the use of predictive toxicity tools in health assessments of existing substances under the canadian environmental prote

1. Presentation for the McKim Workshop on Strategic Approaches for Reducing Data Redundancy in Cancer Assessment May 2010 Presented by K Hughes Existing Substances Risk Assessment Bureau Health Canada The Use Of Predictive Toxicity Tools In Health Assessments Of Existing Substances Under The Canadian Environmental Protection Act

2. Page 2 Outline Background Priority Substances Assessments Prioritization of the Domestic Substances List Screening assessments Looking into the Future Summary

3. Page 3 The Canadian Environmental Protection Act (CEPA) Provides the regulatory framework and process for risk assessment and risk management of chemicals Responsibility is shared by Ministers of Environment & Health Distinction between “new” and “existing” substances with the creation of Domestic Substances List (DSL) in 1991 Existing substances are those that were “in commerce” in Canada in mid 1980s

4. Page 4 Assessment of Existing Substances under CEPA CEPA 1988 Priority Substance List assessments In-depth, complex; 5 year legislated deadlines PSL1 (n=44 substances, released in 1989) PSL2 (n=25 substances, released in 1995) CEPA 1999 Ministers’ mandate substantially expanded Introduced Categorization of entire DSL (n = 23,000) Screening assessments Priority Substances List assessments

5. Page 5 Outcome of Assessments under CEPA Purpose of in depth and screening assessments under CEPA is to determine if a substance would be considered “toxic” as defined in the Act “… is entering or may enter the environment in a quantity or concentration or under conditions that: have or may have an immediate or long-term harmful effect on the environment or its biological diversity; constitute of may constitute a danger to the environment on which life depends; or constitute or may constitute a danger in Canada to human life or health.”

6. Page 6 Timelines & Mandatory Decision Making Timelines often require that conclusions be made for substances for which there are little or no empirical data We must reach a conclusion at each level (prioritization or assessment) – setting aside until more data become available is not an option!

7. Page 7 Use of Predictive Tools in CEPA Assessments Predictive tools are used for all levels of assessment activity (including prioritization) To support an assessment conclusion or prioritization decision To provide principle basis for a conclusion or prioritization decision For grouping substances for class assessments Directing research

8. Page 8 Priority Substances List Assessments Prepared under CEPA 1988 Mostly data-rich substances Longer time period for preparation permitted some data generation Assessment of carcinogenicity involved thorough evaluation of weight of evidence and analysis of mode of action; calculation of potency estimates and exposure/potency indices Limited use of predictive tools; just becoming aware of potential applications More use of exposure & pharmacokinetic models than toxicity models Some extrapolation from similar chemicals Would likely play bigger role in any future in-depth assessments

9. Page 9 Categorization of the DSL Categorization was a priority setting exercise that systematically identified substances that should be subject to further assessment and subsequent risk management if deemed appropriate CEPA 1999 required Ministers of the Environment and Health to categorize the 23,000 substances on the DSL according to specific criteria to identify substances that May present, to individuals in Canada, the greatest potential for exposure; or Are persistent (P) or bioaccumulative (B), in accordance with the regulations, AND inherently toxic to humans or to non-human organisms, as determined by laboratory or other studies

10. Page 10 The Prioritization Process

11. Page 11 CategorizationHuman Health Aspects Early recognition that legislative construct for categorization would not identify all priorities from human health perspective Persistence and bioaccumulation not key determinants of potential to harm human health for all types of substances Other properties more relevant (e.g., reactive compounds) Integrated framework designed that permitted more comprehensive identification of true health priorities Involved simple and complex exposure and hazard tools developed to identify health priorities

12. Page 12 The Simple Hazard ToolSimHaz Applied to entire DSL to identify known high and low hazard substances, based on classifications/assessments of other agencies based on weight of evidence For carcinogenicity, included classifications from IARC, EU, EPA, NTP, other HC programs Appropriate assessments selected based on comprehensiveness of review, peer review process, etc. Data rich substances; no need for direct use of predictive models but may have been used by classifying agency

13. Page 13 The Complex Hazard ToolComHaz Applied to further refine list of substances that were determined likely to pose significant potential for population exposure Hierarchical approach to consideration of: Multiple endpoints relevant to characterization of hazard Sources of relevant information

14. Page 14

15. Page 15

16. Page 16 ComHaz Tool – Endpoint-specific Criteria (subset)

17. Page 17 QSAR Models Used For DSL Prioritization (ComHaz Stage 1)

18. Page 18 Other Predictive Toxicity Tools Used in ComHaz (stage 1) Expert systems (Derek) Chemical categories of concern (based on EPA list) Analogues

19. Page 19 Interpretation of QSAR Results in ComHaz Guidance documentation developed to assist evaluators Acceptance criteria developed for interpretation of predictions of QSAR models in DSL prioritization to ensure consistency in application

20. Page 20 “Validation” for TopKat predictions

21. Page 21 “Validation” for Casetox predictions A prediction is considered positive if following conditions are met: (1) Activity =30 case units (2) Probability = 85% probability or a bottom line conclusion of high activity (3) No unknown fragments/ functionalities (4) At least one valid biophore A prediction is considered to be negative/inactive when: (1) The activity is < 20 Case Units, regardless of the probability (occurs when no valid biophore is present) (2) Concluding statement indicates inactivity and there are no unknown fragments/functionalities present

22. Page 22 “Validation” for Derek predictions Predictions of “plausible”, “probable” or “certain” for a number of relevant endpoints are considered “positive” sufficient for prioritizing a substance on the basis of concern for that endpoint. Other outcomes are considered to be “inconclusive”.

