ESAC PEER REVIEW. Corrected Version

Transcription

1 EUROPEAN COMMISSION DIRECTORATE GENERAL JRC JOINT RESEARCH CENTRE Institute for Health and Consumer Protection European Centre for the Validation of Alternative Methods (ECVAM) ESAC PEER REVIEW Corrected Version Proposal for a new testing strategy (Step-down approach) to reduce the use of fish in acute aquatic toxicity testing Results of the retrospective application of the step-down approach to acute aquatic toxicity data of new and existing chemicals, plant protection products and pharmaceuticals July 2005 (corrected in November 2005)

2 CONTENTS 1. Introduction 2 2. Current testing requirements and fish acute toxicity testing 2 3. Basis of the new testing strategy (Step-down Approach) 3 4. Background to the analysed databases The New Chemicals Database (NCD) The Existing Chemicals Database (IUCLID) Plant protection products: Active substances and metabolites Pharmaceuticals for human use 6 5. Data analysis Data precision Data selection Calculation of fish needed according Annex V C.1/OECD Calculation of fish needed by applying the step-down approach to 9 data sets 5.5. Calculation of possible reduction 9 6. Results and possible reduction The New Chemicals Database (NCD) The Existing Chemicals Database (IUCLID) Plant protection products: Active substances and metabolites Pharmaceuticals for human use Discussion and conclusions References 23 Annex (includes explanations to extracted data and calculation of reductions) 24 1

3 1. Introduction In 2003, ECVAM established a taskforce on ecotoxicology. One aim of the taskforce is to evaluate possibilities to reduce, refine and replace the use of fish in ecotoxicity testing. In 2004, Tom Hutchinson, a member of the taskforce, presented his work on a new testing strategy, which could reduce the number of fish used for acute toxicity testing of pharmaceutical for human use by 73% (Hutchinson et al, 2003). It was decided to evaluate whether this strategy could be applicable for the testing of other substances, e.g. chemicals and plant protection products. During the course of last year, ECVAM and the ECB staff jointly worked on the retrospective analysis of the data in the New Chemicals Database (NCD) and existing chemicals database (IUCLID). In addition, the 91 pharmaceuticals of the Hutchinson et al paper were evaluated applying more stringent data selection criteria. Data on new plant protection products were derived from the reports published on the DG SANCO website. The results of the evaluation of the NCD were recently published (Jeram et al., 2005). The results of the evaluation of the existing chemicals and active substances/metabolites of plant protection products have not been published. The report was compiled by Sonja Jeram (ECVAM), who also carried out the evaluation, in collaboration with Marlies Halder (ECVAM) and Juan Riego Sintes (ECB). Tom Hutchinson (Member of the ECVAM Taskforce), who first described the step-down approach, gave scientific advice and provided the data on the pharmaceuticals. Further scientific advice was obtained from Birgit Sokull-Kluettgen (ECB) and Joaquin Baraibar Fentanes (ECB) and Stefan Scheer (ECB) extracted the data from NCD and IUCLID. 2. Current testing requirements and fish acute toxicity testing According to the current EU regulatory requirements, acute aquatic toxicity is determined by three endpoints, namely algae EC50 72h (Directive 67/548/EEC, Annex V C.3 OECD TG 201), daphnia EC50 48h (Annex V C.2, OECD TG 202), and fish LC50 96h (Annex V C.1, OECD TG 203) and these base set data are stipulated for all substances for which the marketing quantity exceeds one tonne per year per manufacturer (Annex VII A of the Directive). The current fish acute toxicity test determines the median lethal concentration for fish after 96 hours of exposure to a range of concentrations of the test substance (LC50 96h). It recommends the use of a minimum of seven fish per concentration and control(s) (Table 1). In practice, regulatory studies conducted by industry laboratories often use five exposure concentrations plus a dilution water control, giving a total of 42 fish per test. This number excludes any previous range finding test such as using five fish in a single concentration plus five control animals, where the concentration is usually selected on the basis of the physicalchemicals properties (e.g. log Kow or water solubility) of the testing substance, and previous testing of similar substances (read across). The guidelines also give the possibility to perform a limit test. This is normally used for substances being not toxic in pre-testing. The limit test is performed with seven fish (14 fish including the control) at a single concentration, which is often 100 mg/l. No fish mortality at this concentration would mean that no further tests need to be performed and the acute fish test (LC50 >100 mg/l) would not lead to classification of the tested substance. 2

