Refinement and Reduction of Acute Oral Toxicity Testing: A Critical Review of the Use of Cytotoxicity Data

Transcription

1 ATLA 39, , Refinement and Reduction of Acute Oral Toxicity Testing: A Critical Review of the Use of Cytotoxicity Data Arnhild Schrage, Katja Hempel, Markus Schulz, Susanne N. Kolle, Bennard van Ravenzwaay and Robert Landsiedel BASF SE, Experimental Toxicology and Ecology, Ludwigshafen, Germany Summary Acute oral toxicity testing is still required for the classification and labelling of chemicals, agrochemicals and related formulations. There have been increasing efforts over the last two decades to reduce the number of animals needed for this testing, according to the Three Rs concept. To evaluate the utility of an in vitro cytotoxicity test in our routine testing for acute oral toxicity, we have implemented in our laboratory the neutral red uptake (NRU) method, with Balb/c 3T3 fibroblasts after a 48-hour exposure, which was recommended in ICCVAM Report , Initially, we tested 16 substances that had existing in vivo and in vitro data available, to prove our technical proficiency with the in vitro test. Then, testing was performed with 187 test substances, including a broad variety of chemicals, agrochemicals and formulations. The starting dose for acute oral systemic toxicity assays in rats (LD50) was estimated by using the prediction model presented in the ICCVAM validation study, and subsequently compared to the results obtained by in vivo testing performed according to, or similar to, OECD Test Guideline 423. Comparison of all of the 203 predicted LD50 values that were deduced from the in vitro IC50 values, with the in vivo results from oral toxicity studies in rats, resulted in a low overall concordance of 35%. The in vitro cytotoxicity assay achieved a good concordance of 74%, only for the weakly toxic substances (EU-GHS Cat. 4). However, it must be noted that 71% of the substances tested (i.e. 145/203) were classified as being weakly toxic in vitro. We further analysed the utility of the in vitro test for predicting the starting dose for an in vivo study, and the potential reduction in animal usage that this would engender. In this regard, the prediction by the cytotoxicity test was useful for 59% of the substances. However, the use of a standard starting dose of 300mg/kg bw by default (without previous cytotoxicity testing) would have been almost as useful (50%). In contrast, the prediction by an experienced toxicologist was correct for 95% of the substances. However, this was only performed for 40% of the substances, mainly those of no to low toxicity. Calculating the theoretical animal numbers needed in several scenarios supported these results. The additional analysis, considering some physicochemical data (solubility, molecular weight, log P OW ), substance class and mode of action, revealed no specific applicability domains. In summary, the use of the 3T3 NRU cytotoxicity data alone did not sufficiently contribute to refinement and reduction in the acute oral toxicity testing of the substance portfolio tested routinely in our laboratory. Key words: acute oral systemic toxicity, Balb/c 3T3 fibroblasts, cytotoxicity, in vitro, in vivo, neutral red uptake, Three Rs. Address for correspondence: Robert Landsiedel, BASF SE, Experimental Toxicology and Ecology, Ludwigshafen, Germany. robert.landsiedel@basf.com Introduction Historically, acute oral toxicity was the initial test performed for the evaluation of the toxic characteristics of a substance. But today, an acute toxicity study is no longer needed for pharmaceuticals (1), and all acute animal experiments for testing cosmetic products or their contents have been banned within the EU since March 2009, under Directive 2003/15/EC (2). Currently, testing for acute oral toxicity is still required in the toxicological assessment of chemicals and agrochemicals worldwide. In Europe, the regulatory requirement for acute toxicity testing of industrial chemicals is outlined in Regul - ation (EC) No 1907/2006 (the REACH regulation ), with specification of the data requirements in Annexes VII XI (3), and for agrochemicals by Regulation (EC) No 1107/2009, with the specified data requirements listed in Annexes II and III (4). In addition, regulations for biocides and medicinal products may apply. Comparable regulations are in force nearly all over the world, as required by the respective competent authorities in the USA, Japan, China, Brazil, and other countries. As emphasised by Seidle et al. (5), one of the most important goals to be achieved with the results of acute toxicity studies, is the classification and labelling of the tested substances. Therefore, the determination of an exact LD50 value in the study is usually unnecessary. Hence, testing is performed at the upper boundaries of hazard classification levels. With the coming into force of the Globally Harmonised System of

