The ECVAM International Validation Study on In Vitro Tests for Skin Corrosivity. 1. Selection and Distribution of the Test Chemicals

Toxicology in Vitro 12 (1998) 471±482

The ECVAM International Validation Study on In Vitro Tests for Skin Corrosivity. 1. Selection and Distribution of the Test Chemicals M. D. BARRATT1, P. G. BRANTOM2, J. H. FENTEM3*, I. GERNER4, A. P. WALKER5 and A. P. WORTH3

1 SEAC Toxicology Centre (formerly the Environmental Safety Laboratory), Unilever Research, Colworth House, Sharnbrook, Bedford MK44 1LQ, UK, 2BIBRA International, Woodmansterne Road, Carshalton, Surrey SM5 4DS, UK, 3ECVAM, JRC Environment Institute, 21020 Ispra (VA), Italy, 4BgVV, Thielallee 88-92, 14195 Berlin, Germany and 5Apojay Consultancy, 6 Cragside, Whitley Bay, Tyne & Wear NE26 3DU, UK

(Accepted 15 February 1998) AbstractÐAn international validation study on in vitro tests for skin corrosivity was conducted during 1996 and 1997 under the auspices of the European Centre for the Validation of Alternative Methods (ECVAM). The main objectives of the study were to assess the performances of selected in vitro tests in discriminating between: (a) corrosives (C) and non-corrosives (NC), for selected groups of chemicals (e.g. organic acids, phenols) and/or for all chemicals (single chemical entities only); and (b) known R35 (UN packing group I) and R34 (UN packing groups II & III) chemicals. Each test was evaluated for reliability and relevance by using a test set of 60 coded chemicals. In this paper, the test chemicals used in the validation study are identi®ed; they include organic acids (6C/5NC), organic bases (7C/3NC), neutral organics (9NC), phenols (2C/3NC), inorganic acids (6C/1NC), inorganic bases (2C/2NC), inorganic salts (1C/2NC), electrophiles (3C/5NC) and soaps/surfactants (3NC). The in vivo classi®cations and important physicochemical properties (e.g. logP, pKa) of the test chemicals are given. The main criterion for including chemicals in the test set was that their corrosivity classi®cations were based on unequivocal animal data. Where available, structure±activity information was also used to support the corrosivity classi®cations. Despite the small numbers of chemicals in some of the categories, it was felt that the test set chosen represented the best possible for evaluating the performances of the in vitro tests for predicting skin corrosivity, given the limited availability of unequivocal animal data. The prediction of skin corrosivity from pH data was also investigated for those chemicals with extreme pH values (i.e. pHE 2 or e11.5). Nine of the 12 strongly acidic or alkaline chemicals in the test set, which were predicted to be C on the basis of their pH values, had also been found to be C in vivo. # 1998 Elsevier Science Ltd. All rights reserved Abbreviations: C = corrosive; CSSC = Chemicals Selection Sub-committee; DOT = US Department of Transportation; ECETOC = European Centre for Ecotoxicology and Toxicology of Chemicals; ECVAM = European Centre for the Validation of Alternative Methods; EEC = European Economic Community; Mpt = melting point; MT = Management Team; MV = molecular volume; NC = noncorrosive; OECD = Organisation for Economic Cooperation and Development; PC = principal component; PCA = principal components analysis; PII = primary irritation index; PM = prediction model; QSAR = quantitative structure±activity relationship; R34 = causes burns (EU risk phrase); R35 = causes severe burns (EU risk phrase); TER = transcutaneous electrical resistance; UN = United Nations; I, II & III = UN packing groups I, II & III. Keywords: alternative methods; skin corrosivity; structure±activity relationships; test chemicals; validation.

Introduction Skin corrosion is the production of irreversible tissue damage in the skin (OECD, 1992). The ability *Author for correspondence.

of chemicals to induce skin corrosion is an important consideration in establishing procedures for the safe handling, packing and transport of chemicals. For this reason, the determination of skin corrosion potential is included in international regulatory requirements for the testing of chemicals.

0887-2333/98/$19.00+0.00 # 1998 Elsevier Science Ltd. All rights reserved. Printed in Great Britain PII: S0887-2333(98)00021-1

472

M. D. Barratt et al.

The most commonly used approach for assessing skin corrosion is a modi®ed version of the Draize skin irritation test. In the original Draize test (Draize, 1965; Draize et. al., 1944), the test substance was applied for 24 hr, under an occluded patch, to the abraded and intact skin of rabbits. The response was recorded as the extent of erythema, oedema and eschar formation occurring during a period of 72 hr following application of the substance. It was not speci®ed, however, whether the reversibility of the response should be observed. A modi®cation of the Draize test, published in 1981 as OECD Testing Guideline 404, differed from the original test in four main respects: (a) both occlusive and semi-occlusive patches were allowed; (b) patches were applied to the skin for 4 hr rather than for 24 hr; (c) patches were applied to intact skin rather than abraded skin; and (d) in cases where tissue destruction was observed, the inspection period was prolonged to enable the reversibility of the e€ect to be assessed. The 1981 OECD guideline was subsequently adopted by the European Economic Community (EEC) for use in the classi®cation of substances and preparations, by incorporation of the guideline into Annex V of Directive 67/548/EEC (Ocial Journal of the European Communities, 1984). The Directive requires the classi®cation of corrosive chemicals according to the following risk phrases: R35 (``causes severe burns''), if full-thickness destruction of intact skin tissue occurs after the chemical has been applied for 3 min; and R34 (``causes burns''), if full-thickness destruction occurs after it has been applied for up to 4 hr (Ocial Journal of the European Communities, 1983 and 1993). In a typical study, the test material is applied to the shaved skin of albino rabbits, and the production of irreversible full-thickness necrosis is determined by visual inspection of the skin for up to 21 days. Histopathological evaluation may be used to con®rm the changes observed. Regulations originating with the United Nations (UN, 1977) require the labelling of packaged chemicals for international transport purposes. The UN guidelines recommend that corrosives be classi®ed into potency categories, termed ``packing groups''. Packing groups I, II and III are assigned on the basis of the capacity of a chemical, when tested on the intact skin of albino rabbits, to produce skin corrosion following exposures of 3 min, 1 hr or 4 hr, respectively. Packing group I is analogous to R35 in that application of the test material is for 3 min or less; however, full-thickness destruction of intact skin tissue must occur within an observation period of up to 1 hr following removal of the substance for packing group I to be assigned (for EU classi®cation purposes, the observation period is 21 days); packing groups II and III combined are analogous to R34, although the observation periods di€er (14 days for compliance with the UN guide-

