Opportunities and Challenges for Using IVIVE to Improve Decision Making. Weihsueh A. Chiu, PhD Texas A&M University

Transcription

1 Opportunities and Challenges for Using IVIVE to Improve Decision Making Weihsueh A. Chiu, PhD Texas A&M University

2 Conflict of Interest Statement Neither myself nor any of my coauthors, including members of our immediate families, have any financial interest or affiliation with a commercial organization that has a direct or indirect interest in the subject matter of my presentation.

3 Outline: Key Opportunities and Challenges 1. Data generation 2. Variability and susceptibility 3. Uncertainty 4. Domain of applicability 5. New types of in vitro systems 6. Transparency and ease of use

4 Basic approach to IVIVE / RTK: Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

5 Basic approach to IVIVE / RTK: Key Challenges 1 Hepatocytes (10 donor pool) Ln Conc (um) Time (min) Hepatic Clearance Q gut + Q liver Liver Tissue Liver Blood CL metab Gut Lumen Gut Blood Q liver 1. Data generation 2. Variability and susceptibility 3. Uncertainty 4. Domain of applicability 5. New types of in vitro systems 6. Transparency and ease of use k gutabs 3 Rest of Body Body Blood Q gut Q 4 GFR 2 Steady State 5 Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding 6 Source: Barbara Wetmore

6 Key Challenge 1: Data Generation Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

7 Key Challenge 1: Data Generation Hepatocytes (10 donor pool) Plasma (6 donor pool) Ln Conc (um) Time (min) Hepatic Clearance Plasma Protein Binding Higher throughput than in vivo data used for traditional PBPK models. RTK data (publically) available for >400 chemicals (many more than for traditional PBPK).

8 Key Challenge 1: Data Generation Hepatocytes (10 donor pool) Ln Conc (um) Time (min) Hepatic Clearance Not high throughput compared to high throughput in vitro screening assays (>10,000 chemicals) Each compound needs its own analytic method Plasma (6 donor pool) Plasma Protein Binding How to bridge the gap?

9 Solutions: Prioritize Data Generation Prioritize using existing chemicals with RTK data? Number of Values Distribution of Chemical C ss (µm) Cumulative Percent Distribution Information Median 1 µm Upper 90 th %ile 111 µm Upper 95 th %ile 230 µm What would be the MOE given a conservative Css? 0 0 Need to know decision context up front, e.g., In vitro assays to which IVIVE is being applied Exposure estimates for comparison to IVIVE-based PoDs

10 Solutions: Prioritize Data Generation More sophisticated prioritization-schemes Classes of chemicals (e.g., analogous to TTC) QSAR / machine learning Can afford to have more uncertainty due to prioritization context Wambaugh et al. (2015)

11 Key Challenge 2: Variability and Susceptibility Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

12 Key Challenge 2: Variability and Susceptibility Typical RTK data based on pooled samples. No information on variability No information on lifestages Liver Tissue Liver Blood CL metab Gut Lumen Gut Blood k gutabs Q liver SimCYP and similar models: Good information on physiological variability Q gut + Q liver Rest of Body Body Blood Q GFR Q gut Steady State Blood Concentrations Assume clearance variability of ~30%.

13 Context: historical in vivo data on toxicokinetic variability Historical in vivo TK variability data across ~30 chemicals Limitations: Mostly pharmaceuticals Limited sample size for each chemical Mostly healthy volunteers TK Variability Factor (ratio of 95% most sensitive to median human) 4.7 Source: WHO/IPCS (2015) and Dale Hattis

14 Context: historical in vivo data on toxicokinetic variability Historical in vivo TK variability data across ~30 chemicals Source: WHO/IPCS (2015) and Dale Hattis TK Variability Factor (ratio of 95% most sensitive to median human) RTK-based TK variability across ~400 chemicals RTK potentially missing chemicals with high TK variability