2 274 A. Schrage et al. Classif ication and Labelling of Chemicals (GHS; 6), the dose levels tested are usually 2000mg/kg body weight (bw), 300mg/kg bw, 50mg/kg bw and 5mg/kg bw, as described in the Organisation for Economic Co-operation and Development (OECD) Test Guide line (TG) 423: Acute Oral Toxicity Acute Toxic Class Method (7). In the light of the Three Rs concept (Replace ment, Reduction and Refinement), initially described by Russell & Burch (8), refinements to the acute oral toxicity test guideline have been implemented over the last two decades (e.g. OECD TG 423 replacing TG 401; 9), leading to a significant reduction in the animal numbers used. Furthermore, attempts have been made to identify possible alternative methods for the prediction of acute oral toxicity, such as correlating rodent and/or human acute lethal toxicity with in vitro cytotoxicity (10 16). Methods for predicting other important components of acute toxicity, such as type, onset, duration and reversibility of the toxic effects, as well as toxicokinetics and metabolism, are still at the research level (17). In addition, there is ongoing research on quantitative structure activity relationship (QSAR) modelling (18, 19), and in the ACuteTox Integrated Project, with the prospect of an approach that makes use of information from 28-day repeated dose toxicity studies, when it eventually becomes available from ECVAM (20). For evaluating the usefulness and limitations of two specific in vitro cytotoxicity test methods, an international, multi-laboratory validation study was organised by ICCVAM/NICEATM and ECVAM. Three laboratories tested 72 reference substances for their effects on neutral red uptake (NRU) in BALB/c 3T3 mouse fibroblasts (3T3) and normal human epidermal keratinocytes (NHK). The resulting data were used to estimate the starting dose for rodent acute oral toxicity testing, based on linear regressions developed from the Registry of Cytotoxicity (RC) database (10). It was concluded that the 3T3 and NHK NRU test methods were not sufficiently accurate to predict acute oral toxicity for regulatory hazard classification, but that these in vitro methods might be used in a weight-of-evidence approach to determine the starting doses for the current acute oral toxicity protocols (i.e. the Acute Toxic Class [ATC] method). Furthermore, it is likely that the starting doses for substances with certain toxic mechanisms that were not expected to be active in 3T3 or NHK cells (e.g. those that are neurotoxic or cardiotoxic), would be underestimated by these basal cytotoxicity test methods (16). However, computer simulation of the ATC method testing showed that, for the substances tested in the validation study, use of the NRU test methods resulted in an average saving of 5 10% of the animals required per test. Although the 3T3 NRU test method was less reproducible than the NHK NRU test method, it resulted in a slightly greater reduction in the number of animals used, had a greater accuracy for prediction of the GHS acute oral toxicity category from the IC50, and performed better in the revised RC regressions evaluated for the prediction of the LD50 values (16). To contribute to animal welfare, and based on these ICCVAM recommendations, we implemented the Balb/c 3T3 NRU test method in the routine assessment of oral acute toxicity in rats. We collected in vitro and in vivo data for 187 substances. Retrospectively, we analysed the utility of the in vitro method for assessing oral acute toxicity. We estimated its usefulness for predicting the in vivo classification and for predicting the starting dose for the subsequent in vivo test. We also calculated the animal numbers required in each instance, in order to estimate possible reductions in animal use. Additionally, we tried to identify specific applicability domains, hoping to improve the predictivity of the cytotoxicity test. Materials and Methods The data set We collected data on the in vitro cytotoxicity and the acute oral toxicity in rats for 203 substances (16 substances from the Halle Register 2003 [10] and 187 in-house test substances, including a broad variety of chemicals, agrochemicals and formulations thereof), in order to compare the in vitro and in vivo results (Tables 1a and 1b). Of the substances, 70% were agrochemical active ingredients (AI) or formulations: 61 fungicide AI and 19 formulations, nine herbicide AI and nine formulations, 39 insecticide AI and 5 formulations. The remaining 40 chemicals comprised the 16 substances from the Halle Register 2003 and eight mixtures, as well as some polymers and dyes, additives and coatings. The substances from the Halle Register 2003 (Table 1a) were tested only in vitro, as the in vivo data already existed. For the others, both tests were performed with identical batches, except the substances which were tested pre-good Laboratory Practice (GLP), or as otherwise indicated in the table by superscript a. Acute oral toxicity In total, 203 acute oral toxicity studies in rats were compared to the parallel in vitro tests. Data for 16 test substances were obtained from literature (10), whereas 187 test substances were analysed in routinely-performed acute oral toxicity testing in vivo, according to the provisions of the German Animal Welfare Act and the European Council Directive

3 Refinement and reduction of acute oral toxicity testing 275 Table 1: General information and LD50 (acute, oral, rat), IC50 and predicted LD50 values, for the substances from the Halle Register 2003 and for the BASF in-house substances a) Substances with published in vivo and in vitro data, tested in vitro (n = 16) Substance information Halle Register 2003 BASF SE LD50 EU-GHS Predicted Predicted Physical MW (oral, R/M) category IC50 LD50 IC50 LD50 No. Name CAS No. form [g/mol] [mg/kg bw] (oral, R/M) [μg/ml] [mg/kg bw] [μg/ml] [mg/kg bw] 1 Nicotine liquid N-Methyl-N -nitro-n-nitrosoguanidine solid Acrylamide liquid p-cresol solid Phenol a liquid Aniline liquid ,2,4-Trichlorobenzene liquid Salicylic acid solid Acetylsalicylic acid a solid Ibuprofen solid Sodium dodecyl sulphate solid Toluene liquid n.req Diethyl phthalate b liquid n.req Captan solid ,010 n.req Ethanol a liquid ,008 n.req. 17, , ,2-Propandiol b liquid ,017 n.req. 26, , Data are sorted by LD50 value [mg/kg bw]. In vivo and in vitro data from the Halle Register 2003 (10), as well as our own in vitro results (BASF SE), are shown. MW = molecular weight; R/M = rat/mouse; n.req. = not required. a Recommended reference standards (ICCVAM Report 2006; 16); b Similar structure to recommended reference standards.