lines compared with 21 days for EU classi®cation purposes). In October 1993, these UN guidelines were accepted by the US Department of Transportation (DOT, 1991). The use of laboratory animals for corrosivity testing can cause them pain and su€ering. This concern has been addressed in recent amendments to some international testing guidelines/requirements, which allow for the determination of skin corrosion by using alternative, in vitro/ex vivo, methods. For example, the updated 1992 OECD Testing Guideline 404 for acute dermal irritation/corrosion (OECD, 1992) states that `` . . . it may not be necessary to test in vivo materials for which corrosive properties are predicted on the basis of results from in vitro tests''. Similarly, the 18th amendment to Directive 67/548/EEC (Ocial Journal of the European Communities, 1993) states that ``classi®cation can be based on the results of validated in vitro tests''. A ®rst step towards de®ning alternative tests which could be used within the context of these revised test guidelines was taken by conducting a prevalidation study during 1993 and 1994 (Botham et al., 1995). The tests evaluated were the rat skin transcutaneous electrical resistance (TER) assay, CORROSITEXTM (InVitro International, Irvine, CA, USA) and the Skin2TM ZK1350 corrosivity test (Advanced Tissue Sciences, La Jolla, CA, USA). As a follow-up to the prevalidation study, ECVAM has co-ordinated and funded a validation study during 1996 and 1997 (ECVAM, 1995 and 1996). The main objectives of the validation study were to determine whether the tests evaluated in the prevalidation study, and an additional test, EPISKINTM (SADUC±BiomateÂriaux ImedexTM, Chaponost, France), were capable of: (a) discriminating corrosives (C) from non-corrosives (NC) for selected groups of chemicals (e.g. organic acids, phenols) and/or all chemicals (single chemical entities only); and (b) identifying known R35 (UN packing group I) and R34 (UN packing groups II & III) chemicals. The relative performances of the four tests were assessed by conducting each test in three independent laboratories, using a set of 60 coded chemicals. The results of the validation study are presented and discussed in a separate paper, along with an overview of its management and the methodological details for the four tests (Fentem et al., 1998). In this paper, an explanation is given of the basis for selecting and classifying the 60 test chemicals; this includes a discussion on the use of quantitative structure±activity relationship (QSAR) methods. The corrosivity classi®cations of the test chemicals are presented, along with various physicochemical properties. In addition, the prediction of skin corrosivity potential from pH data is considered.

Skin corrosivityÐtest chemicals

Chemicals selection During the planning stage of the validation study, the Management Team (MT) appointed an independent Chemicals Selection Sub-committee (CSSC), whose remit was to make a selection of chemicals which would fairly and objectively test the claims of the alternative methods to be included in the study. The ®rst consideration of the CSSC was to select chemicals for which there were unequivocal animal data on skin corrosivity, and which would therefore be suitable for use in the in vitro±in vivo comparisons. Most of the animal data reviewed were compiled by a working party established by the European Centre for Ecotoxicology and Toxicology of Chemicals (Bagley et al., 1996; ECETOC, 1995), although some additional data were provided, in con®dence, by an industrial company (Table 1). Chemicals with supporting data from these sources were allocated to the following categories: organic acids, organic bases, neutral organics, phenols, inorganic acids, inorganic bases, inorganic salts, electrophiles, and soaps/surfactants (anionic detergents). The chemicals were selected on the basis of their putative mechanisms of irritation/corrosivity. As a result of their greater lipophilicities, organic chemicals generally have greater skin permeabilities than do inorganic chemicals. The corrosivity of organics is postulated to result from the chemical ®rst penetrating the stratum corneum and then killing the epidermal cells beneath; this is, for example, a property of certain acids, bases, phenols, electrophiles and oxidizing agents. By contrast, most non-electrophilic neutral organic chemicals appear to be NC. Inorganic acids, bases and oxidizing agents are expected to have low skin permeabilities by virtue

473

of their high polarities. These chemicals probably cause corrosion by ®rst eroding the stratum corneum. Most inorganic salts appear to be NC. Anionic and cationic surfactant chemicals have low skin permeabilities due to their charged headgroups and their large molecular volumes (MV), while non-ionic surfactants have slightly higher skin permeabilities because of their greater lipophilicities. Both anionic and non-ionic surfactants appear to be NC. The corrosivity of cationic surfactants may result from solubilization of the stratum corneum, followed by destruction of the tissue beneath. Criteria for selecting chemicals The main criterion for including chemicals in the test set was that the corrosivity classi®cations were based on unequivocal animal data. In the case of borderline chemicals, structure±activity information was used to support the C/NC classi®cation. In addition, the following selection criteria were followed as closely as the availability of data allowed: (a) the balance between the number of chemicals in each putative mechanistic category should be maintained; (b) the number of C and NC chemicals in each category should be approximately equal, as should the total number of C and NC chemicals; (c) the range of biological responses displayed by the chemicals within each category should be as wide as possible, to test fairly the discriminating ability of the various alternative methods; and (d) for the NC chemicals, a range of skin irritation potentials should be covered, based on their primary irritation indices (PII) (ECETOC, 1995). Using these criteria, the CSSC drew up a list of 65 chemicals (comprising a main list of 60 and a reserve list of ®ve chemicals), with supporting toxicological test data. The reserve list was generated in

Table 1. Summary of animal data for the 11 chemicals not taken from the ECETOC reference chemicals databank* (information supplied in con®dence by an industrial company) Exposure period No. 1 2 55 4 25 28 32 43 53 57 20

Chemical Organic acids Hexanoic acid Organic bases 1,2-Diaminopropane 1-(2-Aminoethyl)piperazine Inorganic acids Boron tri¯uoride dihydrate Sulfuric acid (10% wt.) Phosphorus tribromide Phosphorus pentachloride Hydrochloric acid (14.4% wt.) Sulfamic acid Phosphoric acid Inorganic salts Iron (III) chloride

3 min

1 hr

4 hr

EU risk phrase

UN packing group

ND$

ND

+%

R34

II/III

+ ÿ}

+ +

+ +

R35 R34

I II

+ ND + + ND ÿ ÿ

+ + ND + + ÿ +

+ ND + + ND ÿ +

R35 Ck R35 R35 R34} NC R34

I Ck I I II/III NC II

ÿ

+

+

R34

II

*ECETOC, 1995; Bagley et al., 1996; $ND = not determined; %+ = skin corrosivity observed; }ÿ = no evidence of skin corrosivity; kC = the corrosivity classi®cation (EU risk phrase, UN packing group) could not be assigned unequivocally because no data were available for a 3 min exposure (an R34/II & III label is more probable); }R34 = classi®cation taken from Annex I of Directive 67/ 548/EEC. Chemical identi®cation numbers are as in Table 2.