23. Page 23 Post-Categorization Assessments Health Canada & Environment Canada identified substances of potential concern from health or environmental perspective (or both) Chemicals Management Plan (CMP) announced to better protect Canadians and the environment from hazardous chemicals CMP included a “Ministerial Challenge” to industry to provide information on the highest priorities from categorization (~200) to shift from predisposition of “toxic” under CEPA Screening level assessments being prepared for the Challenge substances

24. Page 24 Intent of the Challenge The Ministers intend… “to develop or implement measures to assess and manage the risks posed by certain substances” “to identify industrial best practices in order to set benchmarks for risk management, product stewardship, and virtual elimination” The Ministers consider… “Evidence that a substance for which the critical health effect is assumed to have no threshold – i.e. a genotoxic carcinogen – it is assumed that there is a probability of harm to human health at any level of exposure, and therefore indicates that the substance meets the criteria for “toxic” to human health in CEPA 1999” “Evidence that a substance exhibits carcinogenicity, mutagenicity, developmental toxicity, or reproductive toxicity, and a high likelihood of exposure to individuals in Canada, indicates that the substance meets the criteria for “toxic” to human health in CEPA 1999”

25. Page 25 Challenge Screening Assessments Substances nominated as high health priorities were identified by the Simple Hazard tool; therefore had some basis (usually empirical data) for classification by another agency For assessment of carcinogenicity, generally rely on assessment of other agency; no need to re-do weight of evidence determination Update literature to see if story has changed; if a mode of action has been adequately supported

26. Page 26 Challenge Screening Assessments(cont’d.) However, data are not always available for some endpoints (esp. carcinogenicity & genotoxicity) Also need to assess potential health effects of eco priorities, most of which were data poor Predictive tools ((Q)SAR models, analogues) used to identify potential effects

27. Page 27 (Q)SAR Models Used in Challenge Screening Assessments Topkat Casetox Derek Leadscope Model Applier Toxtree OncoLogic OECD Toolbox

28. Page 28 Use of (Q)SAR in Challenge AssessmentsPTPTT (draft) Empirical data limited to acute toxicity and a few tests in Salmonella (negative w & w/o activation) Topkat, Casetox, Derek & Leadscope Model Applier used to predict carcinogenicity, genotoxicity, developmental toxicity and reproductive toxicity Results were predominantly negative Conclusion: substance not considered to be of high concern for health endpoints

29. Page 29 Use of Analogues in Challenge Assessmentsn-Butyl glycidyl ether No empirical cancer data for n-BGE, but positive genotoxicity data Considered several analogues (glycidol, allyl glycidyl ether, phenyl glycidyl ether) Positive cancer data Positive genotoxicity data Conclusion; n-BGE likely genotoxic carcinogen Also ran Topkat, Casetox and Derek on all substances Although not all positive for cancer for all models, all substances predicted similarly in each model (all positive for genotoxicity) Supported choice of analogues

30. Page 30 Post-Challenge Screening Assessments Approximately 3000 more substances to be addressed by 2020 Majority were identified as eco priorities with little or no empirical data relevant to human health effects Necessitates greater reliance on predictive tools Need to increase acceptance of use of predictive tools in regulatory risk assessment

31. Page 31 Class Assessments In light of large number of substances to be assessed, likely to be conducting more class assessments of related chemicals Use tools to help build the classes (Q)SAR models to fill data gaps Read-across to extrapolate/interpolate potential health effects

32. Page 32 ComHaz Stage 2Weight of Evidence Component Development and application of second stage of ComHaz tool in a “rapid screening” context Preliminary Weight of Evidence for genotoxic carcinogenicity

33. Page 33

34. Page 34 ComHaz Stage 2Preliminary Weight of Evidence Component Because of conservative nature of criteria in first stage ComHaz for these endpoints (one positive hit) need to further evaluate Framework for decision making on potential for substance to be a genotoxic carcinogen Involves consideration of multiple lines of evidence (empirical data, (Q)SAR, analogues) as required based on confidence level Scoring system to weight positive and negative results based on relative predictivity of assay (empirical data) or assay upon which (Q)SAR prediction is based Currently being tested on eco priorities Additional refinements could include weighting models, quantifying confidence

35. Page 35 Future Directions (Q)SAR models Expand suite of models (e.g., Times, Caesar, Lazar etc.) Create our own models based on databases containing relevant chemicals (e.g., Pipeline, Leadscope) Develop guidance regarding which models provide more confident predictions for different types of substances Further investigate metabolism models; assist in grouping substances with possible similar modes of action (e.g., OECD Toolbox) Develop more in-house expertise Sharing of data/resources/expertise across agencies

36. Page 36 Future Directions(cont.) Toxicogenomics Investigate how to enhance applicability for use in regulatory risk assessment Expand library of relationships of gene expression profiles to impact on function, toxicity pathways and biological effects (i.e., “what does it mean?”) Anchor technology in “known” chemicals Increase confidence in interpretation Achieves objectives of reducing animal testing and ability to assess more chemicals more quickly Requires collaboration to get there soonest!

37. Page 37 Future Directions(cont.) ToxCast program Collaborate on research to increase understanding and interpretation of results Develop approaches for incorporation into health risk assessment Sponsor inclusion of more industrial chemicals (especially our remaining priorities)

38. Page 38 Summary Predictive toxicity tools play a role in all levels of assessment of existing substances under CEPA Need to further develop tools and more fully integrate into assessment approaches to increase confidence and acceptance in regulatory agencies Exciting new technologies in development!