15 Solutions: Bottom-up Approach for Variability Across Subgroups Wetmore et al., 2014, Toxicol Sci. 142(1): Primary Hepatocytes Ln Conc (um) Time (min) Hepatic Clearance Cl in vitro Plasma C ss General Population Plasma C ss for: CYP1A2 Neonates rcyp1a2 Cl rcyp1a2 CYP3A4 CYP1A2 CYP2E1 CYP3A4 CYP2C9 Asians rcyp3a4 rcyp Cl rcyp3a4 CYP Cl rcyp CYP2D6 CYP2C8 UGT1A4 CYP3A5 CYP2C19 UGT1A1 UGT2B7 CYP2B6 Northern Europeans rugt UGT Cl rugt Intrinsic Clearance Rates Children And so on Source: Barbara Wetmore

16 Solutions: Bottom-up Approach for Variability Across Subgroups Wetmore et al., 2014, Toxicol Sci. 142(1): Chemical Median C ss for Healthy Population 95 th Percentile C ss for Most Sensitive Most Sensitive Much larger than typical adult RTK-based estimates! Estimated HK AF % Contribution of Isozyme Differences to Average HK AF Acetochlor Neonatal Azoxystrobin Neonatal Bensulide Neonatal Carbaryl Neonatal Difenoconazole Renal Insufficiency Fludioxonil Neonatal Haloperidol Neonatal Lovastatin Neonatal Tebupirimfos Renal Insufficiency Plasma C ss General Population Plasma C ss for: Neonates Asians Northern Europeans Children And so on Source: Barbara Wetmore

17 Solutions: Bottom-up Approach for Variability Across Subgroups Wetmore et al., 2014, Toxicol Sci. 142(1): Chemical Median C ss for Healthy Population 95 th Percentile C ss for Most Sensitive Most Sensitive Estimated HK AF Differences largely driven by isozyme differences, not physiology! % Contribution of Isozyme Differences to Average HK AF Acetochlor Neonatal Azoxystrobin Neonatal Bensulide Neonatal Carbaryl Neonatal Difenoconazole Renal Insufficiency Fludioxonil Neonatal Haloperidol Neonatal Lovastatin Neonatal Tebupirimfos Renal Insufficiency Plasma C ss General Population Plasma C ss for: Neonates Asians Northern Europeans Children And so on Source: Barbara Wetmore

18 Solutions: Using chemical-specific variability data Established methods from traditional PBPK modeling Example with Methyleugenol In vitro data using recombinant P450 enzymes Literature data on variability in enzyme content and liver weight Additional in vitro S9 data across ~20 individuals median 2-fold 2.8-fold 2.2-fold 95%ile 3-fold 99%ile Css in males (top) and females (bottom) Re-analysis of data from Al-Subeihi et al. (2012, 2015)

19 Key Challenge 3: Uncertainty Ln Conc (um) ? Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

20 Key Challenge 3: Uncertainty Sometimes more than 10-fold error compared to in vivo TK. Uncertainty could be more formally incorporated into IVIVEbased estimates. Alternatively could adjust MOE benchmark to compensate. Source: John Wambaugh

21 Key Challenge 4: Domain of applicability Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

22 Key Challenge 4: Domain of applicability Simple threecompartment PBPK model Parent compound is the active moiety, not a metabolite Chronic exposure steady state drives toxicity Single compound Previous speakers have provided examples of solutions developing more refined PBPK models for use in IVIVE. Q gut + Q liver Liver Tissue Liver Blood CL metab Rest of Body Body Blood Q GFR Gut Lumen Gut Blood Q liver Q gut k gutabs Steady State Blood Concentrations

23 Key Challenge 5: New in vitro models Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

24 IVIVE for Tissue Chips 2D cell cultures 2D cell cultures with modified substrates Microphysiological systems and 3D models Organ-on-a-chip Lab-on-a-chip -on-a-chip Timothy.ruban en.wikipedia.org Adapted from NRC ESEH Committee Slide

25 IVIVE for Tissue Chips IVIVE has been based on blood/serum concentrations Tissue Chips may require different dosimetry estimates Portal of entry (gut, lung) Extracellular fluids (urine, interstitial) Cellular concentrations may not be directly accessible Additional computational modeling of chip kinetics. Timothy.ruban en.wikipedia.org