474

M. D. Barratt et al. Table 2. Identities, suppliers and pH values of the 60 chemicals used in the ECVAM skin corrosivity validation study

No.

Chemical

CAS no.

Supplier*

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26

Hexanoic acid 1,2-Diaminopropane Carvacrol Boron tri¯uoride dihydrate Methacrolein Phenethyl bromide 3,3'-Dithiodipropionic acid Isopropanol o-Methoxyphenol (Guaiacol) 2,4-Xylidine 2-Phenylethanol Dodecanoic acid (lauric acid) 3-Methoxypropylamine Allyl bromide Dimethyldipropylenetriamine Methyl trimethylacetate Dimethylisopropylamine Potassium hydroxide (10% aq.) Tetrachloroethylene Iron (III) chloride Potassium hydroxide (5% aq.) n-Butyl propionate 2-tert-Butylphenol Sodium carbonate (50% aq.) Sulfuric acid (10% wt.) Isostearic acid

142-62-1 78-90-0 499-75-2 13319-75-0 78-85-3 103-63-9 1119-62-6 67-63-0 90-05-1 95-68-1 60-12-8 143-07-7 5332-73-0 106-95-6 10563-29-8 598-98-1 996-35-0 1310-58-3 127-18-4 7705-08-0 1310-58-3 590-01-2 88-18-6 497-19-8 7664-93-9 30399-84-9

27 28 29

Methyl palmitate Phosphorus tribromide 65/35 Octanoic/decanoic acid

112-39-0 7789-60-8 68937-75-7

30 31 32 33 34

4,4-Methylene-bis-(2,6-di-tert-butylphenol) 2-Bromobutane Phosphorus pentachloride 4-(Methylthio)-benzaldehyde 70/30 Oleine/octanoic acid

118-82-1 78-76-2 10026-13-8 3446-89-7 Ð

35 36 37 38 39 40 41

Hydrogenated tallow amine 2-Methylbutyric acid Sodium undecylenate (33% aq.) Tallow amine 2-Ethoxyethyl methacrylate Octanoic acid (caprylic acid) 20/80 Coconut/palm soap

61788-45-2 600-07-7 3398-33-2 61790-33-8 2370-63-0 124-07-02 Ð

42 43 44 45 46 47

2-Mercaptoethanol, Na salt (45% aq.) Hydrochloric acid (14.4% wt) Benzyl acetone n-Heptylamine Cinnamaldehyde 60/40 Octanoic/decanoic acids

37482-11-4 7647-01-0 2550-26-7 111-68-2 14371-10-9 68937-75-7

48 49 50

Glycol bromoacetate (85%) Eugenol 55/45 Octanoic/decanoic acids

3785-34-0 97-53-0 68937-75-7

51 52 53 54 55 56 57 58 59 60

Methyl laurate Sodium bicarbonate Sulfamic acid Sodium bisul®te 1-(2-Aminoethyl)piperazine 1,9-Decadiene Phosphoric acid 10-Undecenoic acid 4-Amino-1,2,4-triazole Sodium lauryl sulfate (20% aq.)

111-82-0 144-55-8 5329-14-6 7631-90-5 140-31-8 1647-16-1 7664-38-2 112-38-9 584-13-4 151-21-3

Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Elf Atochem Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Unichema International Aldrich Aldrich Unichema International Aldrich Aldrich Aldrich Aldrich Unichema International Elf Atochem Aldrich Elf Atochem Elf Atochem Aldrich Aldrich Unichema International Elf Atochem Aldrich Aldrich Aldrich Aldrich Unichema International SA Sopura Aldrich Unichema International Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich Aldrich

Form$

pH

L L L L L L S L L L L S L S L L L L L S L L L L L L

3.9 8.3 3.9 1.5 3.6 3.6 ND%} 3.6 3.9 5.5 3.6 ND} 10.0 3.9k 8.3 3.9 8.3 13.2 4.5 1.0k 13.1 3.6 3.9 11.7 1.2 3.6

S} L L

ND 1.0 3.9

S L S L L

ND} 3.9 ND** 6.8 3.9

S L L S L L S

ND} 3.6 8.3 ND} 3.9 3.6 8.3k

L L L L L L

12.0 1.5 3.9 8.4 3.9 3.9

L L L

2.0 3.6 3.9

L S S S L L L S} S L

3.9 8.3k 1.5k 5.3k 8.0 3.9 <1$$ ND 5.5k 3.9

*The chemicals suppliers were as follows: Aldrich (Sigma±Aldrich Co. Ltd, Gillingham, UK), Elf Atochem (Paris-la-DeÂfense, France), Unichema International (Unichema Chemicals Ltd, Wirral, UK) and SA Sopura (Courcelles, Belgium). $L = liquid, S = solid; %ND = not determined (unable to measure); }=insoluble in water; k=for 10% aqueous solution; }=liquid at 458C; **=explosive with water; $$=pH too low to determine accurately. Chemicals 1±10 were tested in the ®rst phase of the validation study.

case any of the 60 chemicals on the main list was dicult to obtain, as in fact proved to be the case for two materials. The ®nal list of 60 chemicals (Tables 2 and 3) comprised 11 organic acids (6C/

5NC), 10 organic bases (7C/3NC), nine neutral organics (9NC), ®ve phenols (2C/3NC), seven inorganic acids (6C/1NC), four inorganic bases (2C/2NC), three inorganic salts (1C/2NC), eight

Skin corrosivityÐtest chemicals

475

Table 3. Corrosivity classi®cations of the 60 chemicals used in the ECVAM skin corrosivity validation study EU risk phrase

UN packing group

PII value*

C C C C C C NC NC NC NC NC

R34 R34 R34 R34 R34 R34

II/III II/III II/III II/III II/III II/III

Ð NPC$ >4 4.44 NPC 5.11 0 0.44 4.33 3.78 2.42

Organic bases 1,2-Diaminopropane Dimethyldipropylenetriamine Tallow amine 1-(2-Aminoethyl)piperazine 3-Methoxypropylamine Dimethylisopropylamine n-Heptylamine 2,4-Xylidine (2,4-dimethylaniline) Hydrogenated tallow amine 4-Amino-1,2,4-triazole

C C C C C C C NC NC NC

R35 R35 R35 R34 R34 R34 R34

I I II II II/III II/III II/III

Ð NPC NPC Ð 6.67 5.61 6.67 1.44 3.56 0

8 11 16 19 22 27 44 51 56

Neutral organics Isopropanol 2-Phenylethanol (phenylethylalcohol) Methyl trimethylacetate Tetrachloroethylene n-Butyl propionate Methyl palmitate Benzyl acetone Methyl laurate 1,9-Decadiene