4 Table 1: Comparison of the acute fish toxicity test and the acute threshold (step-down) test (adapted from Hutchinson et al, 2003) Species Acute fish toxicity test (LC50 fish 96h)* Standard freshwater and marine fish species Acute threshold (step-down) test Standard freshwater and marine fish species Age of test organisms Adults or juveniles Adults or juveniles No. of fish per test vessel Minimum of 7 5 No. of test concentrations No. of replicate test vessels per test concentration 5 minimum plus appropriate controls 1 1 No. of fish per study Typically 42 (35 exposed + 7 in control) 1 minimum plus appropriate controls Proposed 10 (5 exposed + 5 in control) Dilution series 1.0, 1.8, 3.2, 5.6, 10 Not applicable Sequential step down series Not applicable Factor of 3.2 step-down Chemical exposure duration 96 hours 96 hours (or less**) Endpoints Mortality within 96 h Mortality or severe morbidity within 96 h** Test acceptability controls Mortality: 1 out of 7 fish Mortality or severe morbidity: 1 out of 5 fish Calculation of LC 50 Probit, Moving Average Angle or Binomial If required, estimated by binomial method of interpolation. * = Annex V C.1 and OECD TG 203 give the possibility to perform a limit test. This is normally used for substances being not toxic in pre-testing. The limit test is performed with seven fish (14 fish including the control) at a single concentration (see above) ** = Hutchinson et al (2003) propose to terminate the test when >= 2 fish die or show severe signs of mortality after 24 h of exposure, thus offering a possible refinement. 3. Basis of the new testing strategy (step-down approach) The step-down approach was developed based on the observation that, for most of the substances, the fish species tested are not more sensitive in acute aquatic toxicity testing than algae or daphnia (Weyers et al., 2000; Hutchinson et al., 2003). The testing would be performed in two steps: As shown in Fig. 1, tests on algae and daphnia are performed first, these results are then compared and the lowest value is set as the upper threshold concentration (UTC). At this concentration, a threshold test is carried out using five test fish and five control fish. In case that no mortality is observed, no further tests are carried out and the acute fish toxicity result (LC50) is reported as greater than (>) the UTC value. In case that mortality is observed, the second test is performed at a step-down concentration using a dilution factor of 3.2. This factor was chosen based on a semi-logarithmic concentration series (1.0, 3.2, 10, 32, etc.) as proposed by Doudoroff et al. (1951) and Hutchinson et al. (2003). The testing continues to lower concentrations until no mortality is observed. The LC50 96 h value can be obtained from all step-down threshold test data by applying the binominal method of interpolation (Rand and Petrocelli, 1985; Stephan, 1977). 3

5 In addition to the reduction, Hutchinson et al propose to terminate the test by humanely killing the fish when more than two fish die or show severe morbidity and thus reducing pain and distress to the fish (refinement alternative). More details on the step-down approach are included in Hutchinson et al. (2003) and Jeram et al. (2005). Both articles are attached to this report. Figure 1: Step-down approach FIRST STEP: Toxicity testing with algae and daphnia Perform Algae inhibition test (C.3 / TG 201) e.g. EC50 72h = 5 mg/l Perform acute toxicity test for daphnia (C.2 / TG 202) e.g. EC50 48h = 50 mg/l The lowest value is selected e.g. 5 mg/l Upper Threshold Concentration (UTC) SECOND STEP: Toxicity testing with fish Fish acute threshold test performed at single test concentration Upper Threshold Concentration (e.g. UTC = 5 mg/l) Test result no mortality mortality Fish acute threshold test performed at step-down concentration i.e. previous testing concentration diluted by factor 3.2 mortality Test result no mortality END of testing 4

6 4. Background to the analysed data 4.1. The New Chemicals Database (NCD) The New Chemicals Database (NCD) includes summary notification dossiers for all notified new substances put on the market after September 1981 according to Directive 67/548/EEC and subsequent amendments (European Commission, 1967, 1992). The NCD is hosted by the European Chemicals Bureau (ECB). Data in the NCD have high quality since toxicity tests are performed under GLP conditions according to the internationally accepted testing guidelines (Annex V or OECD). The database is continuously updated with dossiers received from the Competent Authorities of the EU Member States and Norway. To date the NCD contains over 6000 notification dossiers referring to more than 4000 substances which are fully available for the experts of the Competent Authorities, however, for reasons of confidentiality of data provided by the notifiers, the substance identity cannot be generally disclosed. Accordingly, all data extracted from the database and used in this study were used in a coded anonymous way. In order to avoid the evaluation of multiple data sets for a single substance, only data from the leading dossiers were taken and only in case of their incompleteness, data from additional notifications were used. The leading dossier corresponds usually to the first notification of a new substance. In general, this dossier includes most data and is used for the evaluation of the substance. Additional notifications of the same substance may refer to this dossier. For a total of 1,439 substances, data on acute toxicity for algae, daphnia and fish were available (examples are given in Table 2). When more than one test result is reported for the same endpoint, the lowest value is normally selected for hazard and risk assessment. In the case of the algae test, both biomass (EbC50) and growth rate (ErC50) values were reported. Therefore, also for our analysis, the lower of the two values was used. Table 2: Examples of data sets (NCD) Substance Algae EbC50 72h [mg/l] Algae ErC50 72h [mg/l] Daphnia EC50 48h [mg/l] Fish LC50 96h [mg/l] > > 80.6 > > 0.2 > 0.2 > 100 > > 258 > > 0.4 > > 18 6 > 0.2 > 0.2 > > > > 100 > > 40 5

