250 likes | 402 Views
Summary of Use of Human Data in Cancer Risk Assessment of Chemicals as Illustrated by the Case of 1,3-Butadiene. Dr. Richard Albertini University of Vermont Burlington, VT. Dr. Robert L. Sielken Jr. Sielken & Associates Consulting, Inc. Bryan, TX. Beyond Science and Decisions:
E N D
Summary of Use of Human Data in Cancer Risk Assessment of Chemicals as Illustrated by the Case of 1,3-Butadiene Dr. Richard Albertini University of Vermont Burlington, VT Dr. Robert L. Sielken Jr. Sielken & Associates Consulting, Inc. Bryan, TX Beyond Science and Decisions: From Issue Identification to Dose-Response Assessment Workshop II Crystal City, VA October 11, 12, & 13, 2010 #23a
1,3-BUTADIENE INHALATION EXPOSURES IN HUMANS Approach to Cancer Risk Assessment Uses mechanistic data based on BD’s overall mode of action (MOA). Uses epidemiological data from human occupational exposures. Applies existing methodology for analysis of human cancer-risk from environmental (ambient) exposures. #23a
MODE OF ACTION (MOA) FOR CANCER BD’s MOA for tumor induction is direct DNA reactive mutagenicity of its METABOLITES: epoxybutene (EB), epoxybutane diol (EBD) and diepoxybutane (DEB), Mice are the most efficient (EB and DEB); rats much less so (and humans produce even less DEB than rats). The relative order of the mutagenic potencies of BD’s metabolites is : DEB* >>EB>EBD, with species’ susceptibility to mutagenesis (and cancer) following metabolism. * bi-functional DNA adducts, d.s. breaks in DNA, chromosome level mutations. #23a
BD’s MOA FOR CANCER:Allows analysis of underlying mutagenic mechanisms Different mutational events underlie different malignancies Animal tumors: Most studies have found point mutations in oncogenes Human malignancies: Tumor excesses have been in leukemias that are characterized by chromosomal level mutations. CML: t(9:22) translocation is the necessary and sufficient mutational event. CLL: Structural and/or numerical chromosomal mutations, possible gene mutations. AML: Chromosome and gene mutations required Objective of analysis: Avoid an endpoint that encompasses a variety of cancers with different mutagenic mechanisms. #23a
The problem being addressed is: The evaluation of risk from environmental (ambient) exposures to a chemical using epidemiological data from studies of human occupational exposures University of Alabama at Birmingham (UAB) epidemiological study of North American workers in the styrene-butadiene rubber industry Although the workplace cumulative exposures can be extrapolated to a safe environmental exposure level, the high intensity BD exposures and other co-exposures in the workplace would not be expected to occur in the ambient setting. Thus, to go from an occupational exposure study to an ambient standard, one needs to remove the contributions from the high intensity BD exposures and other co-exposures (exposure covariates). #23a
Statistical Methods for Quantitative Dose-Response Modeling and Excess Risk Characterization Used Cox proportional hazards models with continuous dose and categorical covariates Incorporated human background hazard rates that reflect human variability and knownage-dependent changes in hazard rates Incorporated observed exposure values for each person-year without requiring grouping Used likelihood based statistical techniques for parameter estimation, hypothesis testing, covariate selection, lag determination, and adjustment for high intensity exposures and co-exposures Used sensitivity analyses to evaluate exposure uncertainty Evaluated the incorporation of potentially correlated exposure variables #23a
Method for Quantitative Excess Risk Characterization Used mathematically correct actuarial life-table calculations to incorporate endpoint specific age-dependent background hazard rates, population age-dependent survival probabilities, and U.S. EPA guideline default age-dependent adjustment factors (ADAFs). Population background hazard rates reflected the human variability therein and supplemented the human variability in a large cohort of workers Estimated more relevant PODs like EC(1/100,000) rather than EC(1/10) because the human data are sufficient for an EC(1/100,000) to be in the midst of the observed data and not require overly model-dependent extrapolation Used mortality models to estimate mortality risks but did not estimate incidence risks based on fitted mortality models Used MOA information to guide decisions about extrapolation below PODs #23a