NC NC NC NC NC NC NC NC NC

3 23 9 30 49

Phenols Carvacrol 2-tert-Butylphenol o-Methoxyphenol (Guaiacol) 4,4-Methylene-bis-(2,6-di-tert-butylphenol) Eugenol

C C NC NC NC

R34 R34

II/III II/III

>4 5.67 2.38 0 2.92

4 28 32 25 57 43 53

Inorganic acids Boron tri¯uoride dihydrate Phosphorus tribromide Phosphorus pentachloride Sulfuric acid (10% wt.) Phosphoric acid Hydrochloric acid (14.4% wt) Sulfamic acid

C C C C C C NC

R35 R35 R35 R34/R35% R34 R34

I I I I/II/III% II II/III

Ð Ð Ð Ð Ð Ð Ð

18 42 21 24

Inorganic bases Potassium hydroxide (10% aq.) 2-Mercaptoethanol, Na salt (45% aq.) Potassium hydroxide (5% aq.) Sodium carbonate (50% aq.)

C C NC NC

R34/R35% R34

I/II/III%

NPC NPC 5.22 2.33

20 52 54

Inorganic salts Iron (III) chloride Sodium bicarbonate Sodium bisul®te

C NC NC

R34

II

Ð 0.11 1.0

5 14 48 6 31 33 39 46

Electrophiles Methacrolein Allyl bromide Glycol bromoacetate (85%) Phenethyl bromide 2-Bromobutane 4-(Methylthio)-benzaldehyde 2-Ethoxyethyl methacrylate Cinnamaldehyde

C C C NC NC NC NC NC

R34 R34 R34

II/III II/III II/III

4.11 7.17 7.67 0 2.44 0.89 1.56 3.71

37 41 60

Soaps/surfactants Sodium undecylenate (33% aq.) 20/80 Coconut/palm soap Sodium lauryl sulfate (20% aq.)

NC NC NC

No.

Chemical

1 29 36 40 47 50 7 12 26 34 58

Organic acids Hexanoic acid 65/35 Octanoic/decanoic (capric) acids 2-Methylbutyric acid Octanoic (caprylic) acid 60/40 Octanoic/decanoic acids 55/45 Octanoic/decanoic acids 3,3'-Dithiodipropionic acid Dodecanoic (lauric) acid Isostearic acid 70/30 Oleine/octanoic acid 10-Undecenoic acid

2 15 38 55 13 17 45 10 35 59

C/NC

0.78 0.92/2.22 0 5.67 1.08 4.56 1.21 3.89 3.0

1.67 2.67 6.78

*PII = primary irritation index (ECETOC., 1995; Bagley et al., 1996); $NPC = not possible to calculate; %=the animal data and other supporting information are not suciently comprehensive to enable unequivocal classi®cations to be made, although it is more probable that R34 (II/III) labels are appropriate. Chemical identi®cation numbers are as in Table 2.

476

M. D. Barratt et al.

electrophiles (3C/5NC), and three soaps/surfactants (3NC). Use of QSAR analysis Where possible, QSARs were used for two main reasons: (a) to aid in the selection of appropriate chemicals for inclusion in the test set for the validation study; and (b) to aid in the interpretation of the in vitro data when any of the test chemicals were known to give questionable or borderline classi®cations on the basis of the in vivo results. QSARs are mathematical correlations based on the premise that the properties of a chemical are implicit in its molecular structure. If the mechanism responsible for the activity of a group of chemicals can be elucidated, and the relevant parameters measured or calculated then, in principle, a QSAR can be established. In one type of QSAR analysis, called principal components analysis (PCA), the original variables are transformed into a new set of orthogonal linear combinants called principal components (PC). The transformation ensures that the variance explained is greatest in the ®rst PC, less in the second PC, and so on. PCA enables multi-component datasets to be represented as two-dimensional or three-dimensional plots without signi®cant loss of information. The QSARs described in this paper are based on the putative mechanisms of skin corrosivity described earlier, for example, a chemical must ®rst penetrate the skin, and can then kill the cells beneath the stratum corneum provided that it is suciently cytotoxic. The parameters required to model this mechanism are the same as those that model percutaneous absorption (Barratt, 1995): (a) logP (the logarithm of the octanol±water partition coecient); (b) MV; and (c) melting point (Mpt), in addition to a putative measure of cytotoxicity: pKa (the negative logarithm of the acid dissociation constant). Similar principles were applied previously to a study of the skin irritation and corrosivity potentials of neutral and electrophilic organic chemicals (Barratt, 1996a). In the case of electrophiles, only two of the three parameters for skin permeability were employed: logP and MV. The Mpt, which can be used to compute aqueous solubility when linked with logP (Suzuki, 1991), was omitted, since almost all of the test chemicals were in the liquid state. The parameter used to model the ``reactivity'' component of electrophile action was the dipole moment; this was successfully used as the reactivity parameter in a previous QSAR study of the eye irritation potentials of neutral organic chemicals (Barratt, 1997). PCA was carried out on four datasets for acids (organics and inorganics), organic bases, phenols and electrophiles. These datasets were compiled by selecting chemicals from datasets used for previous QSAR analyses (Barratt, 1996a,b), and then incorporating those chemicals which were in the vali-

dation study test set but not in one of the original datasets. Chemicals that were present in both the validation study test set and one of the original QSAR training sets are listed in Table 4, whereas chemicals that were present in the validation study test set but not in one of the original QSAR training sets are listed in Table 5, together with their physicochemical properties. The vector loadings of the variables in the four analyses are summarized in Table 6, and plots of the ®rst two PCs in each dataset are shown in Figs 1±4. Insucient data were available to construct a predictive QSAR model for the corrosivities of surfactants. It should be noted that chemical ]30 [4,4-methylene-bis-(2,6-di-tert-butylphenol)] was not included in the QSAR analysis for the phenols because its logP and MV values are well outside the parameter space of the training set data. On the basis of these parameters, the skin permeability of this chemical is judged to be rather low, so it is concluded that its corrosivity will also be low. This is consistent with the animal data for this chemical, which indicated no signs of irritation or corrosivity. Chemicals ]29, ]34, ]47 and ]50 were not included in the QSAR analysis for the acids because they are two-component mixtures rather than single chemical entities. Identities and properties of the test chemicals The identities of the 60 test chemicals are presented in Table 2, along with their physical forms and pH values. In Table 3, the corrosivity classi®cations and chemical types of the 60 chemicals are reported. Additional physicochemical parameters for those chemicals not included in the original QSAR training sets are presented in Table 5; the corresponding values for the remaining test chemicals have been published previously (Barratt, 1996a,b).

