lastaccessed 07 Sep 2012).

Hoffmann, J., Heisler, E., Karpinski, S., Losse, J., Thomas, D., Siefken, W., Ahr, H.J.,Vohr, H.W., Fuchs, H.W., 2005. Epidermal-skin-test 1000 (EST-1000) – a newreconstructed epidermis for in vitro skin corrosivity testing. Toxicol. In Vitro 19,925–929.

Jirova, D., Basketter, D., Liebsch, M., Bendova, H., Kejlova, K., Marriott, M.,Kandarova, H., 2010. Comparison of human skin irritation patch test datawith in vitro skin irritation assays and animal data. Contact Dermatitis 62, 109–116.

Kandarova, H., Hayden, P., Klausner, M., Kubilus, J., Kearney, P., Sheasgreen, J.,2009a. In vitro skin irritation testing: improving the sensitivity of the EpiDermskin irritation test protocol. Altern. Lab. Anim. 37, 671–689.

Kandarova, H., Hayden, P., Klausner, M., Kubilus, J., Sheasgreen, J., 2009b. An in vitroskin irritation test (SIT) using the EpiDerm reconstructed human epidermal(RHE) model. J. Vis. Exp. 29, 1366.

Kojima, H., Ando, Y., Idehara, K., Katoh, M., Kosaka, T., Miyaoka, E., Shinoda, S.,Suzuki, T., Yamaguchi, Y., Yoshimura, I., Yuasa, A., Watanabe, Y., Omori, T., 2012.Validation study of the in vitro skin irritation test with the LabCyte EPI-MODEL24. Altern. Lab. Anim. 40, 33–50.

Kolle, S.N., Kandarova, H., Wareing, B., van, R.B., Landsiedel, R., 2011. In-housevalidation of the EpiOcular™ eye irritation test and its combination with thebovine corneal opacity and permeability test for the assessment of ocularirritation. Altern. Lab. Anim. 39, 365–387.

Macfarlane, M., Jones, P., Goebel, C., Dufour, E., Rowland, J., Araki, D., Costabel-Farkas, M., Hewitt, N.J., Hibatallah, J., Kirst, A., McNamee, P., Schellauf, F., Scheel,J., 2009. A tiered approach to the use of alternatives to animal testing for the


safety assessment of cosmetics: skin irritation. Regul. Toxicol. Pharmacol. 54,188–196.

OECD, 2002. OECD Guidelines for the Testing of Chemicals, Test No. 404: AcuteDermal Irritation/Corrosion.

OECD, 2004. OECD Guidelines for the Testing of Chemicals, Test No. 431: In VitroSkin Corrosion: Human Skin Model Test.

OECD, 2010. OECD Guidelines for the Testing of Chemicals, Test No. 439: In VitroSkin Irritation Reconstructed Human Epidermis Test Method.

OECD, 2012. OECD Guideline for the Testing of Chemicals. Draft Propsosal for anUpdate of the Test Guideline 431. In Vitro Skin Corrosion: ReconstructedHuman Epidermis (RhE) Test Method. Available at: http://www.oecd.org/chemicalsafety/testingofchemicals/50305099.pdf (last accessed 07 Sep 2012).

Poumay, Y., Coquette, A., 2007. Modelling the human epidermis in vitro: tools forbasic and applied research. Arch. Dermatol. Res. 298, 361–369.

Poumay, Y., Dupont, F., Marcoux, S., Leclercq-Smekens, M., Herin, M., Coquette, A.,2004. A simple reconstructed human epidermis: preparation of the culturemodel and utilization in in vitro studies. Arch. Dermatol. Res. 296, 203–211.

Robinson, M.K., McFadden, J.P., Basketter, D.A., 2001. Validity and ethics of thehuman 4-h patch test as an alternative method to assess acute skin irritationpotential. Contact Dermatitis 45, 1–12.

Schrage, A., Kolle, S.N., Moreno, M.C., Norman, K., Raabe, H., Curren, R., van, R.B.,Landsiedel, R., 2011. The bovine corneal opacity and permeability test in routineocular irritation testing and its improvement within the limits of OECD testguideline 437. Altern. Lab. Anim. 39, 37–53.

Spielmann, H., Hoffmann, S., Liebsch, M., Botham, P., Fentem, J.H., Eskes, C., Roguet,R., Cotovio, J., Cole, T., Worth, A., Heylings, J., Jones, P., Robles, C., Kandarova, H.,Gamer, A., Remmele, M., Curren, R., Raabe, H., Cockshott, A., Gerner, I., Zuang, V.,2007. The ECVAM international validation study on in vitro tests for acute skinirritation: report on the validity of the EPISKIN and EpiDerm assays and on theSkin Integrity Function Test. Altern. Lab. Anim. 35, 559–601.

UN, 2007. Globally Harmonized System of Classification and Labelling of Chemicals(GHS), second revised ed. United Nations.

UN, 2011. Globally Harmonized System of Classification and Labelling of Chemicals(GHS), fourth revised ed. United Nations.

kolle sullivan 2012

Documents