Quantitative Excess Risk Characterization Even though Sielken & Associates has used thebest available human data and the best available statistical methods and focused on MOA motivated endpoints (CML, CLL, and AML) rather than a diverse grouping of endpoints with different MOAs, the meaningfulness of the quantified excess risk values is still limited by policy restrictions. There may be no risk due to BD low-level environmental exposures: Humans don’t efficiently metabolize BD to the active metabolite DEB Occupational exposures with increased risk had high intensity exposures and also exposures to chemicals other than BD The required dose -- cumulative BD ppm-years – in the dose-response modeling may not be relevant for low-intensity environmental exposures. -- the slope for cumulative ppm-years is NOT statistically significantly positive for CML and AML Required excess risks were quantified assuming a non-threshold model -- modeling restricted to person-years with lower ppm-years implies threshold-like behavior for CLL Required risk characterizations were bounds instead of best estimates Upper bound potencies (like Q1*) and lower bounds (BMDLs) on the benchmark dose (BMD) are based on specialized statistical bounding procedures that are very insensitive to the shape of the dose-response data, and, hence, are a poor basis for comparing different chemicals and risk-management decision making. #23a
Thank You #23a
Backup Slides #23a
CASE STUDY;1,3-Butadiene (BD) Inhalation Exposures in Humans Uses mechanistic data based on BD’s overall mode of action (MOA). Uses epidemiological data from human occupational exposures. Applies existing methodology for analysis of human cancer-risk from environmental (ambient) exposures. #23a
MODE OF ACTION (MOA) FOR CANCER BD itself is BIOLOGICALLY INACTIVE. Its MOA for tumor induction is direct DNA reactive mutagenicity of its METABOLITES. These metabolites of BD are: epoxybutene (EB), epoxybutane diol (EBD) and diepoxybutane (DEB), produced in all species but in different amounts, with large inter-species differences. Mice are the most efficient in oxidative metabolism (EB and DEB); rats much less so (and humans producing even less DEB than rats). #23a
MUTAGENICITY DRIVES CARCINOGENICITY 1. The relative order of the mutagenic (and carcinogenic) potencies of BD’s metabolites is : DEB >>EB>EBD. 2. Species’ sensitivity to BD’s carcinogenicity follows the relative metabolic efficiencies of the species in producing mutagenic metabolites (especially DEB) and the relative mutagenic potencies of these metabolites, with mice being the most sensitive species and rats and humans the more resistant species. 3. The mutagenic and carcinogenic responses in rats and humans suggest a possible threshold response at low BD exposure concentrations. 4. BD exposures expressed as mutagenic equivalents show tumor induction per mutagenic equivalent to be the same in mice and rats. This relationship is likely to hold true for all species, including humans. #23a
DEB (1)The major mutagenic metabolite of BD Forms bi-functional adducts in DNA forming DNA-DNA and DNA-protein cross-links. Strongly mutagenic inducing DNA double -strand breaks that underlie chromosome level mutations (aberrations). DEB concentrations in mice and rats are approximately the same at 3.0 ppm and 62.5 ppm BD inhalation exposures respectively, which are also the LOAELs for somatic mutation induction in the two species, respectively. Although chromosome level mutations have repeatedly been demonstrated in mice exposed to BD, they have never been demonstrated in rats receiving the parent compound, indicating that rats make insufficient amounts of this clastogenic metabolite. #23a
DEB (2)The major mutagenic metabolite of BD 5. DEB accumulations in humans were not quantifiable using a method that did quantify them in mice and rats at BD inhalation exposure levels of 3.0 and 62.5 ppm, respectively. 6. Several large scale occupational studies of BD-exposed workers, although detecting urinary excretion products and EB and EBD derived hemoglobin adducts, have failed to document consistent associations between BD exposures and gene or chromosomal level mutations. 7. Different metabolic genotypes have shown some differences in metabolism in humans but none have indicated genetic susceptibilities to mutagenic effects. #23a
BD’s MOA FOR CARCINOGENICITY:Allows analysis of underlying mutagenic mechanisms Different mutational events underlie different malignancies Animal tumors: Most studies have found point mutations in oncogenes Human malignancies: Tumor excesses have been in leukemias that are characterized by chromosomal level mutations. (Point mutations may be produced by single hits and arise proportional to dose at higher doses, chromosome mutations such as translocations and interstitial deletions require two double strand breaks in the DNA making their occurrence proportional to dose-squared ). #23a