Table 4. Chemicals that were in both the test set for the validation study and one of the original QSAR training sets Class

No.

Organic acids

1 12 40 35 38 3 9

Organic bases Phenols Inorganic acids Electrophiles

53 57 5 14 31 33 39 46 48

Chemical Hexanoic acid Dodecanoic (lauric) acid Octanoic (caprylic) acid Hydrogenated tallow amine Tallow amine Carvacrol o-Methoxyphenol (Guaiacol) Sulfamic acid Phosphoric acid Methacrolein Allyl bromide 2-Bromobutane 4-(Methylthio)benzaldehyde 2-Ethoxyethyl methacrylate Cinnamaldehyde Glycol bromoacetate

Chemical identi®cation numbers are as in Table 2.

Skin corrosivityÐtest chemicals

477

Table 5. Physicochemical parameters of the chemicals not in the original QSAR training sets No.

Chemical

logP*

MV$

7 26 36 58

Organic acids 3,3'-Dithiodipropionic acid Isostearic acid 2-Methylbutyric acid 10-Undecenoic acid

0.136 7.681 0.713 3.934

148.42 258.5 85.129 163.05

159 37 37 37

4.55 4.75 4.85 4.85

2 10 13 15 17 45 55 59

Organic bases 1,2-Diaminopropane 2,4-Xylidine 3-Methoxypropylamine Dimethyldipropylenetriamine Dimethylisopropylamine n-Heptylamine 1-(2-Aminoethyl)piperazine 4-Amino-1,2,4-triazole

ÿ1.272 1.746 ÿ0.502 ÿ0.776 0.915 2.558 ÿ1.250 2.451

68.304 96.939 80.021 145.76 84.098 111.6 111.65 60.333

37 37 37 37 37 37 37 86

10.0 4.84 10.0 10.7 10.3 10.62 10.7 7.8

3.424 10.07

129.61 431.31

37 158

10.2 010

2.86

122.2

37

10.0

3.192

110.91

23 30 49 6

Phenols 2-tert-Butylphenol 4,4-Methylene-bis-(2,6-di-tertbutylphenol)** Eugenol Electrophiles Phenethyl bromide

Mpt%

pKa}

Dipole momentk

1.75

*P = octanol/water partition coecient; $MV = molecular volume (AÊ3); %Mpt = melting point (8C); }pKa = the negative logarithm of the acid-base dissociation constant; kdipole moment (electrophiles only; debyes); **=not included in the QSAR analysis for phenols. Note that values for the organic acids ]29, ]34, ]47 and ]50 are not given; they were not included in the QSAR analysis for acids since they are mixtures not single chemical entities. The inorganic acids ]4, ]28 and ]32 were excluded from the QSAR analysis because they are acid precursors; the inorganic acids ]25 and ]43 were also excluded because they were not at 100% activity. Chemical identi®cation numbers are as in Table 2.

pH determination The pH values of the 60 test chemicals (Table 2) were measured by BIBRA International (Carshalton, Surrey, UK) under the terms of an ECVAM contract. For all samples, the pH was measured initially by using broad-band pH paper, to establish the approximate range of pH. The

sample pH was then measured more precisely by using pH paper calibrated in 0.1 units of pH. For liquids, a small amount of the test material was dispensed into a sample tube and the pH was measured without dilution. For solid materials, a 10% (w/v) mixture in distilled water was prepared and the pH of the resulting solution/suspension was measured.

Table 6. Vector loadings in the principal components analyses of the four datasets (acids, bases, phenols and electrophiles) PC1

PC2

PC3

PC4

Organic acids logP* MV$ Mpt% pKa} Fraction of variance explained Total variance explained

0.627 0.640 ÿ0.012 0.445 0.514 0.514

ÿ0.225 0.047 0.934 0.274 0.266 0.780

ÿ0.284 (0.320 ÿ0.302 0.852 0.176 0.956

0.690 ÿ0.697 0.190 0.035 0.044 1.00

Organic bases logPk MV Mpt pKa Fraction of variance explained Total variance explained

0.651 0.634 ÿ0.305 0.284 0.428 0.428

0.263 0.326 0.635 ÿ0.649 0.283 0.711

ÿ0.090 0.114 0.697 0.702 0.195 0.906

0.706 ÿ0.692 0.134 0.070 0.095 1.00

Phenols logPk MV (whole molecule) Mpt pKa Fraction of variance explained Total variance explained

0.632 0.546 ÿ0.540 0.107 0.507 0.507

ÿ0.129 ÿ0.351 ÿ0.334 0.865 0.283 0.790

ÿ0.033 0.586 0.649 0.484 0.137 0.927

ÿ0.764 0.486 ÿ0.418 ÿ0.078 0.073 1.00

Electrophiles logPk MV (whole molecule) Total dipole (whole molecule) Fraction of variance explained Total variance explained

0.676 0.718 0.172 0.550 0.550

ÿ0.335 0.092 0.938 0.350 0.900

0.657 ÿ0.691 0.302 0.100 1.00

*P = octanol/water partition coecient; $MV = molecular volume (AÊ3); %Mpt = melting point (8C); }pKa = the negative logarithm of the acid-base dissociation constant; k=calculated using the CHEMICALC system (Suzuki and Kudo, 1990).

478

M. D. Barratt et al.

Fig. 1. Plot of the ®rst two principal components of logP, MV, Mpt and pKa for 54 acids (49 organic and ®ve inorganic). Black squares indicate C chemicals in the training set; empty squares indicate NC chemicals in the training set; crosses indicate predictions. Points are numbered as in Table 2.

Supply, coding and distribution of chemicals

Fig. 3. Plot of the ®rst two principal components of logP, MV, Mpt and pKa for 32 phenols. Black squares indicate C chemicals in the training set; empty squares indicate NC chemicals in the training set; crosses indicate predictions. Points are numbered as in Table 2.

Supply

The test chemicals were independently coded and supplied to the participating laboratories by BIBRA International, in accordance with the procedures outlined previously by Brantom et al. (1995).

Chemicals with the same source and speci®cation as those tested in vivo were obtained, wherever feasible. Where this was not possible, a speci®cation as close as possible to that used in the in vivo testing was chosen.

Fig. 2. Plot of the ®rst two principal components of logP, MV, Mpt and pKa for 43 organic bases. Black squares indicate C chemicals in the training set; empty squares indicate NC chemicals in the training set; crosses indicate predictions. Points are numbered as in Table 2.