7 4.2. The Existing Chemicals Database (IUCLID) Aquatic toxicity data for algae, daphnia and fish were extracted from the Existing Chemicals Database (IUCLID) hosted at the ECB. Test quality of data is not always reported. Only few results in the database were obtained in GLP laboratories but most of them according to the internationally accepted testing guidelines. For 316 substances, data for all three endpoints were available and used for the evaluation. In case that several test results were reported per CAS number, the lowest values were selected for the evaluation. In contrast to the NCD database, the results for algae were not specified regarding the biomass or growth rate Plant Production Products: active substances and metabolites (PPP) Aquatic toxicity data for algae, daphnia and fish were extracted from the technical review reports available on the web site of DG SANCO (Health and Consumer Protection). Data were selected from the reports on new plant protection products: Data for active substances were used and in case where the three endpoints were available, data for metabolites were also considered in the evaluation. A total of 68 substances and metabolites were evaluated. The reports do not include information on data quality or the testing guidelines followed. For the algae test, both the biomass and growth rate results are available. For few cases, the test duration differs from the OECD/Annex V guideline/method. As for the NCD data, the lowest value reported was used in the evaluation Pharmaceuticals for human use (PHARM) Acute aquatic toxicity data for algae, daphnia and fish of 91 active pharmaceutical substances were taken from the publication of Hutchinson et al. (2003) in order to analyse them further. The data originate from various pharmaceutical companies and are coded. 6

8 5. Data analysis 5.1. Data precision For a number of substances, included in the evaluation, the test results were not given as precise figures but were reported as greater than (>) or lower than values (<). Since these values might not always be suitable for the evaluation as explained below (data selection), the selection of data was performed in two steps, which allowed a more precise evaluation and resulted in three data sets for NCD, IUCLID, PPP and PHARM (see below) Data selection Taking into account the above mentioned, data were selected in three steps, applying more and more stringent criteria: Data set I Data set I corresponds to the complete set of extracted data. For the analysis, the test results reported as greater than (>) or lower than (<) were treated as if they were precise values. Data set II For data set II, only substances with at least one precise value for daphnia or algae test were selected, which allowed a precise setting of the theoretical UTC (see Fig 1). Additional substances were excluded, when imprecise values for the LC50 fish did not allow the accurate comparison with the UTC value, i.e., the fish test result is given as greater than (>) but the value alone is lower than the UTC, for example LC50>10 mg/l and UTC=20 mg/l. In this case, it is not clear whether fish toxicity is higher or lower than daphnia or algae toxicity. In some cases, it was not possible to calculate the number of necessary step-down tests, since the fish data were given as lower (<) than values. Data set III For data set III, only substances with precise values for both algae and daphnia test results, were selected. Table 3 gives examples of data and their assignment to data sets and Table 4 lists the three data sets for the four data bases. 7

9 Table 3: Selection of substances according to data precision Example Algae EbC50 72h [mg/l] Algae ErC50 72h [mg/l] Daphnia EC50 48h [mg/l] Fish LC50 96h [mg/l] Precise result Substances included in Data sets I, II or III 1 > 0.2 > 0.2 > 100 > 100 None I 2 > 0.2 > 0.2 > > None I > Algae I, II > 80.6 > 80.6 Algae I, II > 258 > 100 Algae I, II 6 > 0.4 > > 18 Daphnia I, II 7 > 100 > > 40 Daphnia I, II > Algae and Daphnia Algae and Daphnia Algae and Daphnia I, II, III I, II, III I, II, III Table 4: Number of substances in Data set I, II and III for the four databases Database Data set I Data set II Data set III NCD IUCLID PPP PHARM

10 5.3 Calculation of fish needed according Testing method C.1 or OECD TG203 For the fish LC50 96h fish (7-10 x 5 concentrations; 7-10 control fish) are used. We based our calculations on 42 fish. The current legislation allows to performing the limit test, which is normally used for substances being not toxic in pre-testing. The limit test is performed with 14 fish (7 test and 7 control fish) at a single concentration, which is often 100mg/L and the result would then be reported as greater than, i.e. LC50 >100mg/L, which would not lead to classification and labelling Calculation of fish needed by applying the step-down approach to data sets 10 (5 test and 5 control) fish are needed per threshold/step-down test Number of necessary step-down tests was calculated by applying the following rules: - For each substance, EC50 daphnia and algae are compared and the lowest EC50 is set as the Upper Threshold Concentration (UTC) and compared to the LC50 fish value: - if UTC < LC50 fish, only the threshold test with 10 fish would have been performed - if UTC > LC50 fish, the threshold test and further step-down tests would have been performed their number was calculated by comparing theoretical stepdown concentrations using a dilution factor of 3.2 until a concentration < LC50 was reached - if UTC = LC50 fish, it was assumed that two tests were performed - if UTC = LC50 and LC50 >100mg/L it was assumed that one test was performed only for data sets I and referred to as a in the excel evaluation tables). Table 5 on the next page illustrates the calculations. 5.5 Calculation of the possible reduction In a first evaluation, the total number of fish needed when all of the substances would have been tested with the acute LC50 fish 96h using 42 fish/substance was compared to the total number of fish needed with the step-down approach. The second evaluation took into account that a limit test (14 fish per substance) might have been performed for those substances where the fish LC50 results were reported as greater than (>) values above 100mg/L. 9