MUTAGENIC MECHANISMS PROVIDE A METHOD FOR CANCER ENDPOINT SELECTION Human Leukemia: • CML: t(9:22) translocation is the necessary and sufficient mutational event. • CLL: Structural and/or numerical chromosomal mutations, possible gene mutations. • AML: Chromosome and gene mutations required Objective of analysis: Avoid an endpoint that encompasses a variety of cancers with different mutagenic mechanisms. #23a
The problem being addressed is: The evaluation of risk from environmental (ambient) exposures to a chemical using epidemiological data from studies of human occupational exposures The current (as well as previous) dose-response modeling and quantitative characterization of environmental excess risks have been hampered by the restrictions that (1) require that environmental excess risks be calculated from cumulative BD ppm-years as opposed to a potentially more biologically-based dose metric; (2) require the use of non-threshold models even though modeling restricted to person-years with lower ppm-years suggest a threshold-like behavior for CLL and the absence of a statistically significant positive slope for cumulative ppm-years for CML and AML; and (3) require upper bounds on risks and lower bounds on points of departure (PODs) instead of only more scientifically defensible best estimates. Upper bound potencies (like Q1*) and lower bounds (BMDLs) on the benchmark dose (BMD) are based on specialized statistical bounding procedures that are very insensitive to the shape of the dose-response data, and, hence, are a poor basis for comparing different chemicals and risk-management decision making. #23a
The problem being addressed is: The evaluation of risk from environmental (ambient) exposures to a chemical using epidemiological data from studies of human occupational exposures University of Alabama at Birmingham (UAB) epidemiological study of North American workers in the styrene-butadiene rubber industry Although the workplace cumulative exposures can be extrapolated to a safe environmental exposure level, the high intensity BD exposures and other co-exposures in the workplace would not be expected to occur in the ambient setting. Thus, to go from an occupational exposure study to an ambient standard, one needs to remove the contributions from the high intensity BD exposures and other co-exposures (exposure covariates). #23a
2.2 Method for Dose-Response Modeling Used models that support quantifying risks. Used Cox proportional hazards models with continuous dose and categorical covariates. Proportional hazards models enabled excess risks to be calculated from human background hazard rates that reflect human variability and known changes in hazard rates with age. Cox models with continuous dose allowed the observed exposure values for each person-year to be directly incorporated without requiring grouping -- since grouping can distort the fitted dose-response relationship. Grouping can sometimes cause severe problems with Poisson modeling as well as RR and SMR analyses – especially when the number of groups is small. Likelihood based statistical techniques are used for parameter estimation, hypothesis testing, covariate selection, lag determination, and adjustment for high intensity exposures and co-exposures #23a
2.2 Method for Dose-Response Modeling (Continued) The dose-response modeling is adjusted for other exposure covariates in the occupational setting in order to make the model more relevant for estimating environmental risks. Understanding the relative contributions of high-intensity exposure events and cumulative exposure is extremely important when attempting to predict cancer risk associated with environmental exposures, where high-intensity exposure events are not expected to occur. Use sensitivity analyses to evaluate exposure uncertainty and avoid unnecessary data restrictions. Techniques for evaluating the incorporation of potentially correlated exposure variables are illustrated. #23a