Fig. 4. Plot of the ®rst two principal components of logP, MV and dipole moment for 43 electrophiles. Black squares indicate C chemicals in the training set; empty squares indicate NC chemicals in the training set; crosses indicate predictions. Points are numbered as in Table 2.

Skin corrosivityÐtest chemicals

Dispensing The following precautions were taken to avoid any possible confusion of one test chemical with another: (a) only one chemical was dispensed at any one time; and (b) all samples were sealed and labelled with unique coded labels before the next chemical was dispensed. After dispensing, samples were returned to appropriate storage conditions until they were packed for despatch. Coding and labelling Each chemical was ®rst allocated a numeric code. Each participating laboratory was also allocated a unique code number (CR1±CR12), so that subsequent analysis of in vitro data would be possible without knowledge of the identity of the laboratory or test material. The number of aliquots of each test material required for the full study was de®ned (requiring precise de®nition of the number of laboratories participating, and the total number of replicates and runs for each test). Random codes were generated by computer software for the total number of samples, so that each sample had a unique number. Distribution The shipping of chemical samples was carried out in compliance with various regulations governing the transport of dangerous substances and health and safety. These regulations generally require that the UN hazard class and packing group are known for all of the materials to be transported. Despite the small quantities involved in the study, there are signi®cant limitations on the chemicals which can be transported in the same container, as well as prohibitions imposed by the International Aviation Transport Authority (IATA) on carrying certain chemicals as air freight. The procedure for handling many of these issues was developed successfully by working with a professional packing and shipping company. Samples were shipped under conditions designed to minimize damage during transit. Each sample was accompanied by a description of its physical state and appearance, and by an indication of the need to store it under dry nitrogen or another inert gas. In addition, each laboratory was warned on a packing slip to treat all chemicals as potentially C and to apply appropriate handling precautions. As most chemicals are expected to be most stable when stored at low temperatures, recipients were advised to store all samples in a refrigerator. As many of the chemicals to be tested are classi®ed as C for transport purposes, there were severe limitations on the shipping. It was necessary for each package to display on the outside a list of the chemicals it contained, and for this list to be removed by the shippers before delivery to the participating laboratory. As a precaution against error,

479

the shipments were all delivered to named recipients at the laboratories participating in the validation study, di€erent from the experimenters, who were asked to con®rm that the packages had been received and that there were no labels identifying the chemicals. At the request of the MT, the test chemicals were distributed in two stages: the ®rst set of 10 chemicals (]1±]10) were sent to the participating laboratories in the middle of June 1996; the second set of 50 chemicals (]11±]60) were sent on 23 September 1996 (Fentem et al., 1998). Health and safety data An emergency procedure was established to allow any participant to obtain the necessary health and safety data in the event of an accident. The procedure involved lodging a sealed envelope containing a full set of the chemical identities, classi®cations and coding information for each laboratory with the West Midlands Poisons Centre (Dudley, West Midlands, UK). All participating laboratories were given the telephone number of the Poisons Centre so that, if an emergency occurred, the code could be broken, in which case it was arranged that the Poisons Centre would inform the MT. This procedure was not invoked during the study. Prediction of corrosivity from pH According to OECD Testing Guideline 404 (OECD, 1992), substances should not be tested in animals for skin irritation or corrosivity if they can be predicted to be C on the basis of their physicochemical properties. In particular, substances exhibiting strong acidity or alkalinity should not be tested; these are predicted to be C from the following prediction model (PM), taken from the OECD guideline: if pH E 2 4 C; if pHe 11.5 4 C. The pH values provided by BIBRA International were used to evaluate this PM with the test chemicals used in the validation study. For the 12 chemicals that had extreme pH values (i.e. pH E 2 or e11.5; Table 2), and thus were predicted to be C by applying the above PM, the animal-based corrosivity classi®cations are given in Table 7. It can be seen that three of the 12 chemicals are false positives when the predicted (from pH) and observed (in vivo) classi®cations are compared: potassium hydroxide (5% aq.), sodium carbonate (50% aq.), and sulfamic acid. For eight of the 60 chemicals, it proved impossible to determine pH values (Table 2). Discussion When constructing the test set for this validation study, it was highly preferable that the chemicals chosen di€ered from those used in the prevalidation study. Only ®ve of the chemicals included in the

480

M. D. Barratt et al. Table 7. Chemicals predicted to be corrosive on the basis of pH No.

Chemical

pH

4 18 20 21 24 25 28 42 43 48 53 57

Boron tri¯uoride dihydrate Potassium hydroxide (10% aq.) Iron (III) chloride Potassium hydroxide (5% aq.) Sodium carbonate (50% aq.) Sulfuric acid (10% wt) Phosphorus tribromide 2-Mercaptoethanol, Na salt (45% aq.) Hydrochloric acid (14.4% wt) Glycol bromoacetate (85%) Sulfamic acid Phosphoric acid

1.5 13.2 1.0 13.1 11.7 1.2 1.0 12.0 1.5 2.0 1.5 <1

Observed classi®cation* C C C NC NC C C C C C NC C

*The corrosivity classi®cation based on animal data. All of the 12 chemicals are predicted to be C on the basis of having extreme pH values (i.e. pHE 2 or e11.5). Chemical identi®cation numbers are as in Table 2.

validation study test set were also present in the prevalidation study test set (Botham et al., 1995), namely: hexanoic acid (]1), dodecanoic (lauric) acid (]12), tallow amine (]38), octanoic (caprylic) acid (]40) and 55/45 octanoic/decanoic acids (]50). Four of these chemicals are relatively weak carboxylic acids; hexanoic acid and octanoic acid are C, probably because their small MVs result in skin permeabilities which are suciently high to penetrate the stratum corneum. The corrosivity classi®cation of tallow amine is discussed later. QSAR analyses As a general rule, non-electrophilic neutral organic chemicals do not appear to be classi®ed as C,

so PCA was not undertaken on this class of chemicals. Certain alcohols are known exceptions to this rule; for example, propargyl alcohol and 2-butyn1,4-diol are listed as C in Annex 1 of Directive 67/ 548/EEC (Ocial Journal of the European Communities, 1993). The reason for this exceptional toxicity is likely to be oxidation of the alcohol to the conjugated aldehyde (Richardson, 1992). Propargyl alcohol and 2-butyn-1,4-diol were therefore treated as pro-electrophiles rather than as neutral organics when constructing the QSAR training sets (Barratt, 1996a). All the other classes contain both C and NC chemicals, so a separate PCA analysis was carried out for each class. The PCA plots generated

Table 8. Comments on particular test chemicals No.