4 276 A. Schrage et al. Table 1: continued b) BASF in-house substances, tested in vitro and in vivo (n = 187) In vitro data In vivo data Recommended Substance information (Balb/c 3T3 NRU) (acute oral rat) starting dose Max. test Preci- Predicted GLP, EU-GHS By in Physical MW conc. pitate IC50 LD50 OECD LD50 category vitro By expert No. Group form [g/mol] [μg/ml] at [μg/ml] [μg/ml] [mg/kg bw] TG 423 [mg/kg bw] [mg/kg bw] testing judgement 17 chemical solid n.kn pre-glp fungicide solid < fungicide solid < insecticide solid about fungicide solid < insecticide solid < fungicide solid < fungicide solid < fungicide solid < fungicide solid < fungicide solid < fungicide liquid < fungicide solid < fungicide solid < insecticide solid > fungicide solid < fungicide solid > 100 > 586 < fungicide solid > 100 > 586 about insecticide solid > 100 > 586 < insecticide solid > chemical solid > chemical liquid > insecticide solid a fungicide solid n.kn > fungicide solid about herbicide solid pre-glp fungicide solid n.kn > fungicide solid > fungicide solid > herbicide solid > fungicide solid > fungicide solid < fungicide solid > 100 > 586 < fungicide solid n.kn > 100 > 586 > Data are sorted by EU-GHS category and predicted LD50 value [mg/kg bw]. Abbreviations: formul. = formulation; n.a. = not applicable; MW = molecular weight (rounded to 50 units); n.kn. = not known; n.req. = not required; pre-glp = The in vivo study was performed before 1989, regarded as a non-glp screening study. a The in vivo study was performed according to OECD TG 401, which was the test guideline previously in force.

5 Refinement and reduction of acute oral toxicity testing 277 Table 1: continued In vitro data In vivo data Recommended Substance information (Balb/c 3T3 NRU) (acute oral rat) starting dose Max. test Preci- Predicted GLP, EU-GHS By in Physical MW conc. pitate IC50 LD50 OECD LD50 category vitro By expert No. Group form [g/mol] [μg/ml] at [μg/ml] [μg/ml] [mg/kg bw] TG 423 [mg/kg bw] [mg/kg bw] testing judgement 51 fungicide solid > 100 > 586 > fungicide solid > 100 > 586 < insecticide liquid > 100 > 586 > fungicide solid > insecticide liquid about chemical liquid , pre-glp fungicide solid > fungicidal formul. liquid n.a > fungicidal formul. liquid n.a > , fungicidal formul. liquid n.a about fungicidal formul. liquid n.a > fungicidal formul. liquid n.a > fungicide solid > insecticide solid > insecticide solid > insecticide liquid > insecticide solid > fungicide solid > insecticide solid > fungicidal formul. solid n.a > insecticide solid > insecticide solid > insecticide solid > insecticide solid > fungicide solid > biocide liquid n.kn > fungicide solid > insecticide solid > insecticide solid > insecticidal formul. liquid n.a > fungicide solid > fungicide solid > fungicide solid > chemical solid n.kn > fungicide solid > Data are sorted by EU-GHS category and predicted LD50 value [mg/kg bw]. Abbreviations: formul. = formulation; n.a. = not applicable; MW = molecular weight (rounded to 50 units); n.kn. = not known; n.req. = not required; pre-glp = The in vivo study was performed before 1989, regarded as a non-glp screening study. athe in vivo study was performed according to OECD TG 401, which was the test guideline previously in force.

6 278 A. Schrage et al. Table 1: continued In vitro data In vivo data Recommended Substance information (Balb/c 3T3 NRU) (acute oral rat) starting dose Max. test Preci- Predicted GLP, EU-GHS By in Physical MW conc. pitate IC50 LD50 OECD LD50 category vitro By expert No. Group form [g/mol] [μg/ml] at [μg/ml] [μg/ml] [mg/kg bw] TG 423 [mg/kg bw] [mg/kg bw] testing judgement 86 fungicide solid > insecticide solid > insecticide solid > insecticide solid > insecticide solid > chemical liquid n.kn > fungicide solid > fungicide solid > fungicide liquid pre-glp ~ chemical solid > herbicidal formul. liquid n.a > fungicide solid > fungicide solid > fungicide solid > fungicide liquid > fungicide solid > insecticide solid > insecticide solid > fungicide liquid > fungicide solid > insecticidal formul. liquid n.a > fungicide solid > 100 > 586 > fungicide solid > 100 > 586 > fungicide solid > 100 > 586 > fungicide solid > 100 > 586 > insecticide solid > 100 > 586 > insecticide solid > 100 > 586 > insecticide liquid > 100 > 586 > insecticide solid > 100 > 586 > fungicide solid > herbicide solid pre-glp ~ fungicidal formul. liquid n.a > insecticide solid > insecticide solid > chemical solid > Data are sorted by EU-GHS category and predicted LD50 value [mg/kg bw]. Abbreviations: formul. = formulation; n.a. = not applicable; MW = molecular weight (rounded to 50 units); n.kn. = not known; n.req. = not required; pre-glp = The in vivo study was performed before 1989, regarded as a non-glp screening study. a The in vivo study was performed according to OECD TG 401, which was the test guideline previously in force.