26 IVIVE for population-based in vitro models Genetically diverse human population Genetically defined sample of population High throughput in vitro model system Structurally diverse chemical population ~1000 individuals cytotoxicity screening ~170 compounds Abdo et al., 2015 Chemical-Specific TD Variability Factor: The factor estimated to protect up to the most sensitive 1% for human toxicodynamic variability for a chemical 26

27 Solutions for Combining TK and TD Variability 2-fold NOAEL RfD = UF A x UF H 2.8-fold UF A =10 UF H =10 UF A-TK =3 UF A-TD =3 UF H-TK =3 UF H-TD =3 10- fold

28 Solutions for Combining TK and TD Variability Combine like Uncertainty Factors? Calculator-simple and conservative Resulting population percentile unclear. Combine probabilistically Spreadsheet-simple assuming lognormality and independence Can specify population percentile 10- fold 2-fold 2.8-fold

29 Addressing TK and TD variability in mixtures Advantage of in vitro screening is that can more easily test chemical mixtures How do you do IVIVE for a mixture? Abdo et al., Environ Int 85: , 2015

30 Solutions for Addressing TK and TD variability in mixtures Abdo et al., Environ Int 85: , 2015

31 Solutions for Addressing TK and TD variability in mixtures Margin of safety =10 5 Predicted exposure Predicted in vivo cytotoxic dose Margin of safety =10 Abdo et al., Environ Int 85: , 2015

32 Pesticides mixture example illustrates solutions to multiple challenges Data generation missing RTK data for some components of the mixture Approach: Median and Worst Case assumptions based on available RTK data Domain of applicability in vitro data based on testing whole mixtures Approach: IVIVE for each component separately, assuming no interactions, and recombined based on mass New in vitro models assessing TD variability across population of cell lines: Approach: IVIVE conducted for each cell line separately, assuming TK and TD are independent

33 Key Challenge 6: Transparency and Ease of Use Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Gut Lumen k gutabs Liver Blood Gut Blood CL metab Q liver Q gut + Q liver Rest of Body Body Blood Q gut Q GFR Steady State Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding Source: Barbara Wetmore

34 Workflow for improving decision making transparency Ideally open-source and transparent Tailored to different audiences Regulatory scientists Primarily users of data and models Simple, easy to use Reproducible and auditable Research scientists and trainees For both users and generators/developers of data and models Reproducible and replicable

35 NTP/NICEATM IVIVE Workflow tools based on KNIME and Jupyter Data In vitro data: AC50s or LECs, multiple assays In vivo data: range of LELs or exposure data (user input or embedded database) Run Models: (rat or human, injection or oral) Model output: EAD (inj or oral) Output Tables Clearance and Fub: (user input or embedded database) Parameters PK/PBPK Modelling Model evaluation Comparative Plots Source: Nicole Kleinsteuer

36 Anticipated for Spring 2017 Source: Nicole Kleinsteuer

37 Summary: Key Opportunities and Challenges 1 Ln Conc (um) Time (min) Hepatocytes (10 donor pool) Hepatic Clearance Liver Tissue Liver Blood CL metab Gut Lumen Gut Blood Q liver k gutabs 3 Q gut + Q liver Rest of Body Body Blood Q gut Q 4 GFR 2 Steady State 5 Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding 6 Source: Barbara Wetmore

38 Summary: Key Opportunities, Challenges, and Potential Solutions 1 Hepatocytes (10 donor pool) Ln Conc (um) Time (min) Hepatic Clearance Q gut + Q liver Liver Tissue Liver Blood CL metab Gut Lumen Gut Blood Q liver 1. Data generation 2. Variability and susceptibility 3. Uncertainty 4. Domain of applicability 5. New types of in vitro systems 6. Transparency and ease of use k gutabs 3 Rest of Body Body Blood Q gut Q 4 GFR 2 Steady State 5 Blood Concentrations Plasma (6 donor pool) Plasma Protein Binding 6 Source: Barbara Wetmore