Chemical

3

Carvacrol

4 5 9

Boron tri¯uoride dihydrate Methacrolein o-Methoxyphenol

18

Potassium hydroxide (5% aq.)

20 23

Iron (III) chloride 2-tert-Butylphenol

25

Sulfuric acid (10% wt)

28 32 33 35

Phosphorus tribromide Phosphorus pentachloride 4-(Methylthio)-benzaldehyde Hydrogenated tallow amine

38

Tallow amine

40

Octanoic acid

42 46 49

2-Mercaptoethanol Cinnamaldehyde Eugenol

54 58

Sodium bisul®te 10-Undecenoic acid

Chemical identi®cation numbers are as in Table 2.

Comments on data supporting classi®cation/general comments Borderline C/NC chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis) Decomposes Reducing agent (may a€ect MTT assay) Borderline NC/C chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis) C, but supporting data do not enable unequivocal classi®cation as either R34 (II/ III) or R35 (I); more probable to be R34 (II/III) Coloured test material Borderline C/NC chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis) C, but supporting data do not enable unequivocal classi®cation as either R34 (II/ III) or R35 (I); more probable to be R34 (II/III) Highly volatileÐdecomposes Produces fumes on contact with waterÐdecomposes Reducing agent (may a€ect MTT assay) Borderline NC/C chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis) The in vivo exposure time may have been greater than 3 min. Necrosis was observed in two of the three rabbits only from day 7 Borderline C/NC chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis) Reducing agent (may a€ect MTT assay) Reducing agent (may a€ect MTT assay) Borderline NC/C chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis) Reducing agent (may a€ect MTT assay) Borderline NC/C chemical, as judged subjectively from the proximity of the chemical to the classi®cation boundary (SAR analysis)

Skin corrosivityÐtest chemicals

(Figs 1±4) typically show a clear separation of C and NC chemicals, enabling those chemicals which lie close to the C/NC borderline to be identi®ed (Table 8). Knowledge of such chemicals is important when evaluating the performances of the four in vitro tests included in the validation study. Among the acids, octanoic acid (]40; C) and 10-undecenoic acid (]58; NC) could be considered borderline, as could the base, hydrogenated tallow amine (]35; NC). As far as the phenols are concerned, the two C phenols, carvacrol (]3) and 2-tert-butylphenol (]23), are close to the C/NC classi®cation boundary. The NC phenols, o-methoxyphenol (]9) and eugenol (]49), are also borderline. 4,4-Methylenebis-(2,6-di-tert-butylphenol) (]30) is not included on the PCA plot for phenols (Fig. 3); it was excluded from the analysis because its logP and MV values lie too far outside the parameter space, but it is considered to be NC. In Fig. 4, borderline electrophiles are dicult to identify because the classi®cation boundary is less discernible.

Properties of individual chemicals In addition to the borderline chemicals, some of the other chemicals in the validation study test set have properties which should be taken into account when evaluating the performances of the four test methods; these chemicals are identi®ed, and comments relating to them are summarized, in Table 8. An example is tallow amine (]38) which, on the basis of the animal data available, is classi®ed as R35/II, but because it is a long-chain fatty amine it could have been dicult to remove from rabbit skin and this may have resulted in an in vivo exposure time considerably in excess of 3 min. Following the application of tallow amine, necrosis was observed in two of the three rabbits tested only from day 7 (ECETOC, 1995). Other examples are NC chemicals which act as reducing agents; these can give a false overestimate of cell viability when tested in the MTT assay, by reducing the MTT. Five reducing agents are present in the test set (Table 8): methacrolein (]5), 4-(methylthio)-benzaldehyde (]33), 2-mercaptoethanol (]42), cinnamaldehyde (]46) and sodium bisul®te (]54). Certain other chemicals could also interfere with the MTT assay. In the case of two C chemicals, 10% (aq.) potassium hydroxide (]18) and 10% (wt.) sulfuric acid (]25), corrosivity classi®cations (EU risk phrases, UN packing groups) could not be assigned unequivocally because data were not available from animal tests involving a 3 min exposure to the test materials; it is thought, however, that an R34/II & III label is considerably more likely for these chemicals. Similarly, 3-min exposure data were missing for 14.4% (wt) hydrochloric acid (]43) but, in this case, an R34/II & III classi®cation was available from Annex 1 of Directive 67/548/EEC (Table 1).

481

Use of pH in the prediction of corrosivity potential The pH values obtained by BIBRA International for 52 of the test chemicals were used to evaluate the PM outlined in OECD Testing Guideline 404 (OECD, 1992). The comparison between predicted (C) and observed (C or NC, based on the animal data) classi®cations for the 12 chemicals with extreme pH values (Table 7) indicated a sensitivity of the PM of 75%. As not all C chemicals have a mechanism of action directly related to pH (the PM was only applicable to 12 of the 52 test chemicals for which pH data were obtained), it is not appropriate to use pH data in isolation for classi®cation purposes (Oliver and Pemberton, 1985). Other physicochemical properties should be taken into account when predicting corrosivity. The need to consider acid/alkaline reserve values, which are a measure of the bu€ering capacities of chemicals, in addition to their pH values, is recognized in OECD Testing Guideline 404 (OECD, 1992) and in the proposed OECD testing strategy for skin irritation/corrosion (OECD, 1996). This could be achieved by using a PM such as the one developed by Young and colleagues (Young and How, 1994; Young et al., 1988). In order to evaluate appropriate PMs, ECVAM intends to carry out a small follow-up study, using the pH data presented in Table 2 and acid/alkaline reserve measurements (to be determined under the terms of an ECVAM contract). In addition to the use of physicochemical data in predicting skin corrosivity potential, the 1992 revision of OECD Testing Guideline 404 recommends the use of data derived from appropriate in vitro tests. It is therefore envisaged that the follow-up investigation will evaluate the combined use of pH, acid/alkaline reserve and in vitro data (from the four tests included in the validation study) for the prediction of skin corrosivity. AcknowledgementsÐThe authors would like to thank the following people for their advice in selecting the chemicals: Graeme Archer (ECVAM, Italy), Hermann-Georg HolzhuÈtter (Humboldt UniversitaÈt, Germany), Annett Janusch (ECVAM, Italy) and Michael SjoÈstroÈm (UmeaÊ University, Sweden). The contribution of Pat Aspin (BIBRA, UK) with respect to the coding and distribution of the chemicals is particularly acknowledged. The willingness of the West Midlands Poisons Centre (Dudley, UK) to hold health and safety data on the coded chemicals was also appreciated. The majority of the ®nancial support for this validation study was provided by the European Commission (ECVAM) under the terms of contract numbers: 5650-9312 XP ISP GB, 11869-96-05 F1ED ISP D, 11879-96-05 F1ED ISP GB, 11881-96-05 F1ED ISP GB, 11883-96-05 F1ED ISP GB, 11910-96-06 F1ED ISP USA, 11913-96-06 F1ED ISP GB, 11915-96-06 F1ED ISP F, 11978-96-06 F1ED ISP F, 11979-96-06 F1ED ISP F, 11991-96-06 F1ED ISP F, 11997-96-07 F1ED ISP GB, 11998-96-07 F1ED ISP D and 11999-96-07 F1ED ISP D. In addition, the contributions of all of the participating organizations are acknowledged.