7 Refinement and reduction of acute oral toxicity testing 279 Table 1: continued In vitro data In vivo data Recommended Substance information (Balb/c 3T3 NRU) (acute oral rat) starting dose Max. test Preci- Predicted GLP, EU-GHS By in Physical MW conc. pitate IC50 LD50 OECD LD50 category vitro By expert No. Group form [g/mol] [μg/ml] at [μg/ml] [μg/ml] [mg/kg bw] TG 423 [mg/kg bw] [mg/kg bw] testing judgement 121 chemical solid > chemical liquid , > herbicide solid n.kn pre-glp ~ fungicide solid > chemical solid > 1000 > > herbicidal formul. liquid n.a > insecticidal formul. liquid n.a > chemical liquid , > herbicide liquid ,660 n.kn a ~ fungicidal formul. liquid n.a > 2000 n.req fungicide solid n.kn a > 5000 n.req fungicidal formul. liquid n.a > 2000 n.req fungicide solid n.kn a > 5000 n.req fungicidal formul. liquid n.a > 2000 n.req fungicidal formul. liquid n.a > 2000 n.req insecticide solid > 2000 n.req chemical solid n.kn > 2000 n.req fungicidal formul. liquid n.a > 2000 n.req fungicide solid n.kn a > 5000 n.req chemical liquid n.kn > 2000 n.req fungicide solid > 2000 n.req fungicidal formul. solid n.a > 2000 n.req fungicide solid > 2000 n.req fungicidal formul. solid n.a > 2000 n.req fungicidal formul. liquid n.a > 2000 n.req mixture liquid n.a > 2000 n.req fungicide solid > 2000 n.req additive solid > 2000 n.req fungicidal formul. solid n.a > 2000 n.req additive liquid n.kn > 2000 n.req chemical liquid > 2000 n.req mixture liquid n.a > 2000 n.req fungicide solid n.kn a > 5000 n.req fungicide solid > 2000 n.req mixture (dye) liquid n.a > 2000 n.req Data are sorted by EU-GHS category and predicted LD50 value [mg/kg bw]. Abbreviations: formul. = formulation; n.a. = not applicable; MW = molecular weight (rounded to 50 units); n.kn. = not known; n.req. = not required; pre-glp = The in vivo study was performed before 1989, regarded as a non-glp screening study. a The in vivo study was performed according to OECD TG 401, which was the test guideline previously in force.

8 280 A. Schrage et al. Table 1: continued In vitro data In vivo data Recommended Substance information (Balb/c 3T3 NRU) (acute oral rat) starting dose Max. test Preci- Predicted GLP, EU-GHS By in Physical MW conc. pitate IC50 LD50 OECD LD50 category vitro By expert No. Group form [g/mol] [μg/ml] at [μg/ml] [μg/ml] [mg/kg bw] TG 423 [mg/kg bw] [mg/kg bw] testing judgement 156 additive solid > 2000 n.req chemical solid > 2000 n.req chemical solid > 100 > > 2000 n.req herbicide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req insecticide solid > 100 > 586 > 2000 n.req architectural coating liquid n.a > 2000 n.req fungicide solid > 2000 n.req fungicidal formul. liquid n.a > 2000 n.req insecticidal formul. liquid n.a > 2000 n.req architectural coating solid n.a > 5000 n.req fungicidal formul. liquid n.a > 2000 n.req mixture (adhesive) solid n.a > 2000 n.req herbicide solid n.kn a > 2000 n.req chemical solid > 2000 n.req herbicidal formul. liquid n.a > 2000 n.req herbicide solid n.kn a > 5000 n.req chemical solid > 2000 n.req architectural coating solid n.a > 2000 n.req additive solid > 1000 > > 2000 n.req additive liquid n.a > 1000 > > 2000 n.req herbicide solid > 1000 > > 2000 n.req herbicidal formul. solid n.a > 1000 > > 2000 n.req insecticidal formul. solid n.a > 1000 > > 2000 n.req mixture liquid n.a > 1000 > > 2000 n.req fungicidal formul. liquid n.a > 2000 n.req architectural coating liquid n.a > 2000 n.req chemical solid n.kn > 2000 n.req herbicidal formul. liquid n.a > 2000 n.req chemical liquid n.a > 2000 n.req Data are sorted by EU-GHS category and predicted LD50 value [mg/kg bw]. Abbreviations: formul. = formulation; n.a. = not applicable; MW = molecular weight (rounded to 50 units); n.kn. = not known; n.req. = not required; pre-glp = The in vivo study was performed before 1989, regarded as a non-glp screening study. a The in vivo study was performed according to OECD TG 401, which was the test guideline previously in force.