11 Table 5: Examples how the number of step-down tests was calculated Algae EbC50 Algae ErC50 Daphnia EC50 Fish LC50 UTC Step1 Fish LC50 vs. UTC Further testing needed Step2 UTC/3.2 Step3 Step2/3.2 Step4 Step3/3.2 No of Tests > no > no > no < YES E < YES > no

12 6. Results and possible reduction The evaluation shows, that for most of the substances tested, fish is the least sensitive species (see Tables 7-10). Table 6 summarises the overall reduction in the use of fish for the four databases. Table 6-9 demonstrate the results for the individual databases and list the numbers of step-down tests needed, the numbers of fish used in conventional and step-down approach and present the percentage of reduction in number of fish used. Table 6: Summary of possible reduction in the number of fish needed [%] for new and existing chemicals, plant protection products and pharmaceuticals NCD IUCLID PPP PHARM Data Set I Data Set II Data Set III Number of substances LC (70.3) Limit test considered 53.6 (58.8) Number of substances LC (67.0) Limit test considered 61.1 (62.2) Number of substances LC (72.7) Limit test considered 60.1 (63.8) Number of substances LC Limit test considered a Values in the brackets correspond to results of evaluation when imprecise fish LC50 values above 100 mg/l (when UTC=LC50) are taken into consideration, assuming that one step-down test would be sufficient in those cases (see 5.4) Results for NCD The overall results for the NCD are summarised in Table 7. For data set I, 214 (14.9%) out of 1439 substances were most toxic to fish resulting in a reduction of 66.5% fish used. During the evaluation of data set I it was noticed that for 229 substances where UTC equalled LC50, LC50 was reported as >100mg/L. Therefore, a further calculation was performed taking into account that in practice only one step-down test would have been needed. The predicted reduction increased to 70.3% and to 58.8% when considering the limit test (see below and 5.4). Evaluation of data set II (848 substances) revealed that fish was the most sensitive species for 125 (14.7%) substances, which would result in a reduction of 71.2% in the number of fish used. In the most restrictive data set III, for 401 out of the 496 substances acute algal EC50 and daphnid EC50 values were lower than or equal to the fish LC50 data meaning that in only in 91 (18.3%) cases fish was the most sensitive species. Based on the total number of fish used in acute LC50 testing (42 fish per test) versus the total number of fish calculated for the threshold approach (10 fish per step-down test), this leads to a 70.3% reduction in the number of fish used for base set acute toxicity testing of NCD substances. As already mentioned, if one considers that LC50 results reported >100mg/L might be derived from a limit test, then the reduction in the number of fish used applying the step-down 11

13 approach would be slightly lower, namely 53.6, 62.9, and 65.9% for the data sets I, II, and III, respectively. Table 7: Results for the New Chemicals Database Data set I 1,439 substances Data set II 848 substances Data set III 496 substances Endpoint selected to set the UTC [%] Algae (72hEC) Daphnia (48hEC50) Algae equal to daphnia result Most sensitive endpoint [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Fish equal to daphnia/algae UTC result Fish (96hLC50) No. of step-down tests needed [No. of substances] a 1 step 915 (1,144) steps 477 (248) steps steps steps Estimated numbers of fish used in acute toxicity test LC50 test LC50 and the limit test considered 60,438 43,638 35,616 27,580 20,832 18,144 Estimated numbers of fish used in acute threshold test approach a 20,260 (17,970) 10,240 6,190 Estimated reduction of fish used [%] a Step-down approach/lc50 test Step-down approach/lc50 and the limit test considered 66.5 (70.3) 53.6 (58.8) a Values in the brackets correspond to results of evaluation when imprecise fish LC50 values above 100 mg/l (when UTC=LC50) are taken into consideration, assuming that one step-down test would be sufficient in those cases (see 5.4). 12