482

M. D. Barratt et al. REFERENCES

Bagley D. M., Gardner J. R., Holland G., Lewis R. W., ReÂgnier J-F., Stringer D. A. and Walker A. P. (1996) Skin irritation: reference chemicals data bank. Toxicology in Vitro 10, 1±6. Barratt M. D. (1995) Quantitative structure±activity relationships for skin permeability. Toxicology in Vitro 9, 27±37. Barratt M. D. (1996a) Quantitative structure±activity relationships for skin irritation and corrosivity of neutral and electrophilic organic chemicals. Toxicology in Vitro 10, 247±256. Barratt M. D. (1996b) Quantitative structure±activity relationships (QSARs) for skin corrosivity of organic acids, bases and phenols: principal components and neural network analysis of extended datasets. Toxicology in Vitro 10, 85±94. Barratt M. D. (1997) QSARs for the eye irritation potential of organic chemicals. Toxicology in Vitro 11, 1±8. Botham P. A., Chamberlain M., Barratt M. D., Curren R. D., Esdaile D. J., Gardner J. R., Gordon V. C., Hildebrand B., Lewis R. W., Liebsch M., Logemann P., Osborne R., Ponec M., ReÂgnier J-F., Steiling W., Walker A. P. and Balls M. (1995) A prevalidation study on in vitro skin corrosivity testing. The report and recommendations of ECVAM workshop 6. ATLA 23, 219± 255. Brantom P. G., Aspin P. and Thompson C. (1995) Supply, coding and distribution of samples for validation studies. ATLA 23, 348±351. DOT (1991) Code of Federal Regulations. In Method of Testing Corrosion to the Skin. Department of Transportation. Transportation Title 49, Part 173. 136, Appendix A. Oce of the Federal Register, National Archives and Records Administration, Washington, DC. Draize J. H. (1965) Appraisal of the Safety of Chemicals in Foods, Drugs and Cosmetics. Association of Food and Drug Ocials of the United States. pp. 46±49. Topeka, Kansas. Draize J. H., Woodard G. and Calvery H. O. (1944) Methods for the study of irritation and toxicity of substances applied topically to the skin and mucous membranes. Journal of Pharmacology and Experimental Therapeutics 82, 377±390. ECETOC (1995) Skin Irritation and Corrosion: Reference Chemicals Data Bank. ECETOC Technical Report No. 66, pp. 247. European Centre for Ecotoxicology and Toxicology of Chemicals, Brussels. ECVAM (1995) ECVAM News & Views. ATLA 23, 750± 751. ECVAM (1996) ECVAM News & Views. ATLA 24, 653± 655. Fentem J. H., Archer G. E. B., Balls M., Botham P. A., Curren R. D., Earl L. K., Esdaile D. J., HolzhuÈtter H-G. and Liebsch M. (1998) The ECVAM international validation study on in vitro tests for skin corrosivity. 2. Results and evaluation by the Management Team. Toxicology in Vitro 12, x±x. OECD (1992) OECD Guideline for Testing of Chemicals, No. 404: Acute Dermal Irritation/Corrosion.

Organisation for Economic Cooperation and Development, pp. 6. Paris. OECD (1996) Final Report of the OECD Workshop on Harmonization of Validation and Acceptance Criteria for Alternative Toxicological Test Methods. In Organisation for Economic Cooperation and Development, pp. 62. Paris. Ocial Journal of the European Communities (1983) Guide to the classi®cation and labelling of dangerous substances and preparations. Criteria for the choice of chemicals indicating special risks (R phrases) and safety advice (S phrases). Ocial Journal of the European Communities L257, 13±20. Ocial Journal of the European Communities (1984) Commission Directive 84/449/EEC of 25 April 1984 adapting to technical progress for the sixth time Council Directive 67/548/EEC on the approximation of laws, regulations and administrative provisions relating to the classi®cation, packaging and labelling of dangerous substances: Skin Irritation. Ocial Journal of the European Communities L251, 106±108. Ocial Journal of the European Communities (1993) Annexes I, II, III and IV to Commission Directive 93/ 21/EEC of 27 April 1993 adapting to technical progress for the 18th time Council Directive 67/548/EEC on the approximation of the laws, regulations and administrative provisions relating to the classi®cation, packaging and labelling of dangerous substances. Ocial Journal of the European Communities L110A, 1±86. Oliver G. J. A. and Pemberton M. A. (1985) An in vitro epidermal slice technique for identifying chemicals with potential for severe cutaneous e€ects. Food and Chemical Toxicology 23, 229±232. Richardson M. L. (1992) The Dictionary of Substances and their E€ects. Royal Society of Chemistry, London Vol. 1, pp. 144±147. Suzuki T. (1991) Development of an automated estimation system for both partition coecient and aqueous solubility. Journal of Computer-Aided Molecular Design 5, 149±166. Suzuki T. and Kudo Y. (1990) Automatic logP estimation based on combined additive modelling methods. Journal of Computer-Aided Molecular Design 4, 155±198. United Nations (1977) United Nations Economic and Social Council. Joint meeting of the RID safety committee and the group of experts on the transportation of dangerous goods. Trans/GE 15/R 274, 2. Young J. R. and How M. J. (1994) Product classi®cation as corrosive or irritant by measuring pH and acid/alkali reserve. In In Vitro Skin ToxicologyÐIrritation, Phototoxicity, Sensitization. Alternative Methods in Toxicology. Edited by A. Rougier, A. M. Goldberg and H. I. Maibach. Vol. 10, pp. 23±27. Mary Ann Liebert, New York. Young J. R., How M. J., Walker A. P. and Worth W. M. H. (1988) Classi®cation as corrosive or irritant to skin of preparations containing acidic or alkaline substances without testing on animals. Toxicology in Vitro 2, 19±26.