9 Refinement and reduction of acute oral toxicity testing 281 Table 1: continued In vitro data In vivo data Recommended Substance information (Balb/c 3T3 NRU) (acute oral rat) starting dose Max. test Preci- Predicted GLP, EU-GHS By in Physical MW conc. pitate IC50 LD50 OECD LD50 category vitro By expert No. Group form [g/mol] [μg/ml] at [μg/ml] [μg/ml] [mg/kg bw] TG 423 [mg/kg bw] [mg/kg bw] testing judgement 191 chemical liquid n.a > 2154 > > 2000 n.req polymer liquid n.a > 2154 > > 2000 n.req mixture (dye) solid n.a > 2154 > > 2000 n.req herbicidal formul. liquid n.a > 2154 > > 2000 n.req herbicidal formul. liquid n.a > 2154 > > 2000 n.req chemical solid > 2154 > > 2000 n.req mixture (yeast extract) liquid n.a > 2154 > > 2000 n.req polymer liquid n.a > 2154 > > 2000 n.req herbicidal formul. solid n.a. 10, > 2000 n.req architectural coating liquid n.a > 2000 n.req herbicidal formul. liquid n.a. 10, > 2000 n.req chemical liquid > 4642 > > 2000 n.req polymer liquid 14,000 10, > 2000 n.req Data are sorted by EU-GHS category and predicted LD50 value [mg/kg bw]. Abbreviations: formul. = formulation; n.a. = not applicable; MW = molecular weight (rounded to 50 units); n.kn. = not known; n.req. = not required; pre-glp = The in vivo study was performed before 1989, regarded as a non-glp screening study. a The in vivo study was performed according to OECD TG 401, which was the test guideline previously in force.

10 282 A. Schrage et al. 86/609/EEC, in our AAALAC certified laboratory or at a partner institute located in Germany. Eighty-seven studies were performed for registration purposes, according to the OECD TG for acute oral toxicity testing, under GLP conditions, and the remaining 100 were screening studies. All in vivo testing was performed for regulatory or other safety evaluation purposes and no additional in vivo studies were specifically performed to provide data for this analysis. Testing according to OECD TG 423 determines the acute oral toxicity of test substances after a single administration in rats (7; 21 23). Briefly, one or several fixed doses of a test substance (5, 50, 300, 2000mg/kg bw) were administered by gavage to young adult Wistar rats (Crl:WI [Han]; Charles River Wiga GmbH, Sulzfeld, Germany) by using a step-wise procedure, with three animals used per step. Starting doses were determined by expert judgement, based on information about the substance class or comparable formulations. Sub - sequent dose levels were based on the decision-tree in Annex 2 of OECD TG 423. For 75 of the 87 substances tested for registration purposes under GLP conditions, the starting doses were proposed by the same expert. These tests were therefore used for comparison of the number of animals needed when the starting dose was determined either by expert judgement or by using the in vitro cytotoxicity test, and the potential reduction in animal use was calculated. Since the guideline is not intended to allow the calculation of a precise LD50, but does allow for determination of a range of exposures where lethality is expected, the estimated LD50 ranges were used for ranking the compounds based on the classification and labelling categories of the GHS. Eight of the 87 substances were tested according to the formerly valid OECD TG 401, which did not recommend the use of fixed doses. Ninety-four test substances were tested for their acute oral toxicity in screening studies (under non- GLP conditions in this case), following, in principle, the OECD TG 423 procedure. We used the same rat strain of the same age, and also used the recommended animal numbers (three per dose level). Administration was via gavage and the observation period lasted for 14 days. Starting dose levels were again determined by expert judgement, based on information about the substance class or comparable formulations, but since these studies were performed for development purposes, the starting or subsequent dose levels were, in some cases, also based on estimated risk assessment (exposure) or were limited by substance availability. In these cases (substances No. 47 and 51 [LD50 > 50mg/kg bw], and 42 of the substances No [LD50 > 300mg/kg bw]), testing was performed up to a cut-off value, and was not continued to either a higher or a lower level, if the test substance was not considered suitable for further development. Six other substances (used for inhouse validation of the cytotoxicity test) were tested before 1989 (pre-glp), and therefore were also regarded as substances tested in screening tests under non-glp conditions. Classification The classifications used for comparison of in vitro and in vivo results were Categories 1 4 and not required for the highest class (Category 5). Based on animal welfare considerations, testing for Category 5 (testing at 5000mg/kg bw) according to EU-GHS, is not performed (GHS: Third revised edition, section [g]; 6). Moreover, there were no test substances falling into Category 1 in our data set, so Category 1 (< 5mg/kg bw) and Cat - egory 2 (5 50mg/kg bw) were grouped together. Balb/c 3T3 NRU cytotoxicity test The NRU test was performed according to the ICCVAM 2006 Report (16), in our GLP-certified laboratory, except that, unless otherwise noted, the materials were obtained from Biochrom, Berlin, Germany or Sigma-Aldrich, Steinheim, Germany. Balb/c 3T3 fibroblasts (clone A31; ECACC, Salisbury, Wiltshire, UK) were cultured for 24 hours in cdmem (Dulbecco s Modified Eagle s Medium [DMEM], complemented with 10% [v/v] newborn calf serum, 4mM L-glutamine, 100IU penicillin and 100μg/ml streptomycin), in 96-well plates at 37 C and 5% (v/v) CO 2. Usually, the maximal concentration for screening test substances was 100μg/ml, and 2154μg/ml for GLP registration test substances. These concentrations corresponded to predicted in vivo LD50 values of 586 and 1836mg/kg bw, respectively. Eight selected concentrations, with six replicates per concentration, were tested to obtain a reliable concentration effect curve. The substances were diluted in DMEM, dimethyl sulphoxide (DMSO), tetrahydrofuran (THF) or ethanol (the latter both from Riedel-de Haen, Seelze, Germany), with a maximal solvent concentration of 0.5% (v/v). After a 48-hour incubation, the medium containing the test substance was removed and the cells were rinsed with PBS, then incubated with 50μg/ml neutral red for 3 hours at 37 C, 5% CO 2. Another rinse with PBS was followed by a 10-minute lysis step with neutral red desorption solution (deionised water with 50% [v/v] ethanol and 1% [v/v] glacial acetic acid; Merck, Darmstadt, Germany). The OD of the sample wells at 550nm, was determined with a Wallac 1420 multilabel counter (Perkin Elmer, Waltham, MA, USA), and the relative cell viability was calculated as a percentage of the negative control (vehicle control =