39 Summary: Key Opportunities, Challenges, and Potential Solutions 1. Data generation 2. Variability and susceptibility 3. Uncertainty 4. Domain of applicability 5. New types of in vitro systems 6. Transparency and ease of use Prioritization schemes for data generation Addressing metabolismrelated variability o Qualitatively addressed, but not quantified Fit-for-purpose PBPK models o Tissue chips Population-based screening Workflow tools being developed

40 References Abdo N, Wetmore BA, Chappell GA, Shea D, Wright FA, Rusyn I. In vitro screening for population variability in toxicity of pesticidecontaining mixtures. Environ Int Dec;85: doi: /j.envint PubMed PMID: ; PubMed Central PMCID: PMC Abdo N, Xia M, Brown CC, Kosyk O, Huang R, Sakamuru S, Zhou YH, Jack JR, Gallins P, Xia K, Li Y, Chiu WA, Motsinger-Reif AA, Austin CP, Tice RR, Rusyn I, Wright FA. Population-based in vitro hazard and concentration-response assessment of chemicals: the 1000 genomes high-throughput screening study. Environ Health Perspect May;123(5): doi: /ehp PubMed PMID: ; PubMed Central PMCID: PMC Al-Subeihi AA, Alhusainy W, Kiwamoto R, Spenkelink B, van Bladeren PJ, Rietjens IM, Punt A. Evaluation of the interindividual human variation in bioactivation of methyleugenol using physiologically based kinetic modeling and Monte Carlo simulations. Toxicol Appl Pharmacol Mar 1;283(2): doi: /j.taap PubMed PMID: Al-Subeihi AA, Spenkelink B, Punt A, Boersma MG, van Bladeren PJ, Rietjens IM. Physiologically based kinetic modeling of bioactivation and detoxification of the alkenylbenzene methyleugenol in human as compared with rat. Toxicol Appl Pharmacol May 1;260(3): doi: /j.taap PubMed PMID: El-Masri H, Kleinstreuer N, Hines RN, Adams L, Tal T, Isaacs K, Wetmore BA, Tan YM. Integration of Life-Stage Physiologically Based Pharmacokinetic Models with Adverse Outcome Pathways and Environmental Exposure Models to Screen for Environmental Hazards. Toxicol Sci Jul;152(1): doi: /toxsci/kfw082. PubMed PMID: ; PubMed Central PMCID: PMC National Academy of Sciences, Committee on Emerging Science for Environmental Health Decisions. Tissue Chips: The Potential of the Tissue Chip for Environmental Health Studies. Wetmore BA, Wambaugh JF, Ferguson SS, Sochaski MA, Rotroff DM, Freeman K, Clewell HJ 3rd, Dix DJ, Andersen ME, Houck KA, Allen B, Judson RS, Singh R, Kavlock RJ, Richard AM, Thomas RS. Integration of dosimetry, exposure, and high-throughput screening data in chemical toxicity assessment. Toxicol Sci Jan;125(1): doi: /toxsci/kfr254. PubMed PMID: Wambaugh JF, Wetmore BA, Pearce R, Strope C, Goldsmith R, Sluka JP, Sedykh A, Tropsha A, Bosgra S, Shah I, Judson R, Thomas RS, Setzer RW. Toxicokinetic Triage for Environmental Chemicals. Toxicol Sci Sep;147(1): doi: /toxsci/kfv118. PubMed PMID: ; PubMed Central PMCID: PMC Wetmore BA, Allen B, Clewell HJ 3rd, Parker T, Wambaugh JF, Almond LM, Sochaski MA, Thomas RS. Incorporating population variability and susceptible subpopulations into dosimetry for high-throughput toxicity testing. Toxicol Sci Nov;142(1): doi: /toxsci/kfu169. PubMed PMID:

41 Acknowledgements Nour Abdo, JUST Anand Gupta (student), TAMU Dale Hattis, Clark University, USA Nicole Kleinsteuer, NTP/NICEATM Ivan Rusyn, TAMU John Wambaugh, US EPA Barbara Wetmore, US EPA Rick Woychik, NIEHS (retired) Fred Wright, NC State Lauren Zeise, California EPA