14 6.2. Results for the IUCLID database The detailed results for the IUCLID database are summarised in Table 8 (page 15). In case that several results were reported for the same CAS number, the lowest values were used for the evaluation. For data set I, 98 (26.9%) out of 316 substances were most toxic to fish resulting in a reduction of 66.5%. If considered that for 7 substances UTC = LC50 and LC50 >100mg/L, the calculated reduction increased to 67.0% and to 62.2% when considering the limit test (see below and 5.4). Evaluation of data set II (269 substances) revealed that fish was the most sensitive species for 86 (31.2%) substances, which would result in a reduction of 66.0% in the number of fish used. In the most restrictive data set III, for 144 out of the 206 substances acute algal EC50 and daphnid EC50 values were lower than or equal to the fish LC50 data meaning that in 62 (30.1%) cases fish was the most sensitive species. Based on the total number of fish used in acute LC50 testing (42 fish per test) versus the total number of fish calculated for the stepdown approach (10 fish per step-down test), this leads to a 67.1% reduction in the number of fish used. If one considers that LC50 results reported >100mg/L might be derived from a limit test, then the reduction in the number of fish used applying the step-down approach would be slightly lower, namely 61.6, 62.8, and 65.0% for the data sets I, II, and III, respectively Results for the plant protection products The detailed results for the plant protection products are summarised in Table 9 (page 16). For data set I, 9 (13.2%) out of 68 substances were most toxic to fish resulting in a reduction of 69.9%. If considered that for 7 substances UTC = LC50 and LC50 >100mg/L, the calculated reduction increased to 72.7% and to 63.8% when considering the limit test (see below and 5.4). Evaluation of data set II (47 substances) revealed that fish was the most sensitive species for 4 (8.5%) substances, which would result in a reduction of 74.2% in the number of fish used. In the most restrictive data set III, for 24 out of the 28 substances acute algal EC50 and daphnid EC50 values were lower than or equal to the fish LC50 data meaning that in 4 (14.3%) cases fish was the most sensitive species. Based on the total number of fish used in acute LC50 testing (42 fish per test) versus the total number of fish calculated for the step-down approach (10 fish per step-down test), this leads to a 72.8% reduction in the number of fish used. If one considers that LC50 results reported >100mg/L might be derived from a limit test, then the reduction in the number of fish used applying the step-down approach would be slightly lower, namely 60.1, 68.9, and 71.4% for the data sets I, II, and III, respectively. 13

15 Table 8: Results for the IUCLID database Data set I 316 substances Data set II 269 substances Data set III 206 substances Endpoint selected to set the UTC [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Most sensitive endpoint [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Fish equal to daphnia/algae UTC result Fish (96hLC50) No. of step-down tests needed [No. of substances] 1 step (threshold test) 218 (225) a steps 75 (68) steps steps steps steps steps Estimated numbers of fish used in acute toxicity test LC50 test LC50 and the limit test considered 13,272 11,592 11,298 10,318 8,652 8,148 Calculated numbers of fish used with the step-down approach 4450 (4,380) 3,840 2,850 Estimated reduction of fish used [%] a Step-down approach/lc50 test Step-down approach/lc50 and the limit test considered 66.5 (67.0) 61.6 (62.2) a Values in the brackets correspond to results of evaluation when imprecise fish LC50 values above 100 mg/l (when UTC=LC50) are taken into consideration, assuming that one step-down test would be sufficient in those cases (see 5.4). 14

16 Table 9: Results for plant protection products Data set I 68 substances Data set II 47 substances Data set III 28 substances Endpoint selected to set the UTC [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Most sensitive endpoint [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Fish equal to daphnia/algae UTC result Fish (96hLC50) No. of step-down tests needed [No. of substances] 1 step (threshold test) 51 (59) a steps 16 (8) 4 4 Estimated numbers of fish used in acute toxicity test LC50 test LC50 and the limit test considered 2,856 2,156 1,974 1,638 1,176 1,120 Calculated numbers of fish used with the step-down approach 860 (780) Estimated reduction of fish used [%] a Step-down approach/lc50 test Step-downapproach/LC50 and the limit test considered 69.9 (72.7) 60.1 (63.8) a Values in the brackets correspond to results of evaluation when imprecise fish LC50 values above 100 mg/l (when UTC=LC50) are taken into consideration, assuming that one step-down test would be sufficient in those cases (see 5.4). 15