2.3 Method for Quantitative Excess Risk Characterization Use mathematically correct life-table calculations to incorporate endpoint specific age-dependent background hazard rates, population age-dependent survival probabilities, and U.S. EPA guideline default age-dependent adjustment factors (ADAFs). BEIR IV Life-Table Methodology: National Research Council (NRC). 1988. Health risks of radon and other internally deposited alpha- emitters. Committee on the biological effects of ionizing radiation. Biological effects of ionizing radiation IV (BEIR IV). Washington DC: National Academy Press. Sielken, Robert L., Jr., and Ciriaco Valdez-Flores (2009). Life-table calculations of excess risk for incidence versus mortality: Ethylene oxide case study.” Regulatory Toxicology and Pharmacology, Vol. 55, pp. 82-89. Sielken, Robert L., Jr. and Ciriaco Valdez-Flores (2009). Calculating excess risk with age-dependent adjustment factors and cumulative doses: Ethylene oxide case study. Regulatory Toxicology and Pharmacology , Vol. 55: 76-81. Actuarial life-table calculations: Sequentially look at age 1, 2, 3, ... Incorporate age-specific inputs: Cause-specific survival probability (background cause-specific hazard rate) Competing risk survival probability (background mortality rate – other causes) Dose (age-specific dose, cumulative dose, lagged dose, dose in window of exposure) With or Without Age-Dependent Adjustment Factors (age-dependent sensitivity) Calculate Added or Extra Risks at different ages at Specified Doses Calculate Doses with Specified Added or Extra Risks at Specified Ages Multiplicative or Additive Proportional Hazards Dose-Response Models Publically available implementation from Sielken & Associates in Microsoft Excel #23a
2.3 Method for Quantitative Excess Risk Characterization (Continued) Incorporation of population background hazard rates includes the human variability therein and supplements the human variability in a large cohort of workers. Estimate more relevant PODs like EC(1/100,000) rather than EC(1/10) when the human data are sufficient for an EC(1/100,000) to be in the midst of the observed data and not require overly model-dependent extrapolation. Use mortality models to estimate mortality risks, use incidence models to estimate incidence risks, but do not use mortality models to estimate incidence risks. Use MOA information to guide decisions about model shape (linear, quadratic, etc.) and extrapolation below PODs. #23a
2.3 Method for Quantitative Excess Risk Characterization (Continued) Historically, cancer risks attributed to the formation of gene mutations have been assumed to be linearly proportional to dose at low doses. Although valid when point mutations are the initiating events, it is not valid when other kinds of mutations, such as large deletions or stable balanced chromosome translocations or interstitial deletions are the initiating events. These require two double strand-breaks arising in two different chromosome regions in the same cell during the same wave of DNA replication to produce the deletions, stable chromosome translocations and/or interstitial deletions that permit the cell to live. The requirement that 2 critical events occur in close proximity in both space and time leads to a quadratic dose-response relationship. Generally, linear is more associated with gene mutations at high doses beyond a threshold, whereas quadratic is more associated with chromosomal aberrations. A third type of chromosome aberration, i.e. numerical aberration with chromosome gain or loss may also be a threshold event due to disruptions of protein mitotic spindles. The form of the “dose component” of the dose-response model would be: c + a(D- τ) + b(D-τ)2, where c = background risk, a = linear term (point mutations), b = quadratic term (two double strand breaks), τ = threshold term, and D = dose. For a malignancy such as CML, the quadratic term would be expected to predominate. #23a
Quantitative Excess Risk Characterization Even though Sielken & Associates has used the best available human data and the best available statistical methods and focused on MOA motivated endpoints (CML, CLL, and AML) rather than a diverse grouping of endpoints with different MOAs, the meaningfulness of the quantified excess risk values is still limited by policy restrictions. There may be no risk due to BD low-level environmental exposures: Humans don’t efficiently metabolize BD to the active metabolite DEB. Occupational exposures with increased risk had high intensity exposures and also exposures to chemicals other than BD. The required dose -- cumulative BD ppm-years – in the dose-response modeling may not be relevant for low-intensity environmental exposures. Required excess risks were quantified assuming a non-threshold model. Required risk characterizations were upper bounds on risks and lower bounds on points of departure (PODs) instead of more scientifically defensible best estimates. #23a