11 Refinement and reduction of acute oral toxicity testing %). Finally, the concentration which led to a 50% reduction in cell growth compared to the control, the IC50 (IC = Inhibiting Concentration, in μg/ml or mmol/l), was estimated. Table 2: Calculation of the number of animals required, according to the starting dose used and the LD50 value Calculations and statistics The predicted LD50 for lethal oral toxicity in rats was initially described for substances with known molecular weights (10): log LD50 [mmol/kg] = log IC50 [mm] In order to develop a prediction model that would be applicable to mixtures or other substances without known molecular weights, the data forming the basis for this millimole regression were converted to a weight basis (16): log LD50 [mg/kg] = log IC50 [μg/ml] For the estimation of the reduction in animal use, the theoretical animal numbers required in the GLP registration studies and non-glp screening studies are shown in Table 2. To address different dose response curves, the animal numbers used for GLP registration studies were not only estimated according to the minimum number of animals needed to obtain the final in vivo result, but also by using a probit model based on a logarithmic dose scale, with slopes of 2 and 8, in line with the ICCVAM 2006 Report (16). Slope 2, in general, led to slight increases in the number of animals required, but did not change the overall conclusion. Slope 8 led to nearly the same results as the minimum animal numbers required for use in the step-wise procedure outlined in OECD TG 423 (Table 2a). Thus, for further evaluation (i.e. for the screening studies), only the latter calculation was used. For the substances that were tested in GLP registration studies, and for which the expert proposal of a starting dose was available (n = 75), we calculated the absolute number of animals required: a) when a default starting dose of 2000, 300 or 50mg/kg bw was used; b) when the starting dose was predicted by using the in vitro cytotoxicity test; or c) when the starting dose was proposed by an experienced toxicologist. The estimated reduction in the number of animals required, was calculated in each case. The expert proposal was made by a toxicologist with more than 15 years of experience in acute toxicity testing and a background in evaluating BASF test substances, by taking into account all available test substance information, i.e. comparable formulations, structure similarities, etc. Nearly all the calculations and graphical presentations were performed with Microsoft Office Excel a) As per OECD TG 423 Starting dose [mg/kg bw] LD50 > [mg/kg bw] > > b) Slope 2 Starting dose [mg/kg bw] LD [mg/kg bw] c) Slope 8 Starting dose [mg/kg bw] LD [mg/kg bw] The number of animals for GLP registration and non- GLP screening studies were estimated by a) calculating the minimum number of animals needed to achieve the final in vivo result, as outlined in TG 423, and also by using a probit model based on a logarithmic dose scale with b) a slope of 2, and c) a slope of 8, as described in the ICCVAM 2006 Report (16) For correlations, Pearson s correlation coefficient and R 2 were determined. The scatterplots for comparing IC50, predicted LD50 and LD50 obtained in vivo (Figures 1b and 2) were conducted with a 30- Day Demo Version of GraphPadPrism 5 (September 2010: The log P OW was calculated with the SPARC on line calculator v4.5 (September 2010: edu/sparc/) for ph 7, as this is close to the ph of the culture medium in vitro (i.e. ph 7.4; data not shown).

12 284 A. Schrage et al. Figure 1: A comparison of the in vitro data with published in vitro results and in vivo hazard categories 10 5 a) IC50 values [μg/ml] from the Halle Register IC50 values [μg/ml] calculated by BASF SE 10 4 b) = Halle Register; = BASF SE predicted LD50 [mg/kg bw] ( 50) 3 (> ) 4 (> ) in vivo hazard categories (LD50 [mg/kg bw]) 5 (> 2000) The reliability of the cytotoxicity test in our laboratory was confirmed with 16 test substances from the Halle Register 2003 (10). a) The correlation of published IC50 values with our own values (Pearson s correlation coefficient = ). b) A comparison of the predicted LD50 values with the in vivo results. The substances were grouped by the in vivo acute, oral hazard categories (31) and then plotted against the respective predicted LD50 values derived from the in vitro IC50 values. The grey shading shows the correct predictions. Results At the time of analysis, a data set of 203 substances, which were tested in the NRU cytotoxicity test and the acute oral toxicity test in rats, was available in our laboratory. Tables 1a and 1b show the in vivo and in vitro results for all the substances, in addition to some substance information, such as physical form, molecular weight and log P OW. Calculating the predicted LD50 value was

13 Refinement and reduction of acute oral toxicity testing 285 Figure 2: A comparison of the predicted LD50 values with the in vivo hazard categories 10 4 a) predicted LD50 [mg/kg bw] ( 50) 3 (> ) 4 (> ) in vivo hazard categories (LD50 [mg/kg bw]) 5 (> 2000) 10 4 b) predicted LD50 [mg/kg bw] ( 50) 3 (> ) 4 (> ) in vivo hazard categories (LD50 [mg/kg bw]) 5 (> 2000) 10 4 c) predicted LD50 [mg/kg bw] ( 50) 3 4 (> ) (> ) in vivo hazard categories (LD50 [mg/kg bw]) 5 (> 2000) a) All substances (n = 203); b) the GLP registration studies (n = 87); c) the non-glp screening studies (n = 100). The substances were grouped by the in vivo acute, oral hazard categories (31) and then plotted against the respective predicted LD50 values derived from the in vitro IC50 values. The grey shading shows the correct predictions.