17 6.4. Results for pharmaceuticals for human use The detailed results for the pharmaceuticals are summarised in Table 10. Data set I corresponds to Data set II and therefore only the results of data set II and III are given. Evaluation of data set II (91 substances) revealed that fish was the most sensitive species for 13 (14.3%) substances, which would result in a reduction of 71.5% in the number of fish used. In the most restrictive data set III, for 56 out of the 68 substances acute algal EC50 and daphnid EC50 values were lower than or equal to the fish LC50 data meaning that in 12 (17.6%) cases fish was the most sensitive species. Based on the total number of fish used in acute LC50 testing (42 fish per test) versus the total number of fish calculated for the stepdown approach (10 fish per step-down test), this leads to a 70.2% reduction in the number of fish used. If one considers that LC50 results reported >100mg/L might be derived from a limit test, then the reduction in the number of fish used applying the step-down approach would be slightly lower, namely 67.7 and 68.0% for the data sets II and III, respectively. Table 10: Results for pharmaceuticals for human use Data set II 91 substances Data set III 68 substances Endpoint selected to set the UTC [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Most sensitive endpoint [%] Algae (72hEC50) Daphnia (48hEC50) Algae equal to daphnia result Fish equal to daphnia/algae UTC result Fish (96hLC50) No. of step-down tests needed [No. of substances] 1 step (threshold test) steps steps 1 1 Estimated numbers of fish used in acute toxicity test LC50 test LC50 and the limit test considered 3,822 3,374 2,856 2,660 Calculated numbers of fish used with the step-down approach 1, Estimated reduction of fish used [%] Step-down approach/lc50 test Step-down approach/lc50 and the limit test considered

18 7. Discussion and conclusions The step-down approach was applied to 1,439 substances extracted from the NCD, 316 substances of the IUCLID database, 68 plant protection products and metabolites, and 91 pharmaceuticals for human use. Since not all of the results of the algae, daphnia and fish tests were given as precise values, we used more and more stringent selection criteria and consequently created three data sets for each group of substances, where the corresponding data sets I had the lowest and data sets III the highest precision. Regarding the calculation of the achievable reduction, we assumed that the LC50 96h test was performed at five concentrations with seven fish per concentration and seven fish per control. We did not consider that up to 10 fish can be used per concentration or that additional fish might have been used for range finding testing preceding the definitive LC50 test. In a second evaluation, we considered that the limit test (using 14 instead of 42 fish) might have been carried out for all of the substances with imprecise LC50 96h values >100 mg/l. This consideration mainly had an impact on data sets I, since these contain most of the imprecise values and the estimated reduction of number of fish used decreased (see Table 6). The possible reductions for data sets II and III for all substances are quite similar and more realistic than the ones for data sets I, since we only evaluated substances for which at least the result of the daphnia or algae test was given as a precise value. During the course of this study the question arose whether the dilution factor of 3.2 chosen by Hutchinson et al (2003) based on the experience at AstraZeneca is suitable or whether it would increase the uncertainty of the LC50 values compared to generally used dilution factor of 2.2. Our calculations of the number of threshold/step-down tests needed are based on dilution factor 3.2 but we performed for the NCD an additional calculation using dilution factor 2.2. The total number of tests needed only slightly increases and the overall reduction of number of fish used has hardly been affected (Table 11). 17

19 Table 11: Comparison of the number of threshold/step-down tests needed with dilution factor 2.2 and dilution factor 3.2 New Chemical Database and calculation of possible reduction (%) according to the figures given in Table 7 NCD Data set I Data set II Data set III Dilution factor No steps total no of tests Numbers of animals: threshold/step-down approach 10/test 20,260 20,920 10,240 10,640 6,190 6,440 14/test* 28,364 29,288 14,156 14,440 8,666 9,016 Numbers of animals current testing strategy LC50 60,438 35,616 20,832 limit test considered Possible reduction (%) 10/test vs LC50 10/test vs limit test considered 14/test vs LC50 14/test vs limit test considered 43,638 27,580 18, * = This calculation was also carried out assuming that 7 test and 7 control fish would be used in the threshold/step down tests (see below) In addition, the question came up, how many fish should be used in the threshold test and subsequent step-down tests. Are the 5 test and 5 control fish (1 fish could die in the control) proposed by Hutchinson et al (2003) (which correspond to the range finding test) enough to achieve a significant confidence that the LC50 is greater than the concentration used in the threshold/step-down test? Table 12 shows that a significant confidence of 96.88% can only be achieved, if none of the test and control fish dies. In this context, it should be kept in mind that OECD TG 203; Annex V C.1 stipulates a confidence of at least 99% for the limit test, i.e. if 7 test and 7 control fish used, none of the fish should die. 18

20 Table 12: Percentage of confidence that LC50 is greater than the Upper Threshold Concentration (and subsequent step-down concentrations) Group size Confidence (%) Mortality n.s n.s. (81.25) n.s. (89.06) n.s. (93.75) n.s. (85.55) n.s. (82.02) n.s. (94.53) n.s. = not significant Even if 7 test and 7 control fish would be used in the threshold (and subsequent step-down) tests our calculations (Table 13) still indicate a remarkable reduction. Table 13: Summary of possible reduction (%) in the number of fish needed for new and existing chemicals, plant protection products and pharmaceuticals if 14 fish are used in step-down approach NCD IUCLID PPP PHARM Data Set I Data Set II Data Set III Number of substances LC (58.4) Limit test considered 35.0 (42.3) Number of substances LC (53.8) Limit test considered 46.3 (47.1) Number of substances LC (61.8) Limit test considered 44.2 (49.4) Number of substances LC Limit test considered a Values in the brackets correspond to results of evaluation when imprecise fish LC50 values above 100 mg/l (where UTC=LC50) are taken into consideration, assuming that one step-down test would be sufficient in those cases (see 5.4). 19

