M22 - Symposium
Recent Insights in Understanding and Expressing Health Risks Posed by Dioxin and Dioxin-Like Chemicals, Part 2Scotland A 10:30 am-Noon
|
| Chair(s): T. Long, R. Tardiff
|
Despite considerable effort made over the past 30 years, the scientific, regulatory, and advocacy communities remain divided over the risk posed by dioxin-like chemicals to humans and the means to best assess and express this risk. The purpose of this symposium is to bring together some of the latest research and scientific information on the human risks from dioxin-like chemicals and methods to determine the magnitude of those risks.
While several risk assessments for the class of dioxins exist (including the massive effort by USEPA), much new science concerning dioxin-like chemicals continues to emerge along with new risk assessment methods and techniques useful for more accurately representing the magnitude of risks from these chemicals. Such information needs to be considered in the scientific and regulatory discourse that surrounds this class of chemicals in order to improve current and future risk assessments of dioxin-like compounds. This symposium presents some recent advances in understanding and evaluating risks from dioxin-like compound that may significantly influence and improve regulatory policy decisions. The topics to be covered include issues in cancer and non-cancer dose-response modeling, Toxicity Equivalency Factors (TEFs), applications of probabilistic modeling of cancer potency factors, pharmacokinetic and dosimetry issues associated with cancer risk, and recent regulatory policy changes that should be considered in dioxin risk assessments.
|
M22.1 Benchmark Dose Methods for Non-Cancer Continuous Data Health Effects in Animals Exposed to Dioxins. Gaylor D.W.*, Aylward L.L.; Gaylor and Associates LLC and Exponent tfl@thesapphiregroup.com
Abstract: A lower confidence limit on a benchmark dose (BMDL) can be used to identify a point-of-departure (PoD) for use in derivation of a reference dose. In a reassessment of dioxin health effects, the U.S. Environmental Protection Agency based BMDs on changes from the background levels of effects estimated to equal 1% of the asymptotic maximum changes from background at large doses estimated from the Hill equation. This somewhat arbitrary effect level is denoted as the relative ED01 that generally is within the background range of variation among unexposed animals and provides no estimate of risk. An alternative approach is recommended in which the excess risk of 1% of animals with abnormal levels can be estimated at the BMD, denoted as the BMD01. Values for the relative (ED01) and risk-based (BMD01) and their lower 95% confidence limits, LED01 and BMDL01, are estimated from published data for a variety of non-cancer effects associated with exposure to dioxin. For the 30 cases of dose-response data investigated, the BMD01 tended to vary around the lowest dose of dioxin tested, whereas the ED01 tended to be about a factor of three below the lowest dose. The BMD01 was more precisely estimated as reflected by narrower confidence limits. BMDL01 values were, on average, more than fivefold higher than the corresponding LED01. The risk-based BMDL01 still provides a conservative choice for the PoD because the BMDL01 tends to be about tenfold lower than the no-observed-effect-level.
|
M22.2 Getting beyond TEFs for dioxin-like compounds: data quality, variability, and uncertain biological relevance . Starr T.B.*; TBS Associates tfl@thesapphiregroup.com
Abstract: A comprehensive testing program with improved experimental designs should be initiated to replace “interim” TEFs eventually with relevant data. Meanwhile, we can progress beyond the current TEF approach. The “dioxin-like” concept should be enlarged to include as many AhR ligands as possible, including “endodioxins”, and alternative dose metrics (body burden, target organ concentrations) also need to be explored. Validity of the parallel dose-response assumption should be assessed statistically, and the power to reject it when it’s absent should be specified. If non-parallelism prevails, the focus should shift to estimating relative potency (REP) near the bottom of the observable response range using models allowing for non-parallelism. All REPs need a common, risk-based definition. Hill modeling has yielded very inconsistent results, in part because response definitions differ markedly across endpoints. Continuous responses could be deemed adverse if they exceed the 99th percentile response in control animals, with REPs defined as ratios of EC10s (the doses that increase the probability of an adverse response from 1% to 11% (added risk = 10%). REP point estimates should give way to REP sampling distributions, which better reflect variability and uncertainty in the underlying data. These distributions should be assessed statistically for heterogeneity, and REPs too dissimilar to be mutually consistent should not be combined. An explicit rationale should be provided for weights used to combine REPs into TEFs. Inverse variance weighting could be used initially, and generalized later to reflect the uncertain biological relevance of different endpoints to specific human health risks.
|
M22.3 Probability Distributions for Carcinogenic Potencies. Crouch E.A.C.*; Cambridge Environmental Inc tfl@thesapphiregroup.com
Abstract: Probabilistic risk assessments provide information that is not available using the usual point estimate approach, including evaluation of variability and uncertainty and of population estimates of risk and their uncertainties in addition to individual estimates. Complete probabilistic cancer risk estimates require probability distributions for uncertainty and variability in cancer potencies. I here provide a consistent, data based, objective methodology for evaluating such distributions, based on animal carcinogenicity experiments, epidemiological investigations, or both combined. I begin with evaluation of single animal experiments, including selection of classes of dose-response shapes for fitting observations and extrapolating to low dose. I extend standard asymptotic likelihood methods to evaluation of full distributions, and illustrate an empirical likelihood with preferable characteristics. Thresholds in the dose-response curve are incorporated by suitably extending the class of dose-response curves. Multiple animal experiments on the same chemical are combined using random-effects models to account for intra-strain, inter-strain, and inter-species variation, and extrapolating an interpretation of such variability to the context of human risk assessment. Prior distributions for cancer potency and inter-species extrapolation are demonstrated based on similar evaluations of large numbers of chemicals, using as an ensemble all chemicals tested for carcinogenesis. The same approach applies to analysis of epidemiological data, although often with different limitations. I discuss these differences, and show how to consistently combine epidemiological evidence with animal data in a probabilistic framework. Finally, I point out that there is nothing unique to the evaluation of distributions; any scientifically based point estimate requires the same information.
|
M22.4 Applying Congener-Specific Cancer Potency Factors and Probability Distributions in Estimating the Risk of a Dioxin Mixture . Kirman C.R., Ferson S.D.*, Long T.F.; The Sapphire Group Inc., Applied Biomathematics, Setauket, NY, Cambridge Environmental Inc. tfl@thesapphiregroup.com
Abstract: The Toxicity Equivalency Factor (TEF) approach for assessing complex mixtures of dioxins and furans has been widely embraced over the last 20 years because of its ease of application, endorsement by regulatory authorities, and lack of alternatives. There remain questions, however, about how accurately the current TEF approach represents the risk posed from exposure to dioxin/furan mixtures. There is also no specific recommendation for exiting the TEF approach now that better, more relevant data are available; consequently, congener-specific toxicity data that would normally be used to assess risk may be ignored in preference to this weighting scheme. To decrease the uncertainty in assessing the potential potency of a mixture it is important to use congener-specific data when available. For several dioxin and furan congeners there are now sufficient data to allow direct assessment outside the TEF approach. We evaluated cancer potency factors (CPFs) for 2,3,4,7,8-pentachlorodibenzofuran and 1,2,3,6,7,8- and 1,2,3,7,8,9-hexachlorodibenzodioxin since there are sufficient carcinogenicity data available for these congeners. A brief review of the data, the independent CPFs and probability distributions developed for these compounds will be given, together with an example analyzing data sets that contain mixtures of dioxins and furans. A comparison between the results of a risk assessment using the current TEF approach and these congener-specific CPFs and probability distributions will be provided to illustrate the potential differences in outcome and the importance of developing and using data-derived CPFs and probability distributions for specific dioxin and furan congeners whenever possible.
|