14 286 A. Schrage et al. generally done by using Halle s mass-based formula, as the molecular weight was not applicable for some test substances, e.g. the formulations. Nevertheless, a molar analysis was performed for the substances with a known molecular weight. Reproducibility and reliability of the cytotoxicity test To demonstrate our technical proficiency with the 3T3 NRU test method, we tested 16 substances with published IC50 and LD50 values (oral, rat/mouse, [mg/kg bw]; 10): one EU-GHS Category 2, three Category 3, seven Category 4, and five unclassified substances (Table 1a). Three of these can also be found in ICCVAM s list of recommended Reference Standards, and two of the substances are structurally similar to two other substances in this list (16). Our IC50 results were in very good correlation with the published IC50 values, with a Pearson s correlation coefficient of (Figure 1a). Comparison with the in vivo results showed a good accuracy of 69% (11/16), with only two under-predicted and three over-predicted substances, i.e. 13% or 19%, respectively (Figure 1b). Overall concordance of the in vitro and the in vivo tests Of all the tested substances, 39% (79/203) were classified as virtually non-toxic in vivo (LD50 > 2000mg/kg bw, no category according to EU-GHS), and 39% (80/203) as weakly toxic (> mg/kg bw, Cat. 4). Only about one fifth of all the substances were identified as moderate to very toxic: 12% (24/203) with an LD50 > mg/kg bw (Cat. 3); and 10% (20/203) with an LD50 50mg/kg bw (Cat. 1 2). In contrast, the in vitro test identified 71% (145/203) of all the tested substances as weakly toxic (Cat. 4). Of the remainder, 4% (8/203) were virtually non-toxic (no category), 24% (48/203) were toxic (Cat. 3), and 1% (2/203) were very toxic (Cat. 1 2). Therefore, the cytotoxicity assay showed a good prediction only for the weakly toxic substances (Cat. 4), with a concordance of 74% (59/80), but not for the other classes, resulting in a low overall concordance of 35% (71/203; Table 3a and Figure 2a). In contrast, the overall concordance for only the 16 substances from the Halle Register 2003 was rather good, at 69% (11/16; Table 3a). Subgrouping the in-house substances by in vivo registration or screening studies showed that the GLP registration testing was mainly for non/weakly toxic substances, whereas the toxic substances were detected mainly in non-glp screening studies, which are performed in an early phase of product development (Figures 2b and 2c). The utility of predicting the in vivo starting dose Selecting a starting dose that matches the in vivo classification of a substance, reduces the use of animals in the actual in vivo study, independent of the study conditions (i.e. registration testing or screening study). As mentioned above, this was the case for only 35% (71/203) of the substances. For substances with an LD mg/kg bw (Cat. 1 4), selecting a starting dose one category higher or lower than the actual in vivo LD50 may still save some animals. In this regard, prediction of the starting dose by using the in vitro cytotoxicity test was useful for 58% (118/203) of the substances (Tables 1a and 1b). However, the use of a standard starting dose of 300mg/kg bw by default (without previous cytotoxicity testing) would have been almost as useful (50%, 102/203). The prediction of the starting dose by an experienced toxicologist was correct for 95% (71/75) of the respective substances. However, this was only performed for 36% of the substances, which were mainly of no toxicity to low toxicity. Estimation of animal savings For GLP registration studies, the minimum number of animals needed for the acute oral tests (ATC Method; OECD TG 423) was calculated to be 516, if the category was always predicted correctly. This required number was compared to that when an expert proposal of the starting dose was available (n = 75). Additionally, we calculated the number of animals we would have used, if we had selected a default starting dose of 2000, 300, or 50mg/kg bw, or if we had made use of the starting dose predicted by the in vitro cytotoxicity test (Table 4 and Figure 3). It can be clearly seen for the GLP registration studies that the use of a starting dose of 2000mg/kg bw by default, resulted in the same number of animals being used per test (6.9) as the calculated minimum number (if the starting dose was correctly chosen, the ATC requires a minimum of six animals for LD50 > 2000mg/kg bw, nine animals for > mg/kg bw and nine animals for > mg/kg bw; 100% predictivity). Thus, this approach results in the use of the lowest number of animals, when it is applied in our laboratory. This is due to the fact that the prevalence of the > 2000mg/kg bw category within the test substances for the GLP registration studies is very high. If the actual starting doses for the animal-based studies were selected by expert judgement, then about the same number of animals per test (7.3) was used. If the starting doses were selected based on the in vitro cytotoxicity test results, then this would have resulted in the use of 9.1 animals per test (a total of

15 Refinement and reduction of acute oral toxicity testing 287 Table 3: A comparison of the in vivo hazard category and the predicted LD50 values of all tested substances, substances from the Halle Register 2003 only, and BASF in-house substances only a) All tested substances (n = 203) LD50 rat, oral [mg/kg bw] 50 > > > 2000 EU-GHS Cat. Cat. 1 2 Cat. 3 Cat. 4 n.req. Total Total Category Predictivity Category OP UP Predicted > % 75% 25% LD50 > % 41% 23% 37% [mg/kg bw] > % 13% 8% 79% % n.calc. n.calc. Total Total 10% 12% 39% 39% Accuracy 0% 25% 74% 8% Toxicity OP 0% 25% 92% Toxicity UP 100% 75% 1% b) Substances from the Halle Register 2003 (n = 16) LD50 rat, oral [mg/kg bw] 50 > > > 2000 EU-GHS Cat. Cat. 1 2 Cat. 3 Cat. 4 n.req. Total Total Category Predictivity Category OP UP Predicted > % 100% 0% LD50 > % 64% 18% 18% [mg/kg bw] > % 67% 0% 33% % n.calc. n.calc. Total Total 6% 19% 44% 31% Accuracy 0% 67% 100% 40% Toxicity OP 0% 0% 60% Toxicity UP 100% 33% 0% The in vivo acute, oral hazard category (31) was compared to the predicted LD50 derived from the in vitro IC50 for all test substances. The absolute numbers of correct, under-estimated or over-estimated predictions are shown. The bold numbers in the table show the number of correct predictions. OP = over-predicted; UP = underpredicted; n.req. = not required; n.calc. = not calculated.

16 288 A. Schrage et al. Table 3: continued c) BASF in-house substances (n = 187) LD50 rat, oral [mg/kg bw] 50 > > > 2000 EU-GHS Cat. Cat. 1 2 Cat. 3 Cat. 4 n.req. Total Total Category Predictivity Category OP UP Predicted > % 67% 33% LD50 > % 39% 23% 38% [mg/kg bw] > % 9% 9% 82% % n.calc. n.calc. Total Total 10% 11% 39% 40% Accuracy 0% 19% 71% 5% Toxicity OP 0% 27% 95% Toxicity UP 100% 81% 1% The in vivo acute, oral hazard category (31) was compared to the predicted LD50 derived from the in vitro IC50 for all test substances. The absolute numbers of correct, under-estimated or over-estimated predictions are shown. The bold numbers in the table show the number of correct predictions. OP = over-predicted; UP = underpredicted; n.req. = not required; n.calc. = not calculated.

17 Refinement and reduction of acute oral toxicity testing 289 Figure 3: An estimate of the number of animals needed for the GLP registration studies number of animals mg/kg bw 300 mg/kg bw 50 mg/kg bw expert judgement cytotoxicity test starting dose set at or determined by For the 75 substances which were tested in the GLP registration studies, and for which expert proposal of the starting doses was available, the total absolute number of animals was calculated when either a default starting dose of 2000, 300 or 50mg/kg bw was used, or when the predicted starting dose was derived from the in vitro cytotoxicity test or proposed by an experienced toxicologist. 165 additional animals used in the in vivo studies for the 75 substances), as compared with the use of a default starting dose of 2000mg/kg bw. However, selecting a default starting dose of 300mg/kg bw or 50mg/kg bw would have increased the number of animals used in the in vivo studies. As shown in Tables 2b and 2c, we also estimated the expected numbers of animals required for some assumed LD50 values, with a slope of 2 or 8, which is in line with the calculations described in the ICCVAM 2006 Report (16). Slope 2, in general, led to slight increases in the calculated animal numbers, but did not change the overall conclusion, while slope 8 led to nearly the same results as when the minimum animal numbers in the stepwise procedure outlined in OECD TG 423 were applied (Table 2a). Thus, for further evaluation (i.e. screening studies), only the minimum animal numbers were used. For the screening studies, in vivo testing in many cases was not continued up to the highest category, but was stopped at a cut-off level. Since 42 of the 100 screened substances fell into the category of > 300mg/kg bw, two scenarios were proposed for the calculation of the animal numbers. In the first scenario, it was assumed that all > 300mg/kg bw substances would have been in the > 2000mg/kg bw category. In the second scenario, it was assumed that these substances would have been in the > mg/kg bw category (Table 5 and Figure 4). By using this estimation model, the animal numbers based on the first scenario were lower when starting with 2000mg/kg bw, while, for the second scenario, they were lower when starting with 300mg/kg bw. Prediction of the starting dose by using the in vitro cytotoxicity assay would have led to higher animal numbers in both scenarios (an additional 34% or 10%, respectively). Specific characteristics and applicability domains As the cytotoxicity test was not, in general, very predictive, we analysed specific characteristics of the test substances, in order to find applicability domains for: a) limitation of the test substance concentration; b) molecular weight (MW)-based predicted LD50, thereby excluding formulations; c) log P OW ; and d) substance class or mode of action (MoA). One parameter that might influence in vitro cytotoxicity testing, is a limitation of the highest effective concentration of a substance that is available for testing, because of its poor solubility. Another reason for limiting the highest concentration for testing, which applied to some of our screening study substances, is that a restricted amount of a test substance was available for analysis. For either one or both of these reasons, the highest concentrations available for use in in vitro testing were limited for 40 of the test substances (Tables 1b and 6). A dose response relationship is necessary for the calculation of IC50 values [μg/ml]. However, no