21 A further option would be to perform a full LC50 test, when fish mortality occurs in the threshold test at UTC concentration. This would correspond to the requirement for the limit test in OECD 203/Annex V C.1, where a full LC50 test should be performed when mortality is observed. It might be interesting in this context to discuss the necessity of a precise LC50 value for classification and labelling and risk assessment or whether the approach chosen for acute oral toxicity (Annex V B.1bis/B.1tris, OECD 420, 423) might be applicable also for acute fish toxicity testing. Another point of discussion was our assumption that if UTC < LC50 only the threshold test would be performed. Especially when the values are in the same range, our assumption might underestimate the number of necessary tests. It was not possible to investigate this in depth since we only had the LC50 values but no information on the steepness of the dose-response curve. Nevertheless, we re-evaluated NCD data set III and multiplied the UTC (and subsequent step-down concentration by factor 10) and only when LC50 > 10*UTC/step-down concentration the testing was stopped. This resulted in 1,191 tests instead of 619 (results are given in table 14). This is a very conservative approach and it should be kept in mind that even with 5 test and 5 control fish, the LC50 would be greater than UTC/step-down concentration with 96.88% confidence (Table 12) if no mortality occurs. In addition, the reduction of number of fish used has been estimated for the NCD data set III using several hypotheses on the number of fish that could have been used according to the current requirements and the step-down approach (Table 14). The results show that for these hypotheses the achievable reduction in the number of fish lies between 8.1% and 79.5%. If the step-down approach would become part of the regulations, it should be evaluated whether the application of humane endpoints (refinement), i.e. terminating the threshold/stepdown tests when fish die or show severe clinical signs as proposed by Hutchinson et al (2003) or at least humanely kill moribund fish, could be implemented. In addition, it should be investigated whether the same group of control fish could be used for the needed series of step-down tests. Our overall conclusion is that the step-down approach would lead to a considerable reduction in number of fish used for regulatory acute aquatic toxicity testing, retaining the same quality of information needed for the classification and labelling and risk assessment, thereby contributing to animal welfare and to reduction in time and costs. 20

22 Table 14: Possible reduction (%) in fish needed for NCD data set III based on 4 hypotheses concerning the number of fish used when substances were tested according to the current strategy and on 5 for the proposed testing strategy Current strategy Total number of fish used in proposed strategy Total number of fish used in current strategy Step-down approach 10 fish/test (e) 14 fish/test (f) 14 fish/test + factor 10 for LC50 - UTC/steps comparison Threshold test at UTC, if mortality occurs LC50 with 42 fish 10 fish/test (i) 14 fish/test (j) 6,190 8,666 16,674 8,950 10,934 LC50 42 fish (a) 20, limit test 14 fish LC50 42 fish (b) 18, LC50 60 fish (c) 29, limit test 14 fish LC50 60 fish (d) 25, To estimate the number of fish that could have been use according to the current requirements four hypotheses were made: a) All of the substances were tested with the LC50 test using 42 fish b) Substances where LC50>100 mg/l is reported were tested with the limit test using14 fish and the others were test with the LC50 test using 42 fish c) All of the substances were tested with the LC50 test using 60 fish d) Substances where LC50>100 mg/l is reported were tested with the limit test using14 fish and the others were test with the LC50 test using 60 fish. To estimate the number of fish that could be used according to the step-down approach five hypotheses were made: e) All of the substances were tested with the step-down approach using 10 fish/test f) All of the substances were tested with the step-down approach using 14 fish/test g) All of the substances were tested with the step-down approach using 14 fish/test + factor 10 for LC50- UTC/steps comparison h) If mortality occurred in the threshold test (10 fish) at UTC, a LC50 test using 42 fish was performed i) If mortality occurred in the threshold test (14 fish) at UTC, a LC50 test using 42 fish was performed 21

23 8. References Doudoroff, P., Anderson, B.G., Burdick, G.E., Galtsoff, P.S., Hart, W.B., Patrick, R., Strong, E.R., Surber, E.W., Van Horn, W.M., Bioassay methods for evaluation of acute toxicity of industrial wastes to fish. Sewage Ind. Wastes 23, European Commission, Council Directive 67/548/EEC of 18 August 1967 on the approximation of the laws, regulations and administrative provisions relating to the classification, packaging and labelling of dangerous substances. Official Journal of European Communities, L96/1. Office for Official Publications of the European Communities. Luxembourg. European Commission, Council Directive 92/32/EEC of 5 June 1992 amending for the 7th time Council Directive 67/548/EEC. Official Journal of European Communities, L 154. Office for Official Publications of the European Communities, Luxembourg. European Commission, DG Health and Consumer Protection. Technical Review Reports for Plant Protection Products. Hutchinson, T.H., Barrett, S., Buzby, M., Constable, D., Hartmann, A., Hayes, E., Huggett, D., Länge, R., Lillicrap, A.D., Straub, J.O., Thompson, R.S., A strategy to reduce the numbers of fish used in acute ecotoxicity testing of pharmaceuticals. Environ. Toxicol. Chem. 22, Jeram, S., Riego Sintes, J.M., Halder, M., Baraibar Fentanes, J., Sokull-Klüttgen, B., Hutchinson, T.H. (2005) A strategy to reduce the use of fish in acute ecotoxicity testing of new chemical substances notified in the European Union. Regulatory Toxicology and Pharmacology 42, Rand, G.M., Petrocelli, S.R., Fundamentals of Aquatic Toxicology, 1st edition. Hemisphere Publishing Corporation, New York, NY, USA. Stephan, C.E., Methods for calculating an LC50. In Mayer FL, Hamelink JL, eds., Aquatic Toxicology and Hazard Evaluation, ASTM STP 634. American Society for the Testing of Materials, Philadelphia, PA, USA, pp Weyers, A., Sokull-Klüttgen, B., Baraibar-Fentanes, J., Vollmer, G., Acute toxicity data: a comprehensive comparison of results of fish, Daphnia and algae tests with new substances notified in the EU. Environ. Toxicol. Chem. 19,

24 Annex The annex gives an overview on the data files going with the report, the abbreviations and colour codes used in these files: 1. Files Extracted data = complete data set extracted from the database are included plus the selected data sets according to the selection criteria. Data set I, II, III for database NCD, IUCLID and PPP Data set II, III for database PHARM For NCD additional evaluations were performed and are available in extra Workbooks (Table 11, 14) colour codes are used in the extracted data files, see 3) Reduction calculations = summaries of the calculations and achievable reductions are provided for NCD, IUCLID, PPP, PHARM for 10/14 fish per threshold/step-down test and for evaluation a (see below abbreviation and 5.4 and 5.5 of the report; not for PHARM). Several calculations (Table 14) were only performed for NCD. 2. Abbreviations: Algae EbC50 72h, The result of standard algae test (OECD 201) reported as algae biomass also marked as b or Ab in Excel Algae ErC50 72h, The result of standard algae test (OECD 201) reported as algae growth rate also marked as r or Ar in Excel Daphnia EC50 48h, The result of standard Daphnia test (OECD 202) reported as immobilisation of Daphnia ECB, European Chemicals Bureau Evaluation a, Evaluation, where the probability that one test instead of two would be enough in cases where the fish LC50 is an imprecise value greater than 100. This evaluation was only performed for datasets I, where signs (>, <) where not taken into consideration. Factor 3.2, Dilution factor allowing the semilogaritmic step-down concentration series Fish LC50 96h, The result of standard acute fish test (OECD 203) reported as lethality of fish IUCLID, The International Uniform ChemicaL Information Database (existing chemicals) Limit test, The fish test which is performed at one concentration using 7-10 test and 7-10 fish. This test is usually performed at 100 mg/l, when toxicity is not expected from all data available or at the saturation concentration when the substance is poorly soluble. Most sensitive, The test species which is the most sensitive among all test species (the lowest E(L)C50 value). NCD, New Chemicals Database PPP, Plant Protection Products (active substances and metabolites are considered in this study) Step1, Fish LC50 result is compared to the UTC. When fish LC50 is greater we assume that this concentration would not be toxic for fish. The test procedure is stopped. When fish 23

25 LC50 is lower than UTC the toxicity for fish at this concentration is expected. The test procedure continues at lower concentration (Step-down). Step2, Step down concentration at which step-down threshold test is performed when toxicity was observed in Step1 test Step3, Further, step-down tests (Step 4 ) are performed at lower concentrations when toxicity is observed in the considered step Step-down approach, The testing strategy where fish tests are performed at different concentrations subsequently and not in parallel. Testing is performed at lower concentration (Dilution 1, 2 ) only in case when toxicity was observed in the previous step. Threshold test, The fish test which is usually performed in the range finding procedure before the definitive test is performed. Five fish are used in test concentration and five in the negative control. UTC, The Upper Threshold Concentration is the highest concentration at which the fish threshold test should be performed. It is derived from algae and daphnia test results. Among the two values the lowest is selected and is equal to the UTC. 3. Colour codes for extracted data (xcel files attached) BROWN, Substances to be excluded DARK GREEN, Values where the result in the data base is reported as =100mg/L. Applicable only for data sets I. DARK YELLOW, cells with imprecise signs/values greater than (>) GRAY, cells with imprecise signs/values greater than (>) PALE YELLOW, Header row PALE YELLOW, Important cases (Comments included) PURPLE, Fish LC50 is the lowest test result among all tested species RED, Exceptional cases (Comments included) RED, Last line